The autophagic machinery ensures nonlytictransmission of mycobacteriaLilli Gerstenmaiera,1, Rachel Pillaa,1, Lydia Herrmanna, Hendrik Herrmanna,2, Monica Pradoa, Geno J. Villafanoa,Margot Kolonkoa, Rudolph Reimerb, Thierry Soldatic, Jason S. Kingd, and Monica Hagedorna,3

aSection Parasitology, Bernhard Nocht Institute for Tropical Medicine, 20359 Hamburg, Germany; bElectronmicroscopy, Heinrich-Pette-Institute, 20251Hamburg, Germany; cDepartment of Biochemistry, University of Geneva, 1211-Geneva, Switzerland; and dDepartment of Biomedical Sciences, University ofSheffield, Sheffield S10 2TN, United Kingdom

Edited by Ralph R. Isberg, Howard Hughes Medical Institute, Tufts University School of Medicine, Boston, MA, and approved January 7, 2015 (received forreview December 9, 2014)

In contrast to mechanisms mediating uptake of intracellular bacterialpathogens, bacterial egress and cell-to-cell transmission are poorlyunderstood. Previously, we showed that the transmission of path-ogenic mycobacteria between phagocytic cells also depends onnonlytic ejection through an F-actin based structure, called theejectosome. How the host cell maintains integrity of its plasmamembrane during the ejection process was unknown. Here, wereveal an unexpected function for the autophagic machinery innonlytic spreading of bacteria. We show that ejecting mycobacteriaare escorted by a distinct polar autophagocytic vacuole. If autophagyis impaired, cell-to-cell transmission is inhibited, the host plasmamembrane becomes compromised and the host cells die. Thesefindings highlight a previously unidentified, highly ordered interac-tion between bacteria and the autophagic pathway and might rep-resent the ancient way to ensure nonlytic egress of bacteria.

autophagy | Dictyostelium discoideum | Mycobacterium marinum | ejection

In recent years, our understanding of the interactions betweenthe host autophagic machinery and intracellular pathogens has

rapidly expanded. These interactions are complex; although, inmany cases, the engagement of autophagy protects the host bycapturing and destroying the pathogen, some bacteria activelysubvert this pathway to promote their own survival (reviewed inref. 1). Autophagy has also been suggested to promote cell-to-cell transmission of Brucella (2, 3), although the molecularmechanisms are unknown.Both Mycobacterium tuberculosis, which causes tuberculosis in

humans, and the closely related species M. marinum have beenshown to interact with the autophagy machinery of their host cell(4–7). After uptake by immune phagocytes, the bacteria arrestphagosomal maturation and convert their vacuole into a replica-tion-permissive compartment. Both bacteria can translocate intothe host cell cytosol dependent on an intact Region-of-Difference-1-locus (RD1) (8–11). The genomic RD1-locus encodes a secre-tion system, ESX-1 (Type-VII secretion system), which has beenassociated with mycobacterial virulence (ref. 12, reviewed in refs.13 and 14). Once in the cytosol, M. marinum becomes ubiqui-tinated (4) likely recruiting adaptor proteins, such as membersof the sequestosome-1 family (SQSTM1), which also bind LC3(microtubule-associated proteins 1A/1B light chains 3A/LC3A and3B/LC3B), here referred to as Atg8, on autophagosomal mem-branes. In this way, bacteria are normally targeted to autophago-somes and killed, but M. marinum efficiently escapes this fate,most probably by shedding the ubiquitinated material as a decoy(4). However, infection by M. tuberculosis can be overcome bystimulating the classic autophagic pathway (15) and autophagy canreduce the bacterial burden in vivo (7).It was previously thought that M. marinum and M. tuberculosis

leave their host cell by inducing necrotic or apoptotic cell death(16). However, we recently showed that these bacteria alsoexit their host cell and spread via an F-actin structure, termedthe ejectosome (17). This form of egress, which is common to

M. tuberculosis and M. marinum in the amoeba Dictyostelium, isnonlytic for the host cell, even though its plasma membrane isperforated at the site of ejection. Previously, we showed thatejectosome formation is dependent on ESAT-6 (Early secretoryantigenic target 6), a secreted virulence factor encoded in theRD1-locus, and the Dictyostelium small GTPase RacH. How-ever, both the structure and mechanistic details of ejectosomefunction remain unknown.Using the Dictyostelium–M. marinum system (9, 17, 18) to

further dissect the mechanism of ejectosome formation andfunction, we demonstrate an unexpected role for autophagicmembranes in both mycobacteria egress and concomitant cell-to-cell transmission.

