Evaluation of new alternative strategies to predict ...

Evaluation of new alternative strategies

to predict neurotoxicity with human based

test systems

Dissertation

zur Erlangung des akademischen Grades eines

Doktors der Naturwissenschaften (Dr. rer. nat.)

vorgelegt von

Anne-Kathrin Krug

an der

Mathematisch-Naturwissenschaftliche Sektion

Fachbereich Biologie

Tag der mündlichen Prüfung: 17.12.2013

1. Referent: Prof. Dr. Marcel Leist

2. Referent: Prof. Dr. Dr. Thomas Hartung

Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-257433

http://nbn-resolving.de/urn:nbn:de:bsz:352-257433

List of publications

I

List of publications

Publications integrated in this thesis:

Results Chapter 1: Krug AK, Stiegler NV, Matt F, Schönenberger F, Merhof D, Leist M

(2013) Evaluation of a human neurite growth assay as specific screen for developmental

neurotoxicants. Accepted (2. May) in Arch Toxicol

Results Chapter 2: Krug AK, Kolde R, Gaspar JA, et al. (2013) Human embryonic stem

cell-derived test systems for developmental neurotoxicity: a transcriptomics approach. Arch

Toxicol 87(1):123-43

Results Chapter 3: Krug AK, Zhao L, Kullmann C, Pöltl D, X, Ivanova V, Förster S,

Jagtap S, Meiser J, Gutbier S, Léparc G, Schildknecht S, Adam M, Hiller K, Farhan H,

Brunner T, Hartung T, Sacchinidis A, Leist M. Transcriptional and metabolic adaptation of

human neurons to the mitochondrial toxicant MPP+. Under review

Publications not integrated in this thesis:

Stiegler NV, Krug AK, Matt F, Leist M (2011) Assessment of chemical-induced

impairment of human neurite outgrowth by multiparametric live cell imaging in high-density

cultures. Toxicol Sci 121(1):73-87

Schoenenberger F, Krug AK, Leist M, Ferrando-May E, Merhof D (2012) An Advanced

Image Processing Approach based on Parallel Growth and Overlap Handling to Quantify

Neurite Growth. Paper presented at the 9th International Workshop on Computational

Systems Biology (WCSB), Ulm

Schildknecht S, Karreman C, Pöltl D, Efremova L, Kullmann C, Gutbier S, Krug AK,

Scholz D, Gerding H, Leist M. Generation of genetically-modified human differentiated cells

for toxicological tests and the study of neurodegenerative diseases. ALTEX 2013 Jun 7

Sisnaiske J, Hausherr V, Krug AK, Zimmer B, Hengstler J, Leist M, van Thriel C.

Specific neurofunctional disturbances triggered by acrylamide in ESC-derived and primary

neurons.

AND

Hausherr V, van Thriel C, Krug, AK, Leist M, Schöbel N. Neurotoxic effects of tri-o-cresyl

phosphate (TOCP) in vitro – a comparison of functional and structural endpoints.

Submitted to the Special Issue of NeuroToxicology devoted to the Proceedings of INA-14

Oral and poster presentations

II

Oral and poster presentations

Oral presentations:

14th International Neurotoxicology Association meeting (INA14),

Neurodevelopmental Basis of Health and Disease, Egmond aan Zee, The Netherlands, 09-13

June 2013 [David Ray Award for best student talk : Integrating transcriptomics and

metabolomics to identify new pathways of toxicity of the parkinsonian toxin MPP+]

Organized conferences:

Insel-Symposium 2012 – Biomedical Research and Scientific Careers, Konstanz,

Germany, 14-15 June 2012 – Team leader of the organization committee (Graduiertenschule

RTG1331) http://www.inselsymposium.uni-konstanz.de/

Poster presentations:

European Congress on Alternatives to Animal testing, European Society for

Alternatives to Animal Testing (EUSAAT), Linz, Austria, 04-06 September 2012 [Poster:

Evaluation of assay requirement to detect specific neurotoxicants in a human cell-based test]

Society of Toxicology (SOT) 2012, San Francisco, USA, 11-15 March 2012 [Poster:

Evaluation of assay requirement to detect specific neurotoxicants in a human cell-based test]

Third International Conference on Alternatives for Developmental Neurotoxicity

(DNT) Testing, Varese, Italy, 10-13 May 2011 [Poster: Detection of toxicants that

specifically impair spontaneous neurite outgrowth in live human neural precursor cells]

http://www.inselsymposium.uni-konstanz.de/

Table of contents

III

Table of contents

A. Summary ................................................................................................................. 4

Zusammenfassung .............................................................................................................. 5

B. General introduction Toxicity testing in the 21st century – of man and animals ......................................... 8

Cytotoxicity in Toxicology ......................................................................................... 9

Challenging test systems by correct compound selection ........................................ 12

High-throughput and high-content screening ........................................................... 13

High-content imaging (HCI) ................................................................................. 15

Toxicogenomics .................................................................................................... 19

Applications of HCI and toxicogenomics in toxicology ....................................... 24

Applications of HCI and toxicogenomics in neurotoxicology.............................. 26

Aims of the thesis ............................................................................................................. 32

C. Results Chapter 1 Evaluation of a human neurite growth assay as specific screen for developmental

neurotoxicants .................................................................................................................. 33

Abstract ..................................................................................................................... 34

Introduction .............................................................................................................. 35

Results and Discussion ............................................................................................. 38

Materials and Methods ............................................................................................. 57

Supplements .............................................................................................................. 60

D. Results Chapter 2 Human embryonic stem cell-derived test systems for developmental neurotoxicity: a

transcriptomics approach ................................................................................................. 64

Abstract ..................................................................................................................... 66

Introduction .............................................................................................................. 67

Results and Discussion ............................................................................................. 70

Materials and Methods ............................................................................................. 89

Supplements .............................................................................................................. 97

E. Results Chapter 3

Transcriptional and metabolic adaptation of human neurons to the mitochondrial

toxicant MPP+ ................................................................................................................ 111

Abstract ................................................................................................................... 112

Introduction ............................................................................................................ 113

Results .................................................................................................................... 116

Discussion ............................................................................................................... 129

Material and Methods ............................................................................................. 132

Supplements ............................................................................................................ 142

F. Concluding discussion......................................................................................... 150

G. Bibliography ........................................................................................................ 162 Record of contribution ................................................................................................... 175

Summary

4

A. Summary

Animal experiments are still the ‘gold standard’ in safety evaluation defined by the

OECD (Organisation for Economic Co-operation and Development) or the US EPA

(Environmental Protection Agency). Millions of animals are used each year to assess the risk

of chemical toxicities for human health. But animal experiments are expensive, time-

consuming and have a restricted prediction capacity regarding human toxicity. Hence the

demand for validated alternative strategies is high. Validated differentiation protocols of

embryonic stem cells or immortalized human organ specific cell lines provide the possibility

to recapitulate human development and to study organ specific toxicity of different

developmental stages (immature to mature) in vitro. In the framework of this doctoral thesis,

we provide insights into the development and evaluation of test systems established

specifically to assess neurodevelopmental toxicity as well as neurotoxicity in vitro.

In a first step we evaluated an assay based on neurite outgrowth assessment to detect

putative developmentally neurotoxic chemicals. This assay was based on a human

mesencephalic neuronal precursor cell line, called LUHMES. In the study, the model has been

challenged for its reliability and consistency using more than 50 compounds and

combinations of them. We proved the applicability of the assay for screening, and suggest that

the test has the potential to be used for identification and potency-ranking of putative

developmental toxicants with regard to effects on neurite growth.

In a second step we used different human stem cell-based test systems to mimic several

stages of the early human neurodevelopment in vitro. We analysed the transcriptome changes

of these test systems after exposure to two developmental toxicants, valproic acid and

methylmercury. Both toxicants induced test system and compound specific transcriptome

changes. A common toxicant specific signature of transcription factor binding sites was

identified for the different test systems, which we suggest as classifier for compound grouping

in future experiments.

In a last step we used a well described model compound 1-methyl-4-phenylpyridinium

(MPP+) to analyse the suitability of Omics combinations to monitor the MPP+ induced

changes on LUHMES. We found early large adaptive metabolome and transcriptome changes

which taken together lead to the identification of novel pathways involved in early MPP+

toxicity. The findings of this thesis contribute to alternative test-strategy development in

neurotoxicity and disclose important considerations when developing in vitro test systems.

Zusammenfassung

5

Zusammenfassung

Tierversuche sind nach wie vor der Goldstandard für die Sicherheitsbewertung, die von

der OECD bzw. der US EPA vorgeschrieben wird. Millionen Tiere werden jedes Jahr

benötigt, um die Gefahr von chemischen Substanzen für die menschliche Gesundheit

abzuschätzen. Aber Tierversuche sind teuer, zeitintensiv und haben eine beschränkte

Voraussagekraft bezüglich menschlicher Toxizität. Daher ist die Nachfrage an validierten

Alternativstrategien hoch. Differenzierungsprotokolle embryonaler Stammzellen oder

humaner Organ spezifischer Zelllinien, ermöglichen die Rekapitulation humaner Entwicklung

und die Untersuchung Organ spezifischer Toxizität während unterschiedlicher

Entwicklungsstadien in vitro. Im Rahmen dieser Doktorarbeit bieten wir Einblick in die

Entwicklung und Evaluierung von Testsystemen, die spezifisch für die Untersuchung

neuronaler Entwicklungstoxizität und Neurotoxizität in vitro hergestellt wurden.

Im ersten Teil dieser Arbeit galt es einen Assay zu bewerten, der auf die Untersuchung

von Neuritenwachstum ausgelegt ist, um mögliche neuronale Entwicklungstoxikantien zu

identifizieren. Der Assay basiert auf humanen neuronalen Vorläuferzellen, den LUHMES.

Das vorgelegte Modell wurde auf seine Verlässlichkeit und Konsistenz getestet, indem mehr

als 50 verschiedene Substanzen und Kombinationen dieser eingesetzt wurden. Wir zeigen die

Anwendbarkeit dieses Assays für Screenings und untermauern sein Potenzial für die

Identifikation möglicher Entwicklungstoxikantien in Hinsicht auf gestörtes

Neuritenwachstum.

Im zweiten Teil verwendeten wir Testsysteme, basierend auf humanen Stammzellen, um

verschiedene Stadien der frühen humanen neuronalen Entwicklung in vitro darzustellen. Wir

analysierten die Veränderungen auf Transkriptionsebene nachdem die Test-Systeme zwei

Entwicklungstoxikantien, Methylquecksilber und Valproinsäure, ausgesetzt waren. Beide

Substanzen induzierten testsystem- und substanzspezifische Veränderungen in der

Transkription. Eine substanzspezifische und Testsystem übergreifende Signatur in

Transkriptionsfaktor-Bindestellen wurden identifiziert, die wir als Klassifikator für zukünftige

Experimente vorschlagen um ähnliche Substanzen zu gruppieren.

Im letzten Schritt benutzten wir die viel beschriebene Substanz 1-Methyl-4-

Phenylpridinium (MPP+) um die Anwendbarkeit von Omics-Kombinationen zu analysieren

und um die MPP+-induzierten Veränderungen in LUHMES zu überprüfen. Wir beobachteten

erhebliche Anpassungen auf metabolischer und transkriptioneller Ebene, die

Zusammenfassung

6

zusammengenommen zur Identifikation neuer Reaktionswege führten, die in die MPP+-

Toxizität involviert sind. Die Erkenntnisse, die aus dieser Arbeit gewonnen wurden, tragen

zur Entwicklung von alternativen Teststrategien bei und weisen auf wichtige Ansichten hin,

wenn solche in vitro Systeme entwickelt werden.

Abbreviations

7

Abbreviations

ASD Autism spectrum disorders

DNT Developmental neurotoxicity

EC50 Half maximal effective concentration

FDR False discovery rate

GO Gene ontology

GSH Glutathione

HCS High-content screening

hESC Human embryonic stem cells

HTS High-throughput screening

MeHg Methylmercury

MoA Mode of action

MPP+ 1-methyl-4-phenylpyridinium

NT Neurotoxicity

OECD Organization for economic co-operation and development

Omics Generic term used for e.g. transcriptomics, metabolomics

PCA Principal component analysis

PoT Pathways of toxicity

REACH Registration, evaluation, authorisation and restriction of chemicals

SoT Signatures of toxicity

TFBS Transcription factor binding site

VPA Valproic acid

General introduction

8

B. General introduction

Toxicity testing in the 21st century – of man and animals

Toxicological profiling of chemicals for use in drugs, food and cosmetics is strongly

dependent on animal experiments. Most guidelines of the OECD for toxicological risk

assessment dictate these experiments for safety evaluation (e.g. OECD guidelines for

reproductive toxicity studies (No. 443), chronic toxicity studies (No. 452), acute inhalation

studies (No. 436), developmental neurotoxicity studies (No. 426) and more). About three

million animals are used each year in Germany for experimental purposes

(http://bit.ly/UHErzE), over 115 million worldwide (Taylor et al 2008). These numbers

highlight the dependency of research on animals for human safety. But fact is - animals are

different. They differ not only in size (“we are no 70 kg rats” (Hartung 2009)), social

behaviour or life-span, but also in their development, metabolism or immune response (Leist

& Hartung 2013, Seok et al 2013). For example 89% of new chemical entities developed in

pharmaceutical industries fail in the clinical trial. Out of these 11% produce human adverse

effects and 8% fail because of differences in the pharmacokinetics of animals and humans

(McKim 2010). Some pharmaceuticals manage to get on the market because animal tests

predicted them misleadingly safe for humans. Such as Zimeldine, an antidepressant, which

was released in 1983. One year after release, it was withdrawn because of severe neurological

side effects (Nilsson 1983). Thalidomide, released in the late 50s, lead to the contergan-

scandal, resulting in several thousand births of malformed children (Newman 1986). In 2006,

TGN1412, an antibody developed to treat multiple sclerosis, lead to severe side effects in the

clinical trial phase I, where all treated men developed a cytokine storm – an overshoot of the

immune system which can lead to multi organ failure. In macaques and mice no such reaction

was observed, as the protein targeted by the antibody has minor amino acid sequence

differences (Attarwala 2010). Not surprisingly, these are only a few examples of the many

found in literature. One famous quote by Hans Ruesch nicely recapitulates the differences

between humans and animals (Ruesch 1982):"Two grams of scopolamine kill a human being,

but dogs and cats can stand hundred times higher dosages. […] Morphine, which calms and

anesthetizes man, causes maniacal excitement in cats and mice. On the other hand our sweet

almond can kill foxes, our common parsley is poisonous to parrots, and our revered penicillin

strikes another favourite laboratory animal dead - the guinea pig." The absence of toxicity in

one species therefore does not necessarily mean that tests within another species would lead

to the same outcome and vice versa – a positive compound could still be absolutely harmless

http://bit.ly/UHErzE


9

for humans (Hartung & McBride 2011). The low predictive capability of animal experiments

underlines the obvious need for a change in toxicological hazard assessment. Administering

increasing doses of chemicals to animals until they drop dead is not only not always relevant

for humans, as mentioned above, it also doesn’t explain why the animals die. In addition, in

times of REACH (registration, evaluation, authorization, and restriction of chemicals), where

68000 chemicals have to be tested for their toxicological properties, the number of animals is

estimated up to 54 million and testing costs of 9.5 billion Euro (Rovida & Hartung 2009).

Therefore, many toxicologists demand a shift from these top-down approaches (phenotypical

analysis of animals, e.g. death) towards more mechanistically-based bottom-up approaches

(studying the mechanisms behind), which means a complete rethinking of safety evaluation

(Krewski et al 2010). One possible alternative is to develop human-based test systems, and

validate their prediction capacity by combining them with bioinformatic analysis and

modelling. These approaches should be used for prioritizing chemicals in a first step, to

reduce the number of chemicals to be tested in animals, and finally in replacing animal

testing. Thereby the 3R-principle (3R = replace, refine, reduce) of Russell and Burch provides

the underlying basis in achieving this (Russell & Burch 1959). In the next chapters it will be

discussed how these bottom-up approaches look like, how they should be implemented and

examples of studies will be listed, which applied these approaches with special emphasis on

neurotoxicology-related aspects.

Cytotoxicity in Toxicology

Considering the human body, there are essential intra- and inter-cellular processes, which

build-up the whole organism. Inter-cellular processes, such as receptor-ligand mediated

reactions, are vitally important. Nevertheless, most of these essential biological pathways

originate or end inside cells. A multicellular organism could therefore be split into its several

organ specific cell types and every cell type could be considered as an independent entity with

its own unique pathways. Those cells are linked to the other cell types and matrix by several

intercellular interactions. Toxicologists make use of this concept and develop and improve

cell culture conditions in order to obtain organ specific systems. A good example for such an

integrated strategy is to create a “human on a chip”, which is the connection of several organ

specific culture systems to mimic a human body (Fig. 1; (Hartung & Zurlo 2012, Huh et al

2011, Marx et al 2012)). Generating large batteries of alternative tests, mapping more or less

the total human body functions is important but has to be performed very carefully. The more

tests are needed to mimic the human body, the more likely a compound will be predicted as a


10

false-positive (Basketter et al 2012). Therefore the current focus of toxicologists is to

establish fast and high-content screenings, to define a chemical not solely by phenotypic

testing but including studies to determine the mechanism behind the phenotype. Once tests

have been evaluated sufficiently, integrating testing strategies have to be developed to avoid

the generation of many false-positives (Hartung et al 2013).

Figure 1: Human-on-a-chip model Modified from (Marx et al 2012). Outlook onto a possible design for maintaining human organ

equivalents in a common blood vasculature on a chip. One part with organ equivalents is displayed

Those can possibly be connected to compartments for nutrition, bile provision to the intestine, urine

and feces removal systems and a sensor layer to control e.g. pO2, pH and temperature.

In contrast to animal-based toxicity measurements, toxicologists want to use in vitro-

based systems mainly to identify perturbations from healthy state by

understanding the toxic mechanism of a chemical ((Blaauboer et al 2012),

http://www.epa.gov/nheerl/articles/2011/Chemical_Safety_Assessments.html). Several

attempts were made to analyse the correlation of acute poisoning in humans (human LC50

values) with acute cytotoxicity in different cell lines (IC50 values) (Ekwall et al 1998a,

Ekwall et al 1998b, Sjostrom et al 2008). In one study, estimated human LC50 values were

compared to in vitro IC50 values, resulting in a correlation of R²=61% for 66 chemicals

(Sjostrom et al 2008). The in vitro IC50 data was based on the mouse fibroblast cell line 3T3

and the neutral red uptake (NRU) assay. According to the authors this correlation relates to

the similar R2 of about 0.55–0.70 that is given when animal in vivo data is used to predict

human toxicity (Sjostrom et al 2008). Nevertheless, comparing in vitro rodent data (3T3) with

in vivo rodent LC50 data, only a slightly better correlation of R²=0.75 was found (Clothier et

al 1987). The analysis of acute cytotoxicity may therefore provide a very simple tool to


11

estimate acute poisoning concentrations for humans, but additional testing is essential to

improve the predictive capacities of in vitro tests, especially for chronic toxicity. In particular,

damage on vital organs, which does not result in cell death, will not be identified. A few

examples, which will not be detectable in an acute cytotoxicity assay, are listed below:

Disturbed signal transmission, such as impaired synaptic transmission. Several

chemicals and drugs are known to induce brain seizures, for example anesthetics (e.g.

cocaine, (Zimmerman 2012)) or anticholinergics (e.g. atropine (Glatstein et al 2013)).

Altered hormone signalling. Chemicals interfering with hormone signalling are

classified as endocrine disruptors and are linked to severe adverse outcomes, such as

tumours, birth defects and developmental disorders (Colborn et al 1993).

Damage on mtDNA (needs several cell divisions to result in toxic outcome). This

effect is for example known for reverse transcriptase inhibitors, used for HIV

treatment (Brinkman et al 1998) and can lead to neuropathy (Canter et al 2010).

Metabolism-dependent toxicity. Several chemicals or drugs need to be metabolized to

oppose a risk for human safety, such as the mycotoxin aflatoxin B1, or to become non-

toxic, such as the antihistamine terfenadine (Li 2009).

Pharmacokinetic differences: a compound might kill brain cells in vitro, but not in

vivo, because it doesn’t reach the brain. Vice versa, a compound might be ineffective

in vitro, because it dissolves poorly, evaporates quickly, or because it is close to its

nominal concentration in vitro, while it accumulates very strongly in one tissue in

vivo. Saccharin, for example, can form crystals in the bladder and therefore cause

bladder cancer, while it is completely innocuous in vitro

Another concept, also based on pure cytotoxicity assessment, is to compare different cell

lines to predict for organ specific toxicity. Unfortunately, no study was able to correctly

classify compounds for organ specificity on the basis of cell death (Gartlon et al 2006, Halle

2003, Lin & Will 2012). For instance, Lin and Will used 273 hepatotoxic compounds, 191

cardiotoxic compounds, 85 nephrotoxic compounds, and 72 compounds with no reported

organ toxicity. They tested the cytotoxicity potential of these compounds in organ-specific

culture systems (HepG2 cells (hepatocellular carcinoma), H9c2 cells (embryonic

myocardium), and NRK-52E (kidney proximal tubule cells). Finally, they concluded that the

cell lines had “relatively equal value in assessing general cytotoxicity” and that “organ

toxicity cannot be accurately predicted using such a simple approach”. Another study tested


12

1353 compounds in 13 human and rodent cell lines and obtained cytotoxicity profiles for 428

compounds (Xia et al 2008). Although some lead to the same EC50 values in the different cell

lines, other compounds resulted in different responses. These differences were not related to

organ specificity. Furthermore, cells of the same species and same tissue showed considerable

differences, as demonstrated in the study by the neuroblastoma line SK-N-SH and its

derivative line SH-SY5Y (Xia et al 2008). It remains elusive why these differences appeared.

Classification of organ specific toxicants will become difficult if two similar cell lines

respond so different to the same substance/noxa. Therefore, it is important to identify and to

define organ-specific functions as endpoints and to develop test systems analysing them. The

aim is to evaluate whether these endpoints are affected when cytotoxicity is absent (Leist et al

2013). Analysing how a chemical changes cell homeostasis before cell death is induced

makes it possible to identify pathways related to organ specific endpoints. Nevertheless, it

remains crucial to measure cell death in parallel as it is possible that organ-specific endpoints

are affected in parallel to cell death or as a consequence of it. If a compound does not

introduce cell death but inhibits organ specific endpoints at certain concentrations, such

concentrations should be used for follow-up mechanistic studies. Typically these follow-up

studies need to be performed across concentration and time, as the concentration-dependent

intensities of disturbed pathways help to differentiate between toxicant effects and

epiphenomena (changes occurring in parallel, but not related to toxicity), whereas the time-

dependent resolution of activated pathways helps to differentiate between the molecular

initiating event and secondary/tertiary responses.

Challenging test systems by correct compound selection

While setting-up test systems for organ specific toxicities, it is important to challenge the

system to evaluate its suitability and predictivity. Recently several groups have discussed

different approaches of test system development (Crofton et al 2011, Kadereit et al 2012,

Leist et al 2010). The review by Kadereit and colleagues (Kadereit et al 2012) summarizes

specifically the step-wise procedure of compound selection for the establishment of

developmental neurotoxicty (DNT) test systems. This selection procedure may very-well be

applied to other fields of toxicology as well and is summarized briefly below. Developing a

test system with a specific endpoint of interest, mimicking an organ-specific phenotype, has

to be validated. This validation can only be performed by using a set of compounds, with

different characteristics. First of all, compounds have to be included in the first training set,

which are known to interfere with the organ and endpoint of interest. Two classes of “positive


13

compounds” are introduced by Kadereit and colleagues (Kadereit et al 2012). The “gold

standard” compounds are chemicals known to be toxic to the organ of interest in humans due

to existing epidemiological studies. “Mechanistic tool compounds” on the other side, are

chemicals known to disrupt organ specific cellular processes. Those compounds should result

in a response in a sensitive test system. To assure that the system has a high specificity,

compounds without known toxicity, such as sugars, should be used as negative controls.

Alternatively tandem compounds could be used. These compounds are structurally very

similar, but one is toxic the other one is not. Thereby the relative differences of both

compounds in the system can be used for testing the specificity/selectivity of the model.

Further negative compounds are substances with a known target, which is shown to be absent

in the test system. The third class, presented by Kadereit and colleagues, which should be

included in the first training set are the “generally cytotoxic compounds”. Those compounds

trigger cell death independent of the cell-type and should not interfere with any organ specific

process. Examples are apoptosis-inducing compounds, such as staurosporine or etoposide. By

this, it should be verified that the positive controls did not interfere with the organ specific

cellular process as secondary effect, because cells were compromised by induced cell death. If

the generally cytotoxic compounds do interfere with the process of interest, one has to

compare both endpoints (cell death and organ specific cellular process) very carefully.

Positive compounds and general cytotoxic compounds have to be compared to analyse

whether the organ specific cellular process is more potently affected (at lower concentrations)

by the positive compounds in comparison to cell death. Having established a test system

which correctly responds to positive compounds and shows no alteration to negative

compounds, larger screenings can be performed to re-evaluate the test system with a test set

of compounds. If a robust test system is developed, the mechanisms behind the positive

compounds can be studied. Thereby pathway inhibitors play an important role (Kadereit et al

2012). If one of these inhibitors prevents the organ specific cellular process it is likely that the

underlying pathway is important and likely a target of toxicants. Off-target-effects of the

pathway inhibitors and inhibition of their targets should ideally be verified. To understand the

mode of action of chemicals, high-content techniques can be applied.

High-throughput and high-content screening

As there are thousands of chemicals which need to be tested (Rovida & Hartung 2009),

test systems are needed, which are fast and give as much information about a chemical as

possible. To describe such test systems, terms like high-throughput and high-content are


14

combined with screening (HTS/HCS), analysis (HTA/HCA) or imaging (HTI/HCI) and

slightly different definitions are present.

High-throughput, for example, is used to describe the number of chemicals which can be

tested in a certain time. In pharmaceutical industries, around 10 000 to 100 000 chemicals

may be screened per day or week (Frearson & Collie 2009, Hughes et al 2011). In contrast,

during development of HTS for toxicity assessment, only small numbers of chemicals are

usually tested. These chemicals serve as proof-of-principle tools to test whether the platform

is suitable for screening big numbers of chemicals. No real definition exists for the number of

compounds which should be tested (Judson et al 2013). HTS typically focus on specific

cellular processes, such as proliferation, differentiation or migration and are in most cases

image-based. As these HTS usually concentrate on only one process, large batteries of tests

would be required to model a whole organism and to completely characterize a chemical.

However, they are extremely helpful to classify compounds. For example, if a compound is

identified of being a hit in one of these assays it should be ranked as a possible toxic hazard,

and follow-up examinations have to be performed to determine the mechanism behind.

High-content, on the other hand, may either refer to the content of primary information

which taken together describe the final endpoint of the assay, or it refers to the content of

generated endpoints in the end. A classical HCA is HCI. Several defined parameters, such as

cell size, cell morphology, dye intensity, and so on are configured in an algorithm which

encodes the final endpoint, e.g. cell viability. HCI can also be a HTA. Another HCA, which is

not high-throughput capable, are Omics. Here the final data is of high-content, as several

hundreds or thousands of endpoints are assessed. These technologies analyse almost complete

sets of specific cellular factors (ome = a totality of some sort), such as proteins, mRNA or

metabolites and determine in a semi-quantitative manner perturbations to control or healthy

state. This approach is nowadays called toxicogenomics and refers to the application of one or

several Omics technologies to understand the toxic mechanisms of chemicals (Waters &

Fostel 2004). The idea is to perform these Omics on known human toxicants and chemically-

related or functionally-related compounds (which e.g. have been classified by HTS) to

identify similar changed patterns for similar compounds to use these “signatures of toxicity”

(SoTs) as classifiers (Bouhifd et al 2013, Hartung et al 2012). To implement such SoTs for

group chemicals, it is very important to compare concentrations of similar strong effect,

which is only possible if organ specific endpoints are carefully assessed, e.g. 20 % reduction

of “x”, but without cell death induction.


15

Figure 2: Morphology

changes during

differentiation of

neuronal cells Modified from (Scholz et

al 2011)

High-content imaging (HCI)

HCI in toxicology is based on an automated fluorescent microscope. It is usually capable

to capture images of several different fluorophores and typically uses a 5x, 10x or 20x

objective. Several plate formats can be imaged, thereby applies, the more wells, the more

conditions can be tested. Both ways are possible, staining of fixed (dead) cells with e.g.

antibodies specific for a protein of interest, or life-cell staining. The technical challenging part

is not the hardware or the staining, but rather the software and algorithm development for data

analysis (Gough & Johnston 2007). The one rule is ‘what you can see, is what you can

measure’. If it is difficult to see differences with the eyes, it will be just as well difficult to

teach a machine what to analyse. Processes which can be analysed by this technique are

therefore morphological changes of cells, increase or decrease in cell number and distribution

of cells, organelles or molecules in comparison with untreated control cells. All processes

depend on a sufficient spatial resolution. Morphological changes include changes in cell size

and cell shape. Increase or decrease of cell number can be readouts for proliferation or cell

viability/cytotoxicity. Distribution of cells is usually assessed when migration is the endpoint.

Also the distribution of single macromolecules can be studied to understand the underlying

signal transduction or interactions with other molecules.

Morphological changes:

Cells, usually analysed for morphological changes are neurons. Developing neurons, for

example, increase in size, change their shape and develop long extensions, called neurites

(Fig. 2). Especially the latter is of interest, as a neuronal progenitor cell develops neurites to

reach its target region when it differentiates towards a mature neuron. Neurite growth can be

easily assessed with high throughput and disturbances in this process reflect a toxicant’s risk

of being developmentally neurotoxic. Different ways are published to actually determine the

growth process. Dependent on the algorithm one can determine cell population characteristics,

e.g. the area which is covered by the growing neurons (Stiegler et al 2011), or the total neurite


16

length (Radio et al 2008), which are especially suitable endpoints when working with high-

density cultures. One can also determine cell-specific characteristics when low density

cultures are used, and a clear allocation of neurites to cell bodies is possible. In this way

number of neurite branches, neurites/cell or neurite length/cell (Price et al 2006, Yeyeodu et

al 2010) can be analysed. Also differences between the growth of axons and dendrites can be

studied (Harrill et al 2013). Neurite development is not only assessed in 2D-cultures but can

also, for instance, be analysed in vivo in transgenic zebrafish expressing GFP coupled to

proteins specific for neurons in a high-throughput manner (Kanungo et al 2011).

Other cells do change their morphology as well. Anti-cancer drugs, often targeting the

cytoskeleton of cells, are reported to induce morphology changes in cancer cells, which can

be used in high-throughput to cluster cancer-drugs into groups according to the changes in the

different cancer cell lines (Caie et al 2010, Loo et al 2007). Additionally, human embryonic

stem cells (hESC) change their morphology quite significantly during differentiation,

although the differentiation is rather screened by marker expression specific for the different

cell lineages (Balmer et al 2012, Sherman et al 2011, Weng et al 2012), than to study

morphology changes (Bauwens et al 2008). Moreover, apoptotic and necrotic cells can

possibly be differentiated by cell morphology, as cells start to swell before necrosis takes

place, whereas in apoptosis cytoplasm shrinkage can be observed (Leist & Jaattela 2001, Price

et al 2006).

Changes in cell number:

Changing cell numbers can either increase or decrease. Decreasing cell numbers are a

read-out for cell death. To analyse such an effect, the number of cells in a certain well is

assessed. The difficulty here is that dead cells have to be distinguished from viable cells as

dead cells normally do not disappear. Therefore, several different staining methods are

available to determine cytotoxicity in a high-content system. In most cases the total cell

number is counted by a DNA staining, such as Hoechst or DAPI, which is occasionally also

used to count apoptotic nuclei as their DNA is condensed and an intensity increase can be

measured (Diaz et al 2003, Sunil et al 2011). Co-staining with other dyes make the read-out

more sensitive. For instance, life-cell staining such as calcein-AM can be used, which pass

plasma membranes but only become fluorescent when cleaved by esterases in living cells

(Schildknecht et al 2009, Stiegler et al 2011). Propidium iodide is membrane impermeable

and only necrotic dead cells become fluorescent (Breier et al 2008, Torres-Guzman et al

2013). Also the TUNEL staining (TdT-mediated dUTP-biotin nick end labelling) is frequently


17

Figure 4: Example of scratch assay by

means of neural crest cell migration (Zimmer et al 2012).

used to evaluate apoptosis, as fragmented

DNA is marked by labelling the terminal ends

of the nucleic acid fragments (Timar 2004). In

the end the different markers need to be

compared to control condition to evaluate

whether there is an increase in cell death.

Fig. 3 displays an example: Cells are

recognized by a positive Hoechst staining (i),

whereas very intense nuclei are diagnosed as

being apoptotic (ii, orange arrow). Calcein-

AM is cleaved in living cells (iii) and cells are

counted as viable if both dyes overlap,

calcein-negative cells on the other side are

identified as dead cells (iv, green arrows).

The number of viable cells can also

increase, which happens when cells proliferate. Usually a Hoechst staining is combined with

BrdU (5-Bromo-2´-Deoxyuridine) or EdU (5-ethynyl-2’-deoxyuridine) staining. These

nucleoside analogues are incorporated into DNA during DNA replication and are therefore

excellent markers of proliferation. Hence, double-positive cells are identified as proliferating

cells (Breier et al 2008, Culbreth et al 2012, Duncan 2004, Walpita et al 2012).

Distribution of cells/organelles:

Another common read-out for high-

throughput screening is migration. During

embryogenesis, wound healing and immune

response, but also under pathological

conditions such as cancer metastases,

migration plays a crucial role. Therefore

interference of chemicals with migration

could lead to severe consequences. Several

diseases are related to altered migration of

cells, such as schizophrenia (Valiente & Marin 2010), asthma (Luster et al 2005) or

inflammatory bowel diseases (Rieder et al 2007). Several cell types are known to migrate and

are used in in vitro models, such as neural crest cells during development (Zimmer et al

Figure 3: Automated cytotoxicity

analysis in a high-throughput

capable manner Modified from (Stiegler et al 2011)

Auto

matically

identify

nucle

i

Identify cells with calcein-positive somata

Auto

matically

identif

yn

ucl

ei

ii

iii

iv

iCalceinH-33342

Identify cells with calcein-positive somata


18

2012), endothelial cells during angiogenesis (Mastyugin et al 2004), leucocytes during

inflammation (Grimsey et al 2012), or metastasizing cancer cells (Nystrom et al 2005). The

way to assess the migration of these cells originated from a method called wound healing

assay or scratch assay (Fig. 4) (Rodriguez et al 2005). Thereby a confluent layer of cells is

scratched to create a cell free area. After a while cells are counted which migrated back into

this area. Several more easy-to-handle approaches are being developed, to make this

technique more high-throughput capable (Gough et al 2011). As mentioned before, not only

cells change their distribution, but also organelles. Common examples are mitochondria,

which are easy to target by either staining with a membrane potential sensitive dye (Attene-

Ramos et al 2013, Sakamuru et al 2012) or e.g. measuring RFP-tagged mitochondria (Fig. 5)

(Schildknecht et al 2013). Also the Golgi is a dynamic apparatus and good antibodies are

available (Farhan et al 2008), so that a high-throughput screening for Golgi would be possible

as well (Healthcare 2010).

Distribution of molecules:

But not only cell behaviour or vesicular transport or movement can be an indicator for

toxic effects as the interaction of macromolecules (e.g. DNA, RNA, proteins) or the misrouted

distribution of these could be a very sensitive indicator as well. To be able to study such

macromolecular interactions in high-throughput the lateral resolution should be excellent and

the signals of the labelled molecules have to be very intense. That’s why not many studies

have been developed focusing on signal-transduction or interactions of molecules in an

automated manner. Nevertheless, there are protocols being developed, e.g. studying receptor

internalization (Grimsey et al 2008, Ross et al 2008) or interactions of molecules by FRET

(fluorescence resonance energy transfer analysis), wherein a donor-acceptor pair reports on

the distance between dyes on the nm scale. FRET glucose sensors are for example used to

Figure 5: Quantification of

mitochondrial transport in

neurites Modified from (Schildknecht et al

2013). Mitochondria are counted in

kymographs (graphical

representations of spatial position

over time)


19

study glucose flux by imaging (Takanaga et al 2008). This technology is pushed towards

better performance and higher throughput also (Kim et al 2011).

Life-cell imaging:

To come closer to “real-life” scenarios and measurements, it will be necessary to analyse

living cells directly. This is for example the case when a frequency of events should be

determined. Several of the above mentioned functional endpoints are usually followed in live-

mode, like the FRET sensors or cell organelle distribution analysis (e.g. mitochondria).

Another functional endpoint dependent on life-cell imaging is Ca²+ signalling. Differences in

fluorescence intensity or fluorescence wavelength of Ca2+ sensitive dyes decode the Ca2+

signalling (Ansher et al 1986). In addition, the well-known embryonic stem cell test (EST) is

adopted to live-imaging (Schaaf et al 2011). Here mESC or hESC are differentiated towards

beating cardiomyocytes (Seiler & Spielmann 2011) and inference with this process is

examined after exposure to various chemicals. Indeed, the other described endpoints, such as

migration (Shih & Yamada 2011) or apoptosis (Puigvert et al 2010), can also be studied in

life-cell mode to understand the kinetics of these processes. The throughput of these assays is

limited, as every test condition has to be followed for a certain time.

Toxicogenomics

As briefly mentioned above, toxicogenomics is an analytical tool to relate the activity of

a toxicant with altered genetic profiles within the cells of interest. The recording of patterns of

altered molecular expression caused by exposure to chemicals is a very sensitive indicator.

The altered expression can take place on several levels, such as changes in the

transcriptome, the proteome or the metabolome. Omics technologies are therefore concepts to

measure these alterations on one of the levels, to facilitate the identification of the mode of

action of the chemical and to shed light on which pathways are involved (Fig. 6). The

approach of combining data of several Omics sources in toxicology is also called systems

toxicology, derived from the field of systems biology (Hartung et al 2012). In the following

the most common techniques are shortly described.


20

Transcriptomics:

Transcriptomics determines the changes of mRNA expression patterns. Two major

technologies exist, the microarray- and the RNA-sequencing method. Both of these methods

make it possible to evaluate several thousands of transcripts at the same time. Extracted RNA

of treated and untreated conditions is converted to cDNA and usually amplified and labelled.

The microarray is based on about 30 000 different probe sets, which are oligomers aligned on

a microchip. These probe sets are specific for the transcripts of the species of interest.

Thereby several probe sets can target the same transcript and serve as an internal quality

control. Different arrays are available, dependent on the species and on the way the probes are

designed. The most commonly used arrays, such as the affymetrix Human U133 Plus 2.0,

have their probes aligned to the 3’ end of the transcripts. In this way most of the molecules are

caught, as mRNA is usually transcribed into cDNA by oligodT primers (Dalma-Weiszhausz

et al 2006). Other versions of the microarrays have spread their probes over the whole length

of the transcript to be able to catch splice variants (Auer et al 2009). An advantage of RNA

sequencing is, that it is not restricted to any probe sets, and the total RNA in the sample is

sequenced. Usually the cDNA is fragmented and small sequencing adaptors are added. Using

different sequencing technologies, such as 454 (Roche Applied Sciences) or Solexa (Illumina,

Inc.), short sequences are obtained (Morozova & Marra 2008). These sequences have to be

aligned according to the reference transcriptome (Wang et al 2009) and splice variants as well

as miRNA can be detected.

Figure 6: Scheme of systems toxicology/toxicogenomics Toxicity related profiling of altered molecular expression is used to identify compound specific

signatures of toxicity (SoT). The integration of these may lead to verification of pathways,

responsible for the toxic outcome.

Analysis of patterns of altered molceular expression

Transcriptomics Proteomics Metabolomics

NMR-spectroscopy

HPLC/GC-MSSILAC iTRAQ

2-D electrophoresis

qPCR

DNA-microarrays

RNA sequencing

Toxicogenomics

Integration of omics-data

SoT (transcripts) SoT (proteins) SoT (metabolites)

Pathways of toxicity


21

Metabolomics:

One step further to a specific process within the cellular metabolism several new

methods have been developed to investigate the sum of metabolites and to analyse the pattern

or the changes in the levels of important cellular metabolites. Metabolomics, the

determination of metabolite levels, patterns or changes, concentrates on a few thousand

molecules in a cell, much less than transcripts or proteins (Hartung et al 2012). Metabolites

range from small molecules such as carbohydrates, amino acids, nucleotides, phospholipids,

steroids, or fatty acids and their derivatives to smaller peptides (Ramirez et al 2013). Thereby

the intracellular metabolites as well as the secreted extracellular metabolites can help to

analyse the disrupted cell homeostasis. The intracellular metabolites may thus be regarded as

the fingerprint of the toxicity pattern, whereas the extracellular metabolites as the footprint. If

the intracellular changes are of interest one has to assure the rapid quenching of enzymatic

activities during the sampling procedure. This is one of the technical challenging parts, as

sampling has to be done fast and reproducible (Cuperlovic-Culf et al 2010). To detect the

metabolites different analytical techniques are available. In most cases nuclear magnetic

resonance spectroscopy (NMR) or mass spectroscopy (MS) are used. In NMR the molecules

do not need to be separated before, whereas in MS the system is coupled e.g. to an upstream

high-performance liquid chromatography (HPLC). Metabolites can also be modified so that

they are more volatile and gas chromatography (GC)-MS can be used. To identify metabolites

more reliable, MS/MS-based fragmentation and analysis can be performed. New approaches

analyse the conversion of metabolites by enzymatic activities in a cell, called fluxomics

(Klein & Heinzle 2012). Usually isotopically labelled reporters, such as glucose are used,

with one or more heavy C-atoms (13C1-6; number in subscript indicate the number of isotopic

C-atoms). By separating all 13C metabolites, the conversion of glucose to down-stream

metabolites can be analysed and conclusions about differences in flux can be drawn (Niittylae

et al 2009).

Proteomics:

In proteomics peptides and proteins in the cells are under investigation. Proteins are

usually considered as the key-players in cell reactions, as for example mRNA not always

correlates with protein translation or post-translational modifications and proteins are also

responsible for the conversion rate of metabolites. Different methodologies exist, some

separating the proteins based on their size and chemical properties, others include a labelling

step of amino acids or peptide fragments. The peptides themselves are then identified by


22

HPLC-MS/MS. To separate at protein level Gel-LC-MS/MS is the preferred fractionating

method usually dependent on two-dimensional gel electrophoresis. Changed proteins are

identified on the gels and bands are cut to further digest and identify the proteins with HPLC-

MS/MS (Rabilloud et al 2010). To separate at the peptide level, SILAC (stable isotope

labelling with amino acids in cell culture) or iTRAQ (isobaric tag for relative and absolute

quantitation) are used. In SILAC, labelled amino acids can be added to the culture medium, so

that one sample is cultured with the normal light amino acid media and the other sample with

isotopic heavy amino acid media (Ong 2012). If media cannot be controlled like this, or when

working with tissue samples, iTRAQ can be applied. Labelling occurs later in the

experimental procedure. After samples have been taken, proteins are digested and isobaric

mass tags are added. Peptides with tags are then analysed with HPLC-MS/MS (Evans et al

2012).

Other “omes”:

Next to the above introduced Omics technologies other fields exist which also deal with

“omes”. One area, which introduced the ending “ome” to research, is the genome. By

projects, such as the human genome project, researchers focused on the identification of

human genes and differences in those, which may be related to disease (Cavalli-Sforza 2005).

More sophisticated “omes” have recently been introduced in an article published by Nature. It

discusses the emerging number of “omes” and presents those, which are worth to remember

according to Baker (2013). The phenome, for example, deals with the collection of

phenotypic abnormalities in humans, diagnosed with certain diseases, to understand the

outcome of those. Omes which build on the above introduced (transcriptome, proteome,

metabolome) are the interactome, the integrome and the toxome. All of them have in

common, that in the end, a map of pathways will be generated, which should guide scientists

to find answers to different questions. People working with the interactome, for example,

want to list all molecular interactions, to understand, e.g. all protein-protein interactions. The

integrome, on the other side, is the development of technologies and algorithms, which enable

the easy integration of data, generated with different Omics technologies, such as

transcriptomics, proteomics and metabolomics. The most relevant for the field of toxicology,

is the human toxome. This field uses the typical Omics technologies in the context of healthy

and toxicant-treated conditions, to reveal underlying pathways of toxicity which lead to the

observed altered phenotype (Hartung & McBride 2011). This field mainly concentrates on


23

alternative test systems, being established to mimic human body functions more closely than

animal experiments.

Statistics in Toxicogenomics

One of the most important procedures when working with Omics methods is the

application of correct statistics (statistical significance) to identify changes from control with

biological significance. The statistical significance helps to find results which are interesting

in relation to the biological question. The procedure consists of several steps (Dunkler et al

2011) and is summarized briefly in the following. In a first step, quality of raw data files,

generated out of the fluorescence data of microchips or out of the total sum of features (peaks)

of an HPLC-MS or NMR measurement, is checked. For instance estimating the overall

intensity of different microarrays or comparing total ion chromatograms of HPLC-MS

generated data. Based on these quality checks some of the replicates may be excluded from

the further analysis, as the technical procedure of sample preparation may vary between

samples, e.g. DNA annealing onto microarray chips failed or pressure of the MS/NMR device

was not stable. Next, data have to be normalized, so that they are comparable from sample to

sample. For microarray data, the RMA (robust multichip average) function is used, with

which e.g. background correction or quantile normalization can be performed. In the case of

ion chromatograms, a pool of all samples can be run in parallel as quality control samples.

These serve as template for the extracted ion chromatograms to assure that the same extracted

peaks are compared with each other (e.g. retention time shifts can be identified and corrected

for). After that, usually a filtering of unspecific components (such as genes or metabolites)

takes place, meaning that a pre-specified cut-off of fold-changes is applied onto the samples

to reduce the data set. The pre-processed, normalized and filtered data is typically visualized

in a principal component analysis (PCA), displaying the individual samples based on

variables that differ between the samples. These PCA allow to visualize data in several

dimensions and to detect patterns and structures within the data-sets. The PCA is therefore

used as quality control (do samples of the same condition cluster together or is a separation of

groups of interest encoded within the data?) and as classifier for generating new groups.

Finally, the statistics are applied. Usually a modified version of the t-test (comparing means

relative to variance), such as the moderated t-test, are employed to identify potential

significantly changed factors in the samples. The sheer number of probe sets on arrays will

always give rise to a respectable number of false positives. For example, a t-test run for each

gene will predict some as significantly regulated even though the variation found is just due to


24

chance. Therefore the moderated t-tests introduce an estimated standard deviation (SD) for the

whole set of components (a pooled SD for all components) and include this in the calculation

(Goni et al 2009). But false-positives will still increase the more t-tests are performed (e.g.

30 000 for a standard microarray). The false discovery rate (FDR) correction helps to reduce

the number of false positives, as it tries to provide a balance between the identification of real

significant factors (high sensitivity) and avoiding false-positive estimations (high specificity).

Several FDR corrections exist, whereby the Benjamini-Hochberg FDR is the most frequent

one. In this step-down method, p-values of all potentially significant components (n) are

ranked from smallest to largest and a stepwise correction of each p-value is performed

[corrected p-value = p-value x (n/n-(rank of p-value)); if < 0.05, factor is significant]

(Agilent_Technologies 2005, Benjamini & Hochberg 1995). Now lists of significant

components can be generated, which can be further analysed and also confirmed with follow-

up experiments.

Omics give snap-shots of the moment the samples were taken and with the help of

bioinformatics and correct statistics one can identify disturbances from baseline. Those

disturbances can be further investigated. Changed transcripts can for instance be analysed by

grouping them according to their biological function, an analysis called gene ontology

enrichment analysis. With open source tools, such as g:profiler (http://biit.cs.ut.ee/gprofiler/),

one can easily determine, if the changed expression patterns can be grouped to biological

processes and if this is in concordance with expectations, which are dependent on the

biological system and chemical compound used (Balmer et al 2012, Weng et al 2012). The

same analysis can be made with proteomics data, too (Carvalho et al 2009). Changed

metabolites on the other hand, can be mapped onto biochemical pathways, to see whether pro-

minent conversions/reactions are present (http://wikipathways.org/index.php/WikiPathways).

Those analyses are suitable to strengthen or to generate new hypotheses of toxicity

mechanisms of chemicals.

Applications of HCI and toxicogenomics in toxicology

As mentioned before, toxicologists want to understand the mode of action (MoA) behind

organ-specific toxicities of chemicals, to be able to predict these outcomes for other chemical

compounds. Several Omics and HTS studies have been carried out in lung-, heart-, kidney- or

liver-specific in vitro models to elucidate the underlying mechanisms of organ-specific

toxicants. Table 1 was generated by using this search profile:


25

Organ Publication in vitro system Method Compoundtype *

Lung (Maertens et

al 2013)

Murine lung epithelial cells Transcriptomics mouse whole genome

microarrays

Tobacco and marijuana smoke condensate

1

(Tan et al

2012)

Primary and immortalized

human bronchial epithelial cells qPCR high throughput screening approach

800 compounds of

MicroSource Natural

Products Library

(Cha et al

2007)

human bronchial epithelial cell

line Proteomics Gel-LC-MS-MS

BSA-coated titanium

dioxide (TiO2) particles

1

Kidney (Wilmes et al

2013)

Cultured human renal epithelial cells (RPTEC/TERT1)

transcriptomic, proteomic and

metabolomic profiling

Cyclosporine A 1

(Wilmes et al

2011)

Human renal proximal tubular

cells Transcriptomics whole genome microarrays

Cadmium, Diquat,

Cyclosporine A

1

(Faiz et al

2011)

Human renal proximal tubular cells

Metabolomics (13)C NMR spectroscopy

CdCl2 1

(Ellis et al

2011)

RPTEC/TERT1 (non-tumour

human renal epithelial cell line) Metabolomics (1)H NMR spectroscopy

nifedipine, potassium

bromate, monuron, D-

mannitol, ochratoxin A, sodium diclofenac

1

Liver (Van

Summeren et

al 2013)

primary mouse hepatocytes Proteomics 2D-gel electrophoresis

acetaminophen,

amiodarone, cyclosporine A

1

(Doktorova et

al 2013)

primary rat

hepatocytes, HepaRG, HepG2, hESC-derived hepatocyte-like

cells

Transcriptomics 15 drugs 2

(Choucha

Snouber et al

2013)

HepG2/C3a cells Metabolomics 1H NMR spectroscopy

flutamide, hydroxyflutamide

1

(Mennecozzi

et al 2013)

HepaRG HCS (cell count, nuclear intensity, nuclear area, ROS intensity)

92 reference chemicals

with known hepatotoxic

activity

3

(van Delft et

al 2012)

HepG2 RNA-Seq

benzo[a]pyrene 1

(Tolosa et al

2012)

HepG2 HCS (nuclear morphology, mitochon-

drial function, intracellular calcium, oxidative stress)

78 different compounds 3

(Donato et al

2012)

HepG2 HCS (lipid content, ROS generation,

mitochondrial membrane

potential, cell viability)

16 drugs 2

(Wang et al

2011b)

porcine primary hepatocytes Transcript-& proteomics porcine genome array, 2D-DIGE-MS

T-2 toxin 1

(Jennen et al

2011)

HepG2 Metabolomics &

transcriptomics

2,3,7,8-tetrachlorodibenzo-

p-dioxin (TCDD)

1

Heart/

Embryo

-toxicity

(Osman et al

2010)

mESC (EST) proteomics monobutyl phthalate 1

(West et al

2010)

hESC (EST) metabolomics 10 non-teratogens, 14

teratogens

2

(van Dartel et

al 2009)

hESC differentiated towards cardiomyocytes (EST)

Transcriptomics monobutyl phthalate 1

(Mioulane et

al 2012)

hESC-derived cardiomyocytes

vs rat neonatal ventricular

cardiomyocytes

HCS cell death

chelerythrine 1

(Schaaf et al

2011)

human engineered heart tissue

(hEHT – ESC-derived) ? HCS automated video-optical

recording of beating cells

E-4031, quinidine,

procainamide, cisapride,

and sertindole

1

All

organs

diXa diXa - The Data Infrastructure for Chemical Safety

http://wwwdev.ebi.ac.uk/fg/dixa/index.html or http://www.dixa-fp7.eu/ diXa data warehouse collects and links to data of toxicogenomics projects, such as TG-Gates (Japanese

project, which is based on transcriptomics studies on human and rat hepatocytes, approximately 130

compounds tested in time and concentration-dependent manner)

Table 1: Toxicogenomics and HCS in toxicology. Listed studies were found by using: lung OR heart/cardio Or kidney OR liver AND in vitro PLUS

Omics AND/OR imaging AND/OR high-throughput AND/OR high-content. Most prominent

compounds are mentioned. *compound numbers tested: 1 = 1-10, 2 = 10-50, 3 = 50-200


26

(lung OR heart/cardio Or kidney OR liver) AND (in vitro PLUS Omics AND/OR imaging

AND/OR high-throughput AND/OR high-content). Publications, which were found for

cardiotoxicity, dealt mostly with the embryonic stem cell test (EST). As the EST is defined as

an in vitro alternative test designed for the prediction of embryotoxicity rather than prediction

of impacts on differentiated heart tissue, it is listed in the table under heart/embryotoxicity.