ResultsCorrelative Microscopy Reveals a Vacuolar Structure at the Distal Poleof Ejecting Bacteria. To better understand the mechanism ofnonlytic bacterial ejection, we examined the ultrastructure of theejectosome. Using a correlative approach, we were able toidentify ejectosomes by fluorescence microscopy (Fig. 1A andFig. S1 A and E), before ultrastructural analysis of serial thinsections by transmission electron microscopy (TEM) (Fig. 1 B–Dand Movie S1). Fig. 1C shows a representative section of a bac-terium at a very late stage of ejection with the proximal pole

Significance

Pathogenic mycobacteria can be transmitted by direct ejectionfrom one host cell to another. However, the mechanism ofejection, and how lysing the host cell is prevented are un-known. This study explains how the host cell remains intactand alive while Mycobacterium marinum breaks through itsplasma membrane during ejection. We show that a membra-neous cup is specifically recruited to the distal pole of ejectingM. marinum. We demonstrate that these membranes areformed by the canonical autophagic pathway, though they donot mature to autophagolysosomes. Disruption of autophagycauses the host cells to become leaky and die during ejection.This dramatically reduces cell-to-cell transmission of the in-fection, demonstrating an important and unexpected role forautophagy in maintaining plasma membrane integrity duringmycobacterial infection.

Author contributions: L.G., R.P., R.R., T.S., and M.H. designed research; L.G., R.P., L.H.,H.H., M.P., G.J.V., M.K., J.S.K., and M.H. performed research; L.G., R.P., H.H., M.P.,G.J.V., M.K., and M.H. analyzed data; R.R., T.S., J.S.K., and M.H. contributed newreagents/analytic tools; and J.S.K. and M.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1L.G. and R.P. contributed equally to this work.2Deceased July 28, 2014.3To whom correspondence should be addressed. Email: hagedorn@bnitm.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1423318112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1423318112 PNAS | Published online February 2, 2015 | E687–E692

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Apr

il 4,

202

0

(white arrowhead) being extracellular and the distal pole (whitearrow) lagging behind within the cell. The plasma membrane,which is ruptured at the site of ejection, is tightly apposed to thebacterium (Fig. 1C’, arrows). Strikingly, in every electron mi-crograph, the distal pole of the bacterium in the course ofejection was tightly enclosed by a vacuolar structure. Both spa-cious (with an electron lucent lumen, Fig. 1D and Fig. S1 F andG), and flat cisternal structures were observed (Fig. S1 B–D), but3D-reconstruction using serial thin sections revealed these cis-ternal structure always formed a vacuole around the distal bac-terial pole (Fig. 1E, Fig. S2, and Movie S2).

The Autophagic Machinery Is Present at the Distal Pole of EjectingBacteria. The appearance of this polar vacuole (Fig. 1E andMovie S2) is reminiscent of an expanding autophagic membrane.We therefore looked for the presence of autophagosomalmarkers around bacteria in the course of ejection. Consistentwith the TEM images, immunofluorescence imaging showed thatAtg8, a protein that specifically incorporates into expandingautophagosomal membranes (19), strongly localized around thedistal pole of ejecting M. marinum (Fig. 2A and Movie S3). ThisAtg8-positive cup was associated withM. marinum at all stages ofejection, from the very early stages when the bacterium is mainlyinside the cell, until the bacterium is completely ejected, wherean Atg8-positive cup still caps the extremity of the almost fullyextracellular bacterium (Fig. 2 B–D).

The presence of Atg8-containing membranes at bacteria dur-ing ejection indicates the specific recruitment of the autophagicmachinery. Selective autophagy is mediated by ubiquitinationof the target and recruitment of adaptor proteins such asSQSTM1 that contain both ubiquitin- and Atg8-binding domains(reviewed in ref. 20). Consistent with this pathway, both ubiq-uitin and DdSQSTM1 (GFP-SQSTM1), the single Dictyosteliumortholog of SQSTM1 accumulated in a pocket around the distalpole of ejecting bacteria, similarly to Atg8 (Fig. 2 E–G). Extra-cellular M. marinum was never associated with Atg8 or GFP-SQSTM1, indicating that the autophagic membrane is retainedinside the host. Importantly, the bacterial localization of bothubiquitin and Atg8 was largely restricted to ejecting bacteria.Although less than 15% of total cytoplasmic bacteria were as-sociated with ubiquitin (Fig. 2 H and I) or Atg8 (Fig. 2 H and J)

C*

D

BA

E

* *

*

C’