Additionally only the most recently (2009-2013, except for one lung-related paper from 2007)

released papers were collected. During development of HTS for toxicity assessment, a small

number of proof-of-principle chemicals are used. The number of tested chemicals is indicated

in the table.

Applications of HCI and toxicogenomics in neurotoxicology

As mentioned before, animal-based tests are expensive, time-consuming and offer low

species-to-species extrapolation predictivity. Hence, the US National Academy of Sciences

discusses the development of in vitro-based assays to generate toxicity profiles for the

thousands of chemicals lacking any hazard information. Especially the field of

neurotoxicology (NT) and its related field of developmental neurotoxicology (DNT) are very

tricky to assess in vivo. Impairment of the adult nervous system can be manifold, e.g.

neuropathy leading to seizure, paralysis or tremor as well as loss of motor-coordination,

sensory deficits, learning and memory deficits often based on impaired communication of

neurons at the synapse and not on neuronal cell death. TG 424 requests the neurotoxicity

study in rodents and involves daily oral dosing of rats for acute, subchronic, or chronic

assessments (28 days, 90 days, or one year and longer). Primary observations include

behavioural assessments and evaluation of nervous system histopathology. In the developing

brain several important processes take place, and minor alterations in any of these processes

may lead to severe outcomes. It is an orchestrated sequence of events including proliferation,

migration, patterning, differentiation, neurite growth, synaptogenesis and myelination as well

as neurotransmitter turnover (Kadereit et al 2012). DNT testing is currently based on the

OECD guideline TG 426, and only a small group of chemicals has been tested according to

this guideline (Grandjean & Landrigan 2006, Makris et al 2009, McCormick et al 2003). The

guideline instructs daily dosing of at least 60 pregnant rats. Offspring are evaluated for

neurologic and behavioural abnormalities. Drawbacks of these neurotoxicity guidelines are

high costs, long duration, low throughput and the questionable prediction capacities for

human neurotoxicity (Bal-Price et al 2008, Leist et al 2012a). Therefore the field of

neurotoxicology is strongly working on alternatives for NT evaluation. Again cell death has to


27

be evaluated together with neuronal specific endpoints. For NT those endpoints are electrical

activity, neurotransmission, axonal transport, receptor and channel activation, enzyme

activity, synaptogenesis/myelination, excitotoxicity, and neuronal–glial interactions (Bal-

Price et al 2010). In the framework of ACuteTox, an EU funded project, over 50 compounds

have been tested for general cytotoxicity and neuronal specific endpoints in primary cells,

neuronal cell lines and 3D neurospheres (Forsby et al 2009). The authors suggested an 0.7 log

unit difference between the EC50 of neuronal specific and cell death endpoints to classify a

compound as neurotoxic alert. As no single endpoint improved the in vitro/in vivo

extrapolation, the combination of several endpoints was suggested. Not many screening

studies have been performed in the field of in vitro neurotoxicity as the field has mainly been

focused on mechanistic studies (Bal-Price et al 2010). The developing brain has to be even

more carefully assessed, as it is more susceptible to toxic insults. This is due to missing

protective mechanisms in the developing foetus or new-borns, such as the blood brain barrier

(BBB) or DNA repair systems, which are either not present or not fully functional yet

(Adinolfi 1985, Saunders 1986). Several test systems exist, which represent one of the

important processes of brain development. But not all important biological

neurodevelopmental processes can be modelled at present that in turn could relate to toxicity

endophenotypes (TEP). TEPs describe the biologically quantifiable altered functionality of

parts of the nervous system, such as altered electrical circuits, due to exposure to a DNT

chemical. They link basic biological processes that are disturbed by a DNT compound with

the final DNT phenotype observed in the organism, such as lowered IQ (Kadereit et al 2012).

In the tables 2 and 3 studies are listed, which deal with the development of NT or DNT

suitable test systems in either high-throughput mode or with focus on toxicity mechanisms by

applying Omics in some sort.


28

High-content imaging (HCI):

Publication in vitro system Endpoint Compounds

(Stern et al

2013)

Human Ntera2

cells

Differentiation (anti βIII-

tubulin staining –

microscope and

fluorescence plate reader)

Migration (dapi staining –

Oris cell Migration Assay;

brightfield – migration

distance from neurospheres)

cytochalasin D, methylmercury

(MeHg), sodium arsenite,

methylazoxymethanol, valproic acid

(VPA), acetaminophen, penicillin g,

sodium glutamate, sodium

nitroprusside, 8-Br-cyclic gMP, (1h-

[1,2,4]-oxadiazolo[4,3-a]quinoxalin-1-

one (ODQ)

Results: Differentiation and migration effects were seen without cell death induction (alamar

blue assay). 5 days of exposure (DoD5-DoD10) revealed significant differences in

differentiation (βIII tubulin decrease) and 48 h exposure (DoD9-DoD11) revealed migration

differences for MeHg, sodium arsenite, methylazoxymethanol and VPA. Acetaminophen

only affected differentiation, ODQ was only assessed for migration. Dynamic range was

shown by increase in migration with sodium nitroprusside and 8-Br-cyclicGMP.

(Cornelissen

et al 2013)

Mouse primary

hippocampal

neurons

Network activity (Calcium

fluxes – fluo-4AM, dapi)

5-HT (serotonin), ethylene glycol

tetra-acetic acid (EGTA), anti-NGF

Results: Peak decay times and burst frequency for acute effects were determined by adding

5-HT or EGTA for 4 min onto the cells before glutamate was added. 5-HT reduces the

median amplitude of calcium bursts whereas the median burst frequency increases. Whereas

EGTA, a calcium chelator, decreases all measured features. Chronic effects were followed

for 3 days (DoD4-DoD7) with antibody against NGF and a dose-dependent decrease in burst

frequency was detected.

(Zimmer et al

2012)

hESC-derived

neural crest

Migration (scratch assay

based on Hoechst and

calcein-AM staining)

> 20 compounds, including negative

controls, end point–specific controls,

general developmental neurotoxicity

compounds, and chemicals known to

specifically impair NC cell migration

in vivo

Results: By testing over 20 compounds with the scratch assay, it could be shown that neural

crest (NC) cells were sensitive to positive compounds, such as MeHg, VPA and thimerosal,

known to inhibit migration. Other proof-of-principle compounds, such as negative and

general cytotoxic compounds also verified the system. The finding that NC were more

susceptible to toxic insults with regard to migration than other migrating cell types, such as

neuronal precursor cells (NEP), HEK or HeLa cells, indicated the promising value of NC in

identifying developmental toxicants.

(Stiegler et al

2011)

LUHMES Neurite growth (field-based

algorithm dependent on

live-cell staining with

calcein-AM and Hoechst)

MeHg, puromycin, flavopiridol,

metamphetamine, menadione,

cycloheximide, bisindolylmaleimid I

(bis1), brefeldin A, CdCl2, sodium

dodecyl sulfate, U0126, Na3VO4,

tween-20, saponin, K2CrO4,

bisbenzimide-H

Results: Compounds were added for 24 h (DoD2-DoD3) and viability as well as neurite

growth were assessed on the same cells. For some compounds a neurite growth effect was

seen without cell death induction (resazurin, Ca-AM positive cells). A clear separation of

neurite growth inhibitors (MeHg, flavopiridol, cycloheximide, bis1, brefeldin A, U0126,

Na3VO4) from unspecific toxicants was possible.

(Harrill et al

2011b)

rodent primary

mixed cortical

cultures

Synaptogenesis (staining

based on MAP2, synapsin,

Hoechst - punctate synapsin

protein in close apposition

to dendrites

KCl, Na3VO4, mevastatin, bumetanide,

tamoxifen, bis1, dipyridamole,

caffeine

Results: The authors present a technique by counting synapsin-positive puncta in MAP2-

identified dendrites as indicator of synaptic formation. Three compounds, KCl, Na3VO4 and


29


Bis1 were shown to interfere with dendrite length and total number of puncta without

inducing cell death. Only KCl showed a difference in the lowest observed effect level

(LOEL) in these two endpoints, although the difference remains small. For the remaining

positive hits it remains elusive, if the reduced number of puncta is a logical effect of the

reduced number of dendrites.

Mundy and

colleagues

2008-2013

PC12 cells (Breier

et al 2008),

cerebellar granular

cells (Radio et al

2010), hN2

(Harrill et al

2010), rat cortical

neurons (Harrill et

al 2011a, Harrill et

al 2013)

Neurite growth (labeling

fixed cells with bIII-tubulin,

dapi and additionally MAP2

(Harril 2013)

K252a, Na3VO4, bis1, U0126, lead,

LiCl, Brefeldin A dexamethasone,

CdCl2, MeHg, VPA, cyclosporine,

vincristine, Dimethyl phthalate,

Ampethamine, lead acetate (PbAc),

retinoic acid (RA), lead, okadaic acid,

diphenhydramine, omeprazole,

diphenylhydantoin,

Results: K252a, Na3VO4, bis1, U0126, PbAc, LiCl, Brefeldin A dexamethasone, RA, CdCl2,

MeHg and vincristine were identified as neurite growth inhibitors after 20-24 h of treatment.

Often the observed neurite effect was associated with a strong impact on cell death.

Dimethyl phthalate and cyclosporine were identified as neurite growth inducers in PC12

cells. Ampethamine was stated as neurite growth inducer in PC12 cells, but did not show

any effect in the subclone of these cells, the N2S cells. In the recent publication of 2013

impacts on subpopulations (dendrites vs. axons) of neurites were evaluated. Only K252a

was clearly shown to differentially impair axon growth in comparison with dendrite growth,

although it must be mentioned that axons were not directly identified, as MAP2 negative

neurites were labeled as axons.

(Moors et al

2009)

human neural

progenitor cell-

derived

neurospheres

Migration (brightfield

images were acquired every

two minutes and cells

outgrowing of plated

neurospheres were counted)

MeHgCl, HgCl2, staurosporine, H2O2

Results: In this study neurospheres were presented as valuable tool to asses DNT in vitro.

Cells were analysed for a functional apoptosis machinery by staurosporine and H2O2. Also

the differentiation was followed-up over time to analyse the behavior of the spheres. They

analysed migration of cells out of the spheres and could show that MeHgCl and HgCl2

inhibited migration at non-cytotoxic concentrations.

(Breier et al

2008)

ReNcell CX cells Proliferation (BrdU

incorporation)

Proliferation inhibitors: Aphidicolin,

Hydroxyurea, Cytosine Arabinoside,

5-Fluorouracil, Ochratoxin,

DNT chemicals: MeHgCl, CdCl2, 5,5-

Diphenylhydantoin, trans-Retinoic

Acid, Dexamethasone, VPA, d-

Amphetamine Sulfate, PbAc

Results: The authors challenged the system for different times (4 h, 24 h, 48 h) with known

proliferation inhibitors and observed a strong decrease in proliferation without impact on

cell viability (propidium iodide exclusion) for the early time points, whereas after 48 h an

increase in propidium iodide was observed. For the DNT-specific controls (treated for 24 h),

only lead acetate and dexamethasone influenced proliferation at non-cytotoxic

concentrations. The other chemicals did not interfere or showed quite similar decreases in

both endpoints (proliferation, viability). For the latter, the two endpoints are not compared to

each other to identify concentrations which may induce cell death but inhibit more

efficiently proliferation. The authors describe the assay as suitable tool for DNT

identification, as non-DNT compounds did not interfere with proliferation.

Table 2: HCI-related studies in (developmental) neurotoxicity Listed studies were found by screening publications with the keywords ‘neurotoxicity’, ‘high-

throughput’ and ‘high-content’. Only the most prominent chemicals are listed, if not more than 20

chemicals were tested.


30

Toxicogenomics:


(Laurenza et al

2013, Nerini-

Molteni et al 2012,

Pallocca et al

2013)

WA09 (Nerini-Molteni)

and NTera2 (Pallocca)

miRNA expression profiling MeHg

Results: In both studies the suitability of miRNA profiling for DNT assessment was

tested. Thereby pre-configured microfluidic cards were used, which allowed the detection

of a few hundred miRNAs on the basis of qPCR. Both cell types, H9 and NTera-2,

showed alterations in the miRNA expression profiles after being exposed to low

concentrations of MeHg. In H9 (treatment for 10 days, 25 nM), MeHg-regulated

miRNAs were involved in the ubiquiting-proteasome pathway, whereas in NTera-2

(treatment for 5 weeks, 400 nM) a relation between regulated miRNAs and axon

guidance, learning and memory processes was identified. It is suggested that miRNA

profiling provides a complementary analysis to mRNA transcriptome studies to fill

missing gaps in toxicity mechanisms.

(Theunissen et al

2012a)

mESC Transcriptomics (mouse

whole genome arrays)

cyproconazole (CYP),

hexaconazole (HEX), VPA

Results: A 11-day mESC neural differentiation protocol (ESTn) was used to treat

differentiating cells with CYP, HEX or VPA, respectively, (DoD3-DoD4) to perform a

concentration-dependent transcriptome study. Data was compared with morphological

alterations (treated for DoD3-DoD6, neurite growth assessment at DoD11). Only VPA

reduced neurite growth significantly more in comparison to the overall viability. Using

transcriptomics, changes were seen at concentrations below those inducing

morphological effects. The authors declare that the omcis technology was far more

sensitive, although completely different exposure scenarios were compared. Several

mode of action hypotheses are generated based on GO term enrichment analysis, e.g.

enrichment of neuron development for VPA and CYP but not for HEX correlated to

known altered neuronal development of VPA and CYP in vivo. Also sterol-related

processes seemed to play a role in the ESTn upon HEX and CYP treatment.

(Palmer et al 2012) WA01 and WA09 hESC Metabolomics (ESI-QTOF-

MS)

EtOH

Results: HESC were differentiated towards embryoid bodies (EBs), neural progenitors

and neurons. The different stages were treated with 0, 0.1, or 0.3% EtOH for 4 days

before supernatant was analysed for metabolic changes. Several metabolites were altered

in the different cell stages, none of them overlapped. Out of these four metabolites were

verified by MS/MS analysis and were suggested as biomarkers, although they did not

show a dose-response regulation. The test system presented was introduced to detect

alteration in alcohol-induced developmental neurotoxicity, leading to fetal alcohol

spectrum disorder (FASD). For example in EBs, 5′-Methylthioadenosine (MTA) was

suggested as biomarker for FASD, as this metabolite directly reflects the synthesis of

polyamines, which have been shown to be deregulated in FASD in vivo.

(Balmer et al

2012)

WA09 hESC qPCR, transcriptomics

(human whole genome

arrays)

VPA, trichostatin A

Results: HESC were differentiated towards neuroectodermal precursors and treated with

the histone deacetylase inhibitors (HDACi) VPA or TSA for DoD0-DoD6, respectively.

Although the authors did not use a HCS platform, and therefore checked only a small

number of differentiation markers (e.g. PAX6, OCT4, NANOG), a new approach of DNT

assessment is presented by combining the altered expressions to epigenetic modifications.

In a follow-up study (not published yet), whole transcriptome changes were compared to

the observed epigenetic changes after different exposure times and wash-out experiments

with HDACi. It was suggested that secondary histone methylation functions as a potential

persistence detector deciding on reversibility or adversity of drug exposure of different

duration.

(Vendrell et al

2010, Vendrell et

al 2007)

mouse primary cultures of

cerebellar granule cells

(CGCs)

Proteomics (2D-Gel MALDI-

TOF-MS)

MeHg


31


Results: Different exposure scenarios were tested for several MeHg concentrations. In the

study of 2007 viability measurements identified a decrease after 10 days of MeHg

exposure at 100 nM. Proteome changes were analysed for 60 nM and 3-ketoacid-

coenzyme A transferase I (key enzyme in acetoactetate catabolism, important for neural

lipid synthesis) was identified to be decreased. In the study of 2010 cells were exposed

for 6, 11 and 16 days of MeHg. At 100 nM changes in phospho-cofilin were observed.

These findings were hypothesized as possible neurotoxic mechanism of MeHg as

decreased p-cofilin changes the actin-behaviour in the cells. The finding of 2007 of

altered 3-ketoacid-coenzyme A transferase I was not confirmed.

(Slotkin et al

2010)

PC12 cells Transcriptomics (whole rat

genome arrays)

Diazinon, dieldrin, Ni2+

PC12 cells were treated 24 h after seeding with 30 µM of each agent for either 24 h or 72

h, respectively. No cytotoxicity was observed. By the use of transcriptomics, the authors

focused on genes involved on cytokine signaling (FGF, NGF, BDNF) as well as wnt

signaling, resulting in 58 genes. Among those, the authors state that similar expression

changes could be observed and that these similarities in neurotrophic alterations could

contribute to similar functional outcomes, although this statement is not checked.

Table 3: Omics-related studies in (developmental) neurotoxicity Listed studies were found by screening publications with the keywords ‘neurotoxicity’ and ‘Omics’

and most recent studies are mentioned (2010-2013).

The studies mentioned above provide a basis for a new toxicity testing in the 21st century.

They highlight the ability to study chemical mode of actions by using new alternative test

systems. Those test systems are either based on organ specific processes, such as neurite

growth or migration and can be used in higher throughput to screen a large number of

chemicals. As mentioned above, a careful evaluation has to take place, as a combination of

tests is inevitable to cover the most important human body functions. Especially the

generation of many false-positives should be avoided, as more and more tests lead to a higher

chance to result in a positive hit in one of these assays just by chance (Basketter et al 2012).

Other tests, based on Omics technologies, are developed to identify the mechanism of a

chemical more precisely. Again, a careful evaluation has to take place, to identify real

disturbances. Tests combined by integrating testing strategies as well as the identification of

pathways of toxicity may provide the future basis for toxicity testing.

Aims of the thesis

32

Aims of the thesis

Animal experiments are expensive, time-consuming and have a low prediction capacity

for humans. For thousands of chemicals toxicity profiles are lacking and the number of

animals needed to generate those, would be enormous. Therefore alternatives to animal

testing are required, especially in the field of neurotoxicology and developmental

neurotoxicology, as the detection of alterations in the brain is very complex in animals

(Llorens et al 2012). Furthermore, the differences of brain development of humans and

rodents (Clancy et al 2007) make it difficult to extrapolate data found in rodents to the human

situation. Therefore alternatives have to be provided which convince on several levels.

Alternative test systems have to be robust, they should offer good sensitivity and specificity,

costs should be manageable, and they should provide insights into the toxic mechanisms of

chemicals (Leist et al 2010).

A current understanding of how this can be achieved is to analyse the pathways

underlying the toxic outcome of positively tested compounds, named pathways of toxicity

(PoT). Mapping all PoTs of the human body, the human toxome project, will help to screen

chemicals more reliably. The work described in this thesis was undertaken to carefully

evaluate new alternative strategies to predict neurotoxicity with human based test systems. In

three steps of growing complexity impacts of known toxicants on different stages of neuronal

development have been evaluated by the use of different endpoints.

The aims of this thesis were:

1. to challenge a test system based on human neuronal precursor cells (LUHMES) to

assess disturbances of neurite outgrowth by using a broad spectrum of chemicals

2. to evaluate the suitability of human embryonic stem cell (hESC) based test systems for

DNT testing by using transcriptomics

3. to analyse the impact of the neurotoxic model compound MPP+ on LUHMES by

integrating transcriptomic and metabolomic technologies with special emphasis on

identifying involved PoTs

Results Chapter 1 – Evaluation of a human neurite growth assay as specific screen for

developmental neurotoxicants

33

C. Results Chapter 1

Evaluation of a human neurite growth assay as specific screen for


Anne K. Krug1, Nina V. Balmer1, Florian Matt1, Felix Schönenberger2,3, Dorit Merhof2,

Marcel Leist1

Affiliations:

1 Doerenkamp-Zbinden Chair for In vitro Toxicology and Biomedicine, University of

Konstanz, D-78457 Konstanz, Germany

2 Interdisciplinary Center for Interactive Data Analysis, Modelling and Visual

Exploration (INCIDE), University of Konstanz

3 Bioimaging Center (BIC), University of Konstanz

Accepted (2. May 2013) in Archives of Toxicology



34

Abstract

Organ-specific in vitro toxicity assays are often highly sensitive, but they lack specificity.

We evaluated here examples of assay features that can affect test specificity, and some

general procedures are suggested on how positive hits in complex biological assays may be

defined. Differentiating human LUHMES cells were used as potential model for

developmental neurotoxicity testing. Forty candidate toxicants were screened, and several hits

were obtained and confirmed. Although the cells had a definitive neuronal phenotype, the use

of a general cell death endpoint in these cultures did not allow specific identification of

neurotoxicants. As alternative approach, neurite growth was measured as an organ-specific

functional endpoint. We found that neurite extension of developing LUHMES was

specifically inhibited by diverse compounds such as colchicine, vincristine, narciclasine,

rotenone, cycloheximide or diquat. These compounds reduced neurite growth at

concentrations that did not compromise cell viability, and neurite growth was affected more

potently than the integrity of developed neurites of mature neurons. A ratio of the EC50

values of neurite growth inhibition and cell death of > 4 provided a robust classifier for

compounds associated with a developmental neurotoxic hazard. Screening of unspecific

toxicants in the test system always yielded ratios < 4. The assay identified also compounds

that accelerated neurite growth, such as the rho kinase pathway modifiers blebbistatin or

thiazovivin. The negative effects of colchicine or rotenone were completely inhibited by a rho

kinase inhibitor. In summary, we suggest that assays using functional endpoints (neurite

growth) can specifically identify and characterize (developmental) neurotoxicants.



35

Introduction

Toxicological test systems do not only require initial conceptualization and basic

description, as in other fields of science. They also necessitate detailed further development

and a lengthy evaluation process. In some domains, such as cosmetics or drug testing, or in

the pre-selection of environmental toxicants for more extensive testing, assays may be used

without formal validation, if there is sufficient evidence for their scientific validity. Some

regulatory authorities, as well as open platforms such as the evidence-based toxicology (EBT)

consortium, provide guidance on method evaluation (Griesinger et al 2009, Hartung 2010).

For instance, documents have been produced on good cell culture practice (GCCP) (Hartung

et al 2002), on guidelines for data presentation (Leist et al 2010) and assay reliability

(Schneider et al 2009), on how to establish a test system for developmental neurotoxicity

(DNT) and on how to select compounds for DNT testing (Kadereit et al 2012). However, until

now only few test systems in the field of neurotoxicity and developmental neurotoxicity have

been developed further on the basis of such guidance documents (Bal-Price et al 2012,

Fritsche et al 2011). More of this type of work is necessary, as it has recently been noted that

several published studies are necessary for an evaluation of a method according to criteria of

evidence-based toxicology (Judson et al 2013, Stephens et al 2013).

DNT often manifests itself in functional disturbances, that may appear hard to model in

vitro (van Thriel et al 2012). However, it is widely assumed (Bal-Price et al 2012, Hogberg et

al 2009, Kadereit et al 2012) that DNT is ultimately the consequence of the disturbance of

relatively basic biological processes, such as differentiation, proliferation, migration and

neurite growth. Therefore several in vitro systems have been established that test the

disturbance of such biological activities by chemicals (Balmer et al 2012, Frimat et al 2010,

Harrill et al 2011b, Radio et al 2008, Schmid et al 2000, Zimmer et al 2012). One endpoint

that has found a lot of attention is neurite outgrowth (Radio & Mundy 2008). This activity is

required during the formation of the nervous system for the development of dendrites and

axons, and it is a precondition for synaptogenesis and cell connectivity. Different neuronal

cell lines of human or rodent origin can be used to study neurite outgrowth and to measure

disturbances after exposure to toxicants (Harrill et al 2011a). A particular challenge for

development of neurite outgrowth assays is the evaluation of test system predictivity by

comparison to in vivo data. As alternative, it has been suggested to focus more on data quality,

and on a broad evaluation of the biological basis of the test and its mechanistic consistency

under many different situations and types of challenge (Leist et al 2012a).



36

From animal studies, it is known that chemicals can affect neurite growth in different

ways. For instance, in utero cocaine exposure reduced the total length of neurites in the locus

coeruleus of rats (Snow et al 2001). The pesticide diazinon impaired neurite outgrowth in the

forebrain and brainstem of rats, exposed to the chemical on postnatal day 1-4 (Slotkin et al

2006). Inhibited neurite formation was also observed after exposure of 7-day-old rat pups to

ethanol (Joshi et al 2006). In contrast, accelerated growth was observed, e.g. after treatment

with the rho kinase (ROCK) inhibitor Y-27632 which enhanced the sprouting of corticospinal

tract (CST) fibers after CST lesion in adult rats (Fournier et al 2003).

Also in humans, disturbed neurite growth is one of the assumed reasons for disorders of

neural development such as autism spectrum disorders (ASD). In adults with ASD, decreased

axonal length has been observed post mortem in the anterior cingulate cortex (Zikopoulos &

Barbas 2010). Moreover, numerous ASD candidate genes are linked to neurite outgrowth and

neurite guidance (Hussman et al 2011).

The number of in vivo studies analysing the altered growth of neurites under toxicant

stress is limited. It is still a technical challenge to visualize neurites in the developing or

mature brain and to measure changes in the growth rate. New technologies for estimating

neurite density in vivo (Vestergaard-Poulsen et al 2011, Zhang et al 2012) are currently under

development but their application for investigating neurite toxicity in vivo still needs to be

refined. A few studies make use of certain anatomical situations more suitable for analysis.

From these, we know that PCBs can affect dendrite growth of dorsal root ganglion (DRG)

neurons (Yang et al 2009), and that different stress conditions affect dendrites in the

hippocampus (McEwen 1999). Moreover, it is generally known that hypothyroidism during

brain development affects neurite connections (Barakat-Walter et al 2000). Apart from these

pioneering neurodevelopmental studies, there is a large body of evidence, that developed

neurites are particularly sensitive targets of chemical toxicity. A large fraction of known

neurotoxicants specifically targets neurites (Spencer et al 2000). In such cases, specific

neurite degeneration is often occurring independent of cell death. Prominent examples are

chemotherapy-induced neuropathies (Quasthoff & Hartung 2002) after treatment with

platinum compounds or alkaloids such as colchicine or vincristine. Another chemical class

known to induce axonal neuropathy are acrylamide and related structures (LoPachin et al

2002). The above findings suggest that neurites, developed or growing, play an important role

in neurons. Therefore, there is a high need for human cell-based test systems that would

provide data faster and easier than the hitherto used animal models.



37

Several test systems have already been developed successfully to assess neurite

outgrowth in low density cultures (Harrill et al 2010, Mitchell et al 2007, Radio et al 2008,

Ramm et al 2003, Yeyeodu et al 2010), whereas the number of reports based on more

interconnected high-density cultures is quite low (Narro et al 2007, Stiegler et al 2011, Wang

et al 2010), mainly due to the difficulties with assigning a specific neurite to a defined cell. A

particular challenge for toxicological test systems for neurite growth is the definition of

specificity of the observed effect for neurite growth. Identification of such compound features

requires that generally cytotoxic effects are distinguished from effects of chemicals that are

specifically affecting neurite growth, but not overall cell survival. In the present study, we

made use of a human cell-based high-density neuronal test system (Stiegler et al 2011) to

further explore the usefulness of simultaneous measurements of viability and neurite growth

to define assay specificity. For this purpose, the system was challenged with a broad range of

chemicals, including a high number of generally cytotoxic compounds. The dataset generated

with the unspecific toxicants was found to be instrumental for the evaluation of assay

performance with respect to the generation of false positives, and for the identification of

interesting true positive hits. The second goal of the study was to provide data on the

performance characteristics and consistency of the assay under different types of challenge.

For instance, we compared the toxicity of chemicals to developing vs. developed neurites, to

answer the question whether a compound specifically inhibited the outgrowth of neurites. We

also challenged the test system with groups of mechanistically, but not chemically, related

compounds. The mechanistic consistency of the assay was further explored by exposing the

test system to compound mixtures expected to behave additively or antagonistically.



38

Results and Discussion

Conditions and acceptance criteria for the use of neurite growth as test

endpoint

LUHMES cells can be differentiated by addition of tetracycline within 5 days to mature

neurons, as evaluated by the expression of neuronal markers, by changes of their morphology,

and by measurements of electrical activity (Scholz et al 2011). It has been shown earlier that

the cells start expanding their neurites on day 2 (d2) and that quantification of the overall

neurite area at day 3 (d3) is a suitable measure of initial neurite growth (Stiegler et al 2011).

After 4-5 days, this growth is saturated, and a ‘mature’ neurite network of relatively constant

size is established (Scholz et al 2011). We used these characteristics here for two different test

protocols: exposure to chemicals from d2-d3, and measurement on d3 as parameter to assess

‘neurite growth’, and exposure to chemicals from day 5 (d5) – day 6 (d6), and measurement

on d6 as parameter to assess ‘neurite toxicity’ (Fig. 1a). In a first, rough approximation, these

two measures were assumed to reflect developmental neurotoxicity (prevention of neurite

formation) vs neurotoxicity, i.e. damaging effects of compounds to already developed

neurites. We are aware of the fact that such a strict classification represents a strong

simplification of reality. Nevertheless, we assume that comparison of the two assays helps to

identify compounds that act by inhibiting the growth (development) of neurites without

having adverse effects on established neurite structures as such.

The hazardous effect of chemotherapeutic alkaloids such as colchicine, vincristine or of

nocodazole, on neurites is well-established. These microtubule disruptors do not only interfere

with the microtubule organization during cell division, but also with the extension of

microtubules during axonal growth (Daniels 1972, Fontaine-Lenoir et al 2006, Geldof et al

1998). They were therefore considered here as a potential positive control to illustrate the

assay algorithm. The measurement of neurite area is based on a life-cell staining of the total

cell cytoplasm with calcein, and imaging of the result on an automated microscope. The

algorithm identifies then all live structures not belonging to a cell body as viable neurites, as

illustrated in (Fig. 1b). The example images of colchicine (5 nM) effects demonstrate clearly

that the compound reduced the neurite area, while the cell bodies were all still viable. An

important feature of the assay is that it allows for a simultaneous quantification of viable cells

(calcein positive cells) out of the total cell number (all Hoechst33342- stained nuclei) and the

assessment of neurite growth (Fig. 1b).



39

Three microtubule inhibitors, are structurally diverse, and that show different affinities

for tubulin, were used to test the sensitivity and reproducibility of the test system. Colchicine,

nocodazole and vincristine were tested over a large concentration range in three different cell

preparations. The data showed reproducibly an inhibition of neurite growth and a potency

ranking that corresponded to the ranking of tubulin affinities (Correia & Lobert 2001)

(Fig. 1c). A key question is, whether these neurite data can be used as such for the

identification of developmental toxicants and/or their ranking in our assay system. Can

compounds such as colchicine and vincristine really be considered developmental toxicants

Fig. 1: Effect of Microtubule-depolymerising agents on neurite growth Cells were replated at day 2 (d2) into 96 well dishes, and toxicants were added 1 h or 3 days later.

At 24 h after the start of the incubation with chemicals, cells were stained with calcein-AM and H-

33342. The number of viable cells/field and the total neurite area/field were automatically detected

and quantified on a high content screening microscope. a) Exposure scheme of LUHMES cells.

Cells were either treated on d2 for 24 h and endpoints were assessed on d3 (developing cells) or

cells were differentiated until d5, treated for 24 h and measured on d6 (mature cells). b) The upper

row shows representative calcein images on the left side and the corresponding neurite area

detected automatically by the imaging algorithm on the right side. The areas identified as neurites

are marked in red; the nuclei of the cells detected by H-33342 staining are indicated by the circles.

The lower row shows corresponding images of cells treated with 5 nM colchicine. Scale bar =

50 µm. c) Quantification of the neurite area of cells treated on d2 with nocodazole, vincristine or

colchicine. d) Colchicine was added to LUHMES on d2. Resazurin reduction was measured 23 h

later. Subsequently, calcein-AM and H-33342 staining was performed to quantify the number of

viable cells and the neurite area. Blue dashed lines indicate the EC20 values for neurite area (1.9

nM) and viability (10 nM), the black solid line the EC50 of neurite area (4 nM). All data points are

means ± SEM from three independent experiments. *p < 0.05 versus untreated control, #p < 0.05

versus viable cells at that concentration.



40

on the basis of such neurite growth data? We felt that such an interpretation would produce

too many false positives, and that additional criteria would be required to increase the

specificity of the assay. This was examined from different angles in the following

experiments.

The main confounding factor of neurite growth tests may be effects of compounds on

overall cell survival (named here: general viability). For instance, simple detergents (not

assumed to be developmental toxicants or neurite toxicants) may produce similar neurite area

curves as the apparently specific microtubule inhibitors. For this reason, we assessed two

general viability endpoints in all experiments in the same wells used for neurite evaluation:

the relative number (= percentage) of viable cells and the capacity to reduce resazurin to

resorufin. Colchicine was chosen again for an exemplary display. The comparison of all

endpoints at many concentrations of the test compound showed that neurite growth is affected

at much lower concentrations than the general viability. The EC20 for the neurite area was

1.9 nM and for viable cells it was 10 nM (resulting in a ratio > 5). The EC50 of the neurite

area was 4 nM and the viability was still 100% at that concentration (Fig. 1d). A reduction of

neurite growth by 50%, without reduction of viability was also found for nocodazole and

vincristine (Suppl. Fig. 1a, b). Thus, on the basis of this complete set of data, all the three

microtubule inhibitors can be considered as developmental neurotoxicants affecting neurite

outgrowth.

To further explore the relationship of neurite growth and cytotoxicity, we chose a small

set of diverse compounds for further testing. Etoposide, a topoisomerase inhibitor anti-cancer

drug and buthionine sulfoximine (BSO), a metabolic inhibitor of glutathione synthesis, were

chosen as chemicals supposed not to interfere with neurites. Both compounds reduced neurite

growth to a significant extent compared to untreated control cells, and the curve shape of

neurite area did not look much different from that found for the microtubule inhibitors. When

the ‘apparent neurite growth inhibition’ was compared with the reduction of viability, it

became evident that the concentration dependencies for both endpoints were the same (Fig.

2a, b). This was not an averaging effect due to the combination of results from different test

runs, but it was observed in each of at least three independent experiments (Suppl. Fig 1c, d).

The same data as for ‘viable cells’ were also found with the resazurin assay. In fact, the two

tests were used for all experiments in this work, but as the results did not significantly differ,

only one of the endpoints is indicated in most figures. We interpret the findings with BSO and



41

etoposide in a way that ‘apparently inhibited neurite growth’ is a secondary consequence of

reduced viability, and we suggest classifying such compounds as ‘unspecific toxicants’.

This concept has a technical and a conceptual implication. Technically speaking,

unspecific toxicants are neither classical negative (having no effect on the endpoints of the

test system) nor positive hits (showing a specific effect). They form a group of their own.

Conceptually, such compounds have to be interpreted as negative, i.e. as not affecting neurite

growth in any specific way. As this may lead to misunderstandings, it requires some further

specification: such a negative statement does not imply that a compound is not a

developmental toxicant. It only implies that positive evidence for such an activity cannot be

found in this assay system. There is no way to determine whether (a) the compound directly

inhibits neurite growth, and in parallel also reduces viability in this particular cell culture

system, or whether (b) it primarily reduces viability, and that reduced neurite growth is found

because of ongoing cell death. In simple terms, the neurite data cannot be interpreted in a

Fig. 2: Comparison of compounds affecting neurite growth specifically or

unspecifically LUHMES cells were treated as in Fig. 1a; all compounds were added on d2 and effects were

measured 24 h later. All data points are means ± SEM from 3 independent experiments. a)

Etoposide. b) Buthionine sulfoximine (BSO). c) Cycloheximide. d) Paraquat. *p < 0.05 versus

untreated control, #p < 0.05 versus viable cells at that concentration.



42

meaningful way, when they are associated with ongoing cell death. As in all test systems, the

increase of specificity (by including comparison to viability) is accompanied by a decrease in

sensitivity (inability to classify compounds as developmentally neurotoxic, when they affect

cell viability). An example illustrates how changes in the test system may alter sensitivity:

theoretically, cells of another system may be more robust, and tolerate concentrations of e.g.

2 µM etoposide without loss of viability. If neurite growth inhibition in those other cells

would be seen at the same concentration as in LUHMES cells (50% at 2 µM), this other test

system would allow the detection of a developmental neurotoxicity potential of etoposide that

is masked in the LUHMES model by parallel cytotoxicity.

Testing of two further compounds with supposed effects on neurites indicated the need

for some quantitative definition of specificity criteria to define a positive test result.

Cycloheximide, an inhibitor of protein biosynthesis with strong effects on peripheral neurites

(Gilley & Coleman 2010), reduced neurite growth significantly at concentrations at which no

effect on viability was observed (Fig. 2c). There was a large ratio of the two endpoints of the

EC20 values and EC50 values. However, the EC50 for general viability was not reached at

testable compound concentrations. For ratio formation, we therefore introduced the rule that

in this case the highest concentration tested would be used for further calculations. Paraquat is

a pesticide with potential toxicity for dopaminergic neurons (McCormack et al 2002) and it

affected neurite growth of LUHMES more potently than general viability (Fig. 2d). However,

we observed some cytotoxicity at all concentrations associated with strongly reduced neurite

growth. This is admittedly a case that may be classified as positive (specific developmental

neurotoxicant) or negative (unspecific toxicant) depending on the rules of the assay

interpretation model. We decided here to focus mainly on the horizontal shift of the curves as

anchor point for interpretations of the LUHMES assay. This criterion may need to be adapted

as more information on the underlying mechanisms becomes available. The general

usefulness for screen purposes was explored in the following with a large number (> 30) of

compounds suspected to affect neurites.

Classification of substances as specific neurite growth inhibitors

To define thresholds of assay specificity, we used unspecific toxicants. Nine compounds

were chosen according to the following rules: (a) they are not known to affect neurite growth,

(b) their known mode of action and their chemical properties make it unlikely that they

specifically affect the biology of neurite growth. The selected chemicals were the glutathione

synthesis inhibitor buthionine sulfoximine (BSO), the mitochondrial uncouplers CCCP and



43

Fig. 3: Separation of specific neurite growth modulators from unspecific

cytotoxicants Cells were treated on d2 as displayed in Fig. 1a, and 24 h later neurite area and viability were

automatically quantified. Compounds were tested at several concentrations, and their EC50 values

for effects on neurite area and cell viability were determined by a non-linear regression sigmoidal

concentration-response curve fit. The EC50 values of the neurite area were plotted against the

EC50 values of general cell viability. First, a reference control group of 9 unspecific toxicants was

measured (dots in grey, names are underlined). These comprised buthionine sulfoximine (BSO),

carbonylcyanide-3-chlorophenylhydrazone (CCCP), 2,4-dinitrophenol (2,4-DNP), etoposide,

bisbenzimide H (H-33352), potassium chromate (K2CrO4), tert-butyl hydroperoxide (tBuOOH),

tween-20 and sodium dodecyl sulfate (SDS). The solid line indicates equal EC50 of viability and

neurite area. The dashed line indicates an EC50 ratio of 4. Data for 40 compounds were classified

according to this threshold value. Orange colour indicates substances classified to act unspecific on

neurite growth: acrylamide, antimycin A, chlorpyrifos, chlorpyrifos oxon, cisplatin, cytochalasin,

fipronil, haloperidol, honokiol, IPA-3, menadione, methamphetamine (METH), mevastatin, 1H-

[1,2,4]oxadiazolo-[4,3-α]quinoxalin-1-one (ODQ), okadaic acid, oligomycin, piericidin, protein

tyrosine phosphatase inhibitor IV (PTP IV), puromycin, simvastatin and SP600125. Substances

classified as specific neurite growth inhibitors (light blue) were: bisindolylmaleimide I (Bis1),

brefeldin A, colchicine, cycloheximide, diquat, flavopiridol, methylmercury (II) chloride (MeHg),

sodium orthovanadate (Na3VO4), narciclasine, nocodazole, paraquat, rotenone, U0126 and

vincristine. Substances that increased the neurite area (dark blue) were: blebbistatin, HA-1077,

H1152, thiazovivin and Y-27632. The ‘neurite EC50’ of these compounds was defined as the

concentration resulting in a half-maximal increase of the neurite area. Data are means ± SD of 3

separate screens.



44

2,4-DNP, the detergents SDS and tween-20, the heavy metal ion K2CrO4, the DNA-

interacting compounds etoposide and H-33352 and the oxidant tertiary butyl-hydroperoxide

(tBuOOH). This set of chemicals was used as ‘unspecific controls’, i.e. to define non-specific

outcomes of the neurite growth inhibition assay. For this purpose, we determined their EC50

values for neurite growth inhibition and for reduction of general viability, and the ratio of

these EC50 values was calculated for each experiment and each compound. The average ±

standard deviation (SD) of all these ratios was 1.4 ± 0.83, i.e. neurites were on average

affected by unspecific compounds at slightly lower concentrations than general viability. For

defining criteria for ‘positive responses’ we used a rule commonly used for many analytical

methods as guidance: we assumed that significant effects (of specific compounds) should be 3

SD away from the baseline (average of unspecific compounds). Thus, we defined a ratio of 4

as threshold/acceptance criterion for compounds we regarded as positive hits of the screen

(Fig. 3). Typical positive controls known from previous studies (U0126, flavopiridol,

brefeldin A, bisindolylmaleimide I and sodium orthovanadate (Na3VO4) (Harrill et al 2010,

Radio et al 2008, Radio & Mundy 2008, Stiegler et al 2011) had EC50 ratios far above 4.

Compounds with an EC50 ratio < 4 were defined as negative. This rule is the pivotal basis for

conferring specificity to the assay, even though it may reduce its sensitivity. A negative

classification in our assay means that there is no positive evidence for a neurite growth

inhibition. It is not evidence of absence of such a property.

Using these criteria, we screened substances, that we found likely to affect neurite growth

because of their assumed primary mode of action or because of reports in the literature (Fig.

3, Suppl. Fig. 2). The tested compounds comprised many biological activity groups like

cytostatic drugs (cisplatin), redox cyclers/pesticides (paraquat, diquat), mitochondrial toxins

(rotenone, antimycin A, oligomycin, piericidin), cytoskeleton toxicants (colchicine, okadaic

acid, nocodazole, vincristine), acetylcholine-esterase inhibitors (chlorpyrifos, chlorpyrifos-

oxon) and other substances like the neurotoxin acrylamide, the guanylylcyclase inhibitor

ODQ, the antipsychotic and possible teratogenic drug haloperidol, the stress kinase inhibitor

SP600125, the HMG-CoA reductase inhibitors simvastatin and mevastatin, an inhibitor of

protein tyrosine phosphatases PTP IV, the RhoA activator narciclasine, a group of rho kinase

(ROCK) inhibitors (H-1152, HA-1077, thiazovivin) and the myosin II inhibitor blebbistatin.

The tested microtubule inhibitors colchicine, nocodazole and vincristine were classified

as neurite growth specific toxicants. In the group of tested pesticides consisting of rotenone,

paraquat, diquat, chlorpyrifos and chlorpyrifos-oxon, a clear positive effect was determined



45

for rotenone, paraquat and diquat. Another group of compounds influencing neurite growth

were Rho/ROCK pathway modifiers. Some of them accelerated the neurite growth, instead of

inhibiting it.

The ratio of the EC50 values of neurite growth versus viability proved to be a useful

classifier for compounds associated with a developmental neurotoxic hazard. The results

shown here are based on average EC50 values derived from three biological replicates

(independent experiments).

For more extensive screens, a more simplified procedure is desirable. Therefore, we

examined how the classification would have looked like for individual experiments. Also the

data points of the positive compounds from individual experiments all fell within the ‘specific

area’ of the scatter plot (Suppl. Fig. 2). The approach taken here is firmly established in the

field of biomolecular screening, as performed in pharmaceutical industry, but it differs from

the traditional reporting of in vitro test systems in toxicology. The more traditional approach

in this field is based on statistical evaluation of a compound effect vs. a negative control. The

specificity definition we have chosen here is easily adaptable to other situations, including

simpler assays with a single endpoint. Measures based in some way on the

variation/confidence limits of the reference group can always provide a useful tool to classify

further tested compounds: either the compounds are within the ‘noise limit’ ( negative

classification), or outside the background noise ( specific hits). In the LUHMES test system

we newly identified 7 specific neurite growth inhibitors (rotenone, narciclasine, colchicine,

vincristine, nocodazole, paraquat, diquat) and 4 neurite growth accelerators (H1152, HA-

1077, thiazovivin, blebbistatin). These results will be displayed and discussed in greater detail

in the following sections.

Specific effects of rotenone, but not other respiratory chain toxins

Rotenone, a complex I inhibitor of the mitochondrial respiratory chain, inhibited neurite

growth significantly at 0.1 µM (Fig. 4a, green solid line), whereas viability was affected only

at ten times higher concentrations. This was initially surprising. To identify potential

artefacts, the original images of the high-content screen were retrieved and evaluated by

trained observers. The effect was fully confirmed, and representative example images show

clearly that neurite area was reduced by low concentrations of rotenone, while the number of

viable cells per field was not affected. Only at higher concentrations, a concentration-

dependent decrease in cell number was observed, and all viable cells were completely devoid



46

of neurites (Fig. 4b). This big difference of effects on neurite growth and viability was

observed in four independent experiments (Fig. 4c). To follow up on this positive hit we

asked the question whether rotenone targets neurites in general or whether it specifically

Fig. 4: Reduction of neurite growth by

rotenone and other respiratory chain

inhibitors LUHMES cells were treated as in Fig. 1a. a)

Rotenone was added in fresh medium either 1

h after replating on d2 or on day 5. After 24 h

incubation, viability and neurites were

measured, and normalized to untreated

controls. Viability curves of d3 and d6 were

similar. Data are means ± SEM from 3

independent experiments. b) Representative

images are shown, in which the automatically

detected neurite area (red) is overlaid over the

calcein images. The position of the nuclei is

marked by a blue outline. The width of the

micrographs shown is 330 µm. Cells were

incubated on d2 for 24 h with the indicated

concentration of rotenone. c) Cells were

treated on d2 with rotenone for 24 h. The data

for viability and neurites are displayed for 4

independent experiments (dashed lines), each

run in technical triplicates (individual error

bars (± SD)). d) Antimycin A or oligomycin

were added to LUHMES after replating for 24

h. Viability and neurite data are means from 2

(oligomycin) and 3 (antimycin A)

independent experiments. All data points are

means ± SEM from at least two independent

experiments. *p < 0.05 versus untreated

control, #p < 0.05 neurite area versus viable

cells at that concentration.



47

influences their growth. Therefore we compared the effects of rotenone (24 h exposure in both

cases) on differentiating LUHMES on day 2 (d2) with its effects on mature cells with a fully

differentiated neurite network on day 5 (d5) (Fig. 4a, green dashed line). The mature neurites

were less sensitive to rotenone. In fact the concentration-dependency of the neurite

degeneration was not significantly different from the one for general viability, when mature

neurons were used as model system (Fig. 4a, orange dashed line). The general cytotoxicity of

rotenone was the same for d2 and d5 cells (Fig. 4a viability for d5, Fig. 4c viability for d2),

and only the sensitivity of the neurites was different. Thus, rotenone is an example for a

compound with a higher toxic potency for the developing neurons than for the developed

cells. This effect was unique for rotenone, as we found no other mitochondrial toxicant with

such an effect on neurites. Complex I inhibitor piericidin, complex V inhibitor oligomycin

(Fig. 4d) as well as the uncouplers of oxidative phosphorylation CCCP or 2,4-DNP had no

impact on neurites at several tested concentrations. For the complex III inhibitor antimycin A

we identified concentrations (25-50 µM) at which neurites were significantly more affected

than viability. But the EC50 ratio of viability to neurite area was only 1.6 (Fig. 4d), whereas

rotenone showed a ratio > 15. According to our rules, antimycin A was classified as negative.

Other reports, using rodent cells (PC 12 cells, primary hippocampal neurons), also

suggest that rotenone has some specific effect on axon formation (Sai et al 2008, Sanchez et al

2007). The mechanism is unknown, but it has been suggested that complex I inhibitory

parkinsonian toxicants may affect dopaminergic neurons by microtubule depolymerization

(Ren et al 2005). Other processes which are also dependent on correct microtubule formation

like migration and proliferation have also been shown to be inhibited by rotenone in

mesencephalic neural stem cells (Ishido & Suzuki 2010). The process of microtubule

formation is indeed crucial for the growth of axons as suggested for instance by our findings

on colchicine and related compounds. To identify the underlying mechanisms of rotenone’s

developmental neurotoxicity more clearly, in depth experiments and additional technical

approaches are needed. As neurons can tolerate a partial depletion of ATP for long times, if

secondary apoptotic processes are blocked (Poltl et al 2012, Volbracht et al 1999), a specific,

cell death independent action of rotenone on young developing neurons seems likely.

Differential chemical effects on neurite growth vs. neurite stability

Our observation that rotenone specifically targets neurite growth (d2 cells), as compared

to neurite stability (d5 cells), suggests that such a distinction may be used more generally to

define the specificity of an assay (or a compound) for neurite growth inhibition. For this



48

purpose, we tested a group of eleven compounds, which had been classified as specific neurite

toxicants in the d2d3 neurite growth assay, on neurite degeneration (d5d6) (Fig. 1a).

Concentration-response curves were obtained for effects on the neurite area and general

viability for d5 cells treated for 24 h. Scatter plots of the effects on EC50 (general viability)

vs. EC50 (neurite area) measured on d3 (Fig. 5a) or d6 (Fig. 5b) showed that most compounds

did not affect the mature neurites in a specific way (without killing the cells). Nine of the

Fig. 5: Comparison of toxicant effects on d3

and d6 Cells were replated at d2 and eleven compounds were

tested with at least five different concentrations on d2

or on d5. After 23 h resazurin reduction was

measured. Subsequently cells were stained with

calcein-AM and 1 H-33342 for 30 min. The number

of viable cells and the neurite area were automatically

detected by Cellomics Array Scan. EC50 values of

neurite area were plotted against the EC50s of

viability. The dashed line indicates equivalent EC50

values of neurite area and viability. In cases of low

cytotoxicity of compounds, the highest concentration

measured was used as ‘EC50 viability’. All data are

means of 3 independent experiments. a) Comparison

of effects on viability and neurites on d3. b)

Comparison of effects on viability and neurites on d6.

c) Scatter plot of different EC50 ratios of the same

compounds as in a/b. i) EC50 ratio of resazurin

reduction of d6 to d3, ii) EC50 ratio of calcein

positive cells to neurite area of d6, iii) EC50 ratio of

neurite area of d6 to d3, iv) EC50 ratio of calcein

positive cells to neurite area of d3.