ActinM. marinum

Fig. 1. A vacuole caps the distal pole of ejecting M. marinum. (A) A Dic-tyostelium cell ejecting a M. marinum bacterium was localized by confocalfluorescence microscopy. The bacterium is shown in red, actin is shown ingreen. (B) Overlay of the corresponding brightfield and transmission elec-tron microscopy (TEM) image after processing. (Scale bar: 2 μm.) (C) TEMimage of the ejecting bacterium. The white arrow indicates the distal pole ofthe ejecting bacterium, the white arrowhead points toward the proximalpole. (Scale bar: 2 μm.) (C’) shows a magnification of the region where thebacterium perforates the plasma membrane. Typical for ejection, the plasmamembrane is protruding and tightly apposed to the ejecting bacterium(indicated by black arrows). (D) High magnification of the distal pole of theejecting bacterium. (Scale bar: 500 nm.) Black arrowheads indicate the vac-uolar membrane apposed to the bacterium, black arrows point to themembrane exposed to the host cell cytosol. The actin-rich ejectosome is in-dicated with an asterisk in all images. (E) 3D-model of the vacuolar pocketaround the pole of an ejecting bacterium. Serial thin sections were imagedby TEM and surface rendered. Bacteria are depicted in red, polar vacuole isdepicted in yellow, and plasma membrane is depicted in green.

Atg8Atg8Atg8

UB

G

GFP-SQSTM1

F

B C D

20

40

60

80

100

0

%as

soci

atio

n

UB Atg8

intracellular ejectingH

AAtg8

UB

I

J

Atg8

GFP

E

Fig. 2. The autophagic machinery is specifically recruited to the distal poleof ejecting M. marinum. (A–D) Fluorescence microscopy of infected cellsfixed and stained with anti-Atg8 showing polar accumulation of Atg8 atdifferent stages of ejection. Early in ejection, Atg8 (green) is found as a largepocket around the bacterium (outlined with a broken white line) (B), whichgets more focused as the bacterium becomes more extracellular (C and D).The outline of the cell is depicted with a red line. (E–G) Fluorescence mi-crograph of an infected cell expressing GFP (E), GFP-SQSTM1 (F), or stainedwith anti-ubiquitin (UB; G). (A and E–G) Bacteria are shown in blue, actin isshown in red, and the respective marker is shown in green. Ejectosomesare indicated by a white arrow, the distal poles of bacteria are highlightedwith a white arrowhead. (H) Quantification of association of ubiquitin andAtg8 with cytoplasmic and ejecting bacteria. n = 3. (I and J) CytoplasmicM. marinum associated with ubiquitin (UB, I) and Atg8 (J). The bacteria areshown in blue, the respective marker is shown in green. (Scale bars: 2 μm.)

E688 | www.pnas.org/cgi/doi/10.1073/pnas.1423318112 Gerstenmaier et al.

Dow

nloa

ded

by g

uest

on

Apr

il 4,

202

0

at 24 h postinfection (hpi), all ejecting bacteria were labeled attheir intracellular pole (Fig. 2H). Furthermore, although thenonejecting cytosolic bacteria recruited ubiquitin, atg8, andGFP-SQSTM1 to patches along their entire surface (Fig. 2 I andJ and Fig. S3), this was restricted to the distal pole during ejec-tion. These observations indicate that the recruitment of theautophagy machinery during ejection is by a specialized pathwaywith a higher level of regulation and organization.

The Canonical Autophagy Machinery Is Required for Cup Formation.To determine whether the recruitment of Atg8-positive mem-brane to ejecting bacteria required the canonical autophagypathway, we infected Dictyostelium mutants with a defective atg1gene (atg1−). Atg1 (Ulk1/2) is a key component of the canon-ical autophagy initiation complex and the Dictyostelium atg1−mutant is strictly impaired in autophagy (21). atg1− null cellswere however still able to form ejectosomes closely resemblingthe actin-rich structures generated in wild-type cells, with thesame frequency (Fig. 3 A and B). In atg1− cells, none of theejecting bacteria were associated with Atg8 (Fig. 3C), althoughhalf of them (46.4% ± 3.6) retained the association with ubiq-uitin at their distal pole (Fig. S4). Consistent with this, TEManalysis in atg1− cells showed a total loss of the vacuolar cupstructures around the distal pole of ejecting bacteria (Fig. 3 Dand E). To test the requirement for other components of thecanonical autophagic pathway, we also tested for the recruitmentof Atg8 to ejecting bacteria in mutants of the ubiquitin-likeconjugation machinery (Atg5 and Atg7) and the phosphatidylinositol 3-kinase (PI3K) complex (Atg6/Beclin1−) (Fig. 4A).We found that atg5 and atg7 were strictly required for asso-

ciation of Atg8 with the ejectosome, whereas disruption ofone of the two Dictyostelium atg6 orthologs gave a significant,although less penetrant defect, consistent with the partial blockin autophagy reported by others (22) (Fig. 4A). We thereforeconclude that the core autophagic machinery is required for theformation of the ejectosome-associated vacuole.