49

eleven compounds were located on the dashed line, indicating identical EC50 values for both

endpoints at d6.

The data obtained in these experiments also allowed to answer the question, whether the

EC50 values for neurites or for general viability were shifted in absolute terms between d3

and d6 cells. Seven of the compounds were much more potent on developing neurites, than on

developed neurites, and the average of the ratios of EC50 (neurites d6)/EC50 (neurites d3)

was 11.4. This means that the functional endpoint of neurite growth is more sensitive towards

toxicant exposure (Fig. 5c and Suppl. Fig. 3a). To test whether developing cells are in general

more sensitive to toxicant exposure than mature cells, we compared the EC50 values of

resazurin reduction. The ratio of this endpoint for the two developmental stages of d6 to d3 is

0.74. This suggests that the general cytotoxicity is independent of the developmental stage of

the cells, and that the younger cells are not less robust than adult cells (Fig. 5c and Suppl. Fig.

3b).

The above data suggest indirectly, that general cytotoxicity data are no good predictor for

neurotoxicity, even though they are obtained from neuronal cultures. To examine this point in

more detail we selected a subgroup of our test compounds. They comprised neurotoxicants

such as MnCl2, acrylamide and trimethyltin chloride, as well as neurite growth inhibitors and

non-neurotoxicants for which literature values could be found in the Halle registry of

cytotoxicity data (Halle 2003). Resazurin reduction of d3 (Suppl. Fig. 4a) and of d6 cells

(Suppl. Fig. 4b) was plotted against the data from the Halle registry, which are based on

average cytotoxicity tests on several non-neuronal human cell lines such as HeLa and

HEK293. The LUHMES cytotoxicity data and the Halle registry values correlated to about

85%. This means that the cytotoxicity of compounds determined in young or mature

LUHMES as test system correlates to a high degree with that observed in other human cell

lines of non-neuronal origin. This corroborates our assumption that human neurotoxicity

cannot be determined by cytotoxicity measurements in human neuronal cell cultures, and that

only a specific functional assay, such as neurite growth, yields specific results. Observations

pointing into a similar direction were also made in other model systems (Gartlon et al 2006).

Independence of key findings from data processing algorithm

Our concept of whole curve comparisons does not allow statements on individual

concentrations of a given compound. Therefore, we were interested how individual test

conditions (defined concentrations of defined compounds) would distribute in a scatter plot



50

that correlates effects on neurites with those on general viability. We produced a scatter plot

of the individual data points for each concentration of a test compound, so that inhibition of

neurite growth and the general cytotoxicity were used as coordinates. The data were plotted

for three groups of compounds: negative controls, unspecific controls and eight neurite

growth inhibitors (Fig. 6). Negative controls were mannitol and acetylsalicylic acid. They did

not affect any endpoint, even though concentrations up to 4 mM for mannitol and 2 mM for

acetylsalicylic acid were chosen (Fig. 6, green dots). Unspecific compounds, like SDS, BSO,

etoposide, oligomycin, tBuOOH, affected neurite growth and viability to a more or less

similar extent at all tested concentrations (Fig. 6, black dots). More detailed analysis shows,

that for such compounds concentrations exist, at which they reduce general cell viability

Fig. 6: Comparison of endpoint ratios (general viability vs. neurite area) of positive

hits and unspecific toxicants at defined concentrations LUHMES cells were treated and measured as in Fig.3. Each concentration for each compound is

represented by one individual dot in the scatter plot. The effects of substances on viability are

plotted against effects on the neurite area. The dashed line indicates equivalent values for neurite

area and viability. Negative controls, such as aspirin and mannitol are marked by green dots. Black

dots display values for unspecific compounds: buthionine sulfoximine (BSO), 2,4-dinitrophenol

(2,4-DNP), etoposide, bisbenzimide H (H-33352), menadione, oligomycin, tert-butyl

hydroperoxide (tBuOOH), tween-20, saponin and sodium dodecyl sulfate (SDS). Data from

specific compounds are marked by blue dots: bisindolylmaleimide I (Bis1), brefeldin A,

colchicine, cycloheximide, MeHg, Na3VO4, nocodazole, paraquat, rotenone, U0126, and

vincristine. The dashed grey box encircles dots which showed a reduction in neurite area of

> 35 % and in viability of ≤ 20 %.



51

significantly (by 30-60%), but neurite area is reduced much more (by up to 35% more).

Antimycin A, mentioned in the paragraph above (Fig. 4d) is also such a compound. It is

important for the understanding of our test approach that the specificity rule we used here

classifies such substances as negative. The compounds classified as specific inhibitors of

neurite growth localized differently in the scatter plot: there were always data points that

showed a clear impact on neurites, with no major influence on viability (Fig. 6, blue dots).

This way of data evaluation (based on different principles than the EC50 ratios) could form an

alternative basis for a specificity rule. It appears to be useful as an option for smaller screens.

Notably, the newly identified neurite growth inhibitors found in our screen would also have

been detected based on these alternative criteria.

As a further control, we also examined, whether other methods to quantify neurite effects

would lead to similar results. For this purpose, we counted the percentage of cells with or

without neurites for several representative experimental conditions, and using the same

images that had been used for the automated neurite area quantification algorithm. The neurite

area endpoint correlated well with the number of cells with neurites obtained by manual

counting (Fig. 7). A smaller number of conditions was also used for automated counting of

neurite-containing cells, based on specifically-developed software (Schoenenberger et al

2012). Also in this case, the endpoints correlated well. We conclude from these comparisons,

Fig. 7: Comparison of the field

based algorithm with a single cell

based readout LUHMES cells were treated as in Fig.

1a. Incubations were started on d2 and

ended 24 h later to assess neurite area

and number of viable cells per field.

Cells were treated with substances

(BFA = brefeldin A, Bis1 =

bisindolylmaleimide I, CHX =

cycloheximide, Flavo = flavopiridol,

Men = menadione) at the

concentrations indicated, and neurite

area was assessed automatically (by

the field-based Cellomics algorithm).

The same images were re-analysed

manually. Every individual cell was

scored for having a neurite extension

that was longer than the corresponding

cell soma diameter or not. Data from

the field-based algorithm (y-axis) were

compared with manually counted cells

with neurites (x-axis).



52

that the toxicological effects we observed for neurite growth inhibitors in the LUHMES assay

(as presented here) are robust, and can be detected by different analytical methods as well.

Detection of compounds that increase the neurite area

An important parameter for each assay is its dynamic range. A particular question is,

whether deviations from normal can be measured into both directions, and which types of

positive controls can be used. The compound Y-27632 has been known to affect LUHMES

neurite growth positively (Stiegler et al 2011). These findings, and other literature data

(Fuentes et al 2008, Kubo et al 2008, Nikolic 2002) pointed to a role of the ROCK pathway in

the control of neurite growth. The pathway is triggered by the activated RhoGTPase RhoA

that binds to the rho kinase (ROCK), and activates it thereby. ROCK phosphorylates myosin

light chain (MLC), and this results in the induction of actin-activated non-muscle myosin II

ATPase. The downstream consequences are a local collapse of the neuritic growth cone and

induction of stress fibers (Kubo et al 2008). An inhibition of this pathway would therefore

lead to an accelerated neurite growth due to less stress fiber formation and a reduced tendency

of growth cone collapse. The role of this signalling cascade for our test system was explored

further by the use of different compounds that affect this pathway. We found that the different

ROCK inhibitors H1152, HA-1077 and thiazovivin as well as the myosin II inhibitor

blebbistatin accelerated neurite growth significantly (Fig. 8a, c-e). The area of the culture dish

covered with neurites was increased by up to 80% (with inhibitor HA-1077). When the same

compounds were used on d5 LUHMES no measurable effect was observed (data not shown).

Inhibition of the ROCK pathway therefore seems to have a particularly prominent role in the

growth process of neurites. Whether compounds leading to accelerated neurite growth should

be interpreted as toxicants is an open issue and should be the subject of further investigations.

In regenerative medicine, and in adults, accelerated outgrowth or preservation of neurites

would rather be considered beneficial (Hansson et al 2000, Schierle et al 1999, Volbracht et al

2001, Volbracht et al 1999, Volbracht et al 2006). There is, however, some published

evidence that uncontrolled elongation of neurites during development may be related to

neurotoxicity: hypertrophic dendritic outgrowth has been observed in parts of the embryonic

prefrontal cortex after cocaine had been administered to pregnant rabbits at gestational stages

(Jones et al 2000, Stanwood et al 2001).

We also investigated potentially toxic effects of the ROCK pathway activation.

Narciclasine, which greatly increases Rho A's activity (Lefranc et al 2009), strongly decreased

neurite growth (Fig. 8b, e). These data underline the mechanistic consistency of the assay, as



53

the achieved results were as expected considering the interference of these compounds with

the ROCK pathway. The fact, that we can detect accelerating as well as inhibitory effects on

neurite growth gives evidence of the broad dynamic range of our growth assay. Possibly the

test system may also be used for pharmacological questions, e.g. for identification of

compounds that facilitate neuroregeneration by accelerating neurite growth.

Fig. 8: Modulation of neurite outgrowth via the ROCK/RhoA pathway At d2 cells were replated into 96 well plates and compounds were added at the concentrations

indicated. At 24 h later, cells were stained with calcein-AM and 1 H-33342 for 30 min at 37° C.

Neurite area and viability were automatically detected using Cellomics array scan. a) Thiazovivin.

b) Narciclasine. c) HA-1077. d) Blebbistatin. e) Representative micrographs i) control, ii)

narciclasine, iii) HA-1077, iv) blebbistatin. The width of each micrograph corresponds to 210 µm.



54

Biological effects of combinations of substances

In the last step of the assay evaluation, we tested the effect of combinations of

compounds. Consistent responses of the test system to at least binary mixtures would indicate

its usefulness for more mechanistic questions and for exploring toxicity intervention.

Moreover, we hoped to find additional evidence for the specificity of the hits discovered in

our initial screen (Fig. 3). In a first set of experiments, we replicated earlier findings on the

combination of the two kinase inhibitors bisindolylmaleimide I (Bis1) and U0126 (Stiegler et

al 2011). At certain drug concentrations, the neurite area could nearly be reduced to zero,

without a significant reduction of cell viability. Moreover, three entirely independent

Fig. 9: Antagonistic and additive effects of different neurite growth modifiers LUHMES cells were treated as in Fig. 1a. Incubations were started on d2 and 24 h later, the

neurite area and the number of viable cells per field were assessed. a) For all data points shown,

2 µM of bisindolylmaleimide I (Bis1) was added. In addition, different concentrations (0 –

12.5 µM) U0126 were added at the same time. The data for Bis1 alone are shown on the left part

of the x-axis. b) Combination of ROCK-inhibitor Y-27632 plus 2 µM Bis1. c) Combination of

ROCK-inhibitor Y-27632 plus 5 nM of Colchicine. d) Combination of ROCK-inhibitor Y-27632

plus 0.1 µM of Rotenone. All data are means ± SEM of 3 independent experiments. Data are

normalized to untreated controls (ctrl). *p < 0.05 versus single compound treatment indicated on

the left part of the x-axis.



55

experiments with the combination of two chemicals gave consistent results (Fig. 9a). We

interpret this as indication for a high reproducibility and robustness of the test system.

A recent pilot study (Stiegler et al 2011) indicated that the ROCK inhibitor Y-27632 is

able to counteract the neurite growth inhibition of the MAP kinase (MAPK) inhibitor U0126

and that U0126 diminished the neurite accelerating effects of Y-27632. Such effects were

now explored on a broader basis. We used the PKC inhibitor Bis1 to reduce neurite growth.

Then, cells were co-exposed to eight different concentrations of the ROCK inhibitor Y-27632.

A concentration of about 1 µM of the ROCK inhibitor brought the neurite area back to 100%

(from a low start level of 60% by using the PKC inhibitor alone), and concentrations of

10 µM increased the neurite area to 130% of untreated controls (Fig. 9b). The potency of Y-

27632 (half maximal effect at about 2 µM) was similar to its potency, when used alone (Fig.

3). Interestingly, Y-27632 also counteracted the growth-decreasing effects of colchicine and

of rotenone; the concentration of the ROCK inhibitor required to show significant effects was

always in a similar narrow concentration range, and the set of experiments yielded highly

reproducible data (Fig. 9c-d).

In these experiments, the toxic effects of rotenone and colchicine were neutralized by a

treatment that supposedly promotes neurite growth, but does not affect the binding of the

toxicants to the primary targets (tubulin or mitochondrial complex I). These findings suggest

that the adverse outcome of toxic compound exposure may not only depend on the assumed

molecular initiating event, but also on many other factors. However, this would require

detailed investigation in a more mechanistically-oriented study. There is in fact evidence from

the literature that the ROCK pathway may affect microtubule stability (Gorovoy et al 2005,

Takesono et al 2010). Some earlier data from other cellular systems suggest a rescuing effect

of Y-27632 after treatment with rotenone (Sanchez et al 2007) or with microtubule

destabilizing compounds, such as colchicine or nocodazole (Keller et al 2002, Niggli 2003,

Zhang et al 2001). These results from our test system corroborate such findings and indicate a

good technical and mechanistic consistency of the test system. Intervention with toxicant

effects would not only be helpful for clarifying the mode of action of DNT compounds, but it

could also be interesting to explore potential rescue strategies after poisoning.

The practical application of toxicological in vitro test systems requires an extensive

characterization of their performance characteristics. Especially the regulatory use of new

animal-free assays has been strictly coupled to a formal validation procedure, as performed



56

e.g. in Europe by the European Centre for Validation of Alternative Methods (ECVAM)

(Corvi et al 2012, Griesinger et al 2010). Before such a time- and resource-consuming

validation is performed, it is now common practice to pre-validate e.g. assay reproducibility

and biological relevance. For high-throughput assays, evaluations similar to a formal pre-

validation have been suggested as routine procedure to assess the usefulness and performance

of the assays (Judson et al 2013). In both cases, this step of assay establishment requires a

broad range of data to be generated, and multiple compounds to be used. This process usually

goes far beyond an initial publication of a new test system (Hartung 2007, Hartung 2010,

Leist et al 2012a). We have attempted here to provide such data and to provide a transparent

and broad description of a test system that may be taken as example for similar approaches

with other test systems.



57

Materials and Methods

Materials and chemicals:

Acrylamide, antimycin A, acetylsalicyl acid, blebbistatin, brefeldin A, buthionine

sulfoximine (BSO), calcein-AM, carbonyl-cyanide-3-chlorophenylhydrazone (CCCP),

chlorpyrifos, cisplatin, colchicine, cycloheximide, cytochalasin B, dibutyryl-cAMP (cAMP),

2,4-dinitrophenol, diquat dibromide, etoposide, fibronectin, fipronil, flavopiridol, hoechst

bisbenzimide H-33342, honokiol, IPA3, potassium chromate (K2CrO4), mannitol, menadione,

methylmercury (II) chloride (MeHg), mevastatin, narciclasine, nocodazole, oligomycin,

paraquat dichloride, puromycin, resazurin sodium salt, rotenone, saponin, sodium

orthovanadate (Na3VO4), SP600125, tert-butyl hydroperoxide (tBuOOH), tetracycline and

vincristine were from Sigma (Steinheim, Germany).

Recombinant human FGF-2 and recombinant human GDNF were from R&D Systems

(Minneapolis). Bisindolylmaleimide I (Bis1), dimethyl sulfoxide (DMSO), 1H-

[1,2,4]oxadiazolo[4,3-α]quinoxalin-1-one (ODQ), okadaic acid potassium salt, PTP inhibitor

IV, H1152, simvastatin and U0126 were from Calbiochem (Darmstadt, Germany). Y-27632

was from Tocris Bioscience (Bristol, UK), tween-20 and sodium dodecyl sulfate (SDS) were

from Roth (Karlsruhe, Germany), HA-1077 from Ascent scientific (Cambridge, UK),

thiazovivin from Selleck (Munich, Germany), chlorpyrifos oxon from Chem. Service inc.

(West Chester, USA), piericidin from Enzo life science (Lörrach, Germany) and

methamphetamine was obtained from Lipomed (Arlesheim, Switzerland). All culture reagents

were from Gibco unless otherwise specified.

Cell culture:

Handling of LUHMES human neuronal precursor cells was performed as previously

described in detail (Lotharius et al 2005, Schildknecht et al 2009, Scholz et al 2011). Briefly

maintenance of LUHMES cells was performed in proliferation medium, consisting of

advanced DMEM/F12 containing 2 mM L-glutamine, 1 x N2 supplement (Invitrogen), and 40

ng/ml FGF-2 in a 5% CO2/95% air atmosphere at 37° C. LUHMES cells were passaged every

other day and kept until passage 20. For differentiation 8 million cells were seeded in a

Nunclon T175 in proliferation medium for 24 h. The next day medium was changed to

differentiation medium (DM II), consisting of advanced DMEM/F12 supplemented with

2 mM L-glutamine, 1 x N2, 2.25 µM tetracycline, 1 mM dibutyryl 3’,5’-cyclic adenosine

monophosphate (cAMP) and 2 ng/ml recombinant human glial cell derived neurotrophic



58

factor (GDNF). 48 h later cells were trypsinised, and seeded in a density of 100000 cells/cm²

on dishes precoated with 50 µg/ml poly-L-ornithine (PLO) and 1 µg/ml fibronectin in

advanced DMEM/F12 containing 2mM L-glutamine, 1 x N2 and 2.25 µM tetracycline but

without cAMP and GDNF (DM).

Standard experimental setup:

To detect effects on neurite growth, cells were seeded at a density of 30,000 cells per

well in 50 µl DM on PLO/fibronectin coated 96-well dishes. Compounds were serially diluted

in DM, and 50 µl were added to the cells 1 h after seeding. Analyses were performed 24 h

after initiation of the treatment. To detect effects on neurite degeneration cells were seeded at

the same density in 100 µl DM. At day 5 (d5) DM was removed and 100 µl of fresh DM with

serially diluted compounds were added. Analyses were performed 24 h or 72 h later. The

maximum DMSO concentration used was 0.33% and had no influence on cell viability or

neurite growth.

Resazurin measurement:

Cell metabolic activity was detected by a resazurin assay (Schildknecht et al 2009).

Briefly, 10 µl resazurin solution were added to the cell culture medium to obtain a final

concentration of 10 µg/ml. After incubation for 30 min at 37° C, the fluorescence signal was

measured at an excitation wavelength of 530 nm, using a 590 nm long-pass filter to record the

emission. Fluorescence values were normalized by setting fluorescence values of untreated

wells as 100% and the values from wells containing less than 5% calcein-positive cells as 0%.

Quantification of neurite outgrowth

Neurite growth was detected as previously described in detail (Stiegler et al 2011).

Briefly, cells were stained with 1 µM calcein-AM and 1 µg/ml H-33342 for 30 min at 37° C.

An Array-Scan VTI HCS Reader (Cellomics, PA) equipped with a Hamamatsu ORCA-ER

camera was used for image acquisition. Ten fields per well were imaged in two channels

using a 20x objective (2 x 2 pixel binning). Excitation/emission wavelengths of 365 ± 50/535

± 45 were used to detect H-33342 in channel 1 and 474 ± 40/535 ± 45 to detect the calcein

signal in channel 2.

Nuclei were identified as objects in channel 1 according to their size, area, shape, and

intensity. The nuclear outlines were expanded by 3.2 µm in each direction, to define a virtual

cell soma area (VCSA) which was bigger than the average cell size to reduce false positive

neurite areas. All calcein-positive pixels of the field were defined as viable cellular structures



59

(VCSs). In an automatic calculation, the VCSAs, defined in the H-33342 channel, were used

as filter in the calcein channel and subtracted from the VCS. The remaining pixels (VCS -

VCSA) in the calcein channel were defined as neurite area.

Statistics and data mining:

Data are presented, and statistical differences were tested by ANOVA with post-hoc tests

as appropriate, using GraphPad Prism 5.0 (Graphpad Software, La Jolla, USA).



60

-1 0 1 2

0

50

100

viable cells

neurite area

log BSO [µM]

Via

bil

ity p

ara

me

ter

[%

of

co

ntr

ol

SD

]

-3 -2 -1 0 1

0

50

100

viable cells

neurite area

log Etoposide [µM]V

iab

ilit

y p

ara

me

ter

[% o

f c

on

tro

l

SD

]

A

B

Fig. S1 Toxicity curves of two

positive compounds,

vincristine and nocodazole,

and of two negative

compounds, etoposide and

BSO.

Cells were replated at day 2 (d2)

and compounds were added in

dilution series in triplicates. 24h

later cells were stained with 1

µM calcein-AM and 1 µg/ml H-

33342 for 30 min at 37 C.

a) mean curve of vincristine

toxicity of three biological

replicates. b) mean curve of

nocodazole toxicity of four

biological replicates. c) single

curves of viability and neurite

area of three independent

experiments of etoposide.

d) single curves of viability and

neurite area of three independent

experiments of BSO.

0

50

100

-4 -3 -2 -1 0

viable cells

neurite area

crtl

log Vincristine [µM]

Via

bilit

y p

ara

mete

r

[% o

f co

ntr

ol

SE

M]

0

50

100

-3 -2 -1 0 1

viable cells

neurite area

crtl

log Nocodazole [µM]

Via

bilit

y p

ara

mete

r

[% o

f co

ntr

ol

SE

M]

C

D

Supplements

Fig. S1: Toxicity curves of two positive compounds, vincristine and nocodazole, and of

two negative compounds, etoposide and BSO.

Cells were replated at day 2 (d2) and compounds were added in dilution series in triplicates. 24h later

cells were stained with 1 µM calcein-AM and 1 µg/ml H-33342 for 30 min at 37° C. a) mean curve of

vincristine toxicity of three biological replicates. b) mean curve of nocodazole toxicity of four

biological replicates. c) single curves of viability and neurite area of three independent experiments of

etoposide. d) single curves of viability and neurite area of three independent experiments of BSO.

-1 0 1 2

0

50

100

viable cells

neurite area

log BSO [µM]

Via

bil

ity p

ara

me

ter

[%

of

co

ntr

ol

SD

]

-3 -2 -1 0 1

0

50

100

viable cells

neurite area

log Etoposide [µM]

Via

bil

ity p

ara

me

ter

[% o

f c

on

tro

l

SD

]

A

B

Fig. S1 Toxicity curves of two

positive compounds,

vincristine and nocodazole,

and of two negative

compounds, etoposide and

BSO.

Cells were replated at day 2 (d2)

and compounds were added in

dilution series in triplicates. 24h

later cells were stained with 1

µM calcein-AM and 1 µg/ml H-

33342 for 30 min at 37 C.

a) mean curve of vincristine

toxicity of three biological

replicates. b) mean curve of

nocodazole toxicity of four

biological replicates. c) single

curves of viability and neurite

area of three independent

experiments of etoposide.

d) single curves of viability and

neurite area of three independent

experiments of BSO.

0

50

100

-4 -3 -2 -1 0

viable cells

neurite area

crtl

log Vincristine [µM]

Via

bilit

y p

ara

mete

r

[% o

f co

ntr

ol

SE

M]

0

50

100

-3 -2 -1 0 1

viable cells

neurite area

crtl

log Nocodazole [µM]

Via

bilit

y p

ara

mete

r

[% o

f co

ntr

ol

SE

M]

C

D



61

Fig. S2: Separation of specific neurite growth modulators (individual experiments) from

unspecific cytotoxicants.

-3 -2 -1 0 1 2 3-3

-2

-1

0

1

2

3

H1152 HA-1077Thiazovivin

Blebbistatin

Y-27632

U0126

Bis I

Na3VO4

MeHg

Brefeldin A

Flavopiridol

Cycloheximide

Diquat

Colchicine

Rotenone

Vincristine

Nocodazole

Paraquat

Narciclasine

SDS

K2CrO4

H-33352CCCP

BSO

Etoposide

2,4-DNP

tBuOOH

log EC 50 neurite area

log

EC

50 v

iab

ilit

y

METH

Puromycin

Tween-20

Okadaic acid

HonokiolCisplatin

ChlorpyrifosChlorpyrifos Oxon

Oligomycin

SP600125

Simvastatin

PTP IV

ODQ

PiericidinAntimycin A

Haloperidol

Menadione

Acrylamide

Cytochalasin

MevastatinFipronil

IPA3

Na3VO4

K2CrO4

Fig. S2 Separation of specific neurite growth modulators (individual experiments)

from unspecific cytotoxicants.

Cells were treated on d2 as displayed in Fig. 1a, and 24 h later neurite area and viability were

automatically quantified. Compounds were tested at several concentrations, and their EC50

values for effects on neurite area and cell viability were determined by a non-linear

regression sigmoidal concentration-response curve fit, and EC50 values of neurite area were

plotted against the determined EC50 values of cell viability. A reference control group of 9

unspecific toxicants comprised buthionine sulfoximine (BSO), carbonylcyanide-3-

chlorophenylhydrazone (CCCP), 2,4-dinitrophenol (2,4-DNP), etoposide, bisbenzimide H

(H-33352), potassium chromate (K2CrO4), tert-butyl hydroperoxide (tBuOOH), tween-20 and

sodium dodecyl sulfate (SDS) (dots in grey, names are underlined). The solid line indicates

an EC50 ratio of 1 for viability to neurite area. The dashed line indicates an EC50 ratio of 4.0

used as specificity cut-off here. Data for 40 compounds were classified according to this

threshold value. Orange colour indicates substances classified to act unspecific on neurite

growth: acrylamide, antimycin A, chlorpyrifos, chlorpyrifos oxon, cisplatin, cytochalasin,

fipronil, haloperidol, honokiol, IPA-3, menadione, methamphetamine (METH), mevastatin,

1H-[1,2,4]oxadiazolo-[4,3-α]quinoxalin-1-one (ODQ), okadaic acid, oligomycin, piericidin,

protein tyrosine phosphatase inhibitor IV (PTP IV), puromycin, simvastatin and SP600125.

Light blue: substances classified as specific neurite growth inhibitors, EC50 values of three

individual experiments are displayed: Bisindolylmaleimide I (Bis1), brefeldin A, colchicine,

cycloheximide, diquat, flavopiridol, methylmercury (II) chloride (MeHg), sodium

orthovanadate (Na3VO4), narciclasine, nocodazole, paraquat, rotenone, U0126 and

vincristine. Dark blue: substances with an augmenting effect on neurite area: blebbistatin,

HA-1077, H1152, thiazovivin and Y-27632. Neurite area EC50s of these compounds were

determined as response halfway between the baseline (100%) and maximum. Grey dashed

lines encircle the individual EC50 values determined for one compound.

Cells were treated on d2 as displayed in Fig. 1a, and 24 h later neurite area and viability were

automatically quantified. Compounds were tested at several concentrations, and their EC50 values for

effects on neurite area and cell viability were determined by a non-linear regression sigmoidal

concentration-response curve fit, and EC50 values of neurite area were plotted against the determined

EC50 values of cell viability. A reference control group of 9 unspecific toxicants comprised

buthionine sulfoximine (BSO), carbonylcyanide-3-chlorophenylhydrazone (CCCP), 2,4-dinitrophenol

(2,4-DNP), etoposide, bisbenzimide H (H-33352), potassium chromate (K2CrO4), tert-butyl

hydroperoxide (tBuOOH), tween-20 and sodium dodecyl sulfate (SDS) (dots in grey, names are

underlined). The solid line indicates an EC50 ratio of 1 for viability to neurite area. The dashed line

indicates an EC50 ratio of 4.0 used as specificity cut-off here. Data for 40 compounds were classified

according to this threshold value. Orange colour indicates substances classified to act unspecific on

neurite growth: acrylamide, antimycin A, chlorpyrifos, chlorpyrifos oxon, cisplatin, cytochalasin,

fipronil, haloperidol, honokiol, IPA-3, menadione, methamphetamine (METH), mevastatin, 1H-

[1,2,4]oxadiazolo-[4,3-α]quinoxalin-1-one (ODQ), okadaic acid, oligomycin, piericidin, protein

tyrosine phosphatase inhibitor IV (PTP IV), puromycin, simvastatin and SP600125. Light blue:

substances classified as specific neurite growth inhibitors, EC50 values of three individual

experiments are displayed: Bisindolylmaleimide I (Bis1), brefeldin A, colchicine, cycloheximide,

diquat, flavopiridol, methylmercury (II) chloride (MeHg), sodium orthovanadate (Na3VO4),

narciclasine, nocodazole, paraquat, rotenone, U0126 and vincristine. Dark blue: substances with an

augmenting effect on neurite area: blebbistatin, HA-1077, H1152, thiazovivin and Y-27632. Neurite

area EC50s of these compounds were determined as response halfway between the baseline (100%)

and maximum. Grey dashed lines encircle the individual EC50 values determined for one compound.



62

Fig. S3 EC50 values of

neuronal precursor cells of

neurite area and resazurin

reduction compared to mature

neurons.

Cells were replated at d2 and

compounds were added to the

culture medium in at least 5

distinct concentrations. For

testing of mature neurons, cells

were also replated at d2 and

compounds were added in fresh

medium at day 5 (d5). After 24

hours neurite area was quantified

yielding concentration-response-

curves. EC50 values were

calculated, using the

concentration-response-curves,

as concentrations at 50% of

neurite area were detected,

respectively. All data are means

of 3 to 4 independent

experiments. Dotted lines mark

equality of x-axis values to y-

axis values. a) Comparison of

EC50 values of neurite area of

developing (d3) and mature

LUHMES cells (d6). The ratio of

all d3 EC50 values to d6 is 11.43

2.7. b) Comparison of EC50

values of resazurin reduction of

d3 and d6 cells. The ratio of all

d3 EC50 values to d6 is 0.74

0.79.

-3 -2 -1 0 1 2 3

-3

-2

-1

0

1

2

3

Bis 1U0126

Cycloheximide

Brefeldin A

Na3VO4

Rotenone

Colchicine

Paraquat

Vincristine

Nocodazole

log EC50 resazurin [d3]

log E

C50

resazuri

n [

d6]

-3 -2 -1 0 1 2 3

-3

-2

-1

0

1

2

3

Bis 1U0126

MeHg

Cycloheximide

Brefeldin A

Na3VO4

Rotenone

Colchicine

Paraquat

Vincristine

Nocodazole

log EC50 neurite area [d3]

log E

C50 n

euri

te a

rea [

d6]

A

B

average ratio:

0.74

average ratio:

11.4

Fig. S3: EC50 values of neuronal precursor cells of neurite area and resazurin reduction

compared to mature neurons. Cells were replated at d2 and compounds were

added to the culture medium in at least 5

distinct concentrations. For testing of mature

neurons, cells were also replated at d2 and

compounds were added in fresh medium at day

5 (d5). After 24 hours neurite area was

quantified yielding concentration-response-

curves. EC50 values were calculated, using the

concentration-response-curves, as

concentrations at 50% of neurite area were

detected, respectively. All data are means of 3

to 4 independent experiments. Dotted lines

mark equality of x-axis values to y-axis values.

a) Comparison of EC50 values of neurite area

of developing (d3) and mature LUHMES cells

(d6). The ratio of all d3 EC50 values to d6 is

11.43 ± 2.7. b) Comparison of EC50 values of

resazurin reduction of d3 and d6 cells. The ratio

of all d3 EC50 values to d6 is 0.74 ± 0.79.



63

Fig. S4: EC50 values of neuronal precursor cells and mature neurons of resazurin

reduction compared to data from non-neuronal cell types. Cells were replated at d2 and compounds were

added to the culture medium in at least 5

distinct concentrations. For testing of mature

neurons, cells were also replated at d2 and

compounds were added in fresh medium at d5.

After 24 hours resazurin reduction was

quantified yielding concentration-response-

curves. EC50 values were calculated, using the

concentration-response-curves, as

concentrations at 50% of resazurin reduction

were detected, respectively. All data are means

of 3 to 4 independent experiments. EC50 values

were plotted against cytotoxicity data of non-

neuronal cell lines derived from the Halle

registry. Dotted lines mark the linear regression

through the data points presented. a) and b)

Comparison of EC50 values of resazurin

reduction of d3 a) and d6 b) LUHMES cells

with collected values of the Halle registry. The

correlation of d3 to Halle registry is R2 = 0.87

and of d6 to Halle registry is R2 = 0.85.

-1 0 1 2 3 4 5

-1

0

1

2

3

4

5

Acetaminophen

Acrylamide

Colchicine

Diethylenglycol

MnCl2

TMTC

PuromycinMeHg

Cycloheximide

SDS

CdCl2

log LC50 resazurin Halle registry

log L

C50 r

esazuri

n L

UH

ME

S d

6

-1 0 1 2 3 4 5

-1

0

1

2

3

4

5

Antimycin A

CdCl2

Cycloheximide

MeHg

Paraquat

Puromycin

SDS

AcetaminophenAcrylamide

Colchicine

Diethylenglycol

MnCl2TMTC

log LC50 resazurin Halle registry

log L

C50 r

esazuri

n L

UH

ME

S d

3

R² = 0.85

R² = 0.87

A

B

Fig. S4 EC50 values of neuronal

precursor cells and mature

neurons of resazurin reduction

compared to data from non-

neuronal cell types.

Cells were replated at d2 and

compounds were added to the

culture medium in at least 5

distinct concentrations. For testing

of mature neurons, cells were also

replated at d2 and compounds were

added in fresh medium at d5. After

24 hours resazurin reduction was

quantified yielding concentration-

response-curves. EC50 values were

calculated, using the concentration-

response-curves, as concentrations

at 50% of resazurin reduction were

detected, respectively. All data are

means of 3 to 4 independent

experiments. EC50 values were

plotted against cytotoxicity data of

non-neuronal cell lines derived

from the Halle registry. Dotted

lines mark the linear regression

through the data points presented.

a) and b) Comparison of EC50

values of resazurin reduction of d3

a) and d6 b) LUHMES cells with

collected values of the Halle

registry. The correlation of d3 to

Halle registry is R2 = 0.87 and of

d6 to Halle registry is R2 = 0.85.

Results Chapter 2 – Human embryonic stem cell-derived test systems for developmental

neurotoxicity: a transcriptomics approach

64

D. Results Chapter 2

Human embryonic stem cell-derived test systems for

developmental neurotoxicity: a transcriptomics approach

Anne K. Krug*1, Raivo Kolde*2ab, John Antonydas Gaspar*3, Eugen Rempel*10, Nina V.

Balmer1, Kesavan Meganathan3, Kinga Vojnits5, Mathurin Baquié6, Tanja Waldmann1,

Roberto Ensenat-Waser5, Smita Jagtap3, Richard Evans7, Stephanie Julien6, Hedi

Peterson6, Dimitra Zagoura5, Suzanne Kadereit1, Daniel Gerhard9, Isaia Sotiriadou3,

Michael Heke3, Karthick Natarajan3, Margit Henry3, Johannes Winkler3, Rosemarie

Marchan4, Luc Stoppini6, Sieto Bosgra8, Joost Westerhout8, Miriam Verwei8, Jaak

Vilo2ab, Andreas Kortenkamp7, Jürgen Hescheler3, Ludwig Hothorn9, Susanne Bremer5,

Christoph van Thriel4, Karl-Heinz Krause6, Jan G. Hengstler#4, Jörg Rahnenführer#10,

Marcel Leist#1, Agapios Sachinidis#3

Affiliations:

1University of Konstanz (UKN), Department of Biology, 78457 Konstanz, Germany

2aOÜ Quretec (Qure), Limited liability Company, 51003 Tartu, Estonia

2bUniversity of Tartu, Institute of Computer Science, 50409 Tartu, Estonia

3University of Cologne (UKK), Center of Physiology and Pathophysiology, Institute of

Neurophysiology, 50931 Cologne, Germany

4Leibniz Research Centre for Working Environment and Human Factors at the Technical

University of Dortmund (IfADo), 44139 Dortmund, Germany

5Commission of the European Communities (JRC), Directorate General Joint Research

Centre, 1049 Brussels, Belgium

6University of Geneva (UNIGE), Department of Pathology and Immunology, Geneva

Medical Faculty, 1211 Geneva 4, Switzerland

7Brunel University (Brunel), Uxbridge, UB8 3PH,United Kingdom

8Nederlandse Organisatie voor Toegepast Natuurwetenschappelijk Onderzoek (TNO),

2628 VK Delft, Netherlands

9Gottfried Wilhelm Leibniz University (LUH), Institute for Biostatistics, 30167

Hannover, Germany

10Technical University, Department of Statistics, Dortmund, Germany

Accepted (24. October 2012) in Archives of Toxicology



65

Abbreviations

BMC Benchmark concentration

CNS Central nervous system

DMA DNA microarray

DNT Developmental neurotoxicity

DoD Day of differentiation

DT Developmental toxicity

ESNATS Embryonic stem-cell based novel alternative test systems

FDR False discovery rate

GO Gene ontology

hESC Human embryonic stem cells

MeHg Methylmercury

NEP Neural ectodermal progenitor cells

OECD Organisation for economic co-operation and development

PBPK Physiology-based pharmacokinetic

PS Probe set

REACH Registration, evaluation, authorisation and restriction of chemicals

RT Reproductive toxicity

TFBS Transcription factor binding site

VPA Valproic acid



66

Abstract

Developmental neurotoxicity (DNT) and many forms of reproductive toxicity (RT) often

manifest themselves in functional deficits that are not necessarily based on cell death, but

rather on minor changes relating to cell differentiation or communication. The fields of

DNT/RT would greatly benefit from in vitro tests that allow the identification of toxicant-

induced changes of the cellular proteostasis, or of its underlying transcriptome network.

Therefore, the ‘human embryonic stem cell (hESC)-derived novel alternative test systems

(ESNATS)’ European commission research project established RT tests based on defined

differentiation protocols of hESC and their progeny. Valproic acid (VPA) and methyl mercury

(MeHg) were used as positive control compounds to address the following fundamental

questions: 1) Does transcriptome analysis allow discrimination of the two compounds? 2)

How does analysis of enriched transcription factor binding sites (TFBS) and of individual

probe sets (PS) distinguish between test systems? 3) Can batch effects be controlled? 4) How

many DNA microarrays (DMA) are needed? 5) Is the highest non-cytotoxic concentration

optimal and relevant for the study of transcriptome changes? VPA triggered vast

transcriptional changes, whereas MeHg altered fewer transcripts. To attenuate batch effects

analysis has been focused on the 500 PS which with highest variability. The test systems

differed significantly in their responses (<20% overlap). Moreover, within one test system,

little overlap between the PS changed by the two compounds has been observed. However,

using TFBS enrichment, a relatively large ‘common response’ to VPA and MeHg could be

distinguished from ‘compound specific’ responses. In conclusion, the ESNATS assay battery

allows classification of human DNT/RT toxicants on the basis of their transcriptome profiles.



67

Introduction

Reproductive toxicity (RT) testing is one of the technically most challenging fields of

toxicology, and there is a huge demand for more cost-effective, faster, and more accurate

assays. RT may be caused by chemicals, drugs, pesticides and other compounds that interfere

with biological processes essential for reproduction, and it is therefore of large societal

concern. It has been estimated that up to 50% of the animals used for testing in the context of

REACH will be required to evaluate RT (Seiler et al 2011). Currently, this type of safety

assessment comprises evaluation of chemical effects on spermatogenesis, oogenesis or the

fertilization process. Another large subfield deals with the disturbances of embryofetal

development, and is generally called developmental toxicity (DT) testing.

In the area of RT testing, evaluation of a single compound requires hundreds of animals.

If testing of nervous system development and long term effects are included, even thousands

of rats/rabbits are required. Animal testing, e.g. following OECD test guidelines 414 (2-

generation reproduction), 426 (developmental neurotoxicity (DNT)) or others, often only

gives indirect indications of toxicity such as changed numbers of embryo-foetal death, altered

foetal weight or the development of anatomical or behavioural abnormalities. To significantly

reduce the use of animals and to get further mechanistic insights, in vitro systems modelling

critical parts of the foetal development are being explored as alternatives (Adler et al 2011,

Basketter et al 2012). For instance the development of initial germ layers from pluripotent

cells, and the specification of organ systems such as the central nervous system (CNS) are

such critical parts of the development.

The CNS is considered to be one of the most frequent targets of systemic toxicity, with

the developing nervous system being particularly susceptible (Klaassen 2010, van Thriel et al

2012). This susceptibility to DNT is due to a finely orchestrated sequence of complex

biological processes, such as proliferation, migration, apoptosis, differentiation, patterning,

neurite outgrowth, synaptogenesis, myelination and neurotransmitter synthesis, which are all

targets of numerous toxic chemicals (Kadereit et al 2012). Despite its high relevance, DNT is

one of the least studied forms of toxicity (Kadereit et al 2012, Makris et al 2009). It is also

particularly difficult to study, because DNT is not necessarily caused by cell death. In fact,

chemically induced changes in the proportions of neural cells, positioning or connectivity may

be sufficient to cause DNT (Kadereit et al 2012, Kuegler et al 2010). Currently DNT is tested

according to OECD TG 426, which requires animals to be exposed during gestation and

lactation, and the resulting offspring to be analysed for gross neurologic and behavioural



68

abnormalities. However, this complex in vivo test system is too laborious and expensive to

allow all the testing needed to provide hazard information for thousands of untested

chemicals.

To bridge this gap, embryonic stem cell (ESC)-based systems are currently being

developed (Kuegler et al 2012, Leist et al 2008a, Weng et al 2012, Zimmer et al 2012). These

systems recapitulate early neuronal development in vitro, including neurulation, patterning,

neurogenesis and gliogenesis. In the present study, five human ESC (hESC) based in vitro

systems, named here after the developing institutions, have been employed. They recapitulate

different phases of early tissue specification and neural development (Fig 1). UKK

recapitulates the multi-lineage differentiation of hESC into ecto-, meso-, and endoderm

(Jagtap et al 2011, Meganathan et al 2012). UKN1 models the stage of neuroectodermal

induction that results in the formation of neural ectodermal progenitor cells (NEP) (Balmer et

al 2012, Chambers et al 2009). JRC reproduces the neural tube formation during early

neurogenesis by the formation of neural rosettes and more mature neural cell types

(Stummann et al 2009). UNIGE models the transition from neural precursor cells to mature

neurons, showing morphological signs of neural differentiation, including neurite extensions.

UKN4 already starts with neuronally-committed precursor cells that undergo the maturation

towards post-mitotic neurons with neurites. These cells were not derived from hESC but from

a human fetal brain (Scholz et al 2011, Stiegler et al 2011).

Differentiating murine ESCs show similar waves of gene expression changes as observed

during murine embryonic development in vivo (Barberi et al 2003, Gaspar et al 2012,

Kadereit et al 2012, Zimmer et al 2011a, Zimmer et al 2011b). Such information is not

available for early human development, but it is generally assumed by analogy that hESC

would reproduce normal human tissue differentiation (Leist et al 2008a). Under this

condition, transcriptome analysis, including bioinformatic processing of the data, appears as

an attractive method to detect perturbations caused by chemicals in the normal wave-like

expression patterns in hESC differentiation systems. Moreover, alterations in the proportions

of cell types, as a consequence of exposure to test compounds, should be detectable by DNA

microarrays (DMA), as shown earlier for other systems (Schmidt et al 2008, Schmidt et al

2012). The treatment period for each test system was chosen according to previously

described effects (Fig. 1). For example, in UKN4 neurite outgrowth starts on day of

differentiation (DoD) 2 and can be measured at DoD3 (Stiegler et al 2011). Therefore DMA

analysis was also performed here under similar incubation conditions. In the same vein, it is



69

known for UKN1 that changes in gene expression are best detectable after treatment from

DoD 0 to 6 (Balmer et al 2012) and accordingly transcriptome analysis was done on DoD6

after 6 days of incubation with test compound.

For test system evaluation, we have chosen valproic acid (VPA) and methyl mercury

(MeHg), two model compounds that trigger RT and DNT in humans and animals (Chen et al

2007, Grandjean & Landrigan 2006, Kadereit et al 2012, Wang et al 2011a). The ability of

VPA to cause developmental neurotoxicity has been recognized since the 1970s. VPA is a

clinically-used anti-epileptic drug that acts as a reversible modifier of enzyme activities. It has

also been shown to cause neural tube defects and to trigger large changes of the cellular

transcriptome through the inhibition of histone deacetylases (Jergil et al 2009, Theunissen et

al 2012a, Werler et al 2011). MeHg also causes neural tube defects (Grandjean & Herz 2011,

Robinson et al 2011). However, the transcriptional changes due to MeHg are more limited

and indirect, as it acts through the unspecific modification of many different proteins, in

addition to triggering oxidative stress (Aschner et al 2007). Despite its unclear mode-of-

action, MeHg is a ‘gold standard’, because human DNT has been particularly well

documented, mainly due to the catastrophic endemics caused by MeHg-contaminated food

(Bakir et al 1973, Choi 1989, Davidson et al 2004, Ekino et al 2007, Harada 1995).

The wide-spread use of transcriptomics endpoints requires clarification of important

technical issues. Therefore, we addressed here the following questions: 1) Does DMA

analysis allow differentiation between distinct classes of toxicants and non-toxicants. If yes,

2) how large is the overlap between the available ESC based test systems (Fig 1), and are they

all required for the identification of DNT compounds? 3) How many independent experiments

are needed? 4) At which optimal concentrations should gene array analyses be performed?

The present study provides unequivocal answers to these questions, and will therefore serve

as a basis for further development of RT assays on the basis of DMA classification

algorithms.



70

Results and Discussion

Detection of different transcriptional responses to the DNT model

compounds, valproic acid and methyl mercury

To explore the dynamics and specificity of the transcriptional response of novel hESC-

based in vitro systems (Fig 1), we chose VPA and MeHg as two positive control toxicants

with described effects on DNT, and D-mannitol as the negative control compound. The three

test compounds were initially evaluated in three of the test systems (UKK, UKN1 and JRC) at

the “maximum tolerated concentration”. This benchmark concentration (BMC) was

determined experimentally for each of the test systems as the highest concentration that

reduced overall cell viability by not more than 10% (Fig. S1). In the case of mannitol, a large

range of concentrations, from 1 µM to 100 mM was used, and no cytotoxicity was detected

(data not shown). For the UKN1 system, the response to mannitol was tested by quantitative

PCR for three toxicant-responsive genes (OCT4, OAX6, FOXG1) (Balmer et al 2012). As no

changes were observed for concentrations up to 40 mM, and data on this compound were

provided by the other test systems, DMSO (2 mM) was chosen as the DMA negative control

for UKN1. The transcriptional alterations triggered by the BMC of the two toxicants

(VPA/MeHg) or by the two negative controls (mannitol/DMSO) were measured in 4-5

independent experiments on Affymetrix DMA, and the genes that were differentially

expressed between culture medium-only controls and test compounds were determined by

modern stringent statistical methods (Limma t-test, Benjamini-Yekutieli false discovery rate

(FDR) correction). The complete set of data is displayed in supplementary Table S1.

For a visual monitoring of the different compound effects, the hundred most regulated

(defined by the lowest FDR-corrected p-values) genes (top 50 for VPA and top 50 for MeHg)

were selected for each test system (Table S1), and their relative expression levels were

displayed as heat maps. For all test systems, striking differences were observed between the

regulation patterns of VPA and MeHg. Clustering analysis showed that VPA samples were

clearly separated from the MeHg samples (Fig. 2). This effect was even more pronounced,

when clustering was performed with the 100 top genes regulated by VPA (Fig. S2A). Under

these conditions, the differences between MeHg and negative controls were small or not

apparent. Therefore, clustering was also performed with the top 100 genes regulated by

MeHg. Under these conditions, MeHg samples were clearly separated from those treated with

D-mannitol/DMSO (Fig. S2B).



71

Fig. 1: Overview over the test systems’ treatment protocols used for microarray

analysis. The five test systems cover different periods and processes relevant to early embryonic/neuronal

development, as indicated to the left. The time arrows indicate when cells were re-plated, medium

was exchanged, toxicants were added, and when analysis was performed. Additional information

is presented below each test system on the type of coating and the medium used in different

experimental phases.



72

The number of significantly altered Affymetrix DMA probe sets (PS) was much higher

for VPA compared to MeHg. The sum of all PS changed by VPA in the test systems UKK,

UKN1 and JRC was 15386; for MeHg the sum was 1246 PS (Table S1, Fig. 3). This striking

difference was observed although both compounds were used at their respective BMC in each

test system. Exposure to the negative controls did not result in any significant changes (Fig.

3). Thus, the extent of the responses of the neurally-differentiating hESC to the different

developmental neurotoxicants appears to be compound-specific. Moreover, the responses to

the two model toxicants differed qualitatively (Fig. 2; Fig. S2). The ability to clearly

distinguish known toxicants suggests that the test systems would distinguish unknown classes

of potential toxicants. It may be speculated that safety liabilities of unknown chemicals for

humans may be predicted by comparing their effects in the test systems with those of known

toxicants and non-toxicants. The technical and statistical basis of the above initial findings,

together with their potential biological and toxicological implications was explored further in

the following extended test battery.

Fig. 2: Differential alterations of gene expression by valproic acid (VPA) and methyl

mercury (MeHg). Three different test systems (UKK, UKN1, JRC) were exposed to VPA (blue label on top of the

heatmap) or MeHg (green label) at their respective bench mark concentration, or to D-mannitol

(red) or DMSO (dark red). The differentially expressed genes (vs untreated controls) were

determined in 4-5 independent experiments (shown as columns of the heatmaps). The similarity of

the gene expression patterns is indicated by the Pearson’s distance dendrogram at the top. The

heatmaps are based on 100 selected genes. These comprise the 50 genes with the lowest adjusted

p-values according to the Limma t-Test for regulation by MeHg, and 50 genes with the lowest

adjusted p-values for VPA. The colours of the heatmap indicate the relative gene regulation level

above (red) or below (yellow) the average for each row.



73

Differential constitutive and toxicant-induced responses of the test battery

One may hypothesise that MeHg showed only relatively weak effects in the initial testing

(UKK, UKN1 and JRC) as all these systems only generate immature cells, and such cells may

be relatively resistant to MeHg. Alternatively, such test systems may lack key targets of

mercury toxicity. Such an assumption would be in agreement with findings in neuronally-

differentiating murine ESC, which were highly sensitive to MeHg during the late neuronal

maturation phase, but relatively insensitive during the initial phase of neural precursor

formation (Zimmer et al 2011b). For a broader coverage of effects during later phases of

neurogenesis, two additional test systems were used (Fig. 1, UNIGE and UKN4). The UNIGE

hESC-based test system covers the developmental phase after neural stem cell formation. The

UKN4 test system was used as reference, as this system is well characterised not only for

transcriptome changes, but in particular for functional and phenotypic effects(Stiegler et al

2011). From the literature, it is known that MeHg inhibits neurite outgrowth in this system,

and transcriptome analysis was performed at a concentration known from previous studies to

affect neurites (Stiegler et al 2011).

Fig. 3: Overview of differentially-expressed genes in all test systems. Positive and negative control compounds were tested in the JRC, UKK, UKN1, UKN4 and

UNIGE test systems. The test concentrations for methyl mercury (MeHg), valproic acid (VPA)

and D-mannitol (Mannitol) are indicated in the white fields. The number of significantly altered

probe sets (PS) is indicated separately for up-regulations (red) and down-regulations (blue). The

results for testing without FDR adjustment are indicated in pale-coloured fields. The results after

FDR adjustment by the Benjamini-Yekutieli method are indicated in white bold numbers. The

highest compound concentration tested corresponded to the BMC of the respective test system.

The highest test concentration (800 nM) was 5 times higher than the BMC (160 nM) for UNIGE

only. nd = not done



74

The extended test battery (UKK, JRC, UKN1, UKN4 and UNIGE) was used for

additional testing. The effects of MeHg were examined in all systems at the respective BMCs,

in addition to one lower concentration (LOW). The latter was determined by dividing the

BMC by a factor of four (Fig. S1). Additional experiments were also performed with VPA.

The compound was tested at two relatively similar concentrations in JRC (to test the

reproducibility of the response). It was also examined at 4-fold different concentrations in

UKK (to test potential concentration dependencies of the response). The number of

differentially expressed probe sets (PS) for each condition is summarised in Fig. 3. This broad

experimental approach showed that the transcriptional response of differentiating hESC to

MeHg is indeed very limited. Also, the test systems using more mature cells (UKN4, UNIGE)

did not show any significant response when stringent FDR corrections were used.

Comparison of the results before and after FDR correction showed the unmistakable need

for appropriate statistical treatment of the data. Although the choice of a 5% significance level

will generate on average 2734 false positives when 54675 PS are analysed (as in this study), it

can at times still be counter-intuitive for toxicologists when none of the more than 2000

identified genes is significant after FDR correction. The effect of FDR correction in the

present study is visualized in the form of volcano plots. This form of display orthogonally

separates the two parameters usually considered important in gene expression analysis: the

fold change and the significance level. As the FDR correction only affects the significance

level, one can see the “volcano” heights being compressed, while the width remains the same.

For instance, in the case of JRC incubated with 273 nM MeHg (BMC concentration) all

apparently-significant PS dropped below the usual significance level (p < 0.05). Also, with

UKK exposed to 500 µM VPA (20% of the BMC), the number of 2524 PS that appeared to be

significantly up-regulated before FDR correction dropped down to four really significant PS

after FDR correction. Notably, the apparent significances were ‘lost’ although several PS

appeared to be ‘regulated’ more than 2-fold, at times even up to 4-fold (Fig. 4, Fig. S3). It

should be noted that the gene expression response occurred within a narrow range of

concentrations. The FDR-corrected data sets showed that the number of regulated probe sets

can change from several thousands to zero within a four-fold concentration range. Even a

lowering of the test concentration by only 20% (relative to the BMC) resulted in a reduction

of the identified PS, at least in one system in which this was tested (JRC).



75

Fig. 4: Correlation of fold-change and significance level of gene expression for

different statistical approaches. Data were generated and calculated for each combination of test system and compound, as

illustrated in Fig. 3. In the volcano-plot diagrams, fold-changes of the compound-induced gene

expression are shown on the x-axis (log2-scale). The y-axis shows negative logarithmic adjusted

p-values of a LIMMA t-test (-log10(p-value)). The p-values were A. FDR adjusted, or B. not FDR

adjusted. The dashed lines show the significance level of p = 0.05. The dotted lines show an

example for the p = 0.000001 significance level for orientation. All other test systems and

compounds are shown in the supplemental material (Fig. S3).



76

However, more than 90% of the PS identified at the low concentration in this assay were

also identified at the high concentration (Fig. 5). This good overlap confirmed a robust and

reproducible test system response. When more stringent conditions were used for filtering,

such as the requirement for a ≥ 4-fold change or for a lower p-value, the good overlap

between the two concentrations was maintained (Fig. 5). Altogether, these data suggest that

the most pronounced and robust transcriptional responses can be measured at toxicant

concentrations, which are close to or at the BMC.

Fig. 5: Overlap of differentially expressed probe sets (PS) at different concentrations. The JRC test system was exposed to VPA at a high (= BMC) and low concentration in 5

independent experiments. The circles of the Venn diagrams show the numbers of PS that were

influenced by the two experimental conditions. The overlap gives the number of genes influenced

both at the low and the high concentration. The fraction of the genes in the overlap (ol) with

respect to all genes altered at the low concentration is indicated above each diagram. The number

at the lower right corners indicates the number of PS not influenced by the test compound at any

concentration. Significance was determined by the LIMMA FDR-adjusted t-Test.

The first column shows results without restriction by the p-value and examines the effect of

restrictions by the fold-change value on the number of PS identified. The second column imposes

the additional restriction that all identified PS should have a p-value below 0.05. The third column

shows the results when only PS with a p-value below 0.01 are selected.



77

To obtain a better overview of how the different test systems are related to one another,

we performed a principal component analysis (PCA) encompassing untreated controls and

non-differentiated H9 hESC, in addition to all treated samples. This approach allowed the

visualization of the overall transcript patterns measured by 190 DMA on a 2-dimensional

PCA space (Fig. 6A). Several conclusions can be drawn from a qualitative analysis of the

PCA presentation: First, all test systems clearly differed from non-differentiated hESC.

Second, all test systems differed from one another, i.e. the variance between the different test

systems was larger than the variance of individual samples within a given test system. Third,

samples from one test system clustered together, whether they had been treated with VPA,

MeHg or solvent. On the other hand, samples treated e.g. with MeHg in different test systems

did not cluster together in this form of data presentation. It is noteworthy, that presentation of

data in form of such a comprehensive PCA does not allow the identification of compound

effects, although large, statistically-significant transcriptome changes occurred (e.g. VPA vs

solvent control). To better visualise compound effects, a different statistical treatment is

required before the data are presented. For instance, the large influence of the different test

systems can be attenuated by subtraction of the corresponding controls before display (see

below and Fig. 7). The distinct clustering of all test systems to a different area of the PCA plot

suggests that the test battery is not redundant. Each individual test system seems to react with

different transcriptome changes, and the combination of the tests may thus provide richer data

than any individual test.This would imply that the different systems would be able to identify

different toxicant effects and thus be complementary in their toxicological information. The

test battery may thus constitute an important step towards the replacement of animal tests by

information-rich human cell-based models (Hartung & Leist 2008, Leist et al 2008b). This

will, however, require further testing and validation (Leist et al 2012a). A second important

observation was the presence of outliers in some samples, which will be investigated in

greater detail in the following section (Fig. 6A).

Control of intra-group variability and batch effects

The PCA indicated that eight of the DMA of UKN1 clustered separately from all other

UKN1 samples. The commonality among the eight DMA was that they were measured on a

different day compared to the other samples. Four corresponded to controls and four to

samples treated with VPA. Thus, the clustering was not treatment-related. A similar situation

was observed for ten samples of UNIGE (Fig. 6A). When only the 500 probe sets with the



78

highest variance were considered for the PCA, the “outliers” moved partially or

completely back; that is they clustered together with the other samples within their test system

(Fig. 6B). This suggested that genes with a low variance had contributed to the outlier effect.

Fig. 6: Identification and correction of DNA microarray (DMA) batch effects. The signal of all PS was determined in five different test systems after incubation with compounds

as in Fig. 3. The data for every experiment plus those of 25 untreated controls and solvent controls

and 21 samples of untreated hESC (dark green circles with light blue filling), were used for

principal component analyses (PCA) of altogether 190 DMA. Data from the different test systems

are colour-coded and each DMA is displayed as a circle in the PCA-plot. Circles filled in yellow

code for DMA that clustered away from their respective main groups, and that were considered

outliers due to a batch effect, as they were measured at another time point compared to the other

samples. The axis labels indicate the percentage of the total variance covered by the respective

axis A. The PCA is based on all PS. B. The PCA is based only on the 500 probe sets with the

highest variance. C. The distribution of the PS fluorescence signals (indicated here as “gene

expression value”) is displayed for all 169 test system DMA of this study (each DMA is

represented by one box of the box plot). The size of the boxes indicates the 25th and 75th

percentile (the lower and upper quartiles, respectively) of the PS. The solid lines in the box

indicate the 50th quantile of the distribution. The height of the box being equal to the difference

between the upper and lower quartiles is called the interquartile range (IQR). The dashed lines

(whiskers) indicate gene expression values within the range of 1.5 IQR from the 25th and 75th

percentile. The dots outside the dashed lines (appearing as solid line due to the print resolution)

represent the outliers within one DMA. The DMA corresponding to the differently clustering

samples in A are indicated by boxes filled with yellow, and they show a higher variance. The test

system colour coding of part A, B, and C is identical.



79

A graphical presentation of the variances of all DMA performed for this study indeed

indicated that the “outliers” had a higher variance of the fluorescence signals, although the

average signals were quite similar to all other DMA (Fig. 6C). These data suggest that the

‘distant clustering’ samples are the consequence of a batch effect.

The presented study is still on-going and even larger numbers of samples will have to be

studied. This makes it impossible to analyse all samples in a single batch. Methods to control

for batch effects will therefore be required. As indicated here, one possibility is to include

only the PS with highest variability between the samples into the analysis. As an alternative

approach, the corresponding control values were subtracted from the compound-treated

samples before the PCA analysis. This form of presentation clearly separated VPA and MeHg

incubated samples, and the results obtained by clustering analysis within the individual test

systems were confirmed, also when this multi-systems approach was chosen (Fig 7A). The

subtraction of the controls resulted in the visualization of treatment effects in the PCA that

were not visible when the non-processed data were used (Fig. 6). When only the 500 PS with

the highest variance - rather than all 54,575 PS - were included, there was a more defined

clustering of the VPA samples compared to the MeHg samples (Fig. 7B). The reduction to

500 PS also resulted in a better clustering of other “distant clustering” samples. A stepwise

reduction of PS showed that 500 PS seems to represent a reasonable choice although even

smaller numbers, e. g. 200 PS, would be possible (Fig S4). An interesting implication of this

observation is that the scattering of samples within one group can be caused by relatively

large numbers of PS with low variability and not necessarily by the PS which show the

highest variance. These “high variance PS” appear to be highly relevant for further analysis.

Robustness analysis: role of the number of biological replicates

In the present study, five biological replicates (independent experiments performed at

different days) were generated for most test conditions. One technical replicate (one DMA)

was analysed per experiment. To study whether lower numbers of DMA would also lead to

similar results in the present data set, we chose a statistical permutation approach that

simulated the situation of choosing only 2, 3, or 4 of the 5 experimental replicates (Note that

each replicate consisted of a matched pair of DMA for control and for treated cells).

For each possible combination of these pairs (here for simplicity called DMA or

replicates), the number of PS that overlapped with the original set of PS was identified. In

addition, new PS that had not been originally identified were also detected. The expectation



80

was that if 5 DMA were redundant, then the percentage of original PS identified with 3 or 4

DMA should also be high, and the number of new PS arising from the new analysis should be

low. This approach was run under different conditions. The significant genes were identified

Fig. 7: Principal component analysis (PCA) of relative gene expression data after

subtraction of solvent controls. A. The signal of all PS was determined in five different test systems (UKK, UKN1, JRC, UKN4

and UNIGE) after incubation with compounds as in Fig. 3. Then, the values for the respective

controls were subtracted from the values of the DMA treated with VPA at the BMC (large blue) or

at the LOW concentration (small blue dots), or MeHg (large and small green dots), or D-mannitol

(red), or DMSO (black). These data were then used for PCA analysis. The lower right panel shows

all data together. The other panels show the data for individual test systems within the same axes

as for all systems. In A. all PS were included, while in B. only the 500 PS with the highest

variance were used. Note for instances the outliers in UNIGE marked by arrows in A, and their

perfect clustering in B.



81

by the less stringent Benjamini-Hochberg FDR correction (Fig. 8) or by the very stringent

Benjamini-Yekutieli correction (Fig. S5). Moreover, either all PS were considered, or only the

ones regulated more than 2-fold (Fig. 8, Fig. S5).

The results showed that there was only a moderate advantage of using 5 DMA instead of

4, when only PS with ≥ 2-fold changes were considered in the current data set. Under this

condition, and using less stringent FDR correction, even 3 DMA would have resulted in the

identification of a large majority of genes. The permutation analysis was also found to be a

suitable tool to test data consistency and robustness of the analysis method used. For most test

systems, removal of any of the 5 DMA (pairs) to generate a new data set based on 4 DMA,

yielded largely similar results. This suggests that all different experiments had generated

largely similar data, although they were performed with different cell cultures on different

days. The situation was different for the MeHg samples from UKN1, where removal of one

specific DMA resulted in the identification of more than twice as many significant PS

compared to the remaining 4 DMA. All combinations of the three remaining DMA that lacked

the apparent “outlier” identified much larger numbers of PS compared to the combinations

that included that specific DMA (pair) (Fig. 8). Such an analysis may therefore be used to

develop statistical techniques for the identification of outliers.

The relationship of cytotoxic response and DNT-specific transcriptome

changes

The choice of toxicant concentrations for gene expression analysis is a critical step. If too

high concentrations are used, cell viability will be compromised. The cell death occurring

under these conditions may result in unspecific ‘toxicity associated’ gene expression

responses. Conversely, the use of too low concentrations of test compounds would result in

false negative responses, and in the inability to identify any alterations of the transcriptome.

The magnitude of the response may be dependent on the concentration of the test compound,

which is especially important when compounds are compared and possibly classified or

ranked according to their specific responses.

Furthermore, information on the concentration-dependence may be used for more

detailed characterisation of compound effects, and possibly for the identification of the

hazardous responses as opposed to counter-regulations and unspecific responses (Theunissen

et al 2012a, Theunissen et al 2012b).



82

In the present study, the BMC of the cytotoxicity test (i.e. the highest non-cytotoxic

concentration) was used as the standard test concentration (Fig S1). Although transcriptional

responses can be triggered by MeHg and VPA at concentrations considerably lower than the

cytotoxic concentration (Balmer et al 2012, Zimmer et al 2011b), we found here that the

majority of responses to MeHg in UKN1 was lost even at only 4-fold lower concentrations

than the BMC. We made similar observations for VPA in other test systems.

Fig. 8: Simulation of different numbers of experiments (pairs of DMA) and their

impact on the numbers of significantly-regulated PS. VPA was tested in the test systems JRC und UKK at its BMC in 5 independent experiments, and

in UKN1 in 4 experiments. MeHg was tested in UKN1 in 5 experiments. The number of

significantly regulated genes (Benjamini-Hochberg FDR correction) was calculated without

further restrictions (left) or with the restrictions that the PS should be regulated more than 2-fold

(right). The numbers of PS are indicated above the dashed black lines, which were set as 100%

reference points. The dark blue bars indicate how many of these PS were identified when different

permutations of 2, 3 or 4 experiments (indicated as grey headings) were used. The light blue bars

indicate how many additional PS were identified, when only subsets of the original 5 (4)

experiments were analysed. For instance, the 5 bars in the panel with the coordinates 4/JRC:VPA

represent the five possible ways of omitting one of the experiments. The 10 bars in the panel with

the coordinates 3/JRC: VPA represent the 10 possible permutations of leaving out 2 of the

experiments and then recalculating the significant PS on the basis of the remaining 3 DMA.



83

In in vivo studies, developmental neurotoxicity is defined as effects on the pups in the

absence of maternal toxicity. A corresponding definition for in vitro test systems would be

‘specific alterations of differentiation in the absence of overt (unspecific) cytotoxicity’.

Fulfilment of this condition was carefully explored, and several features of our data indicate

that measurements at the BMC do in fact allow us to draw conclusions on DNT-specific

disturbances triggered by the test compounds: First, we tested whether known toxic

concentrations (800 nM MeHg in UNIGE; BMC was 160 nM) would lead to unspecific

transcriptional responses (Fig. 3). Also under this condition, no significant PS were identified,

i.e. no cell death genes were triggered. We also examined the effect of accidental variations of

the cytotoxicity from experiment to experiment. The fixed BMC indicated here were

determined from a set of pilot experiments. However, the actual cytotoxicity in the individual

experiments in which mRNA levels were analysed showed some biological variation, which

was documented e.g. for UKN1 and UKN4. Examination of these data showed that the MeHg

concentration used for UKN4 reduced cell viability more than the one used for UKN1.

However, no response was observed in UKN4, while an apparently specific response was

triggered in UKN1. Second, some concentrations used for testing VPA in UKN1 triggered

toxicities of more than 10% (data not shown) in the experiments used for DMA analysis (due

to daily experimental variations in sensitivity), but cell death-related GO terms were not

identified. In contrast, MeHg in the same system did not trigger measurable cytotoxicity, but

GO term analysis indicated an upregulation of genes related to apoptosis and neuronal death.

Thus, the use of compounds at the BMC does not seem to be problematic. In the case of

MeHg, triggering of cytotoxic responses is rather a specific feature of the compound (protein

modifier, trigger of oxidative stress). This may be an explanation for the low or absent

transcriptional responses in the test systems. Third, candidate genes typically related to cell

death, DNA damage and oxidative stress were examined in UKN1. Such genes were not

overrepresented amongst the VPA-regulated genes. Moreover, their extent of regulation did

not correlate with the overall magnitude of regulation in the individual experiments (not

shown). Fourth, it was examined how far the responses to different toxicants overlapped. In

case of a strong component of cytotoxicity, it was expected that typical stress genes were

induced, and similarities would be observed in the regulation pattern of different toxicants.

However, only a small fraction of the overall altered PS overlapped between VPA and MeHg

(as examined in detail below, (Fig. 10)). Even though a ‘common transcription factor

response’ between VPA and MeHg of 16 transcription factors (TFs) was observed, there was



84

still a majority of TFs unique for MeHg or VPA. Thus, two compounds, both used at the

BMC, triggered different responses, with no common cytotoxicity pattern.

In summary, the data indicate that the measurement of transcriptional responses at the

BMC is a reasonable approach, although further studies are required for a better

understanding of a possible ‘common toxicity-associated response’. Our limited set of data

indicates that concentrations beyond the BMC do not necessarily result in an unspecific

transcriptional response reflecting cytotoxicity.

Relationship of the BMC with respect to the in vivo relevant concentration

range

Besides the technical considerations concerning the BMC and cytotoxicity, the relevance

of the chosen concentrations for the in vivo conditions needs to be considered. When in vitro

concentrations differ by more than one order of magnitude from concentrations causing

toxicity in vivo, pathways of toxicity may become activated that are not relevant to the in vivo

situation. Unfortunately, human exposure measurements of DNT compounds are often poorly

documented and concentrations in the brain are only rarely known. Nevertheless, human

relevant concentrations of 0.005-0.5 µM MeHg and 500 -1000 µM VPA have been reported

in a recently published review (Kadereit et al 2012). To obtain a clearer picture, we used

Fig. 9: Physiologically based pharmacokinetic (PBPK) modelling of the positive

control compounds MeHg and VPA. Systemic concentrations of MeHg (total blood concentration, upper panel) and VPA (plasma

concentration, lower panel) in rats following exposure to a developmental neurotoxic dose

predicted by PBPK modelling. A) PBPK simulation of MeHg total blood concentration in rat dams

upon daily oral gavage of 4 mg/kg MeHg on gestation days 6 to 9, the lowest developmentally

neurotoxic dose in Bornhausen et al. (1980). Predicted maximum total blood concentration of

0.9 µM is indicated. Maternal and fetal blood concentrations are considered similar. The fetal total

blood concentration is assumed to be available for fetal brain exposure, and equated to the nominal

concentration in in vitro test media. B) PBPK simulation of VPA plasma concentration in rat dams

upon a bolus intraperitoneal dose of 350 mg/kg, the lowest dose causing relevant effects in Rodier

et al. (1996), resulting in a predicted maximum total blood concentration of 6.6 mM (as indicated).

Comparable concentrations have been found in maternal and fetal plasma. The unbound plasma

concentration in vivo is equated to the unbound concentration in in vitro test media.



85

physiology-based pharmacokinetic (PBPK) modelling to calculate in vivo relevant blood and

brain concentrations from the doses that caused DNT in animal studies (Fig 9; Fig. S6A). Oral

exposure to MeHg of 0.01 mg/kg on gestation days 6 - 9 is predicted to result in a maximum

total blood concentration of 0.9 µM (Fig. 9 A).

Thus, similar nominal concentrations should show activity in vitro, although the actual

amount of MeHg penetrating the cells may additionally depend on cysteine concentrations in

the different media of the test systems. A VPA plasma peak concentration of 6.6 mM is

predicted after a single oral dose of 350 mg/kg. This dose resulted in the same model in DNT

(Rodier et al 1996) (Fig. 9B). For extrapolation of such data to in vitro systems, corrections

for differences in protein binding and lipid partitioning in plasma vs cell culture medium have

to be considered (Fig. S6B). Our calculations suggest that the expected equivalent nominal

concentrations in vitro are 3.3 mM for UKK, 2.7 mM for UKN1 and 0.9 mM for JRC, UKN4,

and UNIGE. These results show that the BMC concentrations used in this study are within the

same order of magnitude as the in vivo concentrations which caused DNT in humans and

animals.

Remarkable overlap of overrepresented TFBS amongst genes influenced by

VPA and MeHg

The main focus of this study was to investigate the technical feasibility of using

transcriptomics as a major endpoint to characterise responses of hESC-based test systems. For

a detailed characterisation of the biological responses of the test systems to the compounds, a

different experimental design would be required. Nevertheless, we performed some initial

comparisons of gene ontologies (GO) and transcription factor binding sites (TFBS) that were

overrepresented amongst the regulated PS. The main aim was to find out whether simple

analysis tools can reveal differences and commonalities of the transcriptome responses. For

this approach, five sets of data were compared: the responses of UKN1, JRC and UKK to

VPA and the responses of UKN1 and UKK to MeHg (all at BMC concentration). To obtain

an overview over the main biological processes affected by co-regulated genes the statistically

overrepresented GO terms were identified and displayed for each test system and condition

(Fig. S7). For instance, the genes down-regulated in each test system by VPA pointed to

effects of the toxicant on RNA processing, and on chromatin modification/histone acetylation.

The latter results are consistent with the known activity of the compound as a histone

deacetylase inhibitor (HDACi). GO terms related to effects on “neural tube formation”

“neuron development” and “embryonic morphogenesis” showed up for different conditions.



86

These findings gave a hint that there may be an overlap of higher order biological responses

across the test systems and compounds. However, we are aware of the fact that the GO term

analysis is a very rough tool, and that GO term annotations of many genes can be problematic

(Weng et al 2012). Therefore, we chose the alternative approach of comparing the overlap of

regulated PS between the test systems with the overrepresentation of 267 human TFBS (as

indirect indicator of higher order linked biological processes).

First, the overlap of test systems treated with the same compound was analysed. VPA

regulated 571 PS in all three test systems (Fig. 10A). Thus, only a relatively minor overlap

occurred on the level of individual PS. The PS for VPA showed enrichment of binding sites

for 56 (JRC), 57 (UKK) and 66 (UKN1) TFs. Twenty-five TFBSs overlapped between all

samples treated with VPA (Fig. 10A), i.e. there was a relatively high overlap of responses on

the level of TFBS. A similar behaviour was observed after treatment with MeHg: less than

10% of the PS overlapped between UKN1 and UKK. Amongst these PS, 46 TFBS (UKN1) or

44 TFBS (UKK) were overrepresented and out of these twenty (> 40%) overlapped (Fig.

10B).

In view of these findings, it was interesting to look at an overlap of transcriptome

changes common to each of the toxicants in all test systems. We identified the PS and TFBS

jointly modified in all three test systems by VPA or in UKN1 and UKK by MeHg. Only 3

(0.5%) of the PS generally altered by VPA were also significantly affected by MeHg (Fig.

10C). In contrast, more than 50% of all TFBS common to MeHg or VPA overlapped also

between the two compounds (Fig. 10C). The large overlap of commonly enriched TFBS

between all test systems and compounds provides evidence for the existence of a set of

‘common transcription factors’ (including e.g. E2F, ETF, SP1 and AP-2 (Fig. S8). The only

TFBS enriched by all VPA treatments, but not MeHg, was the homeobox gene Hmx3 (also

known as NKX5.1). The only TFBS enriched by all MeHg treatments, but not VPA, was the

one for GCM transcriptional regulators (Fig. S8).

Similar comparisons of compound responses were also performed in individual test

systems. For instance, in UKK only 205 PS of the 3892 PS regulated by VPA overlapped with

those affected by MeHg (Fig. 10D). On the level of TFBS, the overlap was much larger, as 22

of the 57 TFBS enriched in the genes regulated by VPA, were also found for MeHg (Fig.

S9A).



87

Treatment of the UKN1 test system with VPA or MeHg resulted in the regulation of

genes associated with 66 TFBS in their promoter in the case of VPA and 46 TFBS in the case

of MeHg. Of these, 29 (comprising e.g., AP-2, EGR, STAT1, HIF-1, AhR, Sp1) were similar

for both compounds, 37 (comprising e.g., HSF-1, IRF-1, PAX5, NKX2-5) were specific for

VPA, and 17 (comprising e.g., ATF4, HOXA4, ZIC2) specific for MeHg (Fig. S9B). Again,

the overlap of TFBS was much larger than the one of individual PS. Only 142 of the 3697

genes regulated by VPA overlapped with those affected by MeHg (Fig. 10E).

Besides the commonly-regulated TFBS, we found for each compound also TFBS that

were specific for the test system and the chemical used. These may be used as signatures for

related chemicals within one class, while the commonly-affected TFBS may give a general

indication of toxicity (Supplementary Table S2). In conclusion, a remarkable observation of

Fig. 10: Overlap of altered genes and of

overrepresented transcription factor (TF)

binding sites between test conditions. Five sets of data, as described in Fig. 3 were used

for further analysis and comparisons: exposure of

UKK and UKN1 to both VPA and MeHg and of

JRC to VPA. All toxicants were used at their

BMC. The numbers of differentially expressed

probe sets (Limma t-test, Benjamin-Yakuteli

adjusted p value < 0.05), and enriched

transcription factor (TF) binding sites (PRIMA, p

value < 0.05) were identified. The data are

presented as pairs of Venn diagrams, with PS to

the left and TFBS to the right. Numbers on the

diagrams show the relevant count for each sector

of the diagram. The following sets of data are

compared: A) responses to VPA treatment in the

JRC, UKK and UKN1 test systems; B) responses

to MeHg treatment in UKK and UKN1 (N.B. for

display rules: 44 TFBS were changed in UKK, 20

of which overlapped with UKN1); C) the circles

marked ‘VPA’ show the number of PS/TFBS

regulated in all three test systems by VPA, the

circles marked ‘MeHg’ show the number of

features co-regulated in UKN1 and UKK by

MeHg; D) responses of UKK alone to MeHg or

VPA; E) responses of UKN1 to MeHg and VPA.



88

the present study is that the TFBS showed an astonishingly large overlap in view of the very

small overlap on the level of the individual genes. Analysis of further compounds is required

to determine whether the emerging concept of a ‘common toxic response TFBS’ and a

‘compound specific TFBS’ is universal.



89

Materials and Methods

Chemicals:

Valproic acid (VPA), mannitol, methylmercury chloride (MeHg) were obtained from

Sigma. Stocks of VPA and mannitol were prepared in water. MeHg was initially dissolved in

10% ethanol. A concentration of 10 mM MeHg in this solvent was used as a master stock. For

experiments, the MeHg solution was pre-diluted 1:1000 in water (final solvent concentration

0.1%) and used as the stock for further dilution with medium. The highest test solvent

concentration used in this study (at 1.5 µM MeHg) was 0.0015% ethanol.

Cell culture maintenance and experimental setup:

UKK: NIH-registered H9 human embryonic stem cells (WA09, WiCell Research

Institute, Madison, WI, USA) were cultured in DMEM-F12, 20% KO serum replacement, 1%

non-essential amino acids, penicillin (100 units/ml), streptomycin (100 µg/ml) and 0.1 mM β-

mercaptoethanol supplemented with 4 ng/ml human recombinant basic fibroblast growth

factor (bFGF) at 37 °C and 5% CO2. The undifferentiated stem cells (hESCs) were routinely

passaged with mechanical dissociation on irradiated mouse embryonic fibroblasts (MEF).

Prior to differentiation, the cells were maintained for five days in 60-mm tissue culture plates

(Nunc, Langenselbold, Germany) coated with a hESC-qualified matrix (BD Biosciences,

California, USA) in TESR1 medium (Stem Cell Technologies, mTESR1 basal medium +

mTESR1 5x supplement). For multilineage differentiation, embryoid bodies (EBs) were

prepared as described previously (Jagtap et al 2011) with minor changes (60 to 70 clumps

were added and bacteriological plates were not coated with pluronic), and the EBs were

maintained for 14 days on a horizontal shaker with or without drug treatment. Toxicant

exposure was performed as indicated in Fig. 1.

UKN1: H9 hESCs (as for UKK) were differentiated by dual SMAD inhibition as

described earlier in detail (Balmer et al 2012, Chambers et al 2009, Weng et al 2012). Briefly,

hESC were plated as single cells at a density of 18 000 cells /cm² in medium previously

conditioned for 24 h with mitomycin C-inactivated mouse embryonic fibroblasts, containing

10 µM ROCK inhibitor Y-27632 and 10 ng/ml bFGF. Medium was changed daily to

conditioned medium containing 10 ng/ml bFGF for 2 days. Differentiation was initiated 3

days after re-plating on day of differentiation (DoD) 0 by changing the medium to knockout

serum replacement medium (KSR) (Knockout DMEM with 15% knockout serum

replacement, 2 mM Glutamax, 0.1 mM MEM non-essential amino acids and 50 µM beta-



90

mercaptoethanol) supplemented with 35 ng/ml noggin, 600 nM dorsomorphin and 10 µM SB-

431642. From DoD4 onwards, KSR was replaced stepwise with N2 medium (DMEM/F12

medium, 1% Glutamax, 1.55 mg/ml glucose, 0.1 mg/ml apotransferrin, 25 µg/ml insulin,

100 µM putrescine, 30 nM selenium and 20 nM progesterone), starting with 25% N2 medium

at DoD4. To assess the chemical effects on RNA expression, the cells were differentiated in

the presence or absence of the chemicals from DoD 0 for 6 days.

JRC: NIH-registered H1 hESCs (WiCell Research Institute, Madison, WI, USA) were

cultured as described previously (Stummann et al 2009). Briefly, cells were maintained in

DMEM-F12, 20% KO serum replacement, 1% non-essential amino acids, penicillin (50

units/ml), streptomycin (50 µg/ml), 0.1 mM β-mercaptoethanol and 2 mM glutamine

supplemented with 4 ng/ml bFGF at 37 °C and 5% CO2. The hESCs were routinely passaged

with mechanical dissociation on irradiated MEFs. Prior to differentiation, hESC were grown

in suspension for 2 days in maintenance medium without bFGF to aggregate the colonies.

Then, neuronal differentiation was initiated by plating the aggregates on 20 µg/ml fibronectin-

coated cell culture plates in neural induction medium, consisting of DMEM/F12 medium

supplemented with 1% non-essential amino acids, 50 U/ml penicillin and 50 µg/ml

streptomycin, 1% “N2 supplement”, 0.04 mg/ml heparin and 0.2 ng/ml bFGF. The attached

colonies formed neural tube-like rosette structures.

UNIGE: For neural differentiation, an aliquot of H9 cells (WA09, WiCell Research

Institute, Madison, WI, USA) was thawed and cultured in suspension in T75 flasks with

N2B27 medium (Life Technologies). From day 2 to 7, cells were incubated in N2B27

medium supplemented with 10 µM anti TGF-beta (Ascent), and 2 µM dorsomorphin (Tocris

Bioscience). From day 8 to 32, medium replacement was performed with N2B27 medium

only. On day 33, generated spheres were dissociated as single cells and cultured in N2B27

medium in poly-ornithine (PLO) and laminin coated 6-well plates. On day 36, cells were

detached and frozen in N2B27 medium in different aliquots. To test neurotoxicity of chemical

compounds, an aliquot was thawed in PLO and laminin coated 6-well plates. Cells were

cultured in a neuronal differentiation medium (ND medium) made of NB medium, B-27

supplement, 2 mM L-Glutamine and penicillin/streptomycin (Life Technologies) as well as 10

ng/ml BDNF, 10 ng/ml recombinant human glial cell-derived neurotrophic factor (GDNF)

(Chemie Brunschwig) and 10 µM ROCK inhibitor (Ascent). After one day of recovery, cells

were incubated with the neurotoxicant in ND medium without ROCK inhibitor for 2 days,

and then material was collected for analysis.



91

UKN4: Lund human mesencephalic cells (LUHMES) were cultured exactly as described

earlier (Scholz et al 2011, Stiegler et al 2011). Briefly, cells were maintained in advanced

DMEM-F12, 1 x “N2 supplement”, 2 mM L-glutamine and 40 ng/ml bFGF at 37°C in a

humidified 95% air/5% CO2 atmosphere on Nunclon™ plastic cell culture flasks, coated with

50 ng/ml PLO and 1 μg/ml fibronectin. Proliferating cells were enzymatically dissociated

with trypsin (138 mM NaCl, 5.4 mM KCl, 6.9 mM NaHCO3, 5.6 mM D-Glucose, 0.54 mM

EDTA, 0.5 g/l trypsin from bovine pancreas type-II-S) and passaged every other day.

For differentiation, 8 x 106 LUHMES were seeded into a T175 flask in proliferation

medium, and differentiation was started after 24 h on day 0 (d0), by changing to advanced

DMEM-F12, 1x “N2 supplement”, 2mM L-glutamine, 1 mM dibutyryl 3’,5’-cyclic adenosine

monophosphate (cAMP), 1 μg/ml tetracycline and 2 ng/ml GDNF. After 2 days of cultivation

in culture flasks, cells were trypsinized and seeded onto PLO/fibronectin-precoated 96-well

plates at a cell density of 30 000/well in advanced DMEM-F12, 1x “N2 supplement”, 2 mM

L-glutamine, 1 μg/ml tetracycline. One hour after replating, cells were exposed to toxicants

for 24 h.

Affymetrix gene chip analysis:

Analysis was performed as described earlier (Balmer et al 2012, Jagtap et al 2011).

Briefly, samples from approximately 5x106 cells were collected using RNAprotect reagent

from Qiagen. The RNA was quantified using a NanoDrop N-1000 spectrophotometer

(NanoDrop, Wilmington, DE, USA), and the integrity of RNA was confirmed with a standard

sense automated gel electrophoresis system (Experion, Bio-Rad, Hercules, CA, USA). The

samples were used for transcriptional profiling when the RNA quality indicator (RQI) number

was > 8. First-strand cDNA was synthesised from 100 ng total RNA using an oligo-dT primer

with an attached T7 promoter sequence, followed by the complementary second strand. The

double-stranded cDNA molecule was used for in vitro transcription (IVT, standard

Affymetrix procedure) using Genechip 3’ IVT Express Kit. During synthesis of the aRNA

(amplified RNA, also commonly referred to as cRNA), a biotinylated nucleotide analog was

incorporated, which serves as a label for the message. After amplification, aRNA was purified

with magnetic beads, and 15 μg of aRNA were fragmented with fragmentation buffer as per

the manufacturer’s instructions. Then, 12.5 μg fragmented aRNA were hybridised with

Affymetrix Human Genome U133 plus 2.0 arrays as per the manufacturer’s instructions. The

chips were placed in a GeneChip Hybridization Oven-645 for 16 h at 60 rpm and 45 ºC. For

staining and washing, Affymetrix HWS kits were used on a Genechip Fluidics Station-450.



92

For scanning, the Affymetrix Gene-Chip Scanner-3000-7G was used, and the image and

quality control assessments were performed with Affymetrix GCOS software. All reagents

and instruments were acquired from Affymetrix (Affymetrix, Santa Clara, CA, USA).The

generated CEL files were used for further statistical analysis. The authors declare that

microarray data were produced according to MIAME guidelines and will be deposited in

ArrayExpress upon acceptance of the manuscript.

Cytotoxicity testing:

In order to determine the cytotoxic range of the chemicals to be tested with the DNA

microarrays (DMA), a resazurin assay was performed in all test systems. The assay is based

on the capability of viable and healthy cells to reduce resazurin to resorufin, which can be

measured by a colorimetric or fluorimetric shift as described earlier (Stiegler et al 2011,

Stummann et al 2009). Exposure time to chemicals and day of analyses for this endpoint was

the same as for the experimental setup of the RNA sampling (Fig. 1). Chemicals were tested

at several concentrations. Each condition was run in technical triplicates in at least three

independent biological experiments. On the day of analysis, cells were incubated with 10

µg/ml resazurin for 30 min to 1 h at 37°C and 5% CO2. To determine the background

fluorescence of resazurin itself, a control with only resazurin in medium was included.

Resorufin was measured at a wavelength of 560Ex/590Em with a fluorescence reader. The

mean background fluorescence of resazurin was subtracted from all experimental data.

Further data processing to identify the lowest non-cytotoxic ‘benchmark concentration’

(BMC) of the chemicals was done as follows: data from each experiment were normalised to

their respective untreated controls (set as 100%). The data were then displayed in

semilogarithmic plots. Data points were connected by a non-linear regression sigmoidal dose-

response curve fit. These curves were averaged, and the average curve was plotted. The BMC

was then determined graphically as the data point on the average curve corresponding to the

90% viability value, or as the last real data point left of this value. The BMC was used as test

concentration for DMA analysis. The “lower test concentration” (LOW) was determined by

dividing the BMC by a factor of four.

In vitro-in vivo extrapolation:

In vitro-in vivo extrapolation (IVIVE) of toxicity data can be achieved using

physiologically based pharmacokinetic (PBPK) modeling (Carrier et al 2001, Forsby &

Blaauboer 2007, Louisse et al 2010, Rotroff et al 2010, Verwei et al 2006, Wetmore et al



93

2012).The extrapolation is based on the implicit assumption that equal concentrations at the

target site in vitro and in vivo lead to equal effects. In this project, in vitro nominal

concentrations equivalent to relevant toxic concentrations in vivo were determined in two

steps. (1) PBPK modeling was used to simulate systemic concentrations corresponding to the

lowest dose level at which neurodevelopmental effects were observed in rats. The acslX

software was used for the simulations (v3.0.1.6; Aegis Technologies, Huntsville AL, USA).

(2) The unbound fraction may differ between in vitro and in vivo systems due to differences in

albumin concentrations and lipid fractions between plasma or extracellular fluid and test

medium. The nominal in vitro concentration Cvitro equivalent to the maximum systemic

concentration in vivoCpl was derived by correcting for these differences by:

pl

vitroplb

plLow

vitroLowplbplvitro

P

Pf

VFK

VFKfCC ,

,

,,

1

11

where fb,pl is the plasma bound fraction, VFL,pl and VFL,vitro are the volume fractions of

lipids in plasma and in vitro, Ppl and Pvitro are the concentrations of albumin in plasma and in

vitro(Gulden & Seibert 2003). Supplementary figure S6B shows the lipid content and albumin

concentrations in the test systems and in rat plasma.

IVIVE of MeHg data. The kinetics of MeHg in rats were previously described using a

detailed PBPK model by (Carrier et al 2001). This PBPK model was used in the current

project to predict systemic concentrations of MeHg after exposure to dosages known to result

in relevant toxic effects in vivo. A comprehensive review of neurodevelopmental toxicity of

MeHg in laboratory animals was published by (Castoldi et al 2008b). The lowest maternal

exposures in rat leading to behavioural and neurophysiological effects in the offspring were

between 0.01 and 0.05 mg/kg/day from gestation day 6 to 9 (Bornhausen et al 1980). MeHg

extensively binds to intra- and extracellular proteins by formation of cysteine complexes. The

MeHg-cysteine complexes readily pass placental and blood-brain barriers by facilitated

transport (Gray 1995). Maternal and fetal blood concentrations were found to be similar (Gray

1995).The total blood concentration was therefore assumed to be available for fetal brain

exposure, and equated to the nominal concentration in vitro.

IVIVE of VPA data. A PBPK model for VPA was developed and calibrated according to

data of (Binkerd et al 1988) and (Kobayashi et al 1991). Model equations and

parameterization are given in the supplemental material (Fig. S6). This model was used to

predict systemic VPA concentrations corresponding to the lowest dose at which



94

neurodevelopmental effects were observed in rats in vivo. A single intraperitoneal dose of

VPA in rat dams of 350 mg/kg was found by (Rodier et al 1996) to cause behavioural and

neuro-morphological effects in the offspring. Oral and intraperitoneal doses lead to

comparable plasma kinetics (Ingram et al 2000). VPA is known to pass the placental barrier in

several species; therefore, comparable VPA concentrations were assumed in maternal and

cord plasma. The unbound concentration in plasma was equated to the unbound test medium

concentrations. For correction of binding, a bound fraction in plasma of 63% was used

(Loscher 1978).

Statistical analysis of gene array data:

The following analyses were performed using the statistical programming language “R -

version 2.15.1” For normalisation of the entire set of 190 Affymetrix gene expression arrays,

the Robust Multi-array Average (RMA) algorithm (Irizarry et al 2003) was used that applies

background correction, log2 transformation, quantile normalisation, and a linear model fit to

the normalised data to obtain a value for each probe set (PS) on each array. To avoid having

to re-normalise future generated data for comparison with the current data, we used the R

package RefPlus (Harbron et al 2007) that allows the user to perform extrapolation strategies

by remembering the normalization parameters. After normalization, gene expression for each

gene at each concentration was adjusted by comparing the expression to the corresponding

control array expression, i.e. the difference between gene expression at each concentration

compared to the control was calculated (paired design).

Differential expression was calculated using the R-package limma (Smyth et al 2005).

Here, the combined information of the complete set of genes is used by an empirical Bayes

adjustment of the variance estimates of single genes. This form of a moderated t-test is

abbreviated here as ‘Limma t-test’. The resulting p-values were multiplicity-adjusted to

control the false discovery rate (FDR) by the Benjamini-Yekutieli procedure. As a result, for

each combination of centre (= test system), compound, and concentration, a gene list was

obtained, with corresponding estimates for log fold change and p-values of the Limma t-test

(unadjusted and FDR-adjusted).

Data display algorithms:

General test quality control was as described (Leist et al 2010). Heatmaps were used to

visualise matrices of gene expression values. Colour encodes the magnitude of the values,

ranging from yellow (low) to red (high). Volcano plots were used to visualise genome-wide



95

differential expression. Gene wise fold-change values (log2 scale) are plotted against

(unadjusted or FDR-adjusted Limma t-test) significance values (negative log10 scale) on the

x-axis and y-axis, respectively. Principal component analysis (PCA) plots were used to

visualise expression data in two dimensions, representing the first two principal components,

i.e. the two orthogonal directions of the data with highest variance. The percentages of the

variances covered are indicated in the figures. The software “R - version 2.15.1” was used for

all calculations and display of PCA and heatmaps (R_Development_Core_Team 2011). The

calculation and display of toxicity curves was done using GraphPad Prism 5.0 (GraphPad

Software, La Jolla, USA). The Venn diagrams for comparison of gene expression, gene

ontology (GO) terms and transcription factor binding sites (TFBS) between test systems were

constructed according to (Chow & Rodgers 2005). The size of circles and areas was chosen

proportional to the number of elements included.

Transcription factor binding site enrichment (TFBSE) was performed using the PRIMA

algorithm ((Elkon et al 2003); http://acgt.cs.tau.ac.il/prima/) provided in the Expander

software suite (version 6.04, (Ulitsky et al 2010); http://acgt.cs.tau.ac.il/expander/). Lists of

significant differentially expressed genes with adjusted p value <0.05 were converted to

Entrez Ids (R package hgu133plus2.db) and duplicates were removed. The PRIMA algorithm

was run with a p-value threshold set to 0.05, no multiple testing correction, a background set

of all human genes (provided in the Expander software suite), and using the TRANSFAC

database (8.2) as the data source for transcription factor binding sites. The PRIMA algorithm

analyses 267 separate TRANSFAC entries. PRIMA results are presented in tables with TF

identifiers provided by PRIMA and their full names, or the overlap between TF enrichments

for different treatments, is shown as Venn diagrams or as network diagrams (Cytoscape;

(Shannon et al 2003, Smoot et al 2011); http://www.cytoscape.org).

For the word clouds of the overrepresented gene ontology groups, a g:Profiler query

(Reimand et al 2007) was initially made, and only results from the biological process and

pathway branches were retained. These were viewed as a subgraph of the whole gene

ontology tree. All categories were deleted that were larger than 1000 genes and smaller than

50 genes. Then, connected components from the remaining graph were identified, and from

each of these, the category with the highest p-value was selected. These were ordered by p-

value and the top 40 are displayed. When displaying the categories, the font sizes were first

scaled to be proportional to the log10 of enrichment p-value. To enable global comparison,

the grey shade of the letters was scaled the same way over all plotting windows.



96

To assess the sensitivity of differential expression analysis with respect to the number of

DMA (= experimental replicates), the following approach was used: For each condition, we

identified the differentially expressed genes based on five pairs of DMA (control vs treated),

which was then used as the reference list. Significant PS were identified in all cases by

Limma t-test, with a p < 0.05 as significance threshold. The Benjamini-Hochberg and the

Benjamini-Yekutieli were used for the FDR correction in different experiments as appropriate

and as specified in the figure legends. All possible permutations of 2, 3 or 4 DMA were

calculated, and the differentially expressed PS of all these conditions were identified (using

the same method as for the reference calculation). Finally, the overlap between the new gene

lists and the reference was calculated, to determine the quantity of the reference that could be

recovered with less DMA.



97

Supplements

Suppl. Table S1 (differentially regulated probe sets (PS) of five test systems with several

conditions) and Suppl. Table S2 (overrepresented TFBS of five test conditions) can be found

online : http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3535399/?report=classic

0

10

20

30

40

50

60

70

80

90

100

110

120

experiment 1

experiment 2

ctrl

experiment 3

experiment 4

experiment 5

1010.10.010.00110-410-510-6

Upper test concentration = 0.05 µM

Lower test concentration = 0.0125 µM

mean

MeHg [µM]

resazu

rin

red

ucti

on

[% o

f co

ntr

ol

SE

M]

0

10

20

30

40

50

60

70

80

90

100

110

120

experiment 1

experiment 2

ctrl

experiment 3

1010.10.01


Upper test concentration = 1 µM

mean

MeHg [µM]

resazu

rin

red

ucti

on

[% o

f co

ntr

ol

SE

M]

UKN4

UKK

0

10

20

30

40

50

60

70

80

90

100

110

experiment 6

experiment 1

experiment 2

experiment 3

experiment 4

experiment 5

crtl 1010.10.01

Upper test concentration = 1.5 µM


mean

MeHg [µM]

resazu

rin

red

ucti

on

[%

of

co

ntr

ol]

UKN1

Fig. S1:

Determination of the test

concentrations for DNA

microarray analysis. Cells were treated with MeHg at the

concentrations indicated, under

conditions described for each test

system in Fig. 1. At the end of the

incubation period, cell viability was

determined by the resazurin

reduction assay. Data are normalized

to untreated control cultures which

were defined as 100%. The data

points are averages ± SD from three

technical replicates. Each of the

experiments was repeated several

times (indicated by different color

codings) with different cell

preparations. The data from the

different biological experiments were

averaged (black line). To determine

the “highest non-cytotoxic

concentration”, the BMC was

determined graphically, taking the

variation of individual experi-mental

systems into account. This “upper

test concentration” (= BMC) of the

drug is indicated by the red dashed

line. The “lower test concentration”

(LOW) was determined by dividing

the BMC by a factor of four. This is

indicated by a blue dashed line.



98

B

MeHgVPAMannitol DMSO

A

JRCUKK UKN1

Fig. S2: Differential alterations of gene expression by valproic acid (VPA) and methyl

mercury (MeHg). Three different test systems (UKK, UKN1, JRC) were exposed to VPA (blue label on top of the

heatmap) or MeHg (green label), at their respective bench mark concentration, or to D-mannitol

(red). The differentially expressed genes (vs untreated controls) were determined in 4-5

independent experiments (shown as lanes of the heatmaps). The similarity of the gene expression

patterns is indicated by the Pearson’s distance dendrogram on top. The heatmaps are based on 100

selected genes.

A. These comprise the 100 genes with the lowest adjusted p-values according to the Limma t-test

for regulation by VPA. B. These comprise the 100 genes with the lowest adjusted p-values

according to the Limma t-test for regulation by MeHg.

The colors of the heatmap (yellow red glow lookup table) indicate the relative gene regulation

level above or below the average for each row.



99

Fig. S3 Volcano plot analysis of gene array data after incubation of the test systems

UKK, JRC, UNIGE-1 and UKN 4 with MeHg. Data were generated and calculated for each combination of test system and compound, as

illustrated in Fig. 3. In the volcano-plot diagrams, fold-changes of the compound-induced gene

expression are shown on the x-axis (log2-scale). The y-axis shows negative logarithmic adjusted p-

values of a LIMMA t-test. (-log10(p-value)). The p-values were A. FDR adjusted, or B. not FDR

adjusted. The dashed lines show the p = 0.05 significance level for optical guidance.

+ FDR

- FDR

JRC – 68 nM UKK – 250 nM UNIGE – 40 nM

Sign

ific

ance

Sign

ific

ance

0

2

6

4

8

10

1/16 1/4 1 4

p = 0.05

p = 0.000001

JRC – 273 nM UKK – 1 µM

UKN4 – 50 nM

UKN4 – 200 nM UNIGE – 160 nM

16

0

2

6

4

8

10

1/16 1/4 1 416 1/16 1/4 1 416 1/16 1/4 1 416

JRC – 68 nM UKK – 250 nM UNIGE – 40 nM

Sign

ific

ance

Sig

nif

ica

nce

0

2

6

4

8

10

1/16 1/4 1 4

JRC – 273 nM UKK – 1 µM

UKN4 – 50 nM

UKN4 – 200 nM UNIGE – 160 nM

16

0

2

6

4

8

10

1/16 1/4 1 416 1/16 1/4 1 416 1/16 1/4 1 416

Fold change Fold change Fold change Fold change

Fold change Fold change Fold change Fold change

A

B



100

Fig. S4 Principle component analysis (PCA) of regulated genes in several test systems

after subtraction of controls The signal of all PS was determined in five different test systems systems (UKK, UKN1, JRC,

UKN4 and UNIGE) after incubation with compounds as in Fig. 3. Then, the values for the

respective controls were subtracted from the values of the DMAs treated with VPA at the BMC

(large blue) or at the LOW concentration (small blue dots), or MeHg (large and small green dots),

or D-mannitol (red) or DMSO (black). These data were then used for PCA analysis. The lower

right panel shows all data together, the other panels show the data for individual test systems within

the same axes as for all systems. Corresponding controls have been subtracted. The number of PS

were now stepwise reduced retaining only the PS with highest variability. A: all probesets

(corresponds to Fig. 3C), B: 5000 probesets, C: 1000 probesets, D: 500 probesets, E: 200 probesets,

F: 100 probesets. Good separation results were still obtained using only 500 probesets. Further

reduction to e.g. 100 probesets did no longer allow good separation.

Figures are displayed on the next three pages



101

A

B

All probesets

5000 probesets

MeHg

VPA

Mannitol

DMSO



102

C

D

1000 probesets

500 probesets



103

E

F 100 probesets

200 probesets



104

probesets (PS) overlapping with reference PS new PS not present in reference PS

Fig. S5: Simulation of different numbers of experiments (pairs of DMAs) and their

impact on the numbers of significantly-regulated PS VPA was tested in the test systems JRC und UKK at its BMC in 5 independent experiments, and in

UKN1 in 4 experiments. MeHg was tested in UKN1 in 5 experiments. The number of significantly

regulated genes (Benjamini-Yekutieli FDR correction) was calculated without further restrictions

(left) or with the restrictions that PS should be regulated more than 2-fold (right). The numbers of

PS are indicated above the dashed black lines and they were set as 100% reference points. The blue

bars indicate how many of these PS were identified when different permutations of 2, 3 or 4

experiments (indicated as grey headings) were used. The light blue bars indicate how many

additional PS were identified, when only subsets of the original 5 (4) experiments were analysed.

For instance, the 5 bars in the panel with the coordinates 4/JRC:VPA represent the five possible

ways of leaving out one of the experiments. The 10 bars in the panel with the coordinates

3/JRC:VPA represent the 10 possible permutations of leaving out 2 of the experiments and then

recalculating the significant PS on the basis of the remaining 3 DMA.



105

Plasma Liver Bile

Vmax

Km + Cliv

ka

Intestine 1

k12

k21

kbiIntestine 2

p.o.

(i.p.)