To characterize this compartment further, we also examined thelocalization of both GFP-Atg18 (ortholog of the mammalianproteins WIPI1/2), and the PI(3)P reporter GFP-2xFYVE. Ca-nonically, both proteins are only recruited to the early, expanding

1

2

3

4

0WT atg1-

%ej

ecto

som

es

A B

D E F

C M. marinumActinAtg8

M. marinumActin

atg1- atg1- WT

Fig. 3. The atg1− mutant forms nonfunctional ejectosomes. (A) Ejectosomesare present in atg1− cells (white arrowheads). Actin is shown in red, bacteriumin blue. (B) Ejectosome frequency is unaffected in atg1− cells compared withwild-type cells (WT), n = 3. Infected atg1− cells stained for Atg8 (C) illustratethe absence of Atg8 recruitment. Actin is in red, Atg8 antibody staining ingreen. (D and E) Transmission electron micrograph of an ejectosome in atg1−cells confirms absence of the distal vacuole. (D) The ejectosome region is in-dicated with white arrowheads. (Scale bar: 1 μm.) (E) The distal pole of theejecting bacterium is cytosolic and not capped by a vacuole. (Scale bar:500 nm.) Material spilling from the ejectosome region is indicated with blackarrows (D and E). (F) A transmission electron micrograph of the distal pole ofan ejecting bacterium from wild-type cells (WT). The vacuole engulfing thebacterium is indicated with black arrows. (Scale bar: 500 nm.)

1x104

5x104

4x104

3x104

2x104

non-inf.

inf. non-inf.

inf.

WT atg1-

***n.s.

Mea

nP

I-flu

ores

cenc

epe

rce

ll

non-inf.

inf.

WT

non-inf.

inf.

atg1-

0

10

20

30

%de

adce

lls

***n.s.H I

G Atg8

F Atg8

Infe

cted

acce

ptor

per

infe

cted

dono

rce

lls

E

6 24 26 32hours post infection

0

0.2

0.4

0.6

0.8WTracH-atg1-

M. marinum RD1Atg8

M. smegmatisAtg8

DC

A B

20

40

60

80

100

%as

soci

atio

n

Atg18

2xFYVE

WT

atg5

-at

g6-

atg7

-

SQSTM1-

20

40

60

80

100

%A

tg8

asso

ciat

ion

***

n.s.

***

***

** *** **

Fig. 4. Elimination of autophagy reduces cell-to-cell transmission and cau-ses cell leakage. (A) Dictyostelium knockout mutants were infected withM. marinum and the recruitment of Atg8 to the distal poles of ejectosomeswas scored in three independent experiments. Statistical significance was cal-culated using the one way ANOVA. (***P ≤ 0.001; n.s., not significant). (B)Wild-type cells expressing GFP-Atg18 and GFP-2xFYVE were infected withM. marinum and the accumulation of the respective marker at the distal poleof ejecting bacteria was scored (n = 3). (C and D) Dictyostelium wild-type cellswere coinfected (Fig. S5) with M. marinum wild-type and M. marinum ΔRD1(C) and M. smegmatis (D), respectively. Ejectosome structures (white arrow-heads) are formed by both, M. marinum ΔRD1 (C, blue) and M. smegmatis(D, blue). Distinct Atg8 (green) localization at the distal pole (white arrows)is observed for both, M. marinum ΔRD1 (C) and M. smegmatis (D) alikeM. marinum wild-type. (E) Cell-to-cell transmission was measured by flowcytometry. Acceptor cells (wild-type) were mixed with infected donor cells(wild-type, racH−, and atg1−) at 6 hpi and transmission measured at 6, 24, 28and 32 hpi. Both racH− and atg1− cells are significantly impaired in trans-mitting bacteria to acceptor cells (n = 3, mean ± SEM; *P ≤ 0.1; **P ≤ 0.01).Two-way ANOVA analysis). (F and G) Ejectosome structures (white arrowhead)formed by atg1− cells showed spilling of cytosolic Atg8 (green) on the extra-cellular side of the ejecting bacterium (indicated by white arrow). (Scale bar:2 μm.) (H) In contrast to wild-type (WT) Dictyostelium cells, atg1− cells showa significant increase in mean PI accumulation upon M. marinum infection(24 hpi), n = 3. Similarly, cell death was significantly increased (I) when atg1−cells were infected with M. marinum, whereas wild-type cells were unaffected(24 hpi). Statistical analysis was performed using one-way ANOVA analysis (n =3–4; mean ± SEM; ***P ≤ 0.001). (Scale bars: 2 μm.)