Vmax

Km + Cliv

khyd

F *

(1 - F) *

Fig. S6A: Schematic representation of the PK model developed for VPA. The following differential equations describe the rates of change of VPA (µg/min) in the model

compartments, where A denote amounts in µg and C denote concentrations in µg/mL:

Plasma compartment

livplinαpl

AkAkAkdt

dA 21122

V

AC

plpl

Bile compartment

liv

livm

malivpl

livA

CK

VAkAk

dt

dA

x2112

liv

livliv

V

AC

Intestinal compartments

bibiliv

livm

mabiAkA

CK

VF

dt

dA

x

11

inhydbibiin

AkAkdt

dA

Liver compartment

11

inhydbibiin

AkAkdt

dA

212

inainhydin

AkAkdt

dA

Values for the following parameters were obtained by fitting to data presented by Binkerd et al. (1988) and Kobayashi (1991):

ka – absorption rate constant; 0.05 min-1

k12 – plasma-to-liver transfer rate constant; 0.274 min-1

k21 – liver-to-plasma transfer rate constant; 0.279 min-1

V – (initial) volume of distribution; 51.6 mL

Vliv – liver compartment volume; 12.3 mL

Vmax – maximum velocity bile excretion; 25.2 µg/mL/min

Km – Michaelis constant bile excretion; 362 µg/mL

F – fraction excreted into bile; 0.18

kbi – rate constant for bile flow to intestine; 0.0033 min-1

khyd – rate constant for hydrolysis of glucuronidated VPA; 0.0062 min-1



106

Medium lipid content

albumin

concentration

[mg/l] [µM]

UKK 120 244.3

UKN1 92 184.7

JRC 2.8 5.7

UNIGE* 2.9 5.8

UKN4 2.9 5.8

Rat plasma 3600 421.0

Human plasma 6000 600

Fig. S6B: Estimated lipid content and albumin concentration in in vitro test media as

well as rat and human plasma Test medium lipid content and albumin concentrations were calculated on the basis of available

supplier information on medium constituents. The data on rat plasma used in the present in vitro-in

vivo correlation have been adopted from Verwei et al. (2006). The original references are Barber et

al. (1990) for albumin, and DeJongh et al (1997) for lipids. Human plasma values are mentioned

for comparison and were taken from Gülden and Seibert (2003). The original data on albumin are

from Lindup et al (1987) and for lipids from Patterson et al (1988). Note that plasma lipid content

is highly dependent on diet. Rat values are assumed to reflect average values on standard chow,

human values are average values under fasting conditions.

* B27 medium composition is not disclosed; by assumption the same albumin and lipid

concentrations as DMEM/F12 were used.

Barber, B.J., Schultz, T.J., Randlett, D.L., 1990. Comparative analysis of protein content in rat

mesenteric tissue, peritoneal fluid and plasma. Am. J. Physiol. Gastrointest. Liver

Physiol. 258, G714–G718.

DeJongh, J., Verhaar, H.J.M., Hermens, J.L.M., 1997. A quantitative property–property r

elationship (QPPR) approach to estimate in vitro tissue-blood partition coefficients or

organic chemicals in rats and humans. Arch. Toxicol. 72, 17–25.

Gulden M, Seibert H (2003) In vitro-in vivo extrapolation: estimation of human serum

concentrations of chemicals equivalent to cytotoxic concentrations in vitro.

Toxicology 189(3):211-22

Lindup, W.E., 1987. Plasma protein binding of drugs: some basic and clinical aspects. In: Bridges,

J.W., Chasseaud, L.F., Gibson, G.G. (Eds.), Progress in Drug Metabolism, vol. 10. Taylor

and Francis, London, pp. 141-185.

Patterson, D.G., Jr., Needham, L.L., Pirkle, J.L., Roberts, D.W., Bagby, J., Garrett, W.A.,

Andrews, J.S., Falk, H., Bernert, J.T., Sampson, E.J., Houk, V.N., 1988. Correlation

between serum and adipose tissue levels of 2,3,7,8-tetrachlorodibenzo-p -dioxin in

50 persons from Missouri. Arch. Environ. Contam. Toxicol. 17, 139-143.

Verwei M, van Burgsteden JA, Krul CA, van de Sandt JJ, Freidig AP (2006) Prediction of in vivo

embryotoxic effect levels with a combination of in vitro studies and PBPK modelling.

Toxicol Lett 165(1):79-87



107

Fig. S7: Overrepresented gene ontology groups

UKK.VPAUp: 1765

Down: 2127

nuclear divisionmitosis

mitotic cell cycle

negative regulation of cellular ...

RNA processing

negative regulation of nitrogen ...

negative regulation of gene expr...

DNA metabolic process regulation of cell cycle positive regulation of tr anscrip...regulation of organelle organiza...regulation of RNA splicing

regulation of gene expression, e ...

attachment of spindle microtub ul...

positiv e regulation of v ascular ...

regulation of DNA metabolic processplasma lipoprotein par ticle clea...

DNA repairnegativ e regulation of cell cycl...

regulation of protein modificati...

cellular response to cadmium ionvesicle−mediated transport

multicellular organismal signaling

transmission of nerve impulse

cellular response to zinc ionresponse to organic nitrogen

synaptic transmission

chondrocyte development

GTP metabolic processorganic substance tr anspor t

neuron differentiation

biological adhesion

cell adhesion

enzyme linked receptor protein s ...

negative regulation of canonical...

regulation of membrane depolariz... response to erythropoietin

UKN1.VPAUp: 1533

Down: 2164

negative regulation of RNA metab...negative regulation of nitrogen ...

negative regulation of macromole...


neural tube development embryonic morphogenesis

regulation of cell de velopment

positive regulation of transcrip...

tube closure

regulation of neurogenesisforebrain development

sensory organ development

regulation of non−canonical Wnt ...

anatomical str ucture formation i...

convergent extension involved in...positive regulation of neuron di...

peptidyl−lysine modification

positive regulation of fat cell ...dorsal/ventral axis specification

regulation of Rho protein signal... cell migrationbiological adhesion

cell adhesion

circulatory system development

cardiovascular system developmentcell morphogenesis involved in d...

neuron developmentepithelium development

neuron projection morphogenesis

collagen fibr il organization

response to inorganic substance

ossificationresponse to endogenous stim ulusnegative regulation of progr amme...

regulation of anatomical str uctu...wound healing

enzyme linked receptor protein s ...

taxis

leukocyte migration

positive regulation of cell comm...

JRC.VPAUp: 3817

Down: 3976

mRNA processingnegative regulation of RNA metab...



chromatin modification positive regulation of transcrip...peptidyl−lysine modification

posttranscriptional regulation o ...

regulation of RNA splicingnuclear−tr anscr ibed mRNA poly(A)... cellular macromolecule catabolic...

negativ e regulation of neuron di...

neural tube de velopment

nuclear−tr anscr ibed mRNA catabol...

brain de velopment negativ e regulation of oligodend...

protein deubiquitinationregulation of pro−B cell diff ere...

negativ e regulation of stem cell...

cerebr al cor tex neuron diff erent...

response to endogenous stimulusresponse to peptide hor mone stim...

nucleobase−containing small mole ...

cellular component mor phogenesis

purine−containing compound catab ...

nucleobase−containing compound c...

purine−containing compound metab ...

amine metabolic process

vesicle−mediated transportcell projection organization

protein N−linked glycosylationcellular amino acid metabolic pr ...

cellular response to cadmium ion

neuron projection mor phogenesis

glycoprotein metabolic processwound healing

protein folding

coagulation

hemostasis

regulation of protein ubiquitina...

UKN1.MeHgUp: 44

Down: 375

organ morphogenesisneuron differentiation


cardiovascular system development

epithelium developmentneuron projection de velopment

response to toxin

neuron apoptosis

cell migration

negative regulation of cell death

positive regulation of developme...

replacement ossificationplatelet activ ationregulation of cell diff erentiation


positive regulation of signaling

response to endogenous stim ulus

platelet degranulation

appendage development

osteoblast differentiation

Regulation direction

Down

Up

Enrichment P−value

10-15

10-7.5

1

UKK.VPAUp: 796

Down: 710

mitotic cell cyclechromosome organization

cell cycle phaseRNA processingDNA metabolic process

regulation of cell cycle

histone H3 acetylation regulation of DNA metabolic process

response to zinc ion

UKN1.VPAUp: 2164

Down: 1533




neural tube development

embryonic morphogenesis


positive regulation of tr anscrip...

tube closureregulation of neurogenesis

forebrain development sensory organ development


anatomical str ucture formation i...convergent extension involved in...

positive regulation of neuron di...


positive regulation of fat cell ...

dorsal/ventral axis specification regulation of Rho protein signal...

cell fate commitment

regulation of smoothened signali...

peptidyl−lysine acetylation

forebrain anterior/posterior pat...

axonal fasciculation

histone H3 acetylationnegative regulation of epithelia...

regulation of planar cell polar i...

negativ e regulation of planar ce ...

positiv e regulation of de velopme ...

chromatin modification

diencephalon de velopment

positiv e regulation of Wnt recep ...

smoothened signaling pathw ay

dendr itic spine de velopment

cell projection mor phogenesis

cell migrationbiological adhesion

cell adhesion



neuron development

epithelium developmentneuron projection mor phogenesis

collagen fibr il organizationresponse to inorganic substance

ossificationresponse to endogenous stimulus

negative regulation of programme...

regulation of anatomical str uctu...

wound healing

enzyme linked receptor protein s...

taxis

leukocyte migration



development of pr imary sexual ch...

regulation of ossification

positive regulation of transport

positive regulation of signal tr ...


skin developmentregulation of canonical Wnt rece ...

regulation of organ mor phogenesis

mast cell prolif eration

muscle str ucture de velopment

regulation of binding

outflo w tract mor phogenesis

negativ e regulation of signal tr ...

regulation of phosphor us metabol...

epithelial to mesench ymal tr ansi...

regulation of phospholipase acti...

monocarbo xylic acid tr anspor t

positiv e regulation of tr anscr ip...

positiv e regulation of cell prol...

JRC.VPAUp: 3976

Down: 3817






peptidyl−lysine modificationposttranscriptional regulation o ...

regulation of RNA splicing

nuclear−tr anscr ibed mRNA poly(A)...

cellular macromolecule catabolic...




brain de velopment

negativ e regulation of oligodend...

protein deubiquitination

regulation of pro−B cell diff ere...



cellular protein catabolic process

response to endogenous stimulusresponse to peptide hormone stim...








protein N−linked glycosylation

cellular amino acid metabolic pr ...


neuron projection morphogenesisglycoprotein metabolic process

wound healing

protein folding

coagulation

hemostasis


cell redox homeostasiscellular lipid metabolic process

electron transpor t chainamino acid tr anspor t

peptidyl−aspar agine modificationcellular response to hor mone sti...

glycoprotein biosynthetic process

positive regulation of protein u...

membrane organization

transmembrane receptor protein t...

protein maturation

phospholipid biosynthetic process

coenzyme biosynthetic process

heparan sulfate proteoglycan bio...

alcohol biosynthetic process

protein comple x biogenesis

post−tr anslational protein modif ...

positiv e regulation of catalytic...

molybdopter in cofactor metabolic...

protein comple x assembly

UKN1.MeHgUp: 375

Down: 44

neuron developmentcentral nervous system development

regenerationresponse to toxin

neuron apoptosis

neuron deathcell projection mor phogenesis

cell projection organization

axonogenesisplatelet activation

response to growth factor stimulusbody morphogenesis


embryonic appendage mor phogenesisregulation of neuron apoptosis

regulation of br anching involved...

bone trabecula morphogenesisbone trabecula formation

head development

nerve development

negative regulation of transmemb...transformed cell apoptosis

appendage morphogenesis


Down

Up


10-15

10-7.5

1

UKK.VPAUp: 796

Down: 710

mitotic cell cyclechromosome organization

cell cycle phaseRNA processingDNA metabolic process

regulation of cell cycle

histone H3 acetylation regulation of DNA metabolic process

response to zinc ion

UKN1.VPAUp: 2164

Down: 1533




neural tube development

embryonic morphogenesis



tube closureregulation of neurogenesis

forebrain development sensory organ development


anatomical str ucture formation i...convergent extension involved in...

positive regulation of neuron di...


positive regulation of fat cell ...

dorsal/ventral axis specification regulation of Rho protein signal...

cell fate commitment

regulation of smoothened signali...

peptidyl−lysine acetylation

forebrain anterior/posterior pat...

axonal fasciculation

histone H3 acetylationnegative regulation of epithelia...

regulation of planar cell polar i...

negativ e regulation of planar ce ...

positiv e regulation of de velopme ...


diencephalon de velopment

positiv e regulation of Wnt recep ...

smoothened signaling pathw ay

dendr itic spine de velopment

cell projection mor phogenesis

cell migrationbiological adhesion

cell adhesion



neuron development

epithelium developmentneuron projection mor phogenesis

collagen fibr il organizationresponse to inorganic substance

ossificationresponse to endogenous stimulus

negative regulation of programme...

regulation of anatomical str uctu...

wound healing

enzyme linked receptor protein s...

taxis

leukocyte migration



development of pr imary sexual ch...

regulation of ossification

positive regulation of transport

positive regulation of signal tr ...


skin developmentregulation of canonical Wnt rece ...

regulation of organ mor phogenesis

mast cell prolif eration

muscle str ucture de velopment

regulation of binding

outflo w tract mor phogenesis

negativ e regulation of signal tr ...

regulation of phosphor us metabol...

epithelial to mesench ymal tr ansi...

regulation of phospholipase acti...

monocarbo xylic acid tr anspor t

positiv e regulation of tr anscr ip...

positiv e regulation of cell prol...

JRC.VPAUp: 3976

Down: 3817






peptidyl−lysine modificationposttranscriptional regulation o ...

regulation of RNA splicing

nuclear−tr anscr ibed mRNA poly(A)...

cellular macromolecule catabolic...




brain de velopment

negativ e regulation of oligodend...

protein deubiquitination

regulation of pro−B cell diff ere...



cellular protein catabolic process

response to endogenous stimulusresponse to peptide hormone stim...








protein N−linked glycosylation

cellular amino acid metabolic pr ...


neuron projection morphogenesisglycoprotein metabolic process

wound healing

protein folding

coagulation

hemostasis


cell redox homeostasiscellular lipid metabolic process

electron transpor t chainamino acid tr anspor t

peptidyl−aspar agine modificationcellular response to hor mone sti...

glycoprotein biosynthetic process

positive regulation of protein u...

membrane organization

transmembrane receptor protein t...

protein maturation

phospholipid biosynthetic process

coenzyme biosynthetic process

heparan sulfate proteoglycan bio...

alcohol biosynthetic process

protein comple x biogenesis

post−tr anslational protein modif ...

positiv e regulation of catalytic...

molybdopter in cofactor metabolic...

protein comple x assembly

UKN1.MeHgUp: 375

Down: 44

neuron developmentcentral nervous system development

regenerationresponse to toxin

neuron apoptosis

neuron deathcell projection mor phogenesis

cell projection organization

axonogenesisplatelet activation

response to growth factor stimulusbody morphogenesis


embryonic appendage mor phogenesisregulation of neuron apoptosis

regulation of br anching involved...

bone trabecula morphogenesisbone trabecula formation

head development

nerve development

negative regulation of transmemb...transformed cell apoptosis

appendage morphogenesis


Down

Up


10-15

10-7.5

1



108

enriched not enriched

VPA MeHg

TFBS UKN1 UKK JRC UKN1 UKK

MOVO-B

SRY

Sp1

E2F

MAZ

EGR

ZF5

HIC1

UF1H3[b]

c-Myc:Max

ZNF219

HIF-1

E2F-1

AP-2

ETF

AhR:Amt

MTF-1

FOXP1

Egr-1

MZF1

Pax-4

STAT1

AP-2[a]

AHRHIF

Hmx3

VDR

GCM

Nkx6-2

Oct-1 c

Fig. S8: Enrichment of transcription factor binding sites (TFBS) amongst toxicant-

regulated genes. The test systems UKN1, JRC and UKK were exposed to VPA (pale red nodes) and UKN1 and

UKK were also exposed to MeHg (pale blue nodes) at their BMCs and the significantly regulated

genes were identified as in Fig. 3. The overrepresented TFBS in these sets of genes were

determined with the PRIMA algorithm. The lines of the diagram connect assays with enriched TF

nodes (p<0.05). TF nodes are coloured according to how many of the assays they were enriched in:

in all five treatments (green), in all VPA treatments (red), in all MeHG treatments (blue).



109

TFBS in both treatments

TF ID* TF full Name**

Oct-1Octamer binding factor 1; also

known as POU2F1: POU class 2

homeobox 1 (HGNC)

AhR:ArntAryl hydrocarbon receptor: Aryl

hydrocarbon receptor nuclear

translocator

AIREAutoimmune regulator

AP-2Activating protein 2

c-Myc:Max

v-myc myelocytomatosis viral

oncogene homolog (avian)

(HGNC): myc-associated factor

X

E2FE2F transcription factor

E2F-1E2F transcription factor 1

EGREarly Growth Responsive factor

ETFEGFR-specific transcription

factor

FAC1Now known as BPTF:

bromodomain PHD finger

transcription factor (HGNC)

FOXP1Forkhead box P1

HIC1Hypermethylated in cancer 1

HIF-1Hypoxia Induced Factor 1

MAZMyc-associated zinc finger

protein

MOVO-BMouse homologue of Drosophila

Ovo protein

MTF-1Metal-regulatory transcription

factor 1

Nkx6-2NK6 homeobox 2

Sp1Specificity protein 1

SRYSex-determining region Y

UF1H3BETAUf1h3beta transcription factor

ZF5Zinc finger protein 5

ZNF219Zinc finger protein 219

MeHg VPA

TFBS (MeHg only)


AREB6ZEB1: zinc finger E-box

binding homeobox 1 (HGNC)

c-Mybv-myb myeloblastosis viral

oncogene homolog (avian)

FOXJ2Forkhead box J2

Freac-3FOXC1: forkhead box

C1(HGNC)

GCMGlial cells missing factor A

Helios_AIKZF2: IKAROS family zinc

finger 2 (Helios) (HGNC)

HFH-1FOXM1: forkhead box M1

(HGNC)

HNF-1Hepatocyte nuclear factor

(HNF) 1 homeobox A

HSF1Heat shock transcription factor

1

IRF1Interferon regulatory factor 1

KAISOZBTB33: zinc finger and BTB

domain containing 33 (HGNC)

Lyf-1Lymphoid transcription factor 1

Pax-5Paired box 5

Pit-1Pituitary-specific factor 1

PLZFPromyelocytic leukemia zinc

finger

STATSignal Transducer and

Activator of Transcription

STAT4Signal Transducer and

Activator of Transcription 4

STATxFamily: signal transducers and

activators of transcription

SZF1-1ZNF589: zinc finger protein

589 (HGNC)

TBX5T-box protein 5

VDRVitamin D receptor

VDR,_CAR

,_PXR

VDR: vitamin D receptor; CAR:

constitutive androstane

receptor; PXR: pregnane X

receptor

TFBS (VPA only)


AFP1Alpha fetoprotein enhancer binding

protein

AhRAryl hydrocarbon receptor

AHRHIFAryl hydrocarbon receptor, hypoxia

inducible factor

AP-2alphaActivating protein 2 alpha

AP-2alphaAActivating protein 2 alphaA

ATFActivating transcription factor

ATF4Activating transcription factor 4 (tax-

responsive enhancer element B67)

Brn-2Brain-2; known as POU3F2: POU class 3

homeobox 2 (HGNC)

CBFCore binding factor

CDP_CR1Cut-like homeodomain protein

Egr-1Early Growth Responsive factor 1

GZF1GDNF-inducible zinc finger protein 1

(HGNC)

Hmx3H6 family homeobox 3

HOXA3Homeobox A3

Ik-1Ikaros 1 transcription factor


IPF1Insulin promoter factor 1

IRF-1Interferon regulatory factor 1

MEF-2Myocyte Enhancer Factor 2

MZF1Myeloid zinc finger 1

NanogNanog homeobox

NF-YNuclear factor Y (Y-box binding factor)

NKX3ANow known as NKX3-1: NK3 homeobox

1 (HGNC)

Nrf-1Nuclear respiratory factor 1

Pax-1Paired box gene 1




S8S8 homeobox

STAT1Signal Transducer and Activator of

Transcription 1

STAT5ASignal Transducer and Activator of

Transcription 5A

Tax/CREBTax: now known as CNTN2: contactin 2

(axonal) (HGNC)/ CREB: cyclic AMP

response element-binding protein

Tst-1Now known as POU3F1: POU class 3

homeobox 1 (HGNC)

USFUpstream stimulatory factor

WhnWinged-helix nude

*Transcription factor (TF) ID provided by

PRIMA/Expander

**source was usually TRANSFAC Public, release 7.0.

When appropriate, updated names from the HUGO Gene

Nomenclature Committee, www.genenames.org, are provided,

indicated by (HGNC)

Fig. S9A: Comparison of MeHg and VPA

responses with respect to transcription factor (TF)

enrichment The test system UKK was exposed to MeHg (1 µM) or

VPA (2 mM), and the significantly regulated probesets

were determined (as reported in Fig. 3). Statistical

overrepresentation of TF-binding sites (TFBS) in the

promoters of the regulated genes was determined with the

PRIMA algorithm for both treatments. TFBS were

grouped into those only found enriched for MeHg

treatment (red box), those only found for VPA treatment

(blue box) and those found to be enriched by both

compounds (purple box).



110

TFBS in both treatments


AhR:ArntAryl hydrocarbon receptor: Aryl

hydrocarbon receptor nuclear

translocator

AHR:HIFAryl hydrocarbon receptor,

hypoxia inducible factor

AP-2Activating protein 2

AP-2alphaActivating protein 2 alpha

CAC-BPCAC-binding protein

c-Myc:Maxv-myc viral oncogene homolog:

myc-associated factor X

E2FE2F transcription factor

E2F-1E2F transcription factor 1

EGREarly Growth Responsive factor

Egr-1Early Growth Responsive factor 1

ETFEGFR-specific transcription

factor

HIC1Hypermethylated in cancer 1

HIF-1Hypoxia Induced Factor 1

MAZMyc-associated zinc finger pr.

MAZRMAZ related factor

MEF-2Myocyte Enhancer Factor 2

MOVO-BHomologue of Ovo protein

MZF1Myeloid zinc finger 1

P300E1A-associated protein p300


RREB-1Ras-responsive element binding

protein 1

Sp1Specificity protein 1

SRYSex-determining region Y

STAT1Signal transducer and activator of

transcription 1

TFII-IGeneral Transcription Factor II-I

UF1H3BETAUf1h3beta transcription factor

VDRVitamin D receptor

ZF5Zinc finger protein 5

ZNF219Zinc finger protein 219

TFBS (VPA only)


Alx-4Aristaless homeobox like 4

aMEF-2myocyte-specific enhancer factor,

alternatively spliced exon

BLIMP1B lymphocyte induced maturation protein

1; now known as PRDM1

CDPCCAAT displacement protein

CDXCaudal type homeobox

CHOP:C/EB

Palpha

CAAT/enhancer binding protein

homologous transcription factor: CCAAT

Enhancer Binding Protein alpha

CHX10Now known as ZSX2: visual system

homeobox 2 (HGNC)

CP2Now known as TFCP2:; transcription

factor CP2 (HGNC)

DBPDNA binding protein

ELF-1Enhancer Lymphocyte Factor 1

FAC1Now known as BPTF: bromodomain PHD

finger transcription factor (HGNC)

FOXO4Forkhead box O4 (HGNC)

FOXP1Forkhead box P1

Freac-3known as FOXC1: forkhead box C1

HEN1Now known as NHLH1: nescient helix

loop helix 1

HFH-1Now known as FOXM1: forkhead box M1

Hmx3H6 family homeobox 3

HSF1Heat shock transcription factor 1




ISREInterferon-stimulated response element

Lyf-1Lymphoid transcription factor 1

MTF-1Metal-regulatory transcription factor 1

myogenin_/_

NF-1

MyoG; myogenin (myogeninc factor 4)

(HGNC) _/_ nuclear factor 1

NF-kB (p50)Nuclear Factor kappa B, p50



NkX6-1NK6 homeobox 1

NRSFNeuron-restrictive silencer factor

Olf-1ZNF423; zinc finger protein 423 (HGNC)

OTXOrthodenticle related homeobox protein 1

Pax-5Paired box 5

PU.1PUrine-box binding factor 1

Sox-5SRY (sex determining region Y)-box 5

TEF-1Transcriptional enhancer factor 1

TTF-1Thyroid transcription factor 1

MeHg VPA

TFBS (MeHg only)


ATFActivating transcription factor

ATF4Activating transcription factor 4

C/EBPdeltaCCAAT-enhancer-binding

protein delta

GCMGlial cells missing factor A

HNF4Hepatocyte nuclear factor 4

HOXA4Homeobox A4

LEF1Lymphoid enhancer-binding

factor 1

LRFLeukemia/lymphoma-related

factor

MAFv-maf musculoaponeurotic

fibrosarcoma oncogene

homolog (avian) (HGNC)


N-Mycv-myc related oncogene,

neuroblastoma derived

Oct-1Octamer binding factor 1; also

known as POU2F1: POU class

2 homeobox 1 (HGNC)

Sp3Stimulating Protein 3

SRFSerum response factor

Tax/CREBCNTN2: contactin 2/CREB:

cyclic AMP response element-

binding protein

TFIIAGeneral transcription factor IIA

Zic2Zinc finger protein of the

cerebellum 2

Fig. S9B: Comparison of MeHg and VPA

responses with respect to transcription factor (TF)

enrichment The test system UKN1 was exposed to MeHg (1.5 µM) or

VPA (0.6 mM), and the significantly regulated probesets

were determined (as reported in Fig. 3). Statistical

overrepresentation of TF-binding sites (TFBS) in the

promoters of the regulated genes was determined with the

PRIMA algorithm for both treatments. TFBS were

grouped into those only found enriched for MeHg

treatment (red box), those only found for VPA treatment

(blue box) and those found to be enriched by both

compounds (purple box).

*Transcription factor (TF)

ID provided by

PRIMA/Expander

**source was usually

TRANSFAC Public, release 7.0.

When appropriate, updated

names from the HUGO Gene

Nomenclature Committee,

www.genenames.org, are

provided, indicated by (HGNC)

Results Chapter 3 – Transcriptional and metabolic adaptation of human neurons to the

mitochondrial toxicant MPP+

111

E. Results Chapter 3

Transcriptional and metabolic adaptation of human neurons to

the mitochondrial toxicant MPP+

Anne K. Krug1, Liang Zhao2, Cornelius Kullmann1, Dominik Pöltl1, Violeta Ivanova3,

Sunniva Förster1, Smita Jagtap4, Johannes Meiser5, Simon Gutbier1, Gérman Leparc6,

Stefan Schildknecht1, Martina Adam1, Karsten Hiller5, Hesso Farhan7, Thomas Brunner8,

Thomas Hartung2, Agapios Sachinidis4, Marcel Leist1

Affiliations:

1 Doerenkamp-Zbinden Chair for In vitro Toxicology and Biomedicine, University of

Konstanz, D-78457 Konstanz, Germany

2Center for Alternatives to Animal Testing (CAAT-US), Johns Hopkins Bloomberg

School of Public Health, Baltimore MD 21205, USA

3 Nycomed Chair for Bioinformatics and Information Mining, University of Konstanz, D-

78457 Konstanz, Germany

4 Center of Physiology and Pathophysiology, Institute of Neurophysiology, University of

Cologne, D-50931 Cologne, Germany

5 Luxembourg Centre for Systems Biomedicine, University of Luxembourg, Campus

Belval, L-4362 Esch-Belval, Luxembourg

6 Boehringer Ingelheim Pharma GmbH & Co. KG Div. Research Germany

7 Biotechnology Institute Thurgau at the University of Konstanz, CH-8280 Kreuzlingen,

Switzerland

8 Chair of Biochemical Pharmacology, University of Konstanz, D-78457 Konstanz,

Germany

Submitted to Cell Death and Differentiation



112

Abstract

The model toxicant 1-methyl-4-phenylpyridinium (MPP+) has been studied extensively

to understand signaling and cell biological events related to final neuronal cell death

execution in Parkinson’s disease. However, little is known about the upstream network of

responses taking place in toxicant-treated cells before a point-of-no return is reached.

Acquisition of such toxicogenomics and biochemical data has been hindered in the past by the

lack of sufficiently homogeneous tissue or cells. We addressed this question by using

LUHMES cells, that can be differentiated to highly-enriched, fully postmitotic human

dopaminergic neurons. Use of this model system allowed for the first time a combined

metabolomics (mass spectrometry) and transcriptomics (microarrays and deep sequencing)

approach to address chemical-induced neurotoxicity. At 18 - 24 h after treatment with 5 µM

of the mitochondrial respiratory chain inhibitor MPP+, cellular ATP levels and mitochondrial

integrity were still close to control levels, but pronounced changes were already seen on the

transcriptome and metabolome level. Bioinformatic analysis suggested the transcription factor

Atf-4 as most likely upstream factor orchestrating these changes, and early increases of this

regulator were indeed detected by Western blot. Combined analysis of data from both

approaches suggested early activation of the transsulfuration pathway as response to oxidative

stress. Intermediates of this pathway affect DNA and lipid methylation, consistent with our

findings of altered chromatin conformation, increases in methionine sulfoxide/S-

adenosylmethionine and altered phospholipid composition. Our data confirm on the one hand

established literature data by an unbiased approach, and on the other hand they suggest

several novel stress-related cellular adaptations that may contribute to the overall cell fate

outcome after MPP+ exposure. In summary, the findings of this study suggest that combined

‘Omics’ analysis can be used in toxicology as unbiased approach to unravel earliest changes,

the balance of which decides on the final cell fate.



113

Introduction

The use of Omics technologies, combined with systems biology reasoning and

quantitative assessment of the network of toxicity pathways are at the heart of world-wide

efforts to develop a new toxicology for the 21st century (Basketter et al 2012, Collins et al

2008, Krewski et al 2010, Leist et al 2008b, Ramirez et al 2013, Tice et al 2013). Re-

examination of established toxicants is essential to test the feasibility of new approaches and

to gain knowledge about how and when they are best applied (Andersen et al 2011, Thomas et

al 2013). For instance, huge sets of Omics data have been obtained on standard

hepatotoxicants in the Japanese TG-GATES project. Another approach has been taken by the

large ToxCast program of the US environmental protection agency (EPA), which has

extensively explored correlations between classical data obtained for known environmental

toxicants and a panel of several hundred biochemical / mechanistic endpoints assessed for the

same set of compounds (Knudsen et al 2013, Knudsen et al 2011, Sipes et al 2011). The

successful use of Omics and systems biology approaches has already been demonstrated in

biomedical fields, such as tumor biology, by the discovery of new pathways and drug targets

not evident from classical examinations (Carreras Puigvert et al 2013, Kwong et al 2012, Lee

et al 2012). Here, also the use of human cell-based systems has been probed instead of rodent

models. Such investigations have not yet been reported in neurotoxicology, but

transcriptomics and metabolomics profiling are being used more and more in related fields

such as developmental toxicology (Balmer et al 2012, Krug et al 2013, Meganathan et al

2012, Theunissen et al 2012a, Zimmer et al 2011a).

One of the best-characterized neurotoxicants is 1-methyl-4-phenylpyridinium (MPP+).

This compound is the active metabolite of methyl-phenyl-tetrahydropyridine (MPTP), which

triggers specific dopaminergic degeneration and parkinsonism not only in mice, but also in

primates, including humans (Langston et al 1984a, Langston et al 1984b). MPP+ is

accumulated in its target cells by the dopamine transporter. Once inside the cells MPP+ is

believed to inhibit complex I of the respiratory chain, and to cause cell death by energy failure

(Bezard & Przedborski 2011, Nicklas et al 1985). Despite hundreds of studies, many aspects

of MPP+ toxicity remain unclear. E.g., the compound also triggered cell death in mouse

mesencephalic neurons lacking a functional complex I (Choi et al 2008). Also, survival of

human dopaminergic cells after accumulation of MPP+ has been shown to be uncoupled from

ATP depletion (Poltl et al 2012), and some neurons were even protected by MPP+ treatment

from apoptosis triggered by other stimuli (Volbracht et al 1999). An alternative primary



114

mechanism contributing to the well-established tool compound’s toxicity may be the

generation of reactive oxygen species (ROS), possibly through altered electron flow towards

bimolecular oxygen at a subunit of complex I (Freeman & Crapo 1982).

Besides the primary upstream mechanisms, MPP+ toxicity has also been linked to a

plethora of downstream steps, comprising protease activation, protein translocations and

phosphorylation events (Saporito et al 2000, Schulz 2006). In this light it is astonishing, that

there is a dearth of studies examining which upstream metabolic and transcriptional/proteomic

changes precede the final decision on cell death. Such information would be required to

quantitatively model and predict toxic cell death (Geenen et al 2012, Kolodkin et al 2012).

Some Omics information is available on MPP+. For instance, a genomic profiling screen,

using yeast deletion strains identified the multivesicular body pathway (late endosomes) as

important for MPP+ mediated toxicity (Doostzadeh et al 2007). Transcriptomics analysis of

MPP+-treated mouse N2a neuroblastoma cells revealed changes in 439 transcripts, related to

transamination processes, transporter expression and G-protein-coupled receptor signaling

(Mazzio & Soliman 2012). A proteomics study of MPP+-exposed N2a suggested changes in

glutamate oxaloacetate transaminase 2 and other mitochondrial proteins (Burte et al 2011).

Transcriptome-mapping in mouse striatum suggested three waves of gene expression

following MPTP treatment: early upregulation of oxidative stress genes (Gadd45, Ddit4),

intermediate (24 h) regulation of pro-inflammatory genes and late responses (72 h)

characterized by stress response pathways (Nrf-2, Atf6, Zic1) (Pattarini et al 2008).

Proteomics and transcriptomics studies of mice treated with MPTP for 7 days (ongoing tissue

degeneration) showed changes in over 500 proteins, many of them associated with dopamine

signaling, mitochondrial dysfunction, protein degradation, calcium signaling, the oxidative

stress response, and apoptosis (Zhang et al 2010).

To our knowledge, combined metabolomics and transcriptomics studies have not been

performed in the field of neurotoxicology. Even when other organs or organisms are

considered, we are only aware of two publications, one dealing with copper toxicity in

earthworms (Bundy et al 2008), and one addressing upstream stress response pathways

triggered by cyclosporine A in kidney cells (Wilmes et al 2013). Thus, the metabolic changes

and the resultant network of early adaptations triggered by MPP+ within dopaminergic

neurons still remain largely unknown. Such information is hard to obtain by analysis of tissue

consisting of lots of different neuronal and glial cell populations. Moreover, the stage of cells,

relative to a complex degeneration process can only be controlled in a very homogeneous cell



115

culture system. To address this issue, and to provide directly information on human cell

behavior, we made use of LUHMES cultures that consist of > 95% fully post-mitotic

dopaminergic neurons (Scholz et al 2011). We generated transcriptomics and metabolomics

data at early time points with the goal to run a combined pathway analysis. Key findings on a

common stress response upstream regulator and unexpected transcriptome changes were

confirmed by classical analytical approaches.



116

Results

Metabolome changes in MPP+-exposed dopaminergic neurons

After six days of differentiation, LUHMES cells are post-mitotic, express an intricate

neurite network and assume a dopaminergic phenotype as characterized by high expression

levels of the marker genes Fox3 (NeuN), tyrosine hydroxylase (TH), SLC18A2 (vesicular

monoamine transporter 2), and SLC6A3 (dopamine transporter, DAT) (Scholz et al 2011). At

this stage, the cells are sensitive to a toxicologically-relevant concentration of MPP+ of 5 µM

(Poltl et al 2012, Schildknecht et al 2009), which was chosen for all experiments of this study.

Cells were generally analyzed on day 8 of differentiation (d8), following exposure to MPP+

for varying times (Fig. 1A). Cell death was assessed by measurements of LDH release,

counting of viable cells (calcein-positive) and quantification of resazurin reduction.

Significant cytotoxicity required at least 48 h of MPP+ exposure, and most cells were dead

after 72 h (Fig. 1 B, C). Cellular ATP and glutathione (GSH) levels were maintained for at

least 24 h, and showed a significant decrease after 36 h of treatment (Fig. 1C). The same time

course was observed for the mitochondrial membrane potential/energized mitochondrial mass

(Fig. 1 E, F). Cell death-associated events, such as the release of cytochrome c into the

cytosol or regulation of Bcl-2 family proteins were not measurable at 24 h (Suppl. Fig. S1 A,

B).

These basic model parameters indicated that toxicant stress was compensated to a large

extent for up to 24-36 h after MPP+ exposure, and after that time key functions could not be

maintained. To broadly characterize the metabolic adaptations prior to cell death events, we

performed an untargeted metabolomics analysis: 190 unique metabolites were significantly

altered, and 59 of them were assigned to molecula structures (Suppl. Table S1). A principal

component analysis (PCA) of the total quantified metabolite patterns indicated large, and

highly reproducible differences between control cells and 24 h samples, and a further distinct

shift was observed for 36 h treatments (Fig. 1D, Suppl Fig. S2). Some of the data

corroborated known effects of MPP+ exposure. For instance, the altered energy metabolism

was indicated by a strong decrease in intracellular glucose (and other sugars) accompanied by

an increase of pyruvate and lactate (Fig. 2). Consumption of phosphocreatine and a parallel

accumulation of creatine suggested an exhaustion of the cellular energy buffer (Fig. 2, Suppl.

Fig. S3). A cellular struggle to maintain energy supplies was also indicated by a gradual

increase of ADP, AMP and adenine, although levels of ATP were more or less maintained for

at least 24 h (Fig. 2). Increases in methionine-sulfoxide (Fig. 2, Suppl. Fig. S3) as well as a



117

decrease in dehydroascorbate confirmed an increased oxidative stress in the system, as had

been suggested in earlier studies based on ROS measurements (Schildknecht et al 2009). The

Figure 1. Time-course of MPP+ induced cell death events and metabolome changes. A)

Experimental scheme for cell differentiation, MPP+ exposure and sampling. In all experiments

of this study, an MPP+ concentration of 5 µM was used, and cells were analyzed on day 8 (d8)

of differentiation (green arrow). Red arrows mark time points used for Omics analysis. Blue

arrows mark time points which were analyzed in follow-up experiments. B+C) Cell viability

data: resazurin reduction and lactate dehydrogenase (LDH) release were measured and calcein-

positive/negative cells were counted. Changes of ATP and total cellular glutathione (GSH) were

measured in parallel cultures and all data were normalized to untreated controls. D) Samples

obtained after 24 or 36 h of treatment with MPP+ or solvent control were analyzed by

quadrupole time-of-flight liquid chromatography-mass spectroscopy (Q-TOF LC-MS). A

principal component analysis (PCA) of all metabolite data (labeled by length of exposure) was

performed and the first two dimensions are displayed. E) Cells were stained with

tetramethylrhodamine ethyl esther (TMRE, green) and calcein-AM (red) to identify energized

mitochondria. Representative micrographs display cells treated with solvent (control) or

MPP+(24 h, 48 h). F) The number of TMRE positive pixels in all neurites of the field was

determined by an unbiased image processing algorithm. Data are means ± SD from 3

independent experiments, and 30 fields per experiment (*: p ≤ 0.05).

Pri

ncip

al com

pon

en

t 2

(14

.1%

)



118

broad metabolite profiling also allowed new insights. For instance, the strong increase of

S-adenosyl-methionine (SAM, Suppl. Fig. S3), S-adenosyl-homocysteine (SAH) and

cystathionine pointed to alterations of the methionine/cysteine metabolism (Fig. 2), possibly

to replenish the redox buffer glutathione. In a targeted analysis including an earlier time point,

we looked therefore specifically for cellular cysteine levels. After 18 h, the levels of this

0 10 20 30

0

50

100

150

200

AdenineADPAMPATP

Norm

aliz

ed In

tensity

Valu

es

[% o

f contr

ol S

D]

0 10 20 30

0

50

100

L-Ala

L-AsnL-Glu

L-Asp

Time [h]

Norm

aliz

ed In

tensity

Valu

es

[% o

f contr

ol

SD

]

0 10 20 30

0

100

200

750

1000L-Cystathionine

SAH

SAM

0 10 20 30

0

100

200

300

L-ArgL-LysL-TrpL-Tyr

Time [h]

0 10 20 30

0

50

100

UDP-GalD-Gluc

D-ErythroseUDP-GlucN

orm

aliz

ed Inte

nsity

Valu

es

[% o

f contr

ol S

D]

0 10 20 30

0

100

200

300

Met-SO

Dehydroascorbate

0 10 20 30

0

200

400

600 L-LactatePyruvate

0 10 20 30

0

50

100

150

200

250

P-CreatineCreatine

2-Oxoisovalerate

0 10 20 30

0

50

100

150

1000

2000

Sarcosine

L-Gly

Time [h]

Figure 2. Metabolic adaptations in MPP+-treated neurons. LUHMES cells were treated with

5 µM MPP+ for different times, and samples were taken at day 8. Metabolite concentrations

were determined by Q-TOF LC-MS in 4 independent experiments. Data were normalized to

untreated controls and are displayed as means ± SD. Metabolites that changed significantly (p ≤

0.05,FDR adjusted) are displayed. D-Gluc = D-glucose, UDP-gal = uridinediphosphate

galactose, UDP-gluc = uridinediphosphate glucose, P-creatine = phosphocreatine, Met-SO =

methionine sulfoxide, SAM = S-adenosylmethionine, SAH = S-adenosylhomocysteine.



119

amino acid were significantly increased, while its oxidized form, cystine, decreased (Fig. 3

A). Thus, potential oxidative stress was strongly compensated at that time, and also at 24 h,

cysteine levels were still 50% higher than in control cells, while cystine was unchanged. This

response was very robust, as it was not only observed in technical replicates, but in 3

independent cell preparations used for these experiments. Also other profound changes in

amino acid homeostasis became evident (Fig. 2). The untargeted metabolomics analysis

showed lower levels of alanine, glutamate, aspartate and asparagine, and degradation of

0

20

40

60

80

100

120

Control MPP+

GBR for 18 h GBR for 18 h

+-

Flu

x in

to T

CA

cycle

[pyru

vate

dehydro

genase a

ctiv

ity]

0 18 24 0 18 240

50

100

150

Cysteine Cystine

Time [h]

Mean c

oncentr

ation

[% o

f contr

ol S

EM

]

0 18 24 0 18 24 0 18 240

50

100

Putrescine Spermidine Spermine

Time [h]

Mean c

oncentr

ation

[% o

f contr

ol S

EM

]

C

A B

Ornithine Arg ↑

Putrescine ↓

Spermine ↓

Spermidine ↓

SAM

SAM

CO2

32:1 34:1 34:2 36:2 40:4 40:5 32:1 34:2 36:1 16:1 17:0 18:10

50

100

150

200Lipid changes after 18 h

Mean c

oncentr

ation

[% o

f contr

ol S

EM

]

Phosphatidylcholines Plasmalogens Lyso-

phosphatidylcholines

**

* *

*

**

D

*

Figure 3. Multiple secondary metabolic changes triggered early after exposure to

MPP+. Cells were exposed to MPP+ (5µM) for different times. A+B) Using a targeted analysis

approach, the absolute levels of A) cysteine (2.98 pmol/106control cells) and cystine (3.6

pmol/106 control cells) as well as of the B) polyamines putrescine (1.06 nmol/106control cells),

spermidine (0.17 nmol/106control cells) and spermine (0.28 nmol/106control cells) were

measured in 3 independent experiments. Data are displayed after normalization to controls. For

background information, a scheme of ornithine-polyamine metabolism is displayed (SAM = S-

adenosylmethionine, Arg = arginine, arrow = direction of regulation by MPP+). C) Using 13C

labeled glucose as substrate, the flux from glycolysis into the TCA cycle was determined. In all

pyruvate dehydrogenase activity was determined by mass-spectrometric measurement of the

ratio of isotope-labeled citrate with two labeled carbon atoms (derived from acetyl-CoA

originating from labeled glucose) and citrate with only 12C. GBR = dopamine transport inhibitor

GBR-12935, 1 µM. D) The absolute cellular concentrations of phosphatidylcholines,

plasmalogens and lyso-phosphatidylcholines were determined in the same experiments as in

A+B. Data were normalized to those of control cells. Numbers below the bars indicate number

of total acyl/carbon atoms and double bonds.



120

branched amino acids was indicated by the strong increase in 2-oxo-isovalerate. Cellular

concentrations of the poorly gluconeogenic aminoacids arginine, lysine, tryptophan, and

tyrosine increased upon MPP+ treatment, and the increase in glycine was matched by a similar

decrease of sarcosine (Fig. 2). More such shifts in metabolism were observed: for instance,

the conversion of ornithine to putrescine appeared decreased, and the decrease of putrescine at

18 and 24 h was reflected, with a delay of 6 h, by a decrease of spermidine and spermine, two

biogenic polyamines derived from it (Fig. 3B). The major shifts were expected for central

energy metabolism. However, alterations in the citric acid cycle (TCA cycle), an assumed

primary mode of action of MPP+, could not be deduced from our metabolomics data. To

control, whether such changes indeed occurred, we used a targeted approach to measure the

effect of MPP+ on the channelling of glycolytic metabolites to the TCA cycle. For this, we

used 13C-labelled glucose, and determined its flux through the pyruvate dehydrogenase step

towards citrate. MPP+ did indeed nearly abolish this reaction, whereas inhibition of MPP+

uptake by a dopamine transporter inhibitor (GBR-12935) had no effect on glucose flux

(Fig. 3C).

Also, many changes outside our interest in the core energy and amino acid metabolism

were observed, the most conspicuous of which were the choline phospholipids. A large

number of phosphatidylcholines, plasmalogens and lysophosphatidylcholines was increased,

but it is at present unclear how such possibly secondary changes relate to toxicity pathways or

cellular stress adaptation (Fig. 3D). To better understand the relevance of metabolite changes,

and to facilitate the use metabolomics information for conclusions on altered pathways, we

complemented these earlier analyses with an orthogonal approach, i.e. transcriptome analysis

under the same experimental conditions.

Transcriptome changes preceding cell death in MPP+-exposed neurons

Three time points – 24, 36 and 48 h – were chosen for affymetrix DNA microarray

analysis to investigate potential transcriptional changes triggered by MPP+ (Table S2).

Altogether 411 probe sets (PS) were changed (FDR corrected p-value ≤ 0.05 and a fold

change ≥ 2). Heatmap presentation, clustering analysis and PCA (Fig. 4A, Suppl. Fig S4)

suggested a high reproducibility of the response across different cell preparations. When the

PS were sorted with respect to the time course of gene regulation, four major clusters were

apparent (Fig. 4 B). Clusters #1 and #2 contained the PS monotonously down- or up-regulated

over time. Cluster #3 contained PS (n = 26) up-regulated only after long exposure, and cluster

#4 contained PS first down-regulated and then compensated back to base level at later time



121

points (n = 34). In a different grouping approach, we identified the PS that were already

significantly down- (cluster #5; n=64) or up-regulated (cluster #6; n=116) at the earliest time

points.

We used bioinformatic approaches to identify biological processes that may be affected

by altered transcripts. Significantly overrepresented gene ontologies (GO) were identified for

the transcript clusters to gain information on adaptive responses and stress pathways

potentially triggered in the cells (Fig. 4C, Suppl. Table S4). The PS of cluster #1 strongly

pointed to changes of chromatin organisation and related processes (mitosis, DNA

conformation/packaging). Amongst the PS of the related/overlapping cluster #5 (early down-

regulation) only one GO term, paraspeckles, was overrepresented. We verified this exemplary

finding on the protein level by immunostaining, and our data corroborated the down-

regulation of the paraspeckle-associated paraspeckles component 1 (PSPC1) protein (Fig. 4D,

Suppl Fig. 1C). Overrepresented GOs within up-regulated genes (clusters #2, #6) pointed to

changes in metabolic processes related to amino acid and carboxylic acid turnover, but

surprisingly not to e.g. glycolysis or the pentose phosphate cycle (Fig. 4C, Suppl. Table S4).

As second approach to explore changes in gene expression, we employed illumina RNA

deep sequencing (Suppl. Table S3). This method identified 376 transcripts to be altered

already after 24 h (FDR corrected p-value ≤ 0.05 and a fold change ≥ 2), and the number

further increased over time (Fig 5A). The genes that had been identified by microarray

analysis were confirmed by deep sequencing, and the quantitative results were correlated to a

high degree (Fig. 5B). The additional transcripts identified by RNAseq, but not microarray

yielded information on the expression of genes encoded by mitochondrial DNA: MPP+

exposure significantly reduced the levels of transcripts of complex I and III subunits of the

respiratory chain, while it did not have an effect on nuclear-encoded subunits of these

complexes (Fig. 5C). To capture all additional information contributed by RNAseq we

identified all overrepresented GOs amongst this data set, and compared them with those of

microarray analysis. The additional ones found amongst the sequencing data pointed to

‘alterations of ion transport’ (amongst up-regulated genes), and to ‘disturbances in electron

transport coupled to ATP synthesis’ and in ‘spindle/microtubule cytoskeleton organization’

(amongst down-regulated transcripts) (Fig. 4E, Suppl. Table S5).



122



123

Confirmation of transcriptome data by detailed PCR analysis

To obtain more information on the time course of transcriptome changes, expression of

several genes identified by the two Omics approaches was followed by RT-qPCR analysis at

early time points after exposure to MPP+ (Fig. 5D). Many of the transcripts were changed as

early as 2-12 h after treatment. The expression of TXNIP (thioredoxin interacting protein 1), a

gene playing a role in oxidative stress, was reduced already at 2 h after treatment. Genes

playing roles in chromosomal stability, like HNRNPM (heterogeneous nuclear

ribonucleoprotein), were also down-regulated after as little as 2 h. Genes involved in adaptive

central metabolism, like ASS1 (argininosuccinate synthase 1) or SHMT2 (serine-hydroxy-

methyl-transferase 2) were significantly up-regulated after 12 h. In particular, genes involved

in cysteine synthesis via the transsulfuration pathway, CTH (cystathionase) and CBS

(cystathionine-β-synthase), were up-regulated at 24 h (Fig. 5D). We also explored

paraspeckles-related genes further, and several of these, PSPC1, SFPQ (splicing factor

proline/glutamine-rich) and HNRNPM were coordinately down-regulated. The corresponding

proteins all contribute to paraspeckle structures that are presumed to play a role in mRNA

retention in the nucleus (MacDougall et al 2013, Venkatakrishnan et al 2013). As these rapid

and distinct regulations had not been expected by us, we used a different, but mechanistically

related damage model: the cells were exposed to the canonical complex I inhibitor rotenone

(100 nM), and very similar transcriptional changes were observed (suppl. Fig S5). Thus, the

transcriptome response we observed for MPP+ may reflect a coordinated response to

mitochondrial inhibition in human neurons.

◄Figure 4. MPP+-induced transcriptome changes and their functional annotation. A) Cells

were treated with MPP+ (5 µM) for different times before samples were taken for DNA

microarray-based transcriptome analysis. Probe sets significantly altered at at least one time point

are displayed (FDR adjusted p-value of ≤ 0.05; fold change values ≥ 2). Colours represent Z-

scores of the row-wise normalized expression values for each probe set. The Spearman

correlations of the samples are indicated above the heatmap. Gene clusters (#1-4) consist of probe

sets with similar expression profiles. B) Graphs display fold changes of the top 80 regulated genes

for cluster #1 and #2 and of all genes of cluster #3 and #4. The black solid line represents the

mean tendency of all genes of the cluster. C) Overrepresented gene ontology (GO) terms are

displayed as wordclouds for every cluster separately. For cluster #1 (down-regulated) and #2 (up-

regulated), only the top 30 GOs with a p-value ≤ 0.001 are displayed (remaining GOs can be

found in Suppl Table S4). For cluster #5 (all genes down-regulated significantly after the 24 h

time point) and #6 (up-regulated at 24 h) all overrepresented GOs are displayed. D)

Representative images of cells with labeled paraspeckles component 1 protein (PSPC1, red) are

displayed. Cells were treated with 5 µM MPP+ for 24 h or 48 h and fixed for immunostaining.

Compared to the nuclear counter-stain (green), PSPC1 strongly decreased over time. E) Cells

were treated as in A, and samples were analyzed by RNA sequencing (RNAseq). Overrepresented

GOs were identified, and the ones that were not contained in the microarray data are displayed. A

complete list is supplied in Suppl. Tab. S5. Calibration of wordcloud displays (indicated in dark

blue): the height of the letters reflects the p-value of the GO.



124

-4 -2 0 2 4

-4

-2

0

2

4

fold change [log2] (microarray )fo

ld c

ha

ng

e [

log

2]

(RN

Ase

q)

BA Spearman correlation = 0.81

p < 0.05

0 20 40

-6

-4

-2

0

2

MT-ND6MT-CO2MT-CYBMT-ND2MT-ND5

MT-ND3MT-ND4MT-ND1MT-ATP6MT-CO3

Time [h]

mR

NA

fo

ld c

ha

ng

e [

log

2]

C

CBS

TXNIP1

TYMS

SHMT2

DDIT4

HNRNPM

MLF1IP

CCNB

PPA2

PSPC1

ASNS

ASS1

CTH

DDIT3

SFPQ

GADD34

NQO1

NOXA

0 12 24 36 48

5

10

15

0.4

0.3

0.2

0.1

0.0

2

0.5

Time [h]

mR

NA

leve

l [f

old

change r

ela

tive

to c

ontr

ol

SE

M]

Centromere protein U (MLF1 interacting protein)

D

NAD(P)H dehydrogenase, quinone 1

Paraspeckles component 1

Splicing factor proline/glutamine-rich

Cyclin B1

DNA damage inducible transcript 3 (CHOP, GADD153)

Serine hydroxymethyl-transferase

Argininosuccinate synthase 1

Asparagine synthase

DNA damage inducible transcript 4

Phorbol-12-myristate-13-acetate-induced protein 1 (PMAIP1)

Cystathionase (cystathionine gamma-lyase)

Cystathionine-β-synthaseGrowth arrest and DNA damage-inducible protein (PPP1R15A)

Thioredoxin-interacting protein

Pyrophosphatase (inorganic) 2

Heterogeneous nuclear ribonucleoprotein M

biosynthetic processes oxidative stress

chromosomal changes/paraspeckles

ER stressmitochondrial function

Colour coding of biological process:

qPCR

Thymidylate synthetase

RNAseq:number of regulated

genes

Exposure time

24 h 36 h 48 h

271 298 362

105 138 185

↑

↓



125

Identification of ATF-4 as superordinate regulator of many transcriptome

changes

Some superordinate biological regulators should be identifiable, if the transcriptome

response reflects indeed a cellular adaptation program in the early phase of cell death. We

used therefore a bioinformatic data mining approach to identify putative upstream regulators.

Microarray data were linked to known regulatory pathways and transcription factor-gene

networks. This analysis yielded the highest probability value for the ER stress-related

transcription factor ATF4 (activating transcription factor 4) as upstream regulator (Fig. 6A).

As the corresponding ATF4 gene was not amongst the hits of the Omics analyses, we used

PCR to examine a potential regulation of its RNA. A 2.5 fold increase by RT-qPCR was

indeed detected (Fig. 6B). Since the activity of ATF4 is mainly regulated on the translational

level, we also examined protein levels. A strong and early increase was identifiable by

western blot, and high levels were maintained for at least 2 days (Fig. 6C). To further verify

the initiation of ATF-4 signalling, we examined some upstream and downstream components

of this pathway. The eukaryotic initiation factor 2α (eIF2α) inhibits the translation of ATF4

mRNA, and this block is released by phosphorylation of eIF2α. We observed here such

phosphorylation at 6-12 h after exposure to MPP+, consistent with the strong rise of ATF4

protein levels during this time (Fig. 6C). As downstream target of ATF4, we examined

GADD34, and this was indeed continuously up-regulated over time (Fig. 6C). Thus, the

biochemical observation of ATF-4 pathway activation on several levels corroborated the

suggestion from our initial bioinformatic analysis and confirmed the validity of this approach

concerning regulatory signalling. To gain further insight also on metabolic pathways,

transcriptomics data were mapped in the next step together with metabolite data onto known

human metabolism.

◄Figure 5. Time course of transcriptome changes identified by RNA sequencing and RT-

qPCR. A) Cells were treated with MPP+ (5 µM) for different times before samples were taken for

RNA sequencing (RNAseq) analysis. Differentially expressed transcripts were identified (FDR

adjusted p-value of ≤ 0.05; fold change values ≥ 2). The numbers of up-regulated genes are

highlighted in red, down-regulated genes in blue. B) Scatter plot of fold changes as determined by

microarray or RNAseq. Each data point corresponds to one MPP+ regulated transcript. C)

Expression values for transcripts coded by mitochondrial genes were selected from RNAseq data

set. Regulated complex I subunits are highlighted in orange. The scheme of complex I illustrates

the location of these subunits (orange) in the protein complex. D) Cells were treated as in A) and

mRNA was prepared after 2-48 h. The samples were analyzed by RT-qPCR for selected marker

genes. Data are means ± SEM of three independent differentiations. The area shaded in grey

marks expression changes of < 2 fold. Colour coding indicates biological processes the genes are

involved with.



126

Identification of transsulfuration changes by combined Omics pathway

analysis

After having analyzed metabolomics data and transcriptomics data separately, they were

used on a multi-omics platform for combined pathway enrichment analysis, and several new

significant regulations emerged (Fig. 7E, Suppl. Fig. S6). Serine emerged as an important

Figure 6. Bioinformatic identification of ATF4 as upstream regulator of transcriptional

up-regulation. A) Bioinformatic analysis with IPA software identified ATF4 as regulator of

genes, which were up-regulated (cluster #2). The genes in cluster #2 that are known to be ATF4

targets are indicated, together with their extent of regulation (relative fold change of 24 h vs. 0

h) according to microarray analysis. Pathways, in which the ATF4 target genes are involved

with, are indicated in dark blue (pw = pathway, AA = amino acid). B) Cells were treated with

MPP+ (5 µM) and ATF4 mRNA levels were analyzed by RT-qPCR (relative to GAPDH

expression) after different times. C) Cell lysates were prepared after different times following

MPP+ treatment. They were analyzed by western blot for key elements of the ATF4 pathway.

Data are representative for 3-4 experiments. eIF2a[pS52]: eukaryotic initiation factor 2 alpha

phosphorylated at serine 52.

3.05.0 3.0

4.32.4

4.6 3.8

3.4

2.8

2.5

2.9

2.4

2.9

2.2

2.1

3.3

2.3

2.8

3.24.3

0 12 24 36 48

0.0

0.2

0.4

0.6

0.8

Time [h]

AT

F4 m

RN

A e

xpre

ssio

n

[rela

tive to G

AP

DH

SE

M]

0 6 12 24 36 48

ATF 4

GAPDH

GADD 34

eIF2a[pS52]

Leads to activationPutative target

Prediction Legend

Enzyme

Growth Factor

Kinase

Other

Transcription Regulator

Translation Regulator

Transmembrane Receptor

Transporter

Phosphatase

5 µM MPP+

50

kDa

37

kDa

100

kDa

Time [h]:

GAPDH

GAPDH

C

B

A

**

**

Serine pw

Ap

op

totic

pw

Catio

nic

AA

carr

ier

tRNA processing



127

knot, although it was only regulated to a minor extent itself. Its downstream metabolites

glycine and cystathionine were up-regulated, as were the transcripts for the corresponding

enzymes (SHMT2 and CBS). Moreover, three sequential enzymes linking glycolytic

intermediates and serine were also induced. This appears to be rather a coordinate response

than a random finding, as all these enzymes are known downstream targets of ATF-4.

Changes of this metabolic pathway were linked e.g. to altered C1 metabolism, and by this

way to DNA turnover. However, the most significant link was to the ‘transsulfuration

pathway’ and some of its upstream elements. The up-regulation of this metabolic route

indicated that rate-limiting glutathione precursors (glycine and cysteine) were generated by

MPP+-exposed cells at accelerated rates. Our data suggest that the increased levels of the

methionine-derived thiols SAM, SAH and homocysteine, as well as the serine biosynthesis

pathway acted as precursors for the transsulfuration process. The biological significance of

this alteration is underlined by the additional finding that also the transporter responsible for

Figure 7. Combined metabolomics-transcriptomics identification of pathways affected by

MPP+. Transcriptomics and metabolomics data of the 24 h treatment sample with 5 µM MPP+

were used for bioinformatic analysis. The net of pathways most significantly overrepresented is

displayed. Enzymes (corresponding mRNA) and metabolites that were up-regulated are

displayed in red. Blue indicates decreased levels. ATF4 targets are encircled in orange

(SLC7A11 has been identified by RNAseq only, the other target genes were identified on both

transcriptomics platforms). Underlying biological processes affected by the indicated changes

are displayed in green. CBS = cystathionine-β-synthase, CTH = cystathionase, DHFR =

dihydrofolatereductase, dTMP = deoxythymidine monophosphate, dUMP = deoxyuridine

monophosphate, GSH = glutathione, MTHFD2 = methylenetetrahydrofolate dehydrogenase,

PHGDH = phosphoglycerate dehydrogenase, PSAT1 = phosphoserine aminotransferase 1,

PSPH = phosphoserine phosphatase, ROS = reactive oxygen species, SHMT2 = serine

hydroxymethyl-transferase, TYMS = thymidylate synthetase.



128

replenishing cellular cysteine from extracellular sources was up-regulated. Finally, we

addressed the question of higher cellular levels of total glutathione on the basis of the

activation of pathways involved in the provision of this redox buffer. A titration of MPP+

concentrations indicated that cellular GSH levels (at 24 h after drug exposure) were indeed

significantly increased (by 10%) at 1 µM; they were hardly affected at 5 µM, and clearly

decreased (by 10%) at 20 µM. Theses studies showed ‘in principle’ the augmentation of GSH

levels, but the effect was partially masked by simultaneous cell stress and an increased

demand. To allow measurements under conditions of less stress, we used a slightly modified

model: neurons that had been differentiated for 5 days only are much less sensitive to MPP+

toxicity (Schildknecht, 2009; Pöltl, 2012). Under these conditions, 1 µM MPP+ raised cellular

GSH levels by about 50% at 24 h. Higher MPP+ concentrations resulted in a smaller increase,

or no changes for concentrations > 5 µM. ATF-4 levels also increased concentration- and

time-dependently over time in the immature cells (Suppl. Fig. S7 E) In summary, these

experiments fully corroborated the early upregulation of GSH supply by MPP+ treatment, and

they provide evidence that the altered transsulfuration pathway, as identified here, indeed

changes cellular properties.



129

Discussion

Although MPP+ has been studied in thousands of papers, there is still little information

on the adaptive cellular changes taking place prior to the point-of-no-return in MPP+ triggered

neuronal death. It is well documented that MPP+ rapidly inhibits complex I of the respiratory

chain in mitochondria and that ATP depletion plays a role in MPP+‘s cytotoxicity. Analyzing

for the first time the metabolic changes of MPP+ treated cells, we identified several metabolic

adaptations, for instance TCA cycle shut-down (Fig. 3D), amino acid metabolism (Fig. 2) as

well as phospho-creatine consumption (Fig. 2), all pointing to alternative energy supply in the

cells (Fig. 8). Also enhanced glycolysis following MPP+ exposure was inferred from

decreased levels of glucose, increased lactate concentrations and our knowledge of the

published literature (Mazzio & Soliman 2003a, Mazzio & Soliman 2003b), but this pathway

did not emerge form bioinformatic analysis of transcriptomics data.

The transcriptional changes determined in our study are based on gene microarray

experiments for different exposure times, and we also confirmed the data by RNAseq and

qPCR. The chromatin changes triggered by MPP+, e.g. change of paraspeckles were entirely

unexpected (Fig. 8) and the role of paraspeckles in the system remains elusive. During the last

decade accumulating data points to a role in RNA retention into the nucleus (Nakagawa &

Hirose 2012). The decrease of paraspeckle factors in our case may therefore lead to higher

expression rates of paraspeckles-targeted mRNAs. Also upcoming gene onotolgies indicating

altered microtubule cytoskeleton were unexpected, but confirmed earlier findings of reduced

mitochondrial mobility in the LUHMES system upon MPP+ exposure (Schildknecht et al

2013). A similar effect was observed in PC12 cells treated with higher concentrations of

MPP+ (Cartelli et al 2010).

Furthermore, the transcription factor ATF4 was identified as upstream regulator.

Notably, ATF-4 itself was not a primary hit, but it was identified by bioinformatic analysis on

the basis of its regulated downstream targets. The validity of these conclusions was then

corroborated by a targeted measurement of ATF-4 mRNA and protein levels and of the

expression levels of other pathway constituents. The identification of ATF4 points towards an

increase in ER stress in the cells, or a high demand in amino acids, respectively (Fig. 8). The

upstream activity of ATF-4 may indeed explain the different outcomes seen in different

models. For example a role of ATF-4 in dopaminergic neuron death has already been

suggested 10 years ago, but the findings did not apply to primary neurons exposed to MPP+

(Holtz & O'Malley 2003). Later, the protein was rather associated with pharmacological

protection (Lotharius et al 2005); then again with enhanced damage (Lange et al 2008), and

most recently with neuroprotection (Sun et al 2013). Several amino acid carriers appeared up-

regulated in the transcriptomics data, whereby SLC7A5 and SLC3A2 (build-up the

heterodimeric glycoprotein-associated transporter CD98), SLC1A1 and SLC7A11 are



130

putative ATF4 target genes. After MPP+ treatment for 18 h cystine is down-regulated, but is

compensated back to baseline at 24 h, which may be due to higher uptake rates through the

increased expression of SLC7A11, a cystine-glutamate transporter. This carrier has already

been shown to be up-regulated in methamphetamine treated LUHMES cells after ATF4 was

induced, resulting in higher cystine-uptake rates (Lotharius et al 2005).

Figure 8. Overview of MPP+-induced adaptations in human dopaminergic neurons. The

findings of this study (*) have been incorporated into a network of adaptive regulations. The

molecular initiating event of MPP+toxicity is inhibition of mitochondrial respiratory chain

complex I. Thus, NADH oxidation is hindered, and this leads to an arrest of the TCA cycle. The

subsequent slowdown of the pyruvate dehydrogenase* leads to an accumulation of pyruvate*

and lactate*. Cellular adaptations lead to increased glucose consumption* and usage of

alternative ATP synthesis sources*. Complex I inhibition also leads to a higher O2-• production,

as indicated by decreased levels of the antioxidant dehydroascorbate* and an increase in

methionine sulfoxide*. The two primary events cause a general increase in GSH demand with

adaptations on metabolite and transcriptome level*. Demand for alternative energy sources and

GSH results in changed amino acid metabolism* and higher expression of their transporters*.

Several of these changes contribute to a rise in ER stress, as indicated by phosphorylation of

eIF2a* and increases of ATF4*. Many of the observed changes in gene regulation may be

attributed to ATF-4 and related transcriptions factors. Evidence for DNA damage response* and

altered lipid metabolism* suggests that many further cellular changes take place long before

energy is depleted. At the point of no return the counter-regulation capacity of the cell is

exhausted, ATP and GSH drop steeply*, programmed cell death pathways (NOXA ↑, PUMA ↑,

Bcl-xL ↓) are activated* and loss of mitochondrial integrity ensues*. Pyr = pyruvate, Lac =

lactate, Gluc = Glucose, TCA = tricarboxylic acid cycle, Dehydroasc = dehydroascorbate,

Met(SO) = methionine sulfoxide, GSH = reduced glutathionine, GSSG = glutathionine

disulfide.



131

Up to now, little systems biology data is available on neurodegeneration and

neurotoxicity. The knowledge concentrates on protein changes, mostly posttranslational

modifications by kinases and proteases. Little is known on combined transcriptional and

metabolic responses, in particular at early time points. We approached a combination of

metabolomics and transcriptomics data and identified several involved pathways in early

cellular adaptation, e.g. the serine metabolism, one carbon metabolism or the transsulfuration

pathway. One-carbon metabolism, involving the folate and methionine cycles, seems to be

shifted towards the methionine cycle as indicated by a rise in S-adenosyl-methionine (SAM)

and S-adenosyl-homocysteine (SAH) and the down-regulation of folate-dependent enzymes,

such as TYMS (thymidylate synthase) or DHFR (dihydrofolate reductase). The identified

pathways are interesting hits regarding the compressed energy supply in the system, as a

recently established in situ model of neuron metabolism predicted neurons to support their

energy demands from glycolysis and reactions from serine synthesis, one carbon metabolism

and the glycine cleavage system upon increases in protein aggregates (Vazquez 2013). Next

to alternative energy supplies, the regulated pathways also support the GSH synthesis.

Increase in SLC7A11, in glycine levels via serine conversion and in cysteine levels through

the transsulfuration pathway activation, indicate the high GSH demand in the cells. The cells

are opposed to high levels of oxidative stress, as indicated by e.g. the increase in methionine

sulfoxide, which might lead to the observed pathway adaptations. GSH regulation was further

verified in less susceptive cells for MPP+ toxicity, and a high initial increase of GSH together

with ATF4 and enzymes of the transsulfuration pathways corroborate our metabolic findings.

Our work introduces the importance of endogenous metabolism for a systems biology

understanding of neurotoxicity. We show that changes of metabolism are a pivotal layer of

cell responses to toxic disturbances. Especially the early, non-symptomatic phase that was the

focus here requires knowledge of the regulation layer of endogenous metabolism. It is of

broad significance that there is a phase in cell stress with highly dynamic adaptations. Most

strategies aimed at preventing neurodegeneration or other diseases target downstream

mechanisms associated with the final adverse effects. A new approach suggested by our work

would be a concentration on the question why cells can survive and cope with the stress, and

to use strengthened adaptive responses to counteract disease.



132

Material and Methods

Dibutyryl-cAMP (cAMP), fibronectin, hoechst bisbenzimide H-33342, resazurin sodium

salt, tetracycline, tetramethylrhodamineethylester (TMRE) and MPP+ were from Sigma

(Steinheim, Germany). Recombinant human FGF-2 and recombinant human GDNF were

from R&D Systems (Minneapolis). Tween-20 and sodium dodecyl sulfate (SDS) were from

Roth (Karlsruhe, Germany). All culture reagents were from Gibco unless otherwise specified.

Cell culture:

Handling of LUHMES human neuronal precursor cells was performed as previously

described in detail (Lotharius et al 2005, Schildknecht et al 2009, Scholz et al 2011). Briefly

LUHMES cells were maintained in proliferation medium, consisting of advanced DMEM/F12

containing 2 mM L-glutamine, 1 x N2 supplement (Invitrogen), and 40 ng/ml FGF-2 in a 5%

CO2/95% air atmosphere at 37° C and were passaged every other day. For differentiation 8

million cells were seeded in a Nunclon T175 in proliferation medium for 24 h. In a following

step medium was changed to differentiation medium (DM II), consisting of advanced

DMEM/F12 supplemented with 2 mM L-glutamine, 1 x N2, 2.25 µM tetracycline, 1 mM

dibutyryl 3’,5’-cyclic adenosine monophosphate (cAMP) and 2 ng/ml recombinant human

glial cell derived neurotrophic factor (GDNF). At 48 h later cells were trypsinised, and seeded

in a density of 184*103 cells/cm² on dishes precoated with 50 µg/ml poly-L-ornithine (PLO)

and 1 µg/ml fibronectin in advanced DMEM/F12 containing 2mM L-glutamine, 1 x N2 and

2.25 µM tetracycline but without cAMP and GDNF (DM). On day 4 of differentiation,

medium was exchanged with tetracycline-free DM.

Standard experimental setup:

Cells were seeded at a density of 350,000 cells per well in 500 µl DM on

PLO/fibronectin coated 24-well dishes. At day 6 of differentiation the time series of MPP

exposure started in a reverse fashion by adding 5 µM MPP to the media at different initiation

times (e.g. treatment for 48 h started on day 6, treatment for 24 h started on day 7). Analyses

were performed on day 8 of differentiation.

Cell viability measurement:

Resazurin: Metabolic activity was detected by a resazurin assay (38). Briefly, resazurin

solution were added to the cell culture medium to obtain a final concentration of 10 µg/ml.

After incubation for 30 min at 37° C, the fluorescence signal was measured at an excitation



133

wavelength of 530 nm, using a 590 nm long-pass filter to record the emission. Fluorescence

values were normalized by setting fluorescence values of untreated wells as 100% and the

values from wells containing less than 5% calcein-positive cells as 0%.

LDH release: LDH activity was detected separately in the supernatant and cell

homogenate. Cells were lysed in PBS / 0.1% Triton X-100 for 2 hours. 20 μl of sample was

added to 180 μl of reaction buffer containing NADH (100 μM) and sodium pyruvate (600

μM) in KPP-buffer. Absorption at 340 nm was measured at 37°C in 1 min intervals over a

period of 15 min. The slope of the absorption intensity was calculated. The ratio of

LDHsupernatant/LDHtotal was calculated using the slopes of supernatant and homogenate.

LDH release was expressed in percent. Control data were substracted from LDH values. Basic

release of untreated cells was about 7% in 24 h.

Calcein-AM/TMRE staining: Calcein-AM staining, labeling live cells, and TMRE

staining, labeling all intact mitochondria, were performed with 1 µM Calcein-AM / 50 nM

TMRE / 1 µg/ml H-33342 for 30min at 37°C. Images were collected in three different

fluorescent channels using an automated microscope (Array-Scan VTI HCS Reader (Thermo

Fisher, PA). Using an imaging software (vHCS SCAN, Thermo Fisher, PA) nuclei were

identified in channel 1 (365±50/461±15 nm) as objects according to their size, area, shape,

and intensity. Calcein signal was detected in channel 2 (475±40/525±15 nm). An algorithm

quantified all calcein positive cells as viable and only H-33342 positive nuclei as “not viable”

cells.

For evaluating the mitochondrial mass, nuclei masks, determined in channel 1, were

expanded and transferred to channel 3. All TMRE positive pixels (575±25/640±35 nm)

outside of these masks were counted as mitochondrial mass.

ATP determination: To determined intracellular ATP, cells grown in 24-well plates were

scratched and sonicated in PBS-buffer and boiled at 95 °C for 10 min followed by

centrifugation at 10,000 g for 5 min for the removal of cell debris. For the detection of ATP

levels, a commercially available ATP assay reaction mixture (Sigma–Aldrich), containing

luciferin and luciferase, was used (Volbracht et al., 1999). Fifty microliters sample and 100 µl

of assay-mix were added to a black 96-well plate. Standards were prepared by serial dilutions

of ATP disodium salt hydrate (Sigma–Aldrich) to obtain concentrations ranging from

1000 nM to 7.8 nM.



134

GSH determination: For glutathione determination cells were washed twice with PBS

and lysed in 200 μl of 1% sulfosalicylic acid (w/v). The lysates were collected, sonicated 5

times and centrifuged at 12,000 ×g for 5 min at 4 °C to remove cell debris. Total glutathione

content was determined by a DTNB (5,5′-Dithiobis(2-nitrobenzoic acid)) reduction assay.

Supernatants were diluted 1:10 in H2O, 100 μl sample was mixed with 100 μl assay mixture

containing 300 μM DTNB, 1 U/ml glutathione-reductase, 400 μM NADPH, 1 mM EDTA in

100 mM sodium phosphate buffer, pH 7.5 (all Sigma). DTNB reduction was measured

photometrically at 405 nm in 5 min intervals over 30 min. GSH standard curves (Sigma) were

performed by serial dilutions ranging from 1000 nM to 7.8 nM, respectively.

Western blot Analysis:

Cells were lysed in RIPA-buffer (50 mM Tris-base, 150 mM NaCl, 1 mM EDTA, 0.25%

sodium deoxycholate, 1% NP40, 1 mM Na3VO4, 50 mM NaF, pH 7.5) containing 1x

protease inhibitor (Roche) and 0.5 % phosphatase inhibitor cocktail 2 (Sigma). Determination

of protein concentration was performed by using a BCA protein assay kit (Pierce/Thermo

Fisher Scientific, Rockford, IL, USA). Thirty-five micrograms of total protein were loaded

onto 12% SDS gels or onto 18% for histone modification analysis. Proteins were transferred

onto nitrocellulose membranes (Amersham, Buckinghamshire, UK). Loading and transfer

were checked by brief Ponceau staining. Washed membranes were blocked with either 5%

milk or 5% BSA, dependent on the primary antibody used, in TBS–Tween (0.1%) for 1 h.

Primary antibodies were incubated at 4° C over night. Following washing steps with TBS–

Tween (0.1%), horseradish peroxidase-conjugated secondary antibodies were incubated for 1

h at RT. For visualization, ECL Western blotting substrate (Pierce/Thermo Fisher Scientific)

was used. For the detailed list of antibodies used, see supplementary Table I on page 1

Immunocytochemistry:

Cells were grown on 13 mm glass cover slips (Menzel, Braunschweig, Germany) in 24-

well plastic cell culture plates (NunclonTM) and fixed with 4% paraformaldehyde. After

incubation with the primary antibody overnight and with the appropriate secondary antibody

for 1 h, Hoechst-33342 (1 µg/ml Molecular Probes) was added for 10 min prior to the final

washing step. Cover slips were mounted on glass slides with Fluorsave reagent (Calbiochem).

For visualization a confocal microscope (Zeis LSM510Meta) was used. For image processing,

Photoshop (Adobe) was used. The antibody against PSPC1 was from Sigma (rabbit, 1:200)



135

As secondary antibody, anti-rabbit Alexa-488 (1:1000, Molecular Probes, Eugene, OR, USA)

was applied.

qPCR:

For reverse transcription quantitative PCR (RT-qPCR) analysis, RNA was extracted with the

PureLink RNA mini Kit (invitrogen, Darmstadt, Germany) according to the manufacturer’s

instructions. For transcript analyses of LUHMES, primers (Eurofins MWG Operon,

Ebersberg, Germany) were designed using AiO (All in One) bioinformatics software

(Karreman 2002) and can be found in supplementary table II. All RT-qPCRs were based on

the SsoFast EvaGreen detection system and were run in a CFX96 Cycler (Biorad, München,

Germany) and analysed with Biorad iCycler software. The threshold cycles (Ct) were

determined for each gene and gene expression levels were calculated as relative expression

compared to GAPDH (2-(ΔCt)) or as fold change relative to control (2-(ΔΔCt)). ΔCt and ΔΔCt

were calculated using following formulas:

ΔCt = Ct(conditionX/gene Y) – Ct(conditionX/GAPDH).

ΔΔCt = ΔCt(conditionX/gene Y) – ΔCt(untreated control/gene Y).

Affymetrix gene chip analysis:

Analysis was performed as described earlier (Krug et al 2013). Briefly, samples from

approximately 5 x106 cells were collected using RNA protect reagent from Qiagen. The RNA

was quantified using a NanoDrop N-1000 spectrophotometer (NanoDrop, Wilmington, DE,

USA), and the integrity of RNA was confirmed with a standard sense automated gel

electrophoresis system (Experion, Bio-Rad, Hercules, CA, USA). The samples were used for

transcriptional profiling when the RNA quality indicator (RQI) number was >8. First-strand

cDNA was synthesized from100 ng total RNA using an oligo-dT primer with an attached T7

promoter sequence, followed by the complementary second strand. The double-stranded

cDNA molecule was used for in vitro transcription (IVT, standard Affymetrix procedure)

using Genechip 30 IVT Express Kit. During synthesis of the aRNA (amplified RNA, also

commonly referred to as cRNA), a biotinylated nucleotide analogue was incorporated, which

serves as a label for the message. After amplification, aRNA was purified with magnetic

beads and 15 µg of aRNA was fragmented with fragmentation buffer as per the

manufacturer’s instructions. Then, 12.5 µg fragmented aRNA was hybridized with

Affymetrix Human Genome U133 plus 2.0 arrays as per the manufacturer’s instructions. The

chips were placed in a GeneChip Hybridization Oven-645 for 16 h at 60 rpm and 45 C. For

staining and washing, AffymetrixHWS kits were used on a Genechip Fluidics Station-450.



136

For scanning, the Affymetrix Gene-Chip Scanner-3000-7G was used, and the image and

quality control assessments were performed with Affymetrix GCOS software. All reagents

and instruments were acquired from Affymetrix (Affymetrix, Santa Clara, CA, USA). The

generated CEL files were used for further statistical analysis. The authors declare that

microarray data were produced according to MIAME guidelines and will be deposited in

ArrayExpress upon acceptance of the manuscript.

Statistics and data mining: The microarray data analysis workflow was assembled using

the Konstanz Information Miner open source software (Berthold et al 2008). The raw data

was preprocessed using Robust Multiarray Analysis (RMA) (Smyth et al 2005). Background

correction, quantile normalization, and summarization were applied to all expression data

samples, using the RMA function from the affy package of Bioconductor (Gautier et al 2004,

Gentleman et al 2004). Low-expression genes with a signal below an intensity of 64 in any

one of the 12 conditions were filtered out. The limma package (R & Bioconductor) was used

to identify differentially expressed genes, with untreated cells set as control group. The

moderated t-statistics was applied and was used for assessing the raw significance of

differentially expressed genes. Then, final p-values were derived by using the Benjamini-

Hochberg method to control the false discovery rate (FDR) (Benjamini & Hochberg 1995)

due to multiple hypothesis testing. Transcripts with FDR adjusted p-value of ≤ 0.05 and a fold

change values ≥ |2| were considered significantly regulated. The hierarchical clustering

analysis was performed as previously described (Berry et al 2010). Average linkage was used

as agglomeration rule for the clustering analysis. Distances based on the Pearson’s correlation

coefficient was used to group together transcripts with similar expression patterns across

samples (rows of the heat map). Distances based on Spearman’s rank correlations of the gene

expression values was used to measure the similarity between samples. Then expression

values within each row were normalized as Z-factors, and color-coded accordingly. The

colors represent Z-scores of the row-wise normalized expression values for each gene,

whereas the highest Z-score is in bright yellow and the lowest in dark red.

RNAsequencing:

Illumina library preparation and sequencing: The sequencing library preparation has

been done using 200ng of total RNA input with the TrueSeq RNA Sample Prep Kit v2-Set B

(RS-122-2002, Illumina Inc, San Diego, CA) producing a 275bp fragment including adapters

in average size. In the final step before sequencing, 8 individual libraries were normalised and



137

pooled together using the adapter indices supplied by the manufacturer. Pooled libraries have

then been clustered on the cBot Instrument form Illumina using the TruSeq SR Cluster Kit v3

- cBot - HS(GD-401-3001, Illumina Inc, San Diego, CA). Sequencing was then performed as

50 bp, single reads and 7 bases index read on an Illumina HiSeq2000 instrument using the

TruSeq SBS Kit HS- v3 (50-cycle) (FC-401-3002, Illumina Inc, San Diego, CA).

Approximately 20-30 million reads per sample were sequenced.

RNA-seq computational analysis: Illumina reads were converted to the industry standard

FASTQ format and aligned to the Human GRCh37 Ensembl 70 reference genome using the

STAR program on default settings (http://www.ncbi.nlm.nih.gov/pubmed/23104886). For

increased alignment accuracy, the STAR genome index was generated to include splice

junction annotations with the options “--sjdbGTFfile

Homo_sapiens.GRCh37.70.primary_assembly.gtf --sjdbOverhang 49”. The SAM output from

the STAR aligner was converted to BAM format using the Picard tools suite

(http://picard.sourceforge.net). For gene expression estimation and differential expression

analysis the programs Cufflinks and Cuffdiff version 2.0 were used with the following options

“-u --max-bundle-frags 1000000 --no-effective-length-correction --compatible-hits-norm”

(http://www.ncbi.nlm.nih.gov/pubmed/23222703). The quality of the RNA-seq experiments

was verified with RNASeQC version 1.17 (http://www.ncbi.nlm.nih.gov/pubmed/22539670).

Pre-processed data from cuffdiff 2.0.2 were further analysed with CummeRbund 2.0.0

(http://compbio.mit.edu/cummeRbund/). A significance threshold of FDR (Benjamini

Hochberg multiple testing) corrected p-values was set at 0.05. The (base 2) log of the fold

change y/x (FKPM) was used as measure for differential gene expression. For comparison

with Microarray data the ENSEMBL gene identifiers were converted to HGNC symbols. To

test the correlation between the two platforms the log2 fold changes of overlapping genes for

the 48 h samples were plotted and the spearman correlation was calculated.

GO enrichment – Wordclouds:

To identify individual gene ontologies (GOs) for the altered genes of the transcriptomics

studies, we entered the gene names into the query of gProfiler (http://biit.cs.ut.ee/gprofiler/).

Only GO terms consisting of less than 1000 genes were used to create GO wordclouds. If

more than 30 GOs were identified, only the top 30 GOs with the lowest p-values were

http://picard.sourceforge.net/



138

displayed. All identified GOs can be found in Suppl Table S4/S5. The wordclouds were

produced on http://www.wordle.net/advanced. Scaling of character size is linearly

proportional to the negative ln of the p-value of the respective GO category.

Metabolomics analysis – untargeted:

After MPP+ treatment, cells were washed with ice cold PBS. Dry-ice cold 80:20

MeOH/water solution was added immediately to the wells. Cells were scraped and collected

in a 1.5 mL Eppendorf tube. Wells were washed again with MeOH/water and this solution

was combined with the previous solution. Tubes were stored at -80 °C for at least two hours

to precipitate the proteins. For metabolite extraction, tubes were placed on dry ice for 15 min

and centrifuged at 14,000 g for 5 min at 4 °C. The supernatant was transferred to a new 1.5

mL tube and placed on dry ice. 300 µL of 80:20 MeOH/water was then added to the pellet

and a second extraction was performed. The combined supernatants were evaporated to

dryness at room temperature in a Speedvac concentrator. The dried samples were

reconstituted with 60 µL of 60% MeOH with 0.1% FA, clarified by centrifugation at 14,000 g

for 5 minutes. The clarified samples were transferred to HPLC vials for LC-MS

measurements.

Liquid chromatography: Chromatographic separations were performed using an Agilent

1260 high performance liquid chromatography system with a wellplate autosampler (Agilent,

Santa Clara, CA). For the reverse phase (RP) separation, a TARGA ® (Higgins, Mountain

View, CA) C18 column (50 x 2.1 mm i.d., 3 µm particle size, 120 Å pore size) was used on

the LC system. The LC parameters for RPLC analysis were as follows: autosampler

temperature, 4 ℃; injection volume, 5 ul; column temperature, 35 °C; flow rate, 0.3ml/min.

The solvents and optimized gradient conditions for LC were: Solvent A, water with 1 mM

ammonium fluoride; Solvent B, 100% acetonitrile; elution gradient: 0 min - 2% B; 20 – 25

min – 98% B; post-run time for equilibration, 5 min in 2% B. For aqueous normal phase

(ANP) separation, a Cogent Diamond Hydride ™ (MicroSol, Eatontown, NJ) column (150 x

2.1 mm i.d., 4 µm particle size, 100 Å pore size) was used for separation of metabolites. The

LC parameters were as follows: autosampler temperature, 4 ℃; injection volume, 5 ul;

column temperature, 35 ℃; flow rate, 0.4ml/min. The solvents and optimized gradient

conditions for LC were: Solvent A, 50% methanol / 50% water / 0.05% formic acid; Solvent

B, 90% acetonitrile with 5 mM ammonium acetate; elution gradient: 0 min - 100% B; 20 – 25

min – 40% B; post-run time for equilibration, 10 min in 100% B. The LC system was coupled

directly to the Q-TOF mass spectrometer. A blank injection was run after every 3 samples and



139

a QC sample was run after every 5 samples to identify the sample carryover and check for

stability.

Mass spectrometry: A 6520 accurate-mass Q-TOF LC-MS system (Agilent, Santa Clara,

CA) equipped with a dual electrospray (ESI) ion source was operated in negative-ion mode

for metabolic profiling. The optimized ESI Q-TOF parameters for MS experiments were: ion

polarity, negative; gas temperature, 325 ℃; drying gas, 10 l/min; nebulizer pressure, 45 psig;

capillary voltage, 4,000 V; fragmentor, 140 V; skimmer, 65 V; mass range, 70 - 1,100 m/z;

acquisition rate, 1.5 spectra/s; instrument state, extended dynamic range (1,700 m/z, 2 GHz).

MS/MS experiments were carried out to confirm the putative identification of metabolites

based on mass accuracy. Nitrogen was employed as collision gas and collision energy was

adjustable from 10 to 40 eV. Spectra were internally mass calibrated in real time by

continuous infusion of a reference mass solution using an isocratic pump connected to a dual

sprayer feeding into an electrospray ionization source. Data were acquired with MassHunter

Acquisition software (Agilent Technologies).

Statistics and data mining: For the data processing and chemometric analysis of the LC-

MS data, the acquired raw data files (.d files) were processed with MassHunter Qualitative

Analysis software (Agilent, version 5.0). Reproducibility of chromatograms was first

inspected by overlaying the Total Ion Chromatograms (TICs) of all samples. Data files that

showed extraneous peaks were excluded for further processing. Initially, putative metabolite

identification was achieved by searching the accurate m/z values of the peaks against an in-

house built database derived from HMDB, KEGG, METLIN and other public databases. At

the same time, the Extracted Ion Chromatograms (EICs) for these matched putative

metabolites were generated by performing Find by Formula function integrated into the

software. The abundance of the EICs was calculated by summing the intensities of all

compound-related peaks (e.g. isotopic peaks, adduct peaks, etc.). The pre-processed data files

were exported as ‘cef’ formatted files, which contain a table of mass and retention time pairs

with associated intensities. These ‘cef’ files were imported into Mass Profiler Professional

software (Agilent, version 12.1) for further data processing. For example, peak alignment,

background noise subtraction and other data reduction processes could be done by using this

software. The optimized parameters for these data processing steps were set as follows:

minimum absolute abundance, 5,000 counts; retention time range, 0-25 min; mass range, 70-

1,100 m/z; minimum of ions, 2; multiple charge state forbidden; retention time window,

1min, mass window, 15 ppm + 2.0mDa; To treat all exacted compounds equally regardless of



140

their intensities, each entity was baselined to median of intensity of all samples. An ANOVA

statistical test (p < 0.05) followed by a Benjamini-Hochberg multiple test correction was

performed with the normalized data for differential analysis. A principle component analysis

(PCA) was used for modeling the difference between the controls and MPP+ treated samples.

PCA is an eigenvector-based unsupervised multivariate analysis. By using this analysis, one

could reduce original large set of inter-correlated variables into a few independent

uncorrelated variables (principal components) while retaining the features that contribute

most to the variance.A significant metabolite list was generated after ANOVA test and used

for later pathway analysis.

MS/MS spectra acquired from reference metabolites were used for confirmation of the

identification of statistically significant metabolites. More specifically, the exact m/z values

and intensities of fragment ions from the acquired MS/MS spectra of putative metabolites

must have a reasonable match with that of reference metabolites or the fragment ions from

public databases (e.g. METLIN, MassBank), if available.

Metabolomics analysis – targeted:

Sample preparation was made in the same manner as for the untargeted analysis with some

minor modification, as there was no second extraction performed. Dried samples were stored

at -80°C and send on dry-ice to BIOCRATES Life Sciences AG to further process the

samples. The frozen cell pellets were resuspended in 60μl chilled phosphate buffer. Cell lysis

was done by freezing the cell suspension in liquid nitrogen and thawing it in an ultrasonic

bath (4°C). This freezing-thawing cycle was repeated 3 times. After this, the cell suspension

was centrifuged at 2°C and the supernatant was directly use for analysis. The targeted

metabolomics approach was based on measurements with the AbsoluteIDQTM p180 kit and

the oxidative status assay (BIOCRATESLife Sciences AG, Innsbruck, Austria). The p180 kit

allows simultaneous quantification of 186 metabolites, consisting of amino acids,

acylcarnitines, sphingomyelins, phosphatidylcholines, hexose (glucose), and biogenic amines.

The fully automated assay was based on PITC (phenylisothiocyanate)-derivatization in the

presence of internal standards followed by FIA-MS/MS (acylcarnitines, lipids, and hexose)

and LC/MS (amino acids, biogenic amines) using an AB SCIEX 4000 QTrap™ mass

spectrometer (AB SCIEX, Darmstadt, Germany) with electrospray ionization. The

experimental metabolomics measurement technique is described in detail by patent US

2007/0004044 (accessible online at http://www.freepatentsonline.com/y2007/0004044.html).

For the oxidative status assay three thiol amino acid redox couples (reduced and oxidized



141

forms of homocysteine, cysteine, glutathione) were assayed by LC/MS/MS using a API

5500™ mass spectrometer (AB Sciex, Darmstadt, Germany). The analytes were separated on

a porous graphitic carbon column (PGC) using gradient elution. The total run time of the

analysis is 15 minutes, injection volume was20 μl. The analytes were quantified by positive

ion tandem electrospray ionization mass spectrometry in multiple reaction monitoring mode

using internal standard calibration.

TCA-Flux analysis - GC/MS Sample Preparation and Procedure:

Cells were grown in six-well plates and treated with 1 µM GBR and 12.5 mM D-glucose-

13C6 on day 6, 30 min later 5 µM MPP+ were added for 18 h.. Cells were washed with 1ml

saline solution and quenched with 0.4 ml - 20 °C methanol. After adding an equal volume of

4 °C cold water, cells were collected with a cell scraper and transferred in tubes containing

0.4 ml -20 °C chloroform. The extracts were vortexed at 1,400 rpm for 20 min at 4 °C and

centrifuged at 16,000×g for 5 min at 4 °C. 0.3 ml of the upper aqueous phase was collected in

specific GC glass vials and evaporated under vacuum at -4°C using a refrigerated CentriVap

Concentrator (Labconco). The interphase was centrifuged with 1 ml -20 °C methanol at

16,000×g for 5 min at 4 °C. Metabolite derivatization was performed using an Agilent

Autosampler. Dried polar metabolites were dissolved in 15μL of 2% methoxyamine

hydrochloride in pyridine at 45 °C. After 30min, an equal volume of 2,2,2-trifluoro-N-methyl-

N-trimethylsilyl-acetamide +1% chloro-trimethyl-silane were added and held for 30 min at 45

°C. GC/MS analysis was performed using an Agilent 6890GC equipped with a 30m DB-

35MS capillary column. The GC was connected to an Agilent 5975C MS operating under

electron impact ionization at 70 eV. The MS source was held at 230 °C and the quadrupole at

150 °C. The detector was operated in scan mode and 1μL of derivatized sample was injected

in splitless mode. Helium was used as carrier gas at a flow rate of 1 mL/min. The GC oven

temperature was held at 80 °C for 6 min and increased to 300 °C at 6 °C/min. After 10 min,

the temperature was increased to 325 °C at 10 °C/min for 4 min. The run time of one sample

was 59 min. To determine the pyruvate dehydrogenase activity, citrate with 13C2 was

measured, as a reflection of the conversion of one glucose molecule into two pyruvate

molecules and finally the conversion of pyruvate to citrate by pyruvate dehydrogenase.

Statistics and data mining:

Cytotoxicity data (ATP, GSH, LDH, resazurin)and qPCR are presented as means of

independent experiments, and statistical differences were tested by ANOVA with post-hoc

tests as appropriate, using GraphPad Prism 5.0 (Graphpad Software, La Jolla, USA).



142

Supplements

Antigen Antibody (supplier; clone) Dilution Blocking

(5%)

Species

CREB-2

(ATF4)

Anti-CREB-2 (C-20): sc-200 (santa cruz) 1:200 BSA rabbit

GADD34 Anti-GADD34 (proteintech) 1:1000 BSA mouse

eIF2a-p Anti-eIF2a [pS52] (invitrogen) 1:1000 BSA rabbit

PSPC1 Anti-PSPC1 (sigma) 1:200 Milk rabbit

GAPDH Anti-GAPDH (Sigma; Clone GAPDH-71.1) 1:5000 BSA mouse

Bcl-xl anti Bcl-xl (CellSignaling) 1:1000 Milk rabbit

Cytochrom C Anti-Cytochcrom C (BD Pharmingen;

Clone 7H8.2C12

1:1000 Milk mouse

meH3K9 Anti trimethyl-histone H3 (Lys9)(millipore) 1:1000 BSA rabbit

meH3K27 Anti-trimethyl-histone H3 (Lys27)(millipore) 1:1000 BSA rabbit

meH3K4 Anti-trimethyl-histone H3 (Lys4)(active

motif)

1:1000 BSA rabbit

anti-mouse anti-mouse HRP antibody (Jackson

Immuno Research)

1:2500 goat

anti-rabbit anti-rabbit HRP antibody (GE Healthcare ) 1:5000 goat

Name Forward sequence Reverse Sequence

ASNS GGGGCTTGGACTCCAGCTTG GAGCCTGAATGCCTTCCTCA

ASS1 TGCTCCCTGGAGGATGCCTG GTGTAGAGACCTGGAGGCGC

ATF2 AGAGCGAAATAGAGCAGCAG CATGGCGGTTACAGGGCAAT

ATF4 GGCTGGCTGTGGATGGGTTG CTCCTGGACTAGGGGGGCAA

CBS TCCTGGGAATGGTGACGCTT GTGCTGTGGTACTGGATCTG

CCNB TGGATGTGCCCCTGCAGAAG CAGTGACTTCCCGACCCAGT

CTH TGGATGATGTGTATGGAGGTACAAACAGG GCCTTCAATGTCAATCACCTTCTGGG

DDIT3 ATGGCAGCTGAGTCATTGCC TCCTCAGTCAGCCAAGCCAG

DDIT4 AGTCCCTGGACAGCAGCAAC AACTGGCTAGGCATCAGCAG

GADD34 GCATCACCCAGGCCCAGGAG AGACGAGCGGGAAGGTGTGG

GAPDH CACCATCTTCCAGGAGCGAGATC GCAGGAGGCATTGCTGATGATC

HNRNPM TGGTGTGGTGGTCCGAGCAG GGACGCTCAGGAGGGAAGAA

MLF1IP TTTGTAAGGCAGCCATCGCC CTGTGGCTCTAACCGAAGCA

NOXA CAGTGCCAACTCAGCACATTG CGCCCAACAGGAACACATTGA

NQO1 TGGAGTCGGACCTCTATGCCA CTTGTGATATTCCAGTTCCCCCTGC

PPA2 TGGAAAGCTACGCTATGTGG GCTTCAGGATCATTCGCATTG

PSPC1 CAGCAGCGTGAGCAGGTTGA CGCCGATGCTCCTCTTCATG

SFPQ TCAGGCAAATCTTTTGCGCC CTCTCTTTGGCGCCTCATTT

SHMT2 CAACCTGGCACTGACTGCTC GATGTCCGCGTGCTTGAAAG

TXNIP1 CATGGCGTGGCAAGAGCCTT CTCAGAGCTGGTTCGGCTGG

TYMS CAGCTTCAGCGAGAACCCAG ACCTCGGCATCCAGCCCAAC

Table I – Antibodies used for western blot or immunostaining

Table II – Primers used for RT-qPCR



143

Figure S1: MPP+-induced changes of amount and location of apoptosis-related proteins LUHMES cell lysates were prepared at different times after exposure to MPP+ as illustrated in Fig.

1A. Proteins were quantified by western blot. To the left, representative blots are shown. To the right,

densitometric quantifications of the respective proteins are displayed as means ± SD of 3 independent

differentiations. Calibration was relative to loading control and untreated cell samples. A) Bcl-xl

levels. B) The cytosolic cytochrom C levels were determined by extraction of soluble cytosolic

proteins after permeabilisation of the outer cell membrane with 50 µg/ml digitonin. This procedure did

not permeabilize the outer mitochondrial membrane. Controls were healthy cells without digitonin

(w/o), healthy cells with digitonin (0) and a positive control of cells exposed to 200 nM staurosporine

for 10 h (STS). C) PSPC1 (paraspeckle component 1).



144

Figure S2: Overview of significantly altered metabolites determined during untargeted

analysis with the accurate-mass Q-TOF LC-MS system Samples of four independent experiments per condition (control, 24 h or 36 h treatment with 5 µM

MPP+) were run with the accurate-mass Q-TOF LC-MS system (Agilent, Santa Clara, CA).

Metabolites were determined by the MassHunter Acquisition software (Agilent Technologies). ‘Areas

under the Curve’ (AUC) for every peak/metabolite of the ion chromatogram were calculated and used

as basis for a relative quantification. AUCs were normalized to the mean of the AUCs of control

samples (untreated cells) of 4 independent experiments. Only metabolites that were significantly

regulated (FDR adjusted p-value of ≤ 0.05) and that could be unambiguously identified, are displayed.

Downregulated metabolites

ATP

Cyc

lic A

DP-rib

ose

Dec

anoic

aci

d

Deh

ydro

asco

rbat

e

Deo

xyurid

ine

D-E

ryth

rose

D-G

luco

se

Glu

tath

ione

Guan

ine

Inosi

ne

L-Ala

nine

L-Asp

arag

ine

L-Glu

tam

ate

L-Pro

line

L-Ser

ine

Mal

eam

ate

N-A

cety

l-L-a

spar

tate

N-A

cety

l-L-g

luta

mat

e

O-A

cety

l-L-h

omose

rine

O-A

cety

lneu

ram

inic

aci

d

Pan

toth

enat

e

Phosp

hocrea

tine

sn-g

lyce

ro-3

-Phosp

hoethan

olam

ine

Sorb

itol

Taurine

UDP-a

lpha-

D-g

alac

tose

UDP-g

luco

se

UDP-N

-ace

tyl-D

-gal

acto

sam

ine

UDP-N

-ace

tyl-D

-Glu

cosa

min

e

2,3-

Dim

ethyl

mal

eate

2,5-

Dio

xopen

tanoat

e

3-Oxo

propan

oate

4-Am

inobuta

noate

0

50

100

150

No

rmalized

In

ten

sit

y V

alu

es

[% o

f co

ntr

ol

SD

]Upregulated metabolites

Aden

ine

ADP

AM

P

Chole

ster

ol sulfa

te

Cre

atin

e

Deo

xyribose

Dih

ydro

ptero

ate

Formyl

-N-a

cety

l-5-m

ethoxy

kynure

namin

e

L-Arg

inin

e

L-Cys

tath

ionin

e

L-Lac

tate

L-Lys

ine

L-Met

hionin

e S-o

xide

L-Phen

ylal

anin

e

L-Try

ptophan

L-Tyr

osine

Pyr

uvate

S-A

denosy

l-L-h

omocy

stei

ne

S-A

denosy

l-L-m

ethio

nine

S-M

ethyl

GSH

Thiam

ine

acet

ic a

cid

Ura

te-3

-rib

onucleo

side

2-Ace

tola

ctat

e

3-(4

-Hyd

roxy

phenyl

)lact

ate

3-M

ethyl

-2-o

xobuta

noic a

cid

4-Hyd

roxy

-4-m

ethyl

gluta

mat

e

0

100

200

300

400

2000

4000

control

24 h

36 h

No

rmalized

In

ten

sit

y V

alu

es

[% o

f c

on

tro

l

SD

]

B

A



145

Creatine

Counts vs. Aquisition Time

Phosphocreatine


L-methionine-S-oxide


S-Adenosyl-L-methionine


Control

MPP+

Figure S3: Example-peaks of metabolites affected by MPP+ treatment Cells were treated with MPP+ or solvent for 24 h. Samples were run with the accurate-mass Q-TOF

LC-MS system. Peak overlays are displayed by the Agilent MassHunter Acquisition software. The

areas under the curve (integral of the peak curve) were used for quantification. Data are examples of

typical primary data in a representative experiment.