Gerstenmaier et al. PNAS | Published online February 2, 2015 | E689

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Apr

il 4,

202

0

isolation membrane, in contrast to Atg8, which remains associatedwith the autophagosome throughout (23, 24). We observed thatAtg18 localized to 38 ± 2.9% and 2xFYVE was present at 8.9 ±4.4% of ejecting bacteria (Fig. 4B). This indicates that the pocketis formed at the site of ejection, by the canonical autophagicpathway. Furthermore, as Atg8 could be observed on >90% ofejecting bacteria (Fig. 2H), this pocket must also be undergoingmaturation to some extent. However, when we labeled the ly-sosomal system by feeding cells fluorescent dextran, we neverobserved an accumulation at ejecting bacteria, indicating thatfusion with endolysosomes does not occur (Fig. S6).To examine how the autophagic machinery is recruited, we

also disrupted the recently identified Dictyostelium ortholog ofSQSTM1, the sole known autophagic adaptor in this organism(22). Surprisingly however, loss of SQSTM1 had no affect onAtg8 recruitment. The alternative autophagy adaptors identifiedin mammals are poorly conserved through evolution and no clearorthologs can be identified in Dictyostelium. Therefore, eitheradditional unknown adaptors must exist, or an alternativemechanism of recruitment is responsible for the recruitment ofthe autophagy machinery.

Recruitment of the Autophagic Machinery Is ESX-1 Independent. Thespecific recruitment of the autophagy machinery to the distal poleof ejecting bacteria raises the question whether polar mycobac-terial virulence factors are involved in this recruitment andwhether this is specific for pathogenic M. marinum. The virulencesecretion system ESX-1 is enriched at the poles of M. marinum(25) and is encoded in the RD1-region. Involvement of this se-cretion system in ejection has been demonstrated before (17). Wetherefore asked whether local secretion activity of ESX-1 at thepole might be involved in the observed activation of autophagy. AsESX-1 is required for the bacteria to escape from phagosomesinto the cytosol before ejection, we established a coinfectionprotocol that used wild-type M. marinum to break phagosomesthat contained M. marinum mutants lacking the RD1-locus(M. marinum ΔRD1) at the same time (assay depicted in Fig. S5).Immunofluorescence microscopy showed that both wild-type andΔRD1 M. marinum, were able to form ejectosomes and recruitAtg8 to their distal poles (Fig. 4C). Therefore, ESX-1 –mediatedESAT-6 secretion is required for ejectosome formation (17), thepolarization of this secretion system and its local activity does notmediate recruitment of the autophagy machinery. To monitorwhether the observed ejection is restricted to pathogenicmycobacteria we performed coinfection with nonpathogenicM. smegmatis. Immunofluorescence micrographs showed thatM. smegmatis was indeed able to form an ejectosome and theautophagy machinery was recruited to its distal pole (Fig. 4D).

Autophagy Is Required for Efficient Nonlytic, Cell-to-Cell Transmission.Although not required for their formation, we next asked whetherautophagy was required for ejectosome function. Ejection is themajor mechanism of cell-to-cell transmission in Dictyostelium (17).Therefore, to determine whether ejectosomes generated in atg1-deficient cells are functional, we used a flow cytometry-based assayto monitor the transmission of GFP-expressing M. marinum frommutant donor cells to mRFP-expressing wild-type acceptor cells(Fig. S7). The acceptor cells were added to infected donor cells(wild-type, atg1−, or racH−) at 6 hpi and transmission was mea-sured starting from 24 hpi onwards. Whereas wild-type cells effi-ciently transmitted the bacteria to the acceptor cells, transmissionwas reduced by 50% in atg1− mutants (Fig. 4E). This decrease intransmission is comparable to the defect in racH− cells that wepreviously showed to be deficient in nonlytic cell-to-cell trans-mission due to their complete inability to form ejectosomes (17).Therefore, although not required for ejectosome biogenesis,autophagy is essential for efficient cell-to-cell transmission.