146

Figure S4: Principal component analysis (PCA) of regulated genes of MPP+ treated

LUHMES cells Cells were treated with MPP+ (5 µM) for different times and samples were prepared for microarray

analysis as described in Fig. 1A. The signal of all probe sets significantly regulated (FDR adjusted p-

value of ≤ 0.05; fold change values ≥ 2) was used for principal component analysis (PCA). The first

two dimensions of the respective data are displayed.

Prin

cip

al co

mpo

nen

t 2 (

17

%)

48 h

Control

36 h

24 h

Pri

ncip

alc

om

po

ne

nt2

: 17

%

Principal component 1: 23 %



147

Figure S5: Transcriptional changes triggered by the complex I inhibitor rotenone in

LUHMES cells Cells were treated with different concentrations of rotenone (20, 100, 500 nM) for different times (24,

48, 72 h) and samples were taken together at day 9 (d9) as indicated in the treatment scheme. A)

General cytotoxicity was evaluated by measurement of LDH release into the medium. GSH and ATP

concentrations were determined in cell lysates. B) Several genes, which had been found to be

regulated in the MPP+ toxicity model, were examined. The mRNA from samples treated with 100 nM

rotenone was qualified by RT-qPCR. All Data are means ± SEM of three independent experiments.

They were normalized to untreated control. For easier comparison, the 48 h data for cytotoxicity and

gene regulation are highlighted by a dashed box.

0 20 40 60

0

1

2

3

4

6

8

10

Time [h]

mR

NA

leve

l [fo

ld c

hang

e r

ela

tive

to

co

ntr

ol

SE

M]

0 24 48 72

0

50

100

0.02 µM0.1 µM0.5 µM

Time [h]

AT

P c

on

ce

ntr

atio

n

[% o

f co

ntr

ol

SE

M]

0 24 48 72

0

50

100

0.02 µM0.1 µM0.5 µM

Time [h]

GS

H c

on

ce

ntr

atio

n

[% o

f co

ntr

ol

SE

M]

0 24 48 72

0

20

40

600.02 µM

0.1 µM

0.5 µM

Time [h]

Ext

race

llula

r L

DH

[% o

f co

ntr

ol

SE

M]

Sampling

d-1 d0 d2 d4 d6 d7 d8 d9

48 h 24 h72 hReplate

Medium changeStart diff.

0.1 µM Rotenone

ATF4 – Activating transcirption factor 4

DDIT3 – DNA damage inducibletranscript3 (CHOP, GADD153)

ASS1 – Argininosuccinate synthase 1

ASNS – Asparagine synthase

NOXA – Phorbol-12-myristate-13-acetate-induced protein 1 (PMAIP1)

CTH – Cystathionase(cystathionine gamma-lyase)

CBS – Cystathionine-β-synthase

TXNIP – Thioredoxin-interacting protein

HNRNPM – Heterogeneous nuclearribonucleoprotein M

TYMS – Thymidylate synthetase

Page 6

biosynthetic processes oxidative stress

chromosomal changes/paraspeckles

ER stressmitochondrial function

Colour coding of biological process:

B

A

** **

**

**

**

*



148

Figure S6: Integrated analysis of metabolomics and transcriptomics data to identify

affected pathways The pathway-analysis is based on transcriptomics and metabolomics data of the cells treated for 24 h

with 5 µM MPP+. The integration of both data sets for multi-omics analysis was performed using

Mass Profiler Professional (MPP) software (version 12.6, Agilent Technologies). The MPP multi-

omics capability allows two different omic experiments to be mapped and seen on the same pathway.

Metabolites, identified by Q-TOF LC-MS and transcripts of the DNA microarray analysis were

combined and reanalyzed together for pathway enrichment. A fold change cut-off for transcripts and

metabolites of 1.5 and a significance threshold of 0.05 (FDR corrected)were used. WikiPathways

served as pathway source. Differentially detected metabolites and genes are highlighted in colour and

have an adjacent heat strip for the relative abundances across the different conditions (red = control

samples, yellow = 24 h samples, blue = 36 h samples, grey = 48 h samples). Enzymes or metabolites

with increased levels are indicated in yellow (cysteine was detected only in the targeted analysis); blue

indicates a down-regulation. (AHCY = adenosylhomocysteinase, AHCYL1 = adenosylhomocysteinase-like 1,

AHCYL2 = adenosylhomocysteinase-like 2, AMT = aminomethyltransferase, BHMT = betaine—homocysteine

S-methyltransferase, CBS = cystathionine-β-synthase, CTH = cystathionase, DHFR = dihydrofolatereductase,

DHFRL1 = dihydrofolatereductase-like 1, DNMT1 = DNA (cytosine-5-)-methyltransferase 1, DNMT3A = DNA

(cytosine-5-)-methyltransferase 3 alpha, DNMT3B = DNA (cytosine-5-)-methyltransferase 3 beta, DNMT3L =

DNA (cytosine-5-)-methyltransferase 3-like, dTMP = deoxythymidine monophosphate, dUMP = deoxyuridine

monophosphate, GCLC = glutamate-cysteine ligase, GCLM = glutamate-cysteine ligase, modifier subunit, GSS

= glutathione synthetase, MAT2B = methionine adenosyltransferase II beta, MAT1A = methionine

adenosyltransferase I alpha, MAT2A = methionine adenosyltransferase II alpha, MTHFD1 =

methylenetetrahydrofolate dehydrogenase 1, MTHFD1L = methylenetetrahydrofolate dehydrogenase 1-like,

MTHFD2 = methylenetetrahydrofolate dehydrogenase 2, MTHFD2L = methylenetetrahydrofolate

dehydrogenase 2-like, MTHFR = methylenetetrahydrofolatereductase, MTR = methionine synthase, PHGDH =

phosphoglyceratedehydrogenase, PSAT1 = phosphoserine aminotransferase 1, PSPH =

phosphoserinephosphatase, ROS = reactive oxygen species, SHMT1 = serine hydroxymethyl-transferase 1,

SHMT2 = serine hydroxymethyl-transferase 2, THF = tetrahydrofolate, TYMS = thymidylatesynthetase)

Regulated metabolites

Regulated genes



149

Figure S7: Separation of MPP+ toxicity and counter-regulation in immature and mature

cells Immature cells were treated with MPP+ during differentiation and sampled at day 5 and mature cells

were sampled at day 8. A) Experimental scheme for cell differentiation, MPP+ exposure and sampling

of differentiating cells. In all experiments with immature cells, samples were analyzed on day 5 (d5) of

differentiation (green arrow). Red arrows mark time points of treatments. B) MPP+ uptake was

measured in cells of different maturity stages. Mature LUHMES displayed a significant higher uptake

velocity. C) Intracellular GSH concentrations were determined for mature LUHMES cells (d8) treated

with various concentrations (0.01; 0.5; 1; 5; 25 µM) of MPP+ for 24 hours. D) RT-qPCR analysis of

the transsulfuration pathway was performed. Cystathionase (CTH) and Cystathionine-β-synthase

(CBS) mRNA levels were evaluated on day 5 of cells treated with 1 µM (blue) and 5 µM (red) MPP+

for different time points. E) LUHMES cell lysates were prepared at day 5 of differentiation. Cells were

exposed for indicated time periods to 1 or 5 µM MPP+. Proteins were separated and transferred by

SDS-Page and western blot. ATF4 was visualized by immunoblotting and GAPDH was used as

loading control. F) Intracellular ATP concentrations of LUHMES treated with 1, 5 or 25 µM MPP+ for

the indicated time were determined. No significant change of the ATP concentrations of untreated

compared to treated LUHMES was observed. G) Intracellular GSH concentrations were determined

for immature LUHMES cells (d5) treated with various concentrations (0.01; 0.5; 1; 5; 25 µM) of

MPP+ for 24 hours. H) Intracellular GSH concentrations were determined for immature LUHMES

cells (d5) treated for various durations with 1 µM MPP+.

0 2 4 6 810

80

100

120

140

20 30 40

Time [h]

GS

H c

on

ce

ntr

atio

n

[% o

f co

ntr

ol

SE

M]

0.01

0

50

100

150

0.5 1.0 10 25crtlµM MPP+

GS

H c

on

ce

ntr

atio

n

[% o

f co

ntr

ol

SE

M]

0 24 48 72

0

50

100

1 µM5 µM25 µM

Time [h]

AT

P c

on

ce

ntr

atio

n

[% o

f co

ntr

ol

SE

M]

0

25

50

1 10 20

80

90

100

110

120

crtl 0.01

MPP+ [µM]

GS

H c

on

ce

ntr

atio

n

[% o

f co

ntr

ol

SD

]

0 12 24 36 48 60 72

0

2

4

6

8

10CTH [5 µM MPP+]

CBS [5 µM MPP+]

CTH/CBS [1 µM MPP+]

Time [h]

mR

NA

le

ve

l

[fo

ld c

ha

nge

re

lative

to

co

ntr

ol

SE

M]

Sampling

d-1 d0 d2 d3 d4 d5

72 h 24 h48 h

ReplateStart diff.

A

C D E

BMPP+ addition

d0 d2 d60

15

30

45

60

75

90 5 µM MPP+

MP

P+

up

take

[pm

ol/1

06 c

ells

/30

min S

D]

*

*

*

**

*

* *

*

d8

24 h

d5

24 h

d5

d5 .

ATF 4

GAPDH

50

kDa

0 24 7248

1 µM MPP+

Time [h]: 0 24 7248

5 µM MPP+

F G H

d5

MPP+ [1 µM]

* *

Concluding discussion

150

F. Concluding discussion

This thesis contains two publications and one submitted manuscript that all discuss their

individual findings in the chapters C, D and E. The following paragraph should therefore

provide a concluding discussion to summarize the major achievements of this thesis and to

discuss general aspects that need to be considered for the evaluation of human-based

alternative test systems.

Concepts of toxicity testing

Two different approaches are basically undertaken to add to the paradigm shift in

toxicology, away from animal testing towards more relevant human-based test systems. One

approach is to identify the underlining mechanism of chemical toxicity.

Different concepts exist:

1. Pathways of toxicity (PoT): A concept with the aim to uncover the human

toxome. This concept mainly focusses on networks that build the cellular

homeostasis and, once disturbed, lead to a different cell fate (Hartung & McBride

2011).

2. Adverse outcome pathways: The OECD has introduced the “adverse outcome

pathways”, whereby an adverse outcome is directly linked to a chemicals

molecular initiating event (http://www.oecd.org/env/ehs/testing/49963554.pdf

and (Ankley et al 2010)).

3. Biomarkers of toxicity: An important tool to understand the mechanism of

chemical toxicity and to extrapolate in vitro data to the in vivo situation are the

biomarkers of toxicity (Blaauboer et al 2012).

The second approach is based on the assessment of phenotypic anchor points, to generate

test systems for toxicity outcomes, which cannot or can hardly be observed in animal

experiments. A newly proposed concept introduced the term “toxicity endophenotypes”,

which is based on test systems that focus on biological processes. Those can be modeled in

vitro in contrast to final phenotypes, like mental retardation, that, in most cases, can barely be

directly assessed (Kadereit et al 2012). Many test systems are thereby capable to screen

several compounds, which is important for the large number of chemicals (Rovida & Hartung

2009) which can hardly be screened by animal experiments. These test systems should

provide less complex, less expensive, and faster assays to prioritize which chemicals should

http://www.oecd.org/env/ehs/testing/49963554.pdf


151

be subjected first to more complex, expensive, and slower guideline assays (Judson et al

2013).

How to evaluate alternative test systems - neurite growth as DNT-specific

endpoint

Several documents are available, that highlight the main criteria a test system has to

fulfil. They describe, for example, good cell culture practice (GCCP) (Coecke et al 2005,

Hartung et al 2002), basic requirements for a test system (Crofton et al 2011, Leist et al 2010,

Leist et al 2013, Leist et al 2012b), the validation of those (Hartung 2007, Hartung et al 2013,

Judson et al 2013) and aim to provide guidelines for a formal process to evaluate the

reliability, relevance, and fitness for purpose of the test systems (Judson et al 2013). Several

requirements have to be met to use an alternative test system in prioritization screenings.

About ten years ago those requirements have been written down as seven modules (Hartung et

al 2004), which recently have been revisited to propose these modules as guidelines for test

system development to streamline the applicability of new tests in prioritization (Judson et al

2013). In the first publication resulting from this thesis, we evaluated an existing neurite

growth assay by challenging it with a broad spectrum of chemicals. By means of the seven

modules, the introduced alternative test system is once more discussed in a broader context, to

underline its suitability for prioritization screenings.

1. Test definitions

a. Test protocol and SOPs

A very precise test protocol has been published recently (Stiegler et al 2011) and explains

in detail the assay as well as software/algorithm settings. A transfer of the presented assay

onto other biological systems or other laboratories may therefore be possible. The basic

principle of the assay is dependent on a life-cell staining (calcein-AM) and a DNA staining by

Hoechst. On the basis of Hoechst-positive nuclei neuronal somata are subtracted from the

images and the remaining calcein positive pixels are counted as neurite (overgrown) area. A

second analysis counts all double-positive nuclei as viable cells. The assay was evaluated on a

96well plate format.

b. Definition of positive and negative controls

Pathways known to control neurite growth are manifold. Several pathway inhibitors as

well as environmental chemicals are known to inhibit neurite growth in vitro and in vivo and

have extensively been studied in the introduced test system. A positive compound in the


152

system is defined as an altered growth process, an inhibition or an acceleration, without cell

death induction. Negative compounds do not interfere with the growth process. They are

different from compounds, which interfere with neurite growth and viability to the same

potency. Those are unspecific cytotoxic compounds.

c. Definition of endpoint

Two endpoint classes exist. Endpoints, which describe the biological system and

endpoints that describe the behavior of the test in the presence of chemicals (Leist et al

2012b). The biological system, the LUHMES cells, have intensively been characterized in the

past by means of differentiation characteristics, morphological changes and functional

readouts (Lotharius et al 2005, Schildknecht et al 2013, Schildknecht et al 2009, Stiegler et al

2011). LUHMES provide a homogeneous and easy-to-control biological system. The

endpoint chosen for toxicological testing, neurite growth, has also been characterized

intensively by assessing the growth over time to identify the perfect time window for

treatment, as well as the impact of cell density on the growth process to reveal the best density

to measure the growth (Stiegler et al 2011).

d. Definition of prediction model and data interpretation procedure

The data interpretation takes place on three levels. First, the percentage of growth

inhibition/acceleration in comparison to untreated control is determined. Second, cell death is

assessed in parallel on the same cells with two different endpoints (resazurin reduction,

calcein-positive cells). Third, both endpoints (viability and neurite growth) are directly

compared to each other as testing is done in a concentration-response manner and the ratio of

the potency-values EC50 is calculated.

e. Explanation of mechanistic basis

Neurite growth is precondition to build a complex neurite network that is characteristic

for the highly developed mammalian nervous system. Several intrinsic (e.g. expression of

receptors) as well as extrinsic factors are important, such as protein kinase C (PKC; (Larsson

2006)), mitogen-activated protein kinases (MAPK; (Schmid et al 2000)), Rho-associated

protein kinase (ROCK; (Kubo et al 2008, Nikolic 2002)) or Akt-signaling (Read & Gorman

2009) and interaction of the differentiating cells with components of the extracellular matrix

(ECM). Actin as well as microtubuli reorganization direct the growth of the neurites. Several

genes, linked to neurite growth and guidance are candidate genes for the development of

autism spectrum disorders (Hussman et al 2011).


153

f. Statement of known limitations, e.g., metabolic capacity

The assay is based on calcein-AM staining, whereby only living cells become

fluorescent. Calcein-AM is cleaved by esterases in the cells, a process, which could be

inhibited by chemicals and therefore interfere with the read-out. The use of GFP- or RFP-

tagged LUHMES will avoid this issue, and their suitability for the assay has recently been

shown (Schildknecht et al 2013, Stiegler et al 2011). Nevertheless, chemicals could

themselves be fluorescent and still interfere with the detection. Biological limitations are the

lack of metabolism, and protection or intensification of toxicity by other cells, such as glia

cells, is not assessed.

g. Training set of chemicals

In the first training phase of the assay a set of chemicals has been used according to the

compound selection criteria for DNT (Kadereit et al 2012). Positive controls, such as U0126

(MAPK inhibitor), bisindolylmaleimide I (PKC inhibitor), Na3VO4 or brefeldin A resulted in

a strong inhibition of growth at concentration without cell death induction, whereas Y-27632

(ROCK inhibitor) resulted in an acceleration. Negative controls (mannitol or acetylsalicylic

acid) did not alter the growth process. Several general cytotoxic compounds (e.g. etoposide or

SDS) affected both endpoints (neurite growth and viability) to similar extends, with an EC50

ratio < 2. In the follow-up study, presented in chapter B, a large number of reference

chemicals (over 50) has been tested. Those chemicals were used to precisely describe the

assay by means of accuracy, precision, detection limits, robustness, specificity and sensitivity

as well as the dynamic range (Leist et al 2013).

h. Provisional domain of applicability

As the OECD guideline 426 for DNT testing is time-consuming and very elaborate,

several alternative test systems are being developed. The current aim is to use the assay in

context with other DNT relevant test systems to prioritize first-in-line chemicals, which have

to be run in guided tests to generate final data for safety decisions.

2. Within-laboratory variability (reliability)

a. Assessment of reproducibility of experimental data in same laboratory – different

operators and different times

More than 10 people performed the assay and used a certain set of assay-control-

chemicals to compare performances with each other. A high reproducibility was achieved, as


154

the cells as well as the assay are easy to handle. Also different cell batches and passages were

compared and resulted in similar outcomes.

3. Transferability (reliability)

a. Assessment of reproducibility of experimental data in second laboratory (different

operator)

Until now, no second laboratory performed the assay yet. This is partly due to missing

equipment in other laboratories (such as the automated fluorescence microscope). Therefore

the following requirements of

4. “Ease of transferability“ and “Between-laboratory variability (reliability) – Assessment of

reproducibility of experimental data in 2-4 laboratories” could not be verified.

5. Predictive capacity (relevance)

a. Assessment of predictive capacity of the prediction model associated with the test system

using a set of test chemicals as opposed to the training chemicals

The accuracy, as mentioned above, was determined by using a reference set of chemicals,

known to interfere with the growth process. In addition to pathway inhibitors, several

pesticides and cancer agents were confirmed in the assay.

6. Applicability domain (relevance)

a. Definition of chemical classes and/or ranges of test method endpoints for which the

model makes reliable predictions

As discussed by Judson and colleagues (Judson et al 2013), it is difficult to make any

assumptions on which chemical classes will be detected and which not, as only a smaller set

(in comparison to real high-throughput studies) of reference chemicals was evaluated. The

current experience with the assay permits the statement that chemicals, which interfere with

microtubule polymerization, pesticides, which result in increased reactive oxygen species and

drugs, which interfere with common neurite growth pathways, such as ROCK inhibitors, were

classified as positive compounds. All of these results were confirmed by literature mining.

7. Performance standards

a. Definition of reference chemicals that can be used to demonstrate the equivalence in

performance between a new test and a previously validated test

Several neurite growth assays are available. The most similar assays are provided by

Mundy and colleagues (Harrill et al 2010, Harrill et al 2011a, Harrill et al 2013, Radio et al


155

2010) and training compounds, such as U0126 and Na3VO4 resulted also in the inhibition of

neurite growth. Advantage of the here introduced assay is the cytotoxicity assessment on the

same cells, the human-based biological system, as well as the low variation between

experiments, which could not be achieved in test systems presented by Mundy et al. A clear

separation of neurite growth modulators from unspecific cytotoxic compounds was possible

and chemically related toxicants resulted in the same output.

The provided assay may therefore be used in prioritization screenings for DNT testing, as

it is easy to handle, relatively fast in performance and translatable to robotic systems. Also

other plate formats, such as 384 well plates, were tested successfully. The combination of this

DNT-specific assay together with other DNT-related test systems provides a powerful

alternative to assess DNT effects (such as mental retardation) by assessing basic biological

processes.

Stem cell-based early recapitulation of neuronal development in vitro –

transcriptomics

As mentioned before, an alternative to add to the toxicology paradigm shift is to reveal

the mechanism of toxicity of the tested chemicals. In the second publication of this thesis we

therefore evaluated the relevance of transcriptomics-based toxicity assessment for DNT

prediction.

Five different hESC-based in vitro systems, which recapitulate different stages of early

neural development, were investigated. The normal transcriptional changes during

differentiation are thought to reproduce normal human tissue differentiation (Carri et al 2013),

and could also be observed for murine ESC and murine embryonic development in vivo

(Abranches et al 2009, Barberi et al 2003, Zimmer et al 2011a). If the normal expression

pattern is disturbed, it could lead to altered proportions of cell types within each system,

which should be identifiable with the transcriptomics approach. In our case we used DNA

mircoarrays. The study was meant to gain experience on two levels. On the one hand it should

be determined whether known DNT compounds, VPA and MeHg, would result in altered

expression patterns. On the other hand we obtained data of 169 microarrays with 54 575

probe sets each and wanted to provide a basic concept of how to deal with that many data.

By treating the cells with non-cytotoxic concentrations of VPA and MeHg, known

human DNT chemicals, several observations were made. First of all, VPA, an antiepileptic

drug leading to the fetal valproate syndrome in children exposed to it in utero, which may


156

manifest itself for example in spina bifida, or autism-spectrum symptoms (Bromley et al

2009, Jentink et al 2010, McVearry et al 2009), resulted in strong altered expression patterns

in all systems in which the drug was tested. MeHg on the other hand, which is also known for

its developmental neurotoxicity (Castoldi et al 2008a, Castoldi et al 2008b) resulted in

significantly fewer transcript changes. Those expression differences were expected, as VPA is

a known histone deacetylase inhibitor, interfering directly with transcription. Whereas MeHg,

on the other side, acts through unspecific protein modifications (Aschner et al 2007) and a

weaker effect was not astonishing. Negative controls were included in the study and did not

result in any changes. Also the overlap of the changed transcripts of both chemicals within a

test system and in comparison with other systems was very small. Therefore the observed

effects of the DNT chemicals appear to be compound and test system specific. Surprisingly,

transcription factor binding sites (TFBS) in the promoter region of the changed transcripts

overlapped strongly for a chemical between different test systems and for both chemicals

within a test system. Based on this, the hypothesis was generated that TFBS which did not

overlap for both chemicals may be used as signatures of toxicity (SoTs) to group with other

related chemicals. Those TFBS which did overlap between the two chemicals may be used as

classifier for general toxicity. Recently clusters of TFBS in so called super-enhancer regions,

associated with genes that control and define cell identity, have been identified (Hnisz et al

2013). Chemical-induced changes of cell identities could therefore possibly be due to changes

in master transcription factors associated for example with super-enhancer regions. Hence, the

analysis of TFBS may be an important tool for toxicity assessment.

The biological data presented above was gained by handling the huge amount of data

very carefully. First of all, we were confronted with the impressive impact of false discovery

rate (FDR) correction on the number of regulated transcripts. For instance, out of initially

10985 significantly regulated probe sets by MeHg for one test system only 419 remained after

FDR correction. In another system, only two probe sets out of 8657 remained.

Because 169 microarrays cannot be operated on one day, some outliers were generated

due to batch effects. Two approaches were tested, allowing a data analysis including outliers.

The first approach was to work only with the 500 probe sets with the highest variance. The

effect was visualized in PCAs, as the former outliers (PCA based on all probe sets) clustered

together with their corresponding microarrays (PCA based on 500 probe sets with highest

variance). In a second approach, the corresponding control values were subtracted from the

compound-treated samples and the data was visualized again in a PCA. The outliers clustered


157

now within their group. The last approach we took, was to simulate the impact of reducing

numbers of microarrays (of the different systems) with different permutations, to reveal

whether the common number of replicates, 5, is necessary, or if less microarrays would result

in the same data. Comparing only probe sets with a fold change > 2, the different

permutations for 4 microarrays identified almost the same set of probe sets. Reducing further

to only 3 microarrays, less common probe sets were found, and several new appeared. An

interesting side-effect was, that by this method, also outliers were quite evidently identified,

as the removal of one microarray in one test system resulted in significantly more identified

probe sets. Taken together this study provides basic concepts how one can work with many

data and how hypothesis can be generated by Omics-based approaches.

Defining pathways of toxicity – MPP+ toxicity

In the third part of this thesis,

data complexity was growing even

further. Two Omics technologies,

transcriptomics and metabolomics,

were applied to a model of

neurodegeneration. LUHMES cells

were treated with the neurotoxin

MPP+ and the sequence of changed

events should be studied. Two

questions should be answered by this

approach: Can we confirm existing

data of MPP+ toxicity with the Omics

approaches? Can we identify novel

stress related cellular adaptations?

The molecular initiating event of MPP+ is well known, it inhibits complex I of the

mitochondrial chain reaction. It is also known that this, sooner or later, leads to cell death. We

wanted to reveal the relation of these two events to understand the cellular adaptations until a

point-of-no-return is reached. As mentioned repeatedly during this thesis, cell death

assessment alone would not be sufficient to understand why a cell is dying. Concentrations

and time points have to be identified at which alterations can be observed independent of cell

death induction. A very simplified graph (Fig. 1) exemplifies the time dependent kinetics after

chemical exposure. T1-T4 are different time points and at T4 cell death is induced, e.g.

Figure 1: Illustration of cellular adaptations to

chemical treatment T1-T4 are different time points, A-E are different

factors inside the cell changing upon toxicant

treatment. Orange line indicates the start of the

treatment

Cha

ng

e fro

m b

ase

line

T1 T2 T3 T4

A

B

C

D

E

Time


158

cytochrom c is released from mitochondria (row E). A logical conclusion is that the cells

entered programmed cell death. But it remains elusive what initiated it, and it is difficult to

extrapolate on upstream events. To understand the initial changes, earlier time points should

be included. Combinig the data with bioinformatics, one may extrapolate on downstream

events by, for instance, the identification of a transcription factor (TF), which may activate

apoptotic marker genes, possibly explaining the observed cell death. The time point to

measure should therefore be chosen carefully. In the case of MPP+ toxicity we were especially

interested in the early changes. We included a time point at 24 h after exposure for the Omics

experiments, as we did not expect many transcriptional changes, especially as MPP+ is a

mitochondrial toxin. To our surprise most of the transcriptional changes were already set at

24 h. Based on this, we step-wise included earlier time points and observed very early

transcriptional changes (e.g. as soon as 2 h after exposure) as well as metabolic alterations. At

these early time points, none of our ‘control’ assays, such as ATP, glutathione (GSH) or

apoptotic marker expression indicated any changes. Applying bioinformatic analysis onto our

transcriptomics data, ATF4 was identified as upstream regulator in the system. Although this

transcription factor was revealed on bioinformatic basis, it could be verified in several

experiments and an early up-regulation on protein level was observed, highlighting the

importance of such analyses.

In general, toxicogenomics studies are suitable to strengthen or to generate new

hypothesis of toxicity mechanisms of chemicals, as shown in the second publication of this

thesis. At the same time, SoTs of the chemicals can be obtained (Bouhifd et al 2013, Hartung

et al 2012). Fig. 2A explains the principle: If an analysis would release consonants, instead of

metabolites, one would get a certain signature for a tested chemical. This signature on its own

can be used to group chemicals of similar signature, but it does not necessarily allow putting

the SoTs in a meaningful context, for example the identification of words or sentences. But a

second analysis will identify vowels. These vowels can be combined with the consonants and

with some bioinformatics, networks such as words and sentences become apparent.

The same principle underlies the integration of Omics data (Fig. 2B). Changed

metabolites (consonants) on their own may represent the SoT of a chemical, but a verification

of coherent pathways involved, is difficult.

A combination with second source data, such as transcripts (vowels), now puts the

observations in a meaningful context. Pathways (words) become visible and eventually the


159

sequence of involved pathways may help to identify the complete mechanism (sentence)

behind.

We performed this data integration for the metabolomics and transcriptomics data and

determined several involved pathways on both omic levels, such as the serine metabolism,

folate and methionine metabolism as well as the transsulfuration pathway.

Pathways involved in the changes of cell homeostasis can be manifold, for instance

pathways that actually lead to toxicity (= pathways of toxicity, PoT), pathways that function

as counter-regulation (= pathways of defense, PoD), pathways that are responsible for the

adaptation to the new cell homeostasis or even epiphenomena, pathways which are regulated

by coincidence and have no role in the toxicity mechanism of a chemical (Ramirez et al

2013). In the case of MPP+ toxicity and the identification of the transsulfuration pathway, we

would rather allocate the term PoD to it as PoT, as it is likely a counter-regulation of the cells.

The transsulfuration pathway is involved in the cysteine synthesis, which is the rate-limiting

amino acid for GSH synthesis. As the cells have a high GSH demand, due to the high levels

of oxidative stress, the pathway activation likely contributes to the GSH synthesis. We

verified this hypothesis in follow-up experiments by decreasing MPP+ concentrations slightly

and revealed a strong increase in GSH after the first 24 h, which is followed by a decline,

Figure 2: Principle of integrated omics data

A) Two groups of apparently random consonants on the left (derived from first analysis method)

and vowels in the middle (derived from second analysis method). Right: the same letters, but now

in a context that allows the reader to identify the encoded sentence, a quote from Descartes

(combined data of first and second analysis). The meaningful sequence of consonants is only

apparent when the context with vowels is given. B) Two groups of enriched metabolites on the left

and enriched transcripts in the middle. Right: Pathways analysis arranges the metabolites and

transcripts in a meaningful context, thus the apparently randomly enriched metabolites are now

linked by transcripts involved in their conversion.


160

when concentrations exceed a threshold. Including early time points, such as 8 h after

exposure, an increase in GSH could be observed even for higher concentrations. Presumably,

the observed metabolic and transcriptional changes are triggered by a fast accumulation of

reactive oxygen species, leading to ATF4 activation which itself may trigger the

transsulfuration pathway and the serine synthesis.

Coming back to our initial two questions: Can we confirm existing data of MPP+ toxicity

with the Omics approaches and can we identify novel stress related cellular adaptations – both

can be answered with a yes. Existing data of altered energy metabolism, increase in reactive

oxygen species and the induction of programmed cell death, once ATP and GSH drop steaply,

were confirmed (Dauer & Przedborski 2003, Vila & Przedborski 2003). New implications of

MPP+ toxicity were also revealed. The relation of ATF4 to MPP+ toxicity is known from

literature (Holtz & O'Malley 2003, Lange et al 2008, Sun et al 2013), but the possible

implication of ATF4 in the underlying metabolic changes, such as increase in thiol-

metabolites, cysteine and glycine and the identification of the transsulfuration pathway, was

newly identified. As generated networks out of Omics are scale-free and the varying strength

of interactions remain indefinite (Hartung et al 2012), the identified pathways need to be

verified to understand the sequence of events. In future experiments, the importance of ATF4

Figure 3: Overview of toxicity testing in the 21st century.

Validated organ specific in vitro systems are used for high throughput and/or high content

screenings to assess toxicity. Collected data have to be integrated to reveal underlying pathways

involved in the toxicity phenotype (PoT). Those PoTs have to be confirmed by targeted analysis,

such as knock-down/in of target proteins, overexpression of those or establishment of reporter cell

lines. Once all PoTs have been collected, in silico modeling can be used to predict toxicity

outcomes in humans.

Integrate high-content data from treated states

Identify putative PoTs by bioinformatic

analysis

Confirm PoTs with targeted analysis

X1.

Apply systems toxicology to organ-

specific in vitro systems

Toxicity testing in the 21st century

PoT-based

toxicity

prediction


161

and related factors in the cells has to be tested by, for instance, knock-down experiments of

involved enzymes or ATF4 itself.

The challenge, some toxicologists want to meet, is to identify all critical pathways,

which, taken together, present the complete human toxome (Hartung & McBride 2011,

Ramirez et al 2013). Some projects in that direction are on their way, for example EPA’s

ToxCast project. Over 650 assays were collected in a battery and over 2000 chemicals are

currently tested to prioritize them and to cluster them according to their SoTs (Kleinstreuer et

al 2011, Knudsen et al 2011, Sipes et al 2013, West et al 2010). The human toxicology

consortium, in contrast, promotes the demanded change in toxicity assessment by pushing

regulatory changes towards the PoT approach by taking diverse stakeholders from different

fields on board of the consortium (Stephens et al 2013). The mapping of the human toxome

involves the identification of putative PoTs and the confirmation of them by targeted analysis.

Having established this, the ultra-goal is to generate virtual cells and organs to predict toxicity

in humans in silico (Fig. 3).

Bibliography

162

G. Bibliography

Abranches E, Silva M, Pradier L, Schulz H, Hummel O, et al. 2009. Neural differentiation of embryonic stem

cells in vitro: a road map to neurogenesis in the embryo. PLoS One 4: e6286

Adinolfi M. 1985. The development of the human blood-CSF-brain barrier. Developmental medicine and child

neurology 27: 532-7

Adler S, Basketter D, Creton S, Pelkonen O, van Benthem J, et al. 2011. Alternative (non-animal) methods for

cosmetics testing: current status and future prospects-2010. Archives of toxicology 85: 367-485

Agilent_Technologies. 2005. Multiple-Testing Corrections.

Andersen ME, Clewell HJ, Carmichael PL, Boekelheide K. 2011. Can case study approaches speed

implementation of the NRC report: "toxicity testing in the 21st century: a vision and a strategy?".

ALTEX 28: 175-82

Ankley GT, Bennett RS, Erickson RJ, Hoff DJ, Hornung MW, et al. 2010. Adverse outcome pathways: a

conceptual framework to support ecotoxicology research and risk assessment. Environ Toxicol Chem

29: 730-41

Ansher SS, Cadet JL, Jakoby WB, Baker JK. 1986. Role of N-methyltransferases in the neurotoxicity associated

with the metabolites of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) and other 4-substituted

pyridines present in the environment. Biochem Pharmacol 35: 3359-63

Aschner M, Syversen T, Souza DO, Rocha JB, Farina M. 2007. Involvement of glutamate and reactive oxygen

species in methylmercury neurotoxicity. Braz J Med Biol Res 40: 285-91

Attarwala H. 2010. TGN1412: From Discovery to Disaster. Journal of young pharmacists : JYP 2: 332-6

Attene-Ramos MS, Huang R, Sakamuru S, Witt KL, Beeson GC, et al. 2013. Systematic Study of Mitochondrial

Toxicity of Environmental Chemicals Using Quantitative High Throughput Screening. Chemical

research in toxicology

Auer H, Newsom DL, Kornacker K. 2009. Expression Profiling Using Affymetrix GeneChip Microarrays.

Methods in molecular biology 509: 35-46

Baker M. 2013. Big biology: The 'omes puzzle. Nature 494: 416-9

Bakir F, Damluji SF, Amin-Zaki L, Murtadha M, Khalidi A, et al. 1973. Methylmercury poisoning in Iraq.

Science 181: 230-41

Bal-Price AK, Coecke S, Costa L, Crofton KM, Fritsche E, et al. 2012. Advancing the science of developmental

neurotoxicity (DNT): testing for better safety evaluation. Altex 29: 202-15

Bal-Price AK, Hogberg HT, Buzanska L, Coecke S. 2010. Relevance of in vitro neurotoxicity testing for

regulatory requirements: challenges to be considered. Neurotoxicology and teratology 32: 36-41

Bal-Price AK, Sunol C, Weiss DG, van Vliet E, Westerink RH, Costa LG. 2008. Application of in vitro

neurotoxicity testing for regulatory purposes: Symposium III summary and research needs.

Neurotoxicology 29: 520-31

Balmer NV, Weng MK, Zimmer B, Ivanova VN, Chambers SM, et al. 2012. Epigenetic changes and disturbed

neural development in a human embryonic stem cell-based model relating to the fetal valproate

syndrome. Hum Mol Genet 21: 4104-14

Barakat-Walter I, Kraftsik R, Kuntzer T, Bogousslavsky J, Magistretti P. 2000. Differential effect of thyroid

hormone deficiency on the growth of calretinin-expressing neurons in rat spinal cord and dorsal root

ganglia. J Comp Neurol 426: 519-33

Barberi T, Klivenyi P, Calingasan NY, Lee H, Kawamata H, et al. 2003. Neural subtype specification of

fertilization and nuclear transfer embryonic stem cells and application in parkinsonian mice. Nat

Biotechnol 21: 1200-7

Basketter DA, Clewell H, Kimber I, Rossi A, Blaauboer B, et al. 2012. A roadmap for the development of

alternative (non-animal) methods for systemic toxicity testing - t4 report*. ALTEX 29: 3-91

Bauwens CL, Peerani R, Niebruegge S, Woodhouse KA, Kumacheva E, et al. 2008. Control of human

embryonic stem cell colony and aggregate size heterogeneity influences differentiation trajectories.

Stem cells 26: 2300-10

Benjamini Y, Hochberg Y. 1995. Controlling the False Discovery Rate - a Practical and Powerful Approach to

Multiple Testing. J Roy Stat Soc B Met 57: 289-300

Berry MP, Graham CM, McNab FW, Xu Z, Bloch SA, et al. 2010. An interferon-inducible neutrophil-driven

blood transcriptional signature in human tuberculosis. Nature 466: 973-7

Berthold MR, Cebron N, Dill F, Gabriel TR, Kotter T, et al. 2008. KNIME: The Konstanz Information Miner.

Stud Class Data Anal: 319-26

Bezard E, Przedborski S. 2011. A tale on animal models of Parkinson's disease. Mov Disord 26: 993-1002

Binkerd PE, Rowland JM, Nau H, Hendrickx AG. 1988. Evaluation of valproic acid (VPA) developmental

toxicity and pharmacokinetics in Sprague-Dawley rats. Fundam Appl Toxicol 11: 485-93

Bibliography

163

Blaauboer BJ, Boekelheide K, Clewell HJ, Daneshian M, Dingemans MM, et al. 2012. The use of biomarkers of

toxicity for integrating in vitro hazard estimates into risk assessment for humans. ALTEX 29: 411-25

Bornhausen M, Musch HR, Greim H. 1980. Operant behavior performance changes in rats after prenatal

methylmercury exposure. Toxicol Appl Pharmacol 56: 305-10

Bouhifd M, Hartung T, Hogberg HT, Kleensang A, Zhao L. 2013. Review: Toxicometabolomics. Journal of

applied toxicology : JAT

Breier JM, Radio NM, Mundy WR, Shafer TJ. 2008. Development of a high-throughput screening assay for

chemical effects on proliferation and viability of immortalized human neural progenitor cells. Toxicol

Sci 105: 119-33

Brinkman K, ter Hofstede HJ, Burger DM, Smeitink JA, Koopmans PP. 1998. Adverse effects of reverse

transcriptase inhibitors: mitochondrial toxicity as common pathway. AIDS 12: 1735-44

Bromley RL, Baker GA, Meador KJ. 2009. Cognitive abilities and behaviour of children exposed to antiepileptic

drugs in utero. Curr Opin Neurol 22: 162-6

Bundy JG, Sidhu JK, Rana F, Spurgeon DJ, Svendsen C, et al. 2008. 'Systems toxicology' approach identifies

coordinated metabolic responses to copper in a terrestrial non-model invertebrate, the earthworm

Lumbricus rubellus. BMC Biol 6: 25

Burte F, De Girolamo LA, Hargreaves AJ, Billett EE. 2011. Alterations in the mitochondrial proteome of

neuroblastoma cells in response to complex 1 inhibition. J Proteome Res 10: 1974-86

Caie PD, Walls RE, Ingleston-Orme A, Daya S, Houslay T, et al. 2010. High-content phenotypic profiling of

drug response signatures across distinct cancer cells. Mol Cancer Ther 9: 1913-26

Canter JA, Robbins GK, Selph D, Clifford DB, Kallianpur AR, et al. 2010. African mitochondrial DNA

subhaplogroups and peripheral neuropathy during antiretroviral therapy. J Infect Dis 201: 1703-7

Carreras Puigvert J, von Stechow L, Siddappa R, Pines A, Bahjat M, et al. 2013. Systems biology approach

identifies the kinase Csnk1a1 as a regulator of the DNA damage response in embryonic stem cells. Sci

Signal 6: ra5

Carri AD, Onorati M, Lelos MJ, Castiglioni V, Faedo A, et al. 2013. Developmentally coordinated extrinsic

signals drive human pluripotent stem cell differentiation toward authentic DARPP-32+ medium-sized

spiny neurons. Development 140: 301-12

Carrier G, Brunet RC, Caza M, Bouchard M. 2001. A toxicokinetic model for predicting the tissue distribution

and elimination of organic and inorganic mercury following exposure to methyl mercury in animals and

humans. I. Development and validation of the model using experimental data in rats. Toxicol Appl

Pharmacol 171: 38-49

Cartelli D, Ronchi C, Maggioni MG, Rodighiero S, Giavini E, Cappelletti G. 2010. Microtubule dysfunction

precedes transport impairment and mitochondria damage in MPP+ -induced neurodegeneration. Journal

of neurochemistry 115: 247-58

Carvalho PC, Fischer JS, Chen EI, Domont GB, Carvalho MG, et al. 2009. GO Explorer: A gene-ontology tool

to aid in the interpretation of shotgun proteomics data. Proteome science 7: 6

Castoldi AF, Johansson C, Onishchenko N, Coccini T, Roda E, et al. 2008a. Human developmental

neurotoxicity of methylmercury: impact of variables and risk modifiers. Regul Toxicol Pharmacol 51:

201-14

Castoldi AF, Onishchenko N, Johansson C, Coccini T, Roda E, et al. 2008b. Neurodevelopmental toxicity of

methylmercury: Laboratory animal data and their contribution to human risk assessment. Regul Toxicol


Cavalli-Sforza LL. 2005. The Human Genome Diversity Project: past, present and future. Nature reviews.

Genetics 6: 333-40

Cha MH, Rhim T, Kim KH, Jang AS, Paik YK, Park CS. 2007. Proteomic identification of macrophage

migration-inhibitory factor upon exposure to TiO2 particles. Molecular & cellular proteomics : MCP 6:

56-63

Chambers SM, Fasano CA, Papapetrou EP, Tomishima M, Sadelain M, Studer L. 2009. Highly efficient neural

conversion of human ES and iPS cells by dual inhibition of SMAD signaling. Nat Biotechnol 27: 275-

80

Chen PS, Wang CC, Bortner CD, Peng GS, Wu X, et al. 2007. Valproic acid and other histone deacetylase

inhibitors induce microglial apoptosis and attenuate lipopolysaccharide-induced dopaminergic

neurotoxicity. Neuroscience 149: 203-12

Choi BH. 1989. The effects of methylmercury on the developing brain. Prog Neurobiol 32: 447-70

Choi WS, Kruse SE, Palmiter RD, Xia Z. 2008. Mitochondrial complex I inhibition is not required for

dopaminergic neuron death induced by rotenone, MPP+, or paraquat. Proc Natl Acad Sci U S A 105:

15136-41

Choucha Snouber L, Bunescu A, Naudot M, Legallais C, Brochot C, et al. 2013. Metabolomics-on-a-chip of

hepatotoxicity induced by anticancer drug flutamide and Its active metabolite hydroxyflutamide using

Bibliography

164

HepG2/C3a microfluidic biochips. Toxicological sciences : an official journal of the Society of

Toxicology 132: 8-20

Chow S, Rodgers P. 2005. Extended Abstract: Constructing Area-Proportional Venn and Euler Diagrams with

Three Circles. Presented at Euler Diagrams Workshop 2005, Paris

Clancy B, Finlay BL, Darlington RB, Anand KJ. 2007. Extrapolating brain development from experimental

species to humans. Neurotoxicology 28: 931-7

Clothier RH, Hulme LM, Smith M, Balls M. 1987. Comparison of the in vitro cytotoxicities and acute in vivo

toxicities of 59 chemicals. Molecular toxicology 1: 571-7

Coecke S, Balls M, Bowe G, Davis J, Gstraunthaler G, et al. 2005. Guidance on good cell culture practice. a

report of the second ECVAM task force on good cell culture practice. Altern Lab Anim 33: 261-87

Colborn T, vom Saal FS, Soto AM. 1993. Developmental effects of endocrine-disrupting chemicals in wildlife

and humans. Environ Health Perspect 101: 378-84

Collins FS, Gray GM, Bucher JR. 2008. Toxicology. Transforming environmental health protection. Science

319: 906-7

Cornelissen F, Verstraelen P, Verbeke T, Pintelon I, Timmermans JP, et al. 2013. Quantitation of chronic and

acute treatment effects on neuronal network activity using image and signal analysis: toward a high-

content assay. J Biomol Screen 18: 807-19

Correia JJ, Lobert S. 2001. Physiochemical aspects of tubulin-interacting antimitotic drugs. Curr Pharm Des 7:

1213-28

Corvi R, Aardema MJ, Gribaldo L, Hayashi M, Hoffmann S, et al. 2012. ECVAM prevalidation study on in vitro

cell transformation assays: general outline and conclusions of the study. Mutat Res 744: 12-9

Crofton KM, Mundy WR, Lein PJ, Bal-Price A, Coecke S, et al. 2011. Developmental neurotoxicity testing:

recommendations for developing alternative methods for the screening and prioritization of chemicals.

ALTEX 28: 9-15

Culbreth ME, Harrill JA, Freudenrich TM, Mundy WR, Shafer TJ. 2012. Comparison of chemical-induced

changes in proliferation and apoptosis in human and mouse neuroprogenitor cells. Neurotoxicology 33:

1499-510

Cuperlovic-Culf M, Barnett DA, Culf AS, Chute I. 2010. Cell culture metabolomics: applications and future

directions. Drug Discov Today 15: 610-21

Dalma-Weiszhausz DD, Warrington J, Tanimoto EY, Miyada CG. 2006. The affymetrix GeneChip platform: an

overview. Methods in enzymology 410: 3-28

Daniels MP. 1972. Colchicine inhibition of nerve fiber formation in vitro. J Cell Biol 53: 164-76

Dauer W, Przedborski S. 2003. Parkinson's disease: mechanisms and models. Neuron 39: 889-909

Davidson PW, Myers GJ, Weiss B. 2004. Mercury exposure and child development outcomes. Pediatrics 113:

1023-9

Diaz G, Liu SS, Isola R, Diana A, Falchi AM. 2003. Mitochondrial localization of reactive oxygen species by

dihydrofluorescein probes. Histochemistry and cell biology 120: 319-25

Doktorova TY, Yildirimman R, Vinken M, Vilardell M, Vanhaecke T, et al. 2013. Transcriptomic responses

generated by hepatocarcinogens in a battery of liver-based in vitro models. Carcinogenesis 34: 1393-

402

Donato MT, Tolosa L, Jimenez N, Castell JV, Gomez-Lechon MJ. 2012. High-content imaging technology for

the evaluation of drug-induced steatosis using a multiparametric cell-based assay. J Biomol Screen 17:

394-400

Doostzadeh J, Davis RW, Giaever GN, Nislow C, Langston JW. 2007. Chemical genomic profiling for

identifying intracellular targets of toxicants producing Parkinson's disease. Toxicol Sci 95: 182-7

Duncan R. 2004. DNA microarray analysis of protozoan parasite gene expression: outcomes correlate with

mechanisms of regulation. Trends in parasitology 20: 211-5

Dunkler D, Sanchez-Cabo F, Heinze G. 2011. Statistical analysis principles for Omics data. Methods in

molecular biology 719: 113-31

Ekino S, Susa M, Ninomiya T, Imamura K, Kitamura T. 2007. Minamata disease revisited: an update on the

acute and chronic manifestations of methyl mercury poisoning. J Neurol Sci 262: 131-44

Ekwall B, Barile FA, Castano A, Clemedson C, Clothier RH, et al. 1998a. MEIC evaluation of acute systemic

toxicity - Part VI. The prediction of human toxicity by rodent LD50 values and results from 61 in vitro

methods. Atla-Altern Lab Anim 26: 617-58

Ekwall E, Clemedson C, Crafoord B, Ekwall B, Hallander S, et al. 1998b. MEIC evaluation of acute systemic

toxicity - Part V. Rodent and human toxicity data for the 50 reference chemicals. Atla-Altern Lab Anim

26: 571-616

Elkon R, Linhart C, Sharan R, Shamir R, Shiloh Y. 2003. Genome-wide in silico identification of transcriptional

regulators controlling the cell cycle in human cells. Genome Res 13: 773-80

Bibliography

165

Ellis JK, Athersuch TJ, Cavill R, Radford R, Slattery C, et al. 2011. Metabolic response to low-level toxicant

exposure in a novel renal tubule epithelial cell system. Molecular bioSystems 7: 247-57

Evans C, Noirel J, Ow SY, Salim M, Pereira-Medrano AG, et al. 2012. An insight into iTRAQ: where do we

stand now? Analytical and bioanalytical chemistry 404: 1011-27

Faiz H, Conjard-Duplany A, Boghossian M, Martin G, Baverel G, Ferrier B. 2011. Cadmium chloride inhibits

lactate gluconeogenesis in isolated human renal proximal tubules: a cellular metabolomic approach with

13C-NMR. Archives of toxicology 85: 1067-77

Farhan H, Reiterer V, Kriz A, Hauri HP, Pavelka M, et al. 2008. Signal-dependent export of GABA transporter 1

from the ER-Golgi intermediate compartment is specified by a C-terminal motif. J Cell Sci 121: 753-61

Fontaine-Lenoir V, Chambraud B, Fellous A, David S, Duchossoy Y, et al. 2006. Microtubule-associated protein

2 (MAP2) is a neurosteroid receptor. Proc Natl Acad Sci U S A 103: 4711-6

Forsby A, Bal-Price AK, Camins A, Coecke S, Fabre N, et al. 2009. Neuronal in vitro models for the estimation

of acute systemic toxicity. Toxicol In Vitro 23: 1564-9

Forsby A, Blaauboer B. 2007. Integration of in vitro neurotoxicity data with biokinetic modelling for the

estimation of in vivo neurotoxicity. Hum Exp Toxicol 26: 333-8

Fournier AE, Takizawa BT, Strittmatter SM. 2003. Rho kinase inhibition enhances axonal regeneration in the

injured CNS. J Neurosci 23: 1416-23

Frearson JA, Collie IT. 2009. HTS and hit finding in academia--from chemical genomics to drug discovery.

Drug Discov Today 14: 1150-8

Freeman BA, Crapo JD. 1982. Free-Radicals and Tissue-Injury. Laboratory Investigation 47: 412-26

Frimat JP, Sisnaiske J, Subbiah S, Menne H, Godoy P, et al. 2010. The network formation assay: a spatially

standardized neurite outgrowth analytical display for neurotoxicity screening. Lab Chip 10: 701-9

Fritsche E, Gassmann K, Schreiber T. 2011. Neurospheres as a model for developmental neurotoxicity testing.

Methods in molecular biology 758: 99-114

Fuentes EO, Leemhuis J, Stark GB, Lang EM. 2008. Rho kinase inhibitors Y27632 and H1152 augment neurite

extension in the presence of cultured Schwann cells. J Brachial Plex Peripher Nerve Inj 3: 19

Gartlon J, Kinsner A, Bal-Price A, Coecke S, Clothier RH. 2006. Evaluation of a proposed in vitro test strategy

using neuronal and non-neuronal cell systems for detecting neurotoxicity. Toxicol In Vitro 20: 1569-81

Gaspar JA, Doss MX, Winkler J, Wagh V, Hescheler J, et al. 2012. Gene expression signatures defining

fundamental biological processes in pluripotent, early, and late differentiated embryonic stem cells.

Stem Cells Dev 21: 2471-84

Gautier L, Cope L, Bolstad BM, Irizarry RA. 2004. affy--analysis of Affymetrix GeneChip data at the probe

level. Bioinformatics 20: 307-15

Geenen S, Taylor PN, Snoep JL, Wilson ID, Kenna JG, Westerhoff HV. 2012. Systems biology tools for

toxicology. Arch Toxicol 86: 1251-71

Geldof AA, Minneboo A, Heimans JJ. 1998. Vinca-alkaloid neurotoxicity measured using an in vitro model. J

Neurooncol 37: 109-13

Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, et al. 2004. Bioconductor: open software

development for computational biology and bioinformatics. Genome Biol 5: R80

Gilley J, Coleman MP. 2010. Endogenous Nmnat2 is an essential survival factor for maintenance of healthy

axons. PLoS Biol 8: e1000300

Glatstein M, Danino D, Wolyniez I, Scolnik D. 2013. Seizures Caused by Ingestion of Atropa Belladonna in a

Homeopathic Medicine in a Previously Well Infant: Case Report and Review of the Literature.

American journal of therapeutics

Goni R, Carcía P, Foissac S. 2009. The qPCR data statistical analysis. Integromics SL: 1-9

Gorovoy M, Niu J, Bernard O, Profirovic J, Minshall R, et al. 2005. LIM kinase 1 coordinates microtubule

stability and actin polymerization in human endothelial cells. J Biol Chem 280: 26533-42

Gough AH, Johnston PA. 2007. Requirements, features, and performance of high content screening platforms.

Methods Mol Biol 356: 41-61

Gough W, Hulkower KI, Lynch R, McGlynn P, Uhlik M, et al. 2011. A Quantitative, Facile, and High-

Throughput Image-Based Cell Migration Method Is a Robust Alternative to the Scratch Assay. Journal

of biomolecular screening 16: 155-63

Grandjean P, Herz KT. 2011. Methylmercury and brain development: imprecision and underestimation of

developmental neurotoxicity in humans. Mt Sinai J Med 78: 107-18

Grandjean P, Landrigan PJ. 2006. Developmental neurotoxicity of industrial chemicals. Lancet 368: 2167-78

Gray DG. 1995. A physiologically based pharmacokinetic model for methyl mercury in the pregnant rat and

fetus. Toxicol Appl Pharmacol 132: 91-102

Griesinger C, Barroso J, Zuang V, Cole T, Genschow E, Liebsch M. 2010. Explanatory Background Document

to the OECD Draft Test Guideline on in vitro Skin Irritation Testing., ed. ECVAM

Bibliography

166

Griesinger C, Hoffmann S, Kinsner A, Coecke S, Hartung T. 2009. Special issue: Evidence-based toxicology

(EBT). Preface. Hum Exp Toxicol 28: 83-6

Grimsey NL, Moodley KS, Glass M, Graham ES. 2012. Sensitive and accurate quantification of human

leukocyte migration using high-content Discovery-1 imaging system and ATPlite assay. J Biomol

Screen 17: 386-93

Grimsey NL, Narayan PJ, Dragunow M, Glass M. 2008. A novel high-throughput assay for the quantitative

assessment of receptor trafficking. Clinical and experimental pharmacology & physiology 35: 1377-82

Gulden M, Seibert H. 2003. In vitro-in vivo extrapolation: estimation of human serum concentrations of

chemicals equivalent to cytotoxic concentrations in vitro. Toxicology 189: 211-22

Halle W. 2003. The Registry of Cytotoxicity: toxicity testing in cell cultures to predict acute toxicity (LD50) and

to reduce testing in animals. Altern Lab Anim 31: 89-198

Hansson O, Castilho RF, Kaminski Schierle GS, Karlsson J, Nicotera P, et al. 2000. Additive effects of caspase

inhibitor and lazaroid on the survival of transplanted rat and human embryonic dopamine neurons. Exp

Neurol 164: 102-11

Harada M. 1995. Minamata disease: methylmercury poisoning in Japan caused by environmental pollution. Crit

Rev Toxicol 25: 1-24

Harbron C, Chang KM, South MC. 2007. RefPlus: an R package extending the RMA Algorithm. Bioinformatics

23: 2493-4

Harrill JA, Freudenrich TM, Machacek DW, Stice SL, Mundy WR. 2010. Quantitative assessment of neurite

outgrowth in human embryonic stem cell-derived hN2 cells using automated high-content image

analysis. Neurotoxicology 31: 277-90

Harrill JA, Freudenrich TM, Robinette BL, Mundy WR. 2011a. Comparative sensitivity of human and rat neural

cultures to chemical-induced inhibition of neurite outgrowth. Toxicol Appl Pharmacol 256: 268-80

Harrill JA, Robinette BL, Freudenrich T, Mundy WR. 2013. Use of high content image analyses to detect

chemical-mediated effects on neurite sub-populations in primary rat cortical neurons. Neurotoxicology

34: 61-73

Harrill JA, Robinette BL, Mundy WR. 2011b. Use of high content image analysis to detect chemical-induced

changes in synaptogenesis in vitro. Toxicol In Vitro 25: 368-87

Hartung T. 2007. Food for thought ... on validation. ALTEX 24: 67-80

Hartung T. 2009. Toxicology for the twenty-first century. Nature 460: 208-12

Hartung T. 2010. Evidence-based toxicology - the toolbox of validation for the 21st century? ALTEX 27: 253-63

Hartung T, Balls M, Bardouille C, Blanck O, Coecke S, et al. 2002. Good Cell Culture Practice. ECVAM Good

Cell Culture Practice Task Force Report 1. Altern Lab Anim 30: 407-14

Hartung T, Bremer S, Casati S, Coecke S, Corvi R, et al. 2004. A modular approach to the ECVAM principles

on test validity. Altern Lab Anim 32: 467-72

Hartung T, Hoffmann S, Stephens M. 2013. Mechanistic validation. Altex 30: 119-30

Hartung T, Leist M. 2008. Food for thought ... on the evolution of toxicology and the phasing out of animal

testing. ALTEX 25: 91-102

Hartung T, McBride M. 2011. Food for Thought ... on mapping the human toxome. Altex 28: 83-93

Hartung T, van Vliet E, Jaworska J, Bonilla L, Skinner N, Thomas R. 2012. Systems toxicology. Altex 29: 119-

28

Hartung T, Zurlo J. 2012. Alternative approaches for medical countermeasures to biological and chemical

terrorism and warfare. Altex 29: 251-60

Healthcare G. 2010. Characterization of drug action on the Golgi complex using high-content analysis on IN

Cell Analyzer 2000.

Hnisz D, Abraham BJ, Lee TI, Lau A, Saint-Andre V, et al. 2013. Super-Enhancers in the Control of Cell

Identity and Disease. Cell

Hogberg HT, Kinsner-Ovaskainen A, Hartung T, Coecke S, Bal-Price AK. 2009. Gene expression as a sensitive

endpoint to evaluate cell differentiation and maturation of the developing central nervous system in

primary cultures of rat cerebellar granule cells (CGCs) exposed to pesticides. Toxicology and applied

pharmacology 235: 268-86

Holtz WA, O'Malley KL. 2003. Parkinsonian mimetics induce aspects of unfolded protein response in death of

dopaminergic neurons. J Biol Chem 278: 19367-77

Hughes JP, Rees S, Kalindjian SB, Philpott KL. 2011. Principles of early drug discovery. Br J Pharmacol 162:

1239-49

Huh D, Hamilton GA, Ingber DE. 2011. From 3D cell culture to organs-on-chips. Trends in cell biology 21: 745-

54

Hussman JP, Chung RH, Griswold AJ, Jaworski JM, Salyakina D, et al. 2011. A noise-reduction GWAS

analysis implicates altered regulation of neurite outgrowth and guidance in autism. Mol Autism 2: 1

Bibliography

167

Ingram JL, Peckham SM, Tisdale B, Rodier PM. 2000. Prenatal exposure of rats to valproic acid reproduces the

cerebellar anomalies associated with autism. Neurotoxicol Teratol 22: 319-24

Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, et al. 2003. Exploration, normalization, and

summaries of high density oligonucleotide array probe level data. Biostatistics 4: 249-64

Ishido M, Suzuki J. 2010. Inhibition by rotenone of mesencephalic neural stem-cell migration in a neurosphere

assay in vitro. Toxicol In Vitro 24: 552-7

Jagtap S, Meganathan K, Gaspar J, Wagh V, Winkler J, et al. 2011. Cytosine arabinoside induces ectoderm and

inhibits mesoderm expression in human embryonic stem cells during multilineage differentiation. Br J


Jennen D, Ruiz-Aracama A, Magkoufopoulou C, Peijnenburg A, Lommen A, et al. 2011. Integrating

transcriptomics and metabonomics to unravel modes-of-action of 2,3,7,8-tetrachlorodibenzo-p-dioxin

(TCDD) in HepG2 cells. BMC systems biology 5: 139

Jentink J, Loane MA, Dolk H, Barisic I, Garne E, et al. 2010. Valproic acid monotherapy in pregnancy and

major congenital malformations. N Engl J Med 362: 2185-93

Jergil M, Kultima K, Gustafson AL, Dencker L, Stigson M. 2009. Valproic acid-induced deregulation in vitro of

genes associated in vivo with neural tube defects. Toxicol Sci 108: 132-48

Jones LB, Stanwood GD, Reinoso BS, Washington RA, Wang HY, et al. 2000. In utero cocaine-induced

dysfunction of dopamine D1 receptor signaling and abnormal differentiation of cerebral cortical

neurons. J Neurosci 20: 4606-14

Joshi S, Guleria RS, Pan J, Bayless KJ, Davis GE, et al. 2006. Ethanol impairs Rho GTPase signaling and

differentiation of cerebellar granule neurons in a rodent model of fetal alcohol syndrome. Cell Mol Life

Sci 63: 2859-70

Judson R, Kavlock R, Martin M, Reif D, Houck K, et al. 2013. Perspectives on validation of high-throughput

assays supporting 21st century toxicity testing ALTEX 30: 51-66

Kadereit S, Zimmer B, van Thriel C, Hengstler JG, Leist M. 2012. Compound selection for in vitro modeling of

developmental neurotoxicity. Front Biosci (Landmark Ed) 17: 2442-60

Kanungo J, Lantz S, Paule MG. 2011. In vivo imaging and quantitative analysis of changes in axon length using

transgenic zebrafish embryos. Neurotoxicology and teratology 33: 618-23

Keller H, Zadeh AD, Eggli P. 2002. Localised depletion of polymerised actin at the front of Walker

carcinosarcoma cells increases the speed of locomotion. Cell Motil Cytoskeleton 53: 189-202

Kim S, Streets AM, Lin RR, Quake SR, Weiss S, Majumdar DS. 2011. High-throughput single-molecule

optofluidic analysis. Nature methods 8: 242-5

Klaassen CD, ed. 2010. Casarett and Doull's Toxicology: The Basic Science of Poisons. 7th Edition: McGraw-

Hill.

Klein S, Heinzle E. 2012. Isotope labeling experiments in metabolomics and fluxomics. Wiley interdisciplinary

reviews. Systems biology and medicine 4: 261-72

Kleinstreuer NC, Smith AM, West PR, Conard KR, Fontaine BR, et al. 2011. Identifying developmental toxicity

pathways for a subset of ToxCast chemicals using human embryonic stem cells and metabolomics.

Toxicol Appl Pharmacol 257: 111-21

Knudsen T, Martin M, Chandler K, Kleinstreuer N, Judson R, Sipes N. 2013. Predictive models and

computational toxicology. Methods in molecular biology 947: 343-74

Knudsen TB, Houck KA, Sipes NS, Singh AV, Judson RS, et al. 2011. Activity profiles of 309 ToxCast

chemicals evaluated across 292 biochemical targets. Toxicology 282: 1-15

Kobayashi S, Takai K, Iga T, Hanano M. 1991. Pharmacokinetic analysis of the disposition of valproate in

pregnant rats. Drug Metab Dispos 19: 972-6

Kolodkin A, Simeonidis E, Balling R, Westerhoff HV. 2012. Understanding complexity in neurodegenerative

diseases: in silico reconstruction of emergence. Front Physiol 3: 291

Krewski D, Acosta D, Jr., Andersen M, Anderson H, Bailar JC, 3rd, et al. 2010. Toxicity testing in the 21st

century: a vision and a strategy. J Toxicol Environ Health B Crit Rev. 13: 51-138.

Krug AK, Kolde R, Gaspar JA, Rempel E, Balmer NV, et al. 2013. Human embryonic stem cell-derived test

systems for developmental neurotoxicity: a transcriptomics approach. Arch Toxicol 87: 123-43

Kubo T, Yamaguchi A, Iwata N, Yamashita T. 2008. The therapeutic effects of Rho-ROCK inhibitors on CNS

disorders. Ther Clin Risk Manag 4: 605-15

Kuegler PB, Baumann BA, Zimmer B, Keller S, Marx A, et al. 2012. GFAP-independent inflammatory

competence and trophic functions of astrocytes generated from murine embryonic stem cells. Glia 60:

218-28

Kuegler PB, Zimmer B, Waldmann T, Baudis B, Ilmjarv S, et al. 2010. Markers of murine embryonic and neural

stem cells, neurons and astrocytes: reference points for developmental neurotoxicity testing. ALTEX 27:

17-42

Bibliography

168

Kwong LN, Costello JC, Liu H, Jiang S, Helms TL, et al. 2012. Oncogenic NRAS signaling differentially

regulates survival and proliferation in melanoma. Nat Med 18: 1503-10

Lange PS, Chavez JC, Pinto JT, Coppola G, Sun CW, et al. 2008. ATF4 is an oxidative stress-inducible,

prodeath transcription factor in neurons in vitro and in vivo. J Exp Med 205: 1227-42

Langston JW, Irwin I, Langston EB, Forno LS. 1984a. 1-Methyl-4-phenylpyridinium ion (MPP+): identification

of a metabolite of MPTP, a toxin selective to the substantia nigra. Neurosci Lett 48: 87-92

Langston JW, Langston EB, Irwin I. 1984b. MPTP-induced parkinsonism in human and non-human primates--

clinical and experimental aspects. Acta Neurol Scand Suppl 100: 49-54

Larsson C. 2006. Protein kinase C and the regulation of the actin cytoskeleton. Cell Signal 18: 276-84

Laurenza I, Pallocca G, Mennecozzi M, Scelfo B, Pamies D, Bal-Price A. 2013. A human pluripotent carcinoma

stem cell-based model for in vitro developmental neurotoxicity testing: Effects of methylmercury, lead

and aluminum evaluated by gene expression studies. International journal of developmental

neuroscience : the official journal of the International Society for Developmental Neuroscience

Lee MJ, Ye AS, Gardino AK, Heijink AM, Sorger PK, et al. 2012. Sequential application of anticancer drugs

enhances cell death by rewiring apoptotic signaling networks. Cell 149: 780-94

Lefranc F, Sauvage S, Van Goietsenoven G, Megalizzi V, Lamoral-Theys D, et al. 2009. Narciclasine, a plant

growth modulator, activates Rho and stress fibers in glioblastoma cells. Mol Cancer Ther 8: 1739-50

Leist M, Bremer S, Brundin P, Hescheler J, Kirkeby A, et al. 2008a. The biological and ethical basis of the use

of human embryonic stem cells for in vitro test systems or cell therapy. ALTEX 25: 163-90

Leist M, Efremova L, Karreman C. 2010. Food for thought ... considerations and guidelines for basic test method

descriptions in toxicology. ALTEX 27: 309-17

Leist M, Hartung T. 2013. Inflammatory findings on species extrapolations: humans are definitely no 70-kg

mice. Altex 30: 227-30

Leist M, Hartung T, Nicotera P. 2008b. The dawning of a new age of toxicology. ALTEX 25: 103-14

Leist M, Hasiwa N, Daneshian M. 2013. Summary and Validation of New Animal Free Toxicity Tests. Altex 2

Leist M, Hasiwa N, Daneshian M, Hartung T. 2012a. Validation and quality control of replacement alternatives

– current status and future challenges. Toxicol Res 1: 8-22

Leist M, Jaattela M. 2001. Four deaths and a funeral: from caspases to alternative mechanisms. Nat Rev Mol Cell

Biol 2: 589-98

Leist M, Lidbury BA, Yang C, Hayden PJ, Kelm JM, et al. 2012b. Novel technologies and an overall strategy to

allow hazard assessment and risk prediction of chemicals, cosmetics, and drugs with animal-free

methods. ALTEX 29: 373-88

Li AP. 2009. The use of the Integrated Discrete Multiple Organ Co-culture (IdMOC) system for the evaluation

of multiple organ toxicity. Altern Lab Anim 37: 377-85

Lin Z, Will Y. 2012. Evaluation of drugs with specific organ toxicities in organ-specific cell lines. Toxicological

sciences : an official journal of the Society of Toxicology 126: 114-27

Llorens J, Li AA, Ceccatelli S, Sunol C. 2012. Strategies and tools for preventing neurotoxicity: to test, to

predict and how to do it. Neurotoxicology 33: 796-804

Loo LH, Wu LF, Altschuler SJ. 2007. Image-based multivariate profiling of drug responses from single cells.

Nature methods 4: 445-53

LoPachin RM, Ross JF, Reid ML, Das S, Mansukhani S, Lehning EJ. 2002. Neurological evaluation of toxic

axonopathies in rats: acrylamide and 2,5-hexanedione. Neurotoxicology 23: 95-110

Loscher W. 1978. Serum protein binding and pharmacokinetics of valproate in man, dog, rat and mouse. J

Pharmacol Exp Ther 204: 255-61

Lotharius J, Falsig J, van Beek J, Payne S, Dringen R, et al. 2005. Progressive degeneration of human

mesencephalic neuron-derived cells triggered by dopamine-dependent oxidative stress is dependent on

the mixed-lineage kinase pathway. J Neurosci 25: 6329-42

Louisse J, de Jong E, van de Sandt JJ, Blaauboer BJ, Woutersen RA, et al. 2010. The use of in vitro toxicity data

and physiologically based kinetic modeling to predict dose-response curves for in vivo developmental

toxicity of glycol ethers in rat and man. Toxicol Sci 118: 470-84

Luster AD, Alon R, von Andrian UH. 2005. Immune cell migration in inflammation: present and future

therapeutic targets. Nature immunology 6: 1182-90

MacDougall AS, McCann KS, Gellner G, Turkington R. 2013. Diversity loss with persistent human disturbance

increases vulnerability to ecosystem collapse. Nature 494: 86-9

Maertens RM, White PA, Williams A, Yauk CL. 2013. A global toxicogenomic analysis investigating the

mechanistic differences between tobacco and marijuana smoke condensates in vitro. Toxicology 308:

60-73

Makris SL, Raffaele K, Allen S, Bowers WJ, Hass U, et al. 2009. A retrospective performance assessment of the

developmental neurotoxicity study in support of OECD test guideline 426. Environ Health Perspect

117: 17-25

Bibliography

169

Marx U, Walles H, Hoffmann S, Lindner G, Horland R, et al. 2012. 'Human-on-a-chip' developments: a

translational cutting-edge alternative to systemic safety assessment and efficiency evaluation of

substances in laboratory animals and man? Alternatives to laboratory animals : ATLA 40: 235-57

Mastyugin V, McWhinnie E, Labow M, Buxton F. 2004. A quantitative high-throughput endothelial cell

migration assay. Journal of biomolecular screening 9: 712-8

Mazzio E, Soliman KF. 2003a. D-(+)-glucose rescue against 1-methyl-4-phenylpyridinium toxicity through

anaerobic glycolysis in neuroblastoma cells. Brain Res 962: 48-60

Mazzio E, Soliman KF. 2003b. The role of glycolysis and gluconeogenesis in the cytoprotection of

neuroblastoma cells against 1-methyl 4-phenylpyridinium ion toxicity. Neurotoxicology 24: 137-47

Mazzio E, Soliman KF. 2012. Whole genome expression profile in neuroblastoma cells exposed to 1-methyl-4-

phenylpyridine. Neurotoxicology 33: 1156-69

McCormack AL, Thiruchelvam M, Manning-Bog AB, Thiffault C, Langston JW, et al. 2002. Environmental risk

factors and Parkinson's disease: selective degeneration of nigral dopaminergic neurons caused by the

herbicide paraquat. Neurobiol Dis 10: 119-27

McCormick SM, Frye SR, Eskin SG, Teng CL, Lu CM, et al. 2003. Microarray analysis of shear stressed

endothelial cells. Biorheology 40: 5-11

McEwen BS. 1999. Stress and hippocampal plasticity. Annu Rev Neurosci 22: 105-22

McKim JM, Jr. 2010. Building a tiered approach to in vitro predictive toxicity screening: a focus on assays with

in vivo relevance. Combinatorial chemistry & high throughput screening 13: 188-206

McVearry KM, Gaillard WD, VanMeter J, Meador KJ. 2009. A prospective study of cognitive fluency and

originality in children exposed in utero to carbamazepine, lamotrigine, or valproate monotherapy.

Epilepsy Behav 16: 609-16

Meganathan K, Jagtap S, Wagh V, Winkler J, Gaspar JA, et al. 2012. Identification of Thalidomide-Specific

Transcriptomics and Proteomics Signatures during Differentiation of Human Embryonic Stem Cells.

PLoS One 7: e44228

Mennecozzi M, Landesmann B, Harris GA, Liska R, Whelan M. 2013. Hepatotoxicity Screening Taking a

Mode-Of-Action Approach Using HepaRG Cells and HCA Altex

Mioulane M, Foldes G, Ali NN, Schneider MD, Harding SE. 2012. Development of high content imaging

methods for cell death detection in human pluripotent stem cell-derived cardiomyocytes. J Cardiovasc

Transl Res 5: 593-604

Mitchell PJ, Hanson JC, Quets-Nguyen AT, Bergeron M, Smith RC. 2007. A quantitative method for analysis of

in vitro neurite outgrowth. J Neurosci Methods 164: 350-62

Moors M, Rockel TD, Abel J, Cline JE, Gassmann K, et al. 2009. Human neurospheres as three-dimensional

cellular systems for developmental neurotoxicity testing. Environmental health perspectives 117: 1131-

8

Morozova O, Marra MA. 2008. Applications of next-generation sequencing technologies in functional genomics.

Genomics 92: 255-64

Nakagawa S, Hirose T. 2012. Paraspeckle nuclear bodies--useful uselessness? Cell Mol Life Sci 69: 3027-36

Narro ML, Yang F, Kraft R, Wenk C, Efrat A, Restifo LL. 2007. NeuronMetrics: software for semi-automated

processing of cultured neuron images. Brain Res 1138: 57-75

Nerini-Molteni S, Mennecozzi M, Fabbri M, Sacco MG, Vojnits K, et al. 2012. MicroRNA profiling as a tool for

pathway analysis in a human in vitro model for neural development. Current medicinal chemistry 19:

6214-23

Newman CG. 1986. The thalidomide syndrome: risks of exposure and spectrum of malformations. Clinics in

perinatology 13: 555-73

Nicklas WJ, Vyas I, Heikkila RE. 1985. Inhibition of NADH-linked oxidation in brain mitochondria by 1-

methyl-4-phenyl-pyridine, a metabolite of the neurotoxin, 1-methyl-4-phenyl-1,2,5,6-

tetrahydropyridine. Life Sci 36: 2503-8

Niggli V. 2003. Microtubule-disruption-induced and chemotactic-peptide-induced migration of human

neutrophils: implications for differential sets of signalling pathways. J Cell Sci 116: 813-22

Niittylae T, Chaudhuri B, Sauer U, Frommer WB. 2009. Comparison of quantitative metabolite imaging tools

and carbon-13 techniques for fluxomics. Methods in molecular biology 553: 355-72

Nikolic M. 2002. The role of Rho GTPases and associated kinases in regulating neurite outgrowth. Int J Biochem

Cell Biol 34: 731-45

Nilsson BS. 1983. Adverse reactions in connection with zimeldine treatment--a review. Acta psychiatrica

Scandinavica. Supplementum 308: 115-9

Nystrom ML, Thomas GJ, Stone M, Mackenzie IC, Hart IR, Marshall JF. 2005. Development of a quantitative

method to analyse tumour cell invasion in organotypic culture. The Journal of pathology 205: 468-75

Ong SE. 2012. The expanding field of SILAC. Analytical and bioanalytical chemistry 404: 967-76

Bibliography

170

Osman AM, van Dartel DA, Zwart E, Blokland M, Pennings JL, Piersma AH. 2010. Proteome profiling of

mouse embryonic stem cells to define markers for cell differentiation and embryotoxicity. Reproductive

toxicology 30: 322-32

Pallocca G, Fabbri M, Sacco MG, Gribaldo L, Pamies D, et al. 2013. miRNA expression profiling in a human

stem cell-based model as a tool for developmental neurotoxicity testing. Cell biology and toxicology 29:

239-57

Palmer JA, Poenitzsch AM, Smith SM, Conard KR, West PR, Cezar GG. 2012. Metabolic biomarkers of

prenatal alcohol exposure in human embryonic stem cell-derived neural lineages. Alcoholism, clinical

and experimental research 36: 1314-24

Pattarini R, Rong Y, Qu C, Morgan JI. 2008. Distinct mechanisms of 1-methyl-4-phenyl-1,2,3,6-

tetrahydropyrimidine resistance revealed by transcriptome mapping in mouse striatum. Neuroscience

155: 1174-94

Poltl D, Schildknecht S, Karreman C, Leist M. 2012. Uncoupling of ATP-depletion and cell death in human

dopaminergic neurons. Neurotoxicology 33: 769-79

Price RD, Oe T, Yamaji T, Matsuoka N. 2006. A simple, flexible, nonfluorescent system for the automated

screening of neurite outgrowth. J Biomol Screen 11: 155-64

Puigvert JC, de Bont H, van de Water B, Danen EH. 2010. High-throughput live cell imaging of apoptosis.

Current protocols in cell biology / editorial board, Juan S. Bonifacino ... [et al.] Chapter 18: Unit 18 10

1-13

Quasthoff S, Hartung HP. 2002. Chemotherapy-induced peripheral neuropathy. J Neurol 249: 9-17

R_Development_Core_Team. 2011. A language and environment for statistical computing. R Foundation for

Statistical Computing, Vienna, Austria

Rabilloud T, Chevallet M, Luche S, Lelong C. 2010. Two-dimensional gel electrophoresis in proteomics: Past,

present and future. Journal of proteomics 73: 2064-77

Radio NM, Breier JM, Shafer TJ, Mundy WR. 2008. Assessment of chemical effects on neurite outgrowth in

PC12 cells using high content screening. Toxicol Sci 105: 106-18

Radio NM, Freudenrich TM, Robinette BL, Crofton KM, Mundy WR. 2010. Comparison of PC12 and cerebellar

granule cell cultures for evaluating neurite outgrowth using high content analysis. Neurotoxicol Teratol

32: 25-35

Radio NM, Mundy WR. 2008. Developmental neurotoxicity testing in vitro: models for assessing chemical

effects on neurite outgrowth. Neurotoxicology 29: 361-76

Ramirez T, Daneshian M, Kamp H, Bois FY, Clench MR, et al. 2013. Metabolomics in toxicology and

preclinical research. ALTEX 30: 209-25

Ramm P, Alexandrov Y, Cholewinski A, Cybuch Y, Nadon R, Soltys BJ. 2003. Automated screening of neurite

outgrowth. J Biomol Screen 8: 7-18

Read DE, Gorman AM. 2009. Involvement of Akt in neurite outgrowth. Cell Mol Life Sci 66: 2975-84

Reimand J, Kull M, Peterson H, Hansen J, Vilo J. 2007. g:Profiler--a web-based toolset for functional profiling

of gene lists from large-scale experiments. Nucleic Acids Res 35: W193-200

Ren Y, Liu W, Jiang H, Jiang Q, Feng J. 2005. Selective vulnerability of dopaminergic neurons to microtubule

depolymerization. J Biol Chem 280: 34105-12

Rieder F, Brenmoehl J, Leeb S, Scholmerich J, Rogler G. 2007. Wound healing and fibrosis in intestinal disease.

Gut 56: 130-9

Robinson JF, Theunissen PT, van Dartel DA, Pennings JL, Faustman EM, Piersma AH. 2011. Comparison of

MeHg-induced toxicogenomic responses across in vivo and in vitro models used in developmental

toxicology. Reprod Toxicol 32: 180-8

Rodier PM, Ingram JL, Tisdale B, Nelson S, Romano J. 1996. Embryological origin for autism: developmental

anomalies of the cranial nerve motor nuclei. J Comp Neurol 370: 247-61

Rodriguez LG, Wu X, Guan JL. 2005. Wound-healing assay. Methods in molecular biology 294: 23-9

Ross DA, Lee S, Reiser V, Xue J, Alves K, et al. 2008. Multiplexed assays by high-content imaging for

assessment of GPCR activity. J Biomol Screen 13: 449-55

Rotroff DM, Wetmore BA, Dix DJ, Ferguson SS, Clewell HJ, et al. 2010. Incorporating human dosimetry and

exposure into high-throughput in vitro toxicity screening. Toxicol Sci 117: 348-58

Rovida C, Hartung T. 2009. Re-evaluation of animal numbers and costs for in vivo tests to accomplish REACH

legislation requirements for chemicals - a report by the transatlantic think tank for toxicology (t(4)).

Altex 26: 187-208

Ruesch H. 1982. The Naked Empress or the Great Medical Fraud. Cancer Control Society.

Russell WMS, Burch RL. 1959. The Principles of Humane Experimental Technique. London: Methuen

Sai Y, Wu Q, Le W, Ye F, Li Y, Dong Z. 2008. Rotenone-induced PC12 cell toxicity is caused by oxidative

stress resulting from altered dopamine metabolism. Toxicol In Vitro 22: 1461-8

Bibliography

171

Sakamuru S, Li X, Attene-Ramos MS, Huang R, Lu J, et al. 2012. Application of a homogenous membrane

potential assay to assess mitochondrial function. Physiological genomics 44: 495-503

Sanchez M, Gastaldi L, Remedi M, Caceres A, Landa C. 2007. Rotenone-Induced Toxicity is Mediated by Rho-

GTPases in Hippocampal Neurons. Toxicological Sciences 104: 352-61

Saporito MS, Thomas BA, Scott RW. 2000. MPTP activates c-Jun NH(2)-terminal kinase (JNK) and its

upstream regulatory kinase MKK4 in nigrostriatal neurons in vivo. J Neurochem 75: 1200-8

Saunders NR. 1986. Development of human blood-CSF-brain barrier. Developmental medicine and child

neurology 28: 261-3

Schaaf S, Shibamiya A, Mewe M, Eder A, Stohr A, et al. 2011. Human engineered heart tissue as a versatile tool

in basic research and preclinical toxicology. PloS one 6: e26397

Schierle GS, Hansson O, Leist M, Nicotera P, Widner H, Brundin P. 1999. Caspase inhibition reduces apoptosis

and increases survival of nigral transplants. Nat Med 5: 97-100

Schildknecht S, Karreman C, Poltl D, Efremova L, Kullmann C, et al. 2013. Generation of genetically-modified

human differentiated cells for toxicological tests and the study of neurodegenerative diseases. Altex

Schildknecht S, Poltl D, Nagel DM, Matt F, Scholz D, et al. 2009. Requirement of a dopaminergic neuronal

phenotype for toxicity of low concentrations of 1-methyl-4-phenylpyridinium to human cells. Toxicol

Appl Pharmacol 241: 23-35

Schmid RS, Pruitt WM, Maness PF. 2000. A MAP kinase-signaling pathway mediates neurite outgrowth on L1

and requires Src-dependent endocytosis. J Neurosci 20: 4177-88

Schmidt M, Bohm D, von Torne C, Steiner E, Puhl A, et al. 2008. The humoral immune system has a key

prognostic impact in node-negative breast cancer. Cancer Res 68: 5405-13

Schmidt M, Hellwig B, Hammad S, Othman A, Lohr M, et al. 2012. A comprehensive analysis of human gene

expression profiles identifies stromal immunoglobulin kappa C as a compatible prognostic marker in

human solid tumors. Clin Cancer Res 18: 2695-703

Schneider K, Schwarz M, Burkholder I, Kopp-Schneider A, Edler L, et al. 2009. "ToxRTool", a new tool to

assess the reliability of toxicological data. Toxicol Lett 189: 138-44

Schoenenberger F, Krug AK, Leist M, Ferrando-May E, Merhof D. 2012. An Advanced Image Processing

Approach based on Parallel Growth and Overlap Handling to Quantify Neurite Growth. Presented at

9th International Workshop on Computational Systems Biology (WCSB), Ulm

Scholz D, Poltl D, Genewsky A, Weng M, Waldmann T, et al. 2011. Rapid, complete and large-scale generation

of post-mitotic neurons from the human LUHMES cell line. Journal of neurochemistry 119: 957-71

Schulz JB. 2006. Anti-apoptotic gene therapy in Parkinson's disease. Journal of neural transmission.

Supplementum: 467-76

Seiler A, Oelgeschlager M, Liebsch M, Pirow R, Riebeling C, et al. 2011. Developmental toxicity testing in the

21st century: the sword of Damocles shattered by embryonic stem cell assays? Arch Toxicol 85: 1361-

72

Seiler AE, Spielmann H. 2011. The validated embryonic stem cell test to predict embryotoxicity in vitro. Nature

protocols 6: 961-78

Seok J, Warren HS, Cuenca AG, Mindrinos MN, Baker HV, et al. 2013. Genomic responses in mouse models

poorly mimic human inflammatory diseases. Proceedings of the National Academy of Sciences of the

United States of America 110: 3507-12

Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, et al. 2003. Cytoscape: a software environment for

integrated models of biomolecular interaction networks. Genome Res 13: 2498-504

Sherman SP, Alva JA, Thakore-Shah K, Pyle AD. 2011. Human pluripotent stem cells: the development of high-

content screening strategies. Methods Mol Biol 767: 283-95

Shih W, Yamada S. 2011. Live-cell imaging of migrating cells expressing fluorescently-tagged proteins in a

three-dimensional matrix. Journal of visualized experiments : JoVE

Sipes NS, Martin MT, Kothiya P, Reif DM, Judson RS, et al. 2013. Profiling 976 ToxCast chemicals across 331

enzymatic and receptor signaling assays. Chem Res Toxicol 26: 878-95

Sipes NS, Martin MT, Reif DM, Kleinstreuer NC, Judson RS, et al. 2011. Predictive models of prenatal

developmental toxicity from ToxCast high-throughput screening data. Toxicol Sci 124: 109-27

Sjostrom M, Kolman A, Clemedson C, Clothier R. 2008. Estimation of human blood LC50 values for use in

modeling of in vitro-in vivo data of the ACuteTox project. Toxicol In Vitro 22: 1405-11

Slotkin TA, Levin ED, Seidler FJ. 2006. Comparative developmental neurotoxicity of organophosphate

insecticides: effects on brain development are separable from systemic toxicity. Environ Health

Perspect 114: 746-51

Slotkin TA, Seidler FJ, Fumagalli F. 2010. Unrelated developmental neurotoxicants elicit similar transcriptional

profiles for effects on neurotrophic factors and their receptors in an in vitro model. Neurotoxicology and

teratology 32: 42-51

Bibliography

172

Smoot ME, Ono K, Ruscheinski J, Wang PL, Ideker T. 2011. Cytoscape 2.8: new features for data integration

and network visualization. Bioinformatics 27: 431-2

Smyth GK, Gentleman R, Carey V, Dudoit S, Irizarry R, Huber W. 2005. Limma: linear models for microarray

data. Bioinformatics and Computational Biology Solutions using R and Bioconductor, Springer, New

York: 397—420

Snow DM, Smith JD, Booze RM, Welch MA, Mactutus CF. 2001. Cocaine decreases cell survival and inhibits

neurite extension of rat locus coeruleus neurons. Neurotoxicol Teratol 23: 225-34

Spencer PS, Schaumburg HH, Ludolph AC. 2000. Experimental and clinical neurotoxicology. New York:

Oxford University Press. xl, 1310 p. pp.

Stanwood GD, Washington RA, Shumsky JS, Levitt P. 2001. Prenatal cocaine exposure produces consistent

developmental alterations in dopamine-rich regions of the cerebral cortex. Neuroscience 106: 5-14

Stephens ML, Andersen M, Becker RA, Betts K, Boekelheide K, et al. 2013. Evidence-based toxicology for the

21st century: Opportunities and challenges. ALTEX 30: 74-104

Stern M, Gierse A, Tan S, Bicker G. 2013. Human Ntera2 cells as a predictive in vitro test system for

developmental neurotoxicity. Archives of toxicology

Stiegler NV, Krug AK, Matt F, Leist M. 2011. Assessment of chemical-induced impairment of human neurite

outgrowth by multiparametric live cell imaging in high-density cultures. Toxicol Sci 121: 73-87

Stummann TC, Hareng L, Bremer S. 2009. Hazard assessment of methylmercury toxicity to neuronal induction

in embryogenesis using human embryonic stem cells. Toxicology 257: 117-26

Sun X, Liu J, Crary JF, Malagelada C, Sulzer D, et al. 2013. ATF4 protects against neuronal death in cellular

Parkinson's disease models by maintaining levels of parkin. J Neurosci 33: 2398-407

Sunil D, Isloor AM, Shetty P, Satyamoorthy K, Prasad ASB. 2011. Synthesis, characterization, antioxidant, and

anticancer studies of 6-[3-(4-chlorophenyl)-1H-pyrazol-4-yl]-3-[(2-

naphthyloxy)methyl][1,2,4]triazolo[3,4-b][1,3,4]thiadiazole in HepG2 cell lines. Med Chem Res 20:

1074-80

Takanaga H, Chaudhuri B, Frommer WB. 2008. GLUT1 and GLUT9 as major contributors to glucose influx in

HepG2 cells identified by a high sensitivity intramolecular FRET glucose sensor. Biochimica et

biophysica acta 1778: 1091-9

Takesono A, Heasman SJ, Wojciak-Stothard B, Garg R, Ridley AJ. 2010. Microtubules regulate migratory

polarity through Rho/ROCK signaling in T cells. PLoS One 5: e8774

Tan XL, Marquardt G, Massimi AB, Shi M, Han W, Spivack SD. 2012. High-throughput library screening

identifies two novel NQO1 inducers in human lung cells. American journal of respiratory cell and

molecular biology 46: 365-71

Taylor K, Gordon N, Langley G, Higgins W. 2008. Estimates for worldwide laboratory animal use in 2005.

Alternatives to laboratory animals : ATLA 36: 327-42

Theunissen PT, Robinson JF, Pennings JL, de Jong E, Claessen SM, et al. 2012a. Transcriptomic concentration-

response evaluation of valproic acid, cyproconazole, and hexaconazole in the neural embryonic stem

cell test (ESTn). Toxicological sciences : an official journal of the Society of Toxicology 125: 430-8

Theunissen PT, Robinson JF, Pennings JL, van Herwijnen MH, Kleinjans JC, Piersma AH. 2012b. Compound-

specific effects of diverse neurodevelopmental toxicants on global gene expression in the neural

embryonic stem cell test (ESTn). Toxicology and applied pharmacology 262: 330-40

Thomas RS, Wesselkamper SC, Wang NC, Zhao QJ, Petersen DD, et al. 2013. Temporal Concordance Between

Apical and Transcriptional Points of Departure for Chemical Risk Assessment. Toxicol Sci

Tice RR, Austin CP, Kavlock RJ, Bucher JR. 2013. Improving the Human Hazard Characterization of

Chemicals: A Tox21 Update. Environ Health Perspect

Timar J. 2004. [Molecular mechanism of tumor progression. From Krompecher to the DNA microarray].

Magyar onkologia 48: 3-11

Tolosa L, Pinto S, Donato MT, Lahoz A, Castell JV, et al. 2012. Development of a multiparametric cell-based

protocol to screen and classify the hepatotoxicity potential of drugs. Toxicological sciences : an official

journal of the Society of Toxicology 127: 187-98

Torres-Guzman R, Chu S, Velasco JA, Lallena MJ. 2013. Multiparametric cell-based assay for the evaluation of

transcription inhibition by high-content imaging. J Biomol Screen 18: 556-66

Ulitsky I, Maron-Katz A, Shavit S, Sagir D, Linhart C, et al. 2010. Expander: from expression microarrays to

networks and functions. Nat Protoc 5: 303-22

Valiente M, Marin O. 2010. Neuronal migration mechanisms in development and disease. Current opinion in

neurobiology 20: 68-78

van Dartel DA, Pennings JL, Hendriksen PJ, van Schooten FJ, Piersma AH. 2009. Early gene expression

changes during embryonic stem cell differentiation into cardiomyocytes and their modulation by

monobutyl phthalate. Reproductive toxicology 27: 93-102

Bibliography

173

van Delft J, Gaj S, Lienhard M, Albrecht MW, Kirpiy A, et al. 2012. RNA-Seq provides new insights in the

transcriptome responses induced by the carcinogen benzo[a]pyrene. Toxicological sciences : an official

journal of the Society of Toxicology 130: 427-39

Van Summeren A, Renes J, Lizarraga D, Bouwman FG, Noben JP, et al. 2013. Screening for drug-induced

hepatotoxicity in primary mouse hepatocytes using acetaminophen, amiodarone, and cyclosporin a as

model compounds: an omics-guided approach. Omics : a journal of integrative biology 17: 71-83

van Thriel C, Westerink RH, Beste C, Bale AS, Lein PJ, Leist M. 2012. Translating neurobehavioural endpoints

of developmental neurotoxicity tests into in vitro assays and readouts. Neurotoxicology 33: 911-24

Vazquez A. 2013. Metabolic states following accumulation of intracellular aggregates: implications for

neurodegenerative diseases. PloS one 8: e63822

Vendrell I, Carrascal M, Campos F, Abian J, Sunol C. 2010. Methylmercury disrupts the balance between

phosphorylated and non-phosphorylated cofilin in primary cultures of mice cerebellar granule cells. A

proteomic study. Toxicology and applied pharmacology 242: 109-18

Vendrell I, Carrascal M, Vilaro MT, Abian J, Rodriguez-Farre E, Sunol C. 2007. Cell viability and proteomic

analysis in cultured neurons exposed to methylmercury. Human & experimental toxicology 26: 263-72

Venkatakrishnan AJ, Deupi X, Lebon G, Tate CG, Schertler GF, Babu MM. 2013. Molecular signatures of G-

protein-coupled receptors. Nature 494: 185-94

Verwei M, van Burgsteden JA, Krul CA, van de Sandt JJ, Freidig AP. 2006. Prediction of in vivo embryotoxic

effect levels with a combination of in vitro studies and PBPK modelling. Toxicol Lett 165: 79-87

Vestergaard-Poulsen P, Wegener G, Hansen B, Bjarkam CR, Blackband SJ, et al. 2011. Diffusion-weighted MRI

and quantitative biophysical modeling of hippocampal neurite loss in chronic stress. PLoS One 6:

e20653

Vila M, Przedborski S. 2003. Targeting programmed cell death in neurodegenerative diseases. Nat Rev Neurosci

4: 365-75

Volbracht C, Leist M, Kolb SA, Nicotera P. 2001. Apoptosis in caspase-inhibited neurons. Mol Med 7: 36-48

Volbracht C, Leist M, Nicotera P. 1999. ATP controls neuronal apoptosis triggered by microtubule breakdown

or potassium deprivation. Mol Med 5: 477-89

Volbracht C, van Beek J, Zhu C, Blomgren K, Leist M. 2006. Neuroprotective properties of memantine in

different in vitro and in vivo models of excitotoxicity. Eur J Neurosci 23: 2611-22

Walpita D, Hasaka T, Spoonamore J, Vetere A, Takane KK, et al. 2012. A human islet cell culture system for

high-throughput screening. Journal of biomolecular screening 17: 509-18

Wang C, Luan Z, Yang Y, Wang Z, Cui Y, Gu G. 2011a. Valproic acid induces apoptosis in differentiating

hippocampal neurons by the release of tumor necrosis factor-alpha from activated astrocytes. Neurosci

Lett 497: 122-7

Wang D, Lagerstrom R, Sun C, Bishof L, Valotton P, Gotte M. 2010. HCA-vision: Automated neurite outgrowth

analysis. J Biomol Screen 15: 1165-70

Wang J, Jiang J, Zhang H, Wang J, Cai H, et al. 2011b. Integrated transcriptional and proteomic analysis with in

vitro biochemical assay reveal the important role of CYP3A46 in T-2 toxin hydroxylation in porcine

primary hepatocytes. Molecular & cellular proteomics : MCP 10: M111 008748

Wang Z, Gerstein M, Snyder M. 2009. RNA-Seq: a revolutionary tool for transcriptomics. Nature reviews.

Genetics 10: 57-63

Waters MD, Fostel JM. 2004. Toxicogenomics and systems toxicology: aims and prospects. Nature reviews.

Genetics 5: 936-48

Weng MK, Zimmer B, Poltl D, Broeg MP, Ivanova V, et al. 2012. Extensive Transcriptional Regulation of

Chromatin Modifiers during Human Neurodevelopment. PLoS One 7: e36708

Werler MM, Ahrens KA, Bosco JL, Mitchell AA, Anderka MT, et al. 2011. Use of antiepileptic medications in

pregnancy in relation to risks of birth defects. Ann Epidemiol 21: 842-50

West PR, Weir AM, Smith AM, Donley EL, Cezar GG. 2010. Predicting human developmental toxicity of

pharmaceuticals using human embryonic stem cells and metabolomics. Toxicol Appl Pharmacol 247:

18-27

Wetmore BA, Wambaugh JF, Ferguson SS, Sochaski MA, Rotroff DM, et al. 2012. Integration of dosimetry,

exposure, and high-throughput screening data in chemical toxicity assessment. Toxicol Sci 125: 157-74

Wilmes A, Crean D, Aydin S, Pfaller W, Jennings P, Leonard MO. 2011. Identification and dissection of the

Nrf2 mediated oxidative stress pathway in human renal proximal tubule toxicity. Toxicology in vitro :

an international journal published in association with BIBRA 25: 613-22

Wilmes A, Limonciel A, Aschauer L, Moenks K, Bielow C, et al. 2013. Application of integrated transcriptomic,

proteomic and metabolomic profiling for the delineation of mechanisms of drug induced cell stress. J

Proteomics 79: 180-94

Xia M, Huang R, Witt KL, Southall N, Fostel J, et al. 2008. Compound cytotoxicity profiling using quantitative

high-throughput screening. Environmental health perspectives 116: 284-91

Bibliography

174

Yang D, Kim KH, Phimister A, Bachstetter AD, Ward TR, et al. 2009. Developmental exposure to

polychlorinated biphenyls interferes with experience-dependent dendritic plasticity and ryanodine

receptor expression in weanling rats. Environ Health Perspect 117: 426-35

Yeyeodu ST, Witherspoon SM, Gilyazova N, Ibeanu GC. 2010. A rapid, inexpensive high throughput screen

method for neurite outgrowth. Curr Chem Genomics 4: 74-83

Zhang D, Wang Z, Jin N, Li L, Rhoades RA, et al. 2001. Microtubule disruption modulates the Rho-kinase

pathway in vascular smooth muscle. J Muscle Res Cell Motil 22: 193-200

Zhang H, Schneider T, Wheeler-Kingshott CA, Alexander DC. 2012. NODDI: practical in vivo neurite

orientation dispersion and density imaging of the human brain. Neuroimage 61: 1000-16

Zhang X, Zhou JY, Chin MH, Schepmoes AA, Petyuk VA, et al. 2010. Region-specific protein abundance

changes in the brain of MPTP-induced Parkinson's disease mouse model. J Proteome Res 9: 1496-509

Zikopoulos B, Barbas H. 2010. Changes in prefrontal axons may disrupt the network in autism. J Neurosci 30:

14595-609

Zimmer B, Kuegler PB, Baudis B, Genewsky A, Tanavde V, et al. 2011a. Coordinated waves of gene expression

during neuronal differentiation of embryonic stem cells as basis for novel approaches to developmental

neurotoxicity testing. Cell Death Differ 18: 383-95

Zimmer B, Lee G, Balmer NV, Meganathan K, Sachinidis A, et al. 2012. Evaluation of developmental toxicants

and signaling pathways in a functional test based on the migration of human neural crest cells.

Environmental health perspectives 120: 1116-22

Zimmer B, Schildknecht S, Kuegler PB, Tanavde V, Kadereit S, Leist M. 2011b. Sensitivity of dopaminergic

neuron differentiation from stem cells to chronic low-dose methylmercury exposure. Toxicol Sci 121:

357-67

Zimmerman JL. 2012. Cocaine intoxication. Critical care clinics 28: 517-26

Record of contribution

175

Record of contribution

Results Chapter 1

I designed, performed and analysed most of the experiments. Nina V. Balmer performed

the experiments for figure 2C and some of her data is included in figure 3 and supplementary

figure S2. She and Florian Matt performed the experiments for supplementary figure S4. I

prepared the remaining figures. I wrote the manuscript in collaboration with Marcel Leist.

The chapter is published in Archives of Toxicology

Results Chapter 2

The experiments were performed at four universities by the people stated in brackets:

UKK (University of Cologne – Kesavan Meganathan), UKN1 (University of Konstanz – Nina

V. Balmer), JRC (Joint Research Center, Brussels – Kinga Vojnits), UNIGE (University of

Geneva – Mathurin Baquié) and UKN4 (University of Konstanz – myself). The whole

genome transcriptome analysis for all test systems was performed by Smita Jagtap at the

University of Cologne. Data analysis and figures were prepared by Raivo Kolde, John A.

Gaspar, Eugen Rempel and myself. I further edited all figures and prepared the supplementary

tables and figures. Marcel Leist and I wrote the manuscript in collaboration with Tanja

Waldmann and Agapios Sachinidis.

The chapter is published in Archives of Toxicology

Results Chapter 3

I designed, performed and analysed all biochemical experiments as well as the targeted

metabolomics approach. Cornelius Kullmann and Dominik Pöltl performed the RNA

collection for the transcriptome studies and the metabolite extraction for the untargeted

metabolomics study. Liang Zhao performed the analysis of the untargeted metabolomics

samples. The whole genome transcriptome analysis was performed by Smita Jagtap at the

University of Cologne and Cornelius Kullman and I further analysed the results provided with

the help of Violeta Ivanova. Data analysis of the RNA sequencing data was performed by

Sunniva Förster. I prepared all the figures and wrote the manuscript in collaboration with

Marcel Leist.

The chapter is submitted to Cell Death and Differentiation

176

Danksagung

Als erstes und ganz besonders möchte ich mich bei meinem Doktorvater Marcel Leist

bedanken. In den letzten drei Jahren hatte ich nicht nur viele Freiheiten mich zu verwirklichen

und eigene Ziele zu verfolgen, sondern konnte mich auch immer auf die Unterstützung,

Hilfestellung oder Diskussionsbereitschaft deinerseits verlassen. Für die Möglichkeit, drei

Monate nach Baltimore zu gehen, bin ich dir besonders dankbar.

I also like to thank the CAAT-US-Team in Baltimore, in particular Thomas Hartung and

Helena Hogberg. I enjoyed every day of my stay. I had a lot of fun, especially at all the

Conferences and ice-hockey matches (let’s go caps!) and of course at discussing our projects

during coffee breaks.

Ich bedanke mich außerdem speziell noch mal bei Thomas Hartung, für die Übernahme der

Zweitgutachtertätigkeit.

Natürlich geht ein großes Dankeschön an die AG Leist, den Doktoranden des RTG-1331 und

alle Studenten, die den Aufenthalt für mich in Konstanz zu einer schönen Zeit haben werden

lassen. Speziell danke ich:

• der LUHMES Gruppe –Domi, Diana, Matze, Simon und Mila – wir waren ein super

Team!

• Den Mädels (Lisa, Hanne und Giorgie) und Christiaan auf Zimmer Z905 – es gibt kein

besseres Büro!

• an alle ehemaligen sowie derzeitigen Kollegen, die immer für lecker Kuchen und viel

Spaß an gemeinsamen Abenden gesorgt haben!

Außerdem möchte ich mich bei meiner Familie bedanken – besonders bei meinen Eltern,

Harald und Karin – die immer für mich da waren, meinem Bruder Sebastian mit Familie, und

allen, die mein Fernweh nach München erträglich gemacht haben.

Mein größter Dank geht an meinen Thomas. Pass auf – ich komme!!

Evaluation of new alternative strategies to predict ...

Documents

Transcript of Evaluation of new alternative strategies to predict ...