Although atg1− cells still form a tightly localized actin ringaround ejecting bacteria, we frequently observed by light-mi-croscopy cytoplasmic material spilling out on the extracellularside of the ejectosome (Fig. 4 F and G). Extracellular materialaround the ejecting bacterium was also observed by EM (Fig. 3 Dand E, black arrows). This was never observed in wild-type cells,suggesting that the ejectosomes in atg1− cells may not be closingas tightly around the ejecting bacteria. We therefore used thenuclear accumulation of the membrane impermeable dye pro-pidium iodide (PI) to determine the membrane integrity ofinfected autophagy-deficient cells. After a 10 min exposure ofinfected cells (24 hpi) to PI, 40 ± 3.2% of live atg1− cells showeda significantly higher mean fluorescence than noninfected atg1−cells indicating that the cells become partially permeable to thedye (Fig. 4H and Fig. S8). In contrast, the mean fluorescence ofwild-type cells increased only moderately (10 ± 6%) upon in-fection. Cell leakiness will also eventually result in cell death.Although other ejectosome-independent roles of autophagy mayalso contribute to this, when we quantified the number of deadcells at 24 hpi using the PI-staining and the characteristic for-ward/side scatter of dead cells (Fig. S8) we found significantlymore dead cells when autophagy is blocked (Fig. 4I).

DiscussionIn this paper, we describe a previously unidentified interactionbetween pathogenic M. marinum and the autophagic machinery.In contrast to previously described host-pathogen interactions,this is specific to ejecting bacteria and is required for the trans-mission of bacteria to a naïve host cell. Although this mechanismof transmission clearly benefits the bacteria, it might be con-sidered to also benefit the host, as the polar autophagic vac-uole apparently prevents plasma membrane leakage and hostcell death.Recently, a number of studies have demonstrated autophagy-

independent roles for several autophagy-related proteins. Weshow that the recruitment of the double-membrane cup to ejectingbacteria requires proteins from the canonical autophagy pathway.Importantly, this includes the Atg1 initiation complex, whichis dispensible for the autophagy-independent recruitment ofAtg8/LC3 to the single membranes of phagosomes and macro-pinosomes in mammalian cells (26). In addition, the requirementfor the PI3K complex subunit Atg6 and the ubiquitin-like con-jugation machinery components Atg5 and Atg7, as well as thepresence of PI(3)P and Atg18 at the bacterial pole indicate thatthe ejectosome-associated vesicle is formed at the site of ejectionby the canonical autophagy machinery. However, although thepresence of GFP-Atg18 on only a subset of these vesicles indi-cates that they undergo some maturation, the lack of dextrandelivery to this compartment implies that it does not undergofusion with the endolysosomal system and therefore is unlikely tobecome degradative.The autophagic membrane is tightly restricted to the distal

pole of ejectosome-associated bacteria, however, it is unclearhow this polar recruitment is orchestrated. It has previously beenreported that the autophagy machinery can be recruited todamaged membranes via galectins (27) and the protrusion ofa bacterium through the plasma membrane may well be inter-preted by the cell as plasma membrane wounding. However, atearly stages of ejection, the distal pole of these rod-shapedbacteria is several microns away from the plasma membrane(Fig. S1C), and it is therefore possible that the recruitment ofthis vacuole is directed by the bacteria themselves in some way.The virulence secretion system ESX-1 localizes to the poles of

M. marinum (25) and is encoded in the RD1-region. Involvementof this secretion system in ejection has been demonstrated before(17) in particular the secreted factor ESAT-6. However, theresults of our coinfection assay show that neither the RD1 regionnor M. marinum-specific virulence factors are responsible for

E690 | www.pnas.org/cgi/doi/10.1073/pnas.1423318112 Gerstenmaier et al.

Dow

nloa

ded

by g

uest

on

Apr

il 4,

202

0

distal vacuole generation, which would rather favor a host-drivenprocess. At a molecular level, the accumulation of ubiquitin andSQSTM1 suggest recruitment of the autophagy machinery byfactors which are directly present on the bacterial surface. Theobservation that bacteria ubiquitination is partly reduced whenatg1 is disrupted (Fig. S4) also implies that the autophagicmembrane somehow helps stabilize the ubiquitination. The re-cruitment of Atg8 even in the absence of SQSTM1, suggests theinvolvement of additional autophagy adaptors. In mammaliancells, multiple adaptors have been shown to mediate xenophagyincluding NDP52 and optineurin (28, 29). Although orthologs ofthese proteins have yet to be found in lower eukaryotes, giventhe universal challenge of containing intracellular pathogens(particularly if you are a professional phagocyte) it seems highlylikely that additional adaptors will exist.Previously identified roles for autophagy during infection are

aimed at capturing the bacteria within a vacuole, either to kill it, orallow it to replicate. In contrast, during ejection the autophagicmembrane is required both to maintain host cell plasma mem-brane integrity and promote cell-to-cell transmission. Plasmamembrane damage can be sealed by lysosomal fusion in a calcium-dependent fashion (30). Recently, a new calcium-dependentwound healing mechanism has been discovered. The endosomalsorting complexes required for transport (ESCRT) machineryhas been demonstrated to catalyze the shedding of small vesiclescarrying membrane wounds (31). Even though no molecularmechanism has been proposed, autophagosome interactions withthe plasma membrane have been suggested to allow the releaseof virus particles (32) and the cell-to-cell transmission of Brucella(2). Recently, using a fluorescent timer approach, it was shownthat the autophagic pathway contributes to the shedding ofmicrovesicles that harbor infectious coxsackievirusB and con-tribute to the spreading of the virus from cell to cell (33).Autophagy has also been shown to promote the cell-to-cell

transmission of other pathogens, although the molecular mech-anisms are unknown (2, 3). In contrast to our observations, inwhich the ejecting bacteria are cytosolic, Starr and colleaguesshowed that the Brucella replicating vacuole converts into a vac-uole with autophagic features, and that this is independent ofAtg5 and Atg7. Interestingly, they also show that this conversionis necessary for efficient bacterial transmission. Furthermore,they also report no increase in cell death upon Brucella release,suggesting a nonlytic form of egress, reminiscent of ejectosome-mediated transmission.Here, we show that the autophagic machinery is necessary for

nonlytic cell-to-cell transmission of M. marinum. We proposethat the membrane generated by the autophagic machinery atthe distal pole of ejecting bacteria helps seal the membranewound generated by the ejection (Fig. S9). The loss of the polarautophagic membrane leads to increased permeability at theejectosome and death of the host cell.The range of interactions between intracellular pathogens and

the autophagy machinery continues to expand. The previouslyunidentified and unexpected role for autophagy highlighted inour study indicates the complexity of host-pathogen interactionsthat have evolved over millions of years, and how pathogenvirulence and host defense pathways have reached a form ofequilibrium that benefits both organisms.

Materials and MethodsDictyostelium Cell Culture.Wild-type Dictyostelium cells (Ax2) were axenicallycultivated in HL5c medium (Formedium) at 22 °C. The Dictyostelium atg1−cells were described (34) and racH− cells from F. Rivero (University of Hull,Hull, United Kingdom). Dictyostelium mutants atg5−, atg6−, and atg7−were received from dictyBase (www.dictybase.org). Dictyostelium mutantSQSTM1- was generated using the pKOSG-IBA-Dicty1 system from Stargate.

Mycobacteria Cultivation, Strains, and Plasmids. Mycobacteria were cultivatedwith agitation at 32 °C in Middlebrook 7H9 (Difco), supplemented with10% (vol/vol) OADC (oleic acid, albumin, dextrose, catalase) (Becton Dickinson),5% (vol/vol) glycerol and 0.2% Tween80 (Sigma Aldrich). M. marinum wildtype (M-strain) and the msp12::GFP plasmid were kindly provided byL. Ramakrishnan (Washington University, Seattle) and the pCHERRY3 plas-mid (No. 24659; ref. 35; Tanya Parish, Infectious Disease Research Institute,Queen Mary’s School of Medicine, University of Washington, Seattle) wassupplied by Addgene. Transformed bacteria were selected either withhygromycin (pCHERRY) or kanamycin (msp12::GFP) at 50 μg/mL.

Phalloidin and Antibodies. Rabbit polyclonal antibody was raised against full-length recombinant Atg8, FK2H (directed against ubiquitinylated conjugates)was bought from Enzo (BML-PW0150), and anti-RFP from Chromotek (5F8).Secondary antibodies were goat anti-rabbit or goat anti-mouse/chicken IgGcoupled to Alexa 488, Alexa 568, or Alexa 647 (Invitrogen). Phalloidin stainingwas performed with Alexa Fluor488-, Alexa Fluor568-, or Alexa Fluor647-phalloidin (Molecular Probes).

Infection Assay. Mycobacteria were grown to an OD600 =1 (∼5 × 108 cells permL) and clumps disrupted by passing through 50 μm CellTrics-filters (Partec)and blunt 25-gauge needles (Dispomed, Neoject). Adherent Dictyosteliumwild type or mutant cells were grown overnight without antibiotics toa density of 80–100%. Bacteria were added at a multiplicity of infection(MOI) of 50–100 and centrifuged on the Dictyostelium cells (500 × g, fourtimes for 5 min each). The cells were left to phagocytose for additional10–30 min before free bacteria were washed off with HL5c. To inhibit ex-tracellular proliferation of bacteria the cells were subsequently resuspendedin HL5c supplemented with 5 μg/mL streptomycin and maintained at 25 °C.The moment when the bacteria were added to the attached Dictyosteliumcells was defined as 0 hpi.

Immunofluorescence and Phalloidin Staining. Dictyostelium cells were infectedas above with GFP or mCHERRY expressingM. marinum. At 24 hpi, cells werecentrifuged on poly-lysine coated coverslips (500 × g, 5 min) and fixed eitherin methanol (−80 °C, 1 h) or with Soerensen buffer (14.7 mM KH2PO4,2.5 mM NaHPO4, pH 6.3) containing 4% paraformaldehyde and stained asdescribed (9, 36). The fluorescence images were documented using anOlympus IX81 confocal microscope with a 60× 1.35 NA or 100× oil immersionobjective and Fluoview software v1.7b. Recording parameters for fields of1024 × 1024 pixels with appropriate electronic zoom (6–12×) were 3× lineaveraging (Kalman). To adjust the brightness and contrast of completeimages ImageJ (imagej.nih.gov/ij/) was used.

Correlative Light and Electron Microscopy (CLEM). The combination of lightand electron microscopy was used to analyze the ultrastructure of ejecto-somes. Adherent Dictyostelium cells expressing Lifeact-GFP were infectedas described above with mCHERRY expressing M. marinum. Infected cellswere plated in a sterile 35 mm μ-dish with an imprinted grid (Ibidi). After24 h post infection (hpi), the cells were fixed for two hours with 2% para-formaldehyde (Electron Microscopy Sciences) and 0.25% glutaraldehyde(Electron Microscopy Sciences) in HL5c medium. Fixed cells were washed andstained for 30 min with Alexa Fluor488-phalloidin. Localization and fluo-rescence images of ejectosome structures were documented using theOlympus IX81 confocal microscope as described above.

Subsequently, the cells were fixed with 2.5% glutaraldehyde in 1× PBS(137 mM NaCl, 2.7 mM KCl, pH 7.4) overnight. After washing the samplestwice, they were incubated with OsO4 (1% in PBS, 30 min), washed againand finally incubated with gallic acid (1% in 0.5× PBS, 30 min). After sampledehydration in a rising ethanol series they were embedded in 50% Epoxyresin (Roth, 8619.2) in 100% Ethanol overnight at room temperature andpolymerized at 60 °C with fresh 100% Epoxy resin for 6 h. Consecutive ul-trathin sectioning was performed with a Leica EM UC7 μL tramicrotome andthe samples were examined with a Tecnai Spirit transmission electron mi-croscope at 80 kV (FEI). For 3D reconstruction the image stack was alignedwith ImageJ (stackreg; imagej.nih.gov/ij/) and reconstructed in Imaris (6.2).Movies were exported from Imaris and rendered in AdobeAfterEffectsand AdobeMediaEncoder.

Transmission Assay. All Dictyostelium strains (Ax2, racH−, and atg1−, donorcells) were infected as described above with GFP-expressing M. marinum.Infection rates were measured by flow cytometry (Accuri C6, BD Biosciences),and normalized to each other by dilution with the appropriate strain. At 6 hpithe infected donor cells were mixed with RFP-expressing acceptor Dictyos-telium cells at a 1:5 ratio. At each timepoint (6, 24, 28, and 32 hpi), cells were

Gerstenmaier et al. PNAS | Published online February 2, 2015 | E691

CELL

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Apr

il 4,

202

0

fixed with 4% paraformaldehyde in Soerensen buffer and immunostained withan anti-RFP antibody to enhance the red signal. Subsequently, 250,000 cellswere measured by flow cytometry. The results were analyzed and plottedwith Flowjo 7.6.3 (TreeStar). To quantify transmission of M. marinum fromdonor to acceptor cells the ratio of infected acceptor versus infected donorcells was calculated.

Permeability Test with Propidium Iodide (PI). Dictyostelium cells were infectedwith GFP-expressing bacteria as described above. At 24 hpi 2 × 106 cells werecentrifuged (5 min, 500 × g) and resuspended in 500 μL Soerensen sorbitolbuffer (120 mM sorbitol). PI (Sigma Aldrich, stock solution: 0.5 mg/mL) wasadded at a dilution of 1:10 and fluorescence (FL3) of 250,000 cells wasmeasured (530 nm excitation, 610 nm emission) by flow cytometry (AccuriC6, BD Biosciences). As a positive control for dead and leaky cells, Dictyos-telium cells were heat killed (60 °C, 10 min), stained, and measured asdescribed above.

To obtain the percentages of permeable cells which are right-shifted theshown, data were normalized between compared data sets and areas werecomputed using the density estimate of channel FL3 (PI). Area estimates forthe overlap region between two curves (noninfected and infected) wascomputed and used to obtain the fraction of the total area shifted right ofthe overlap. All computations were preformed in R (R Core Team (2014). R:A language and environment for statistical computing. R Foundation forStatistical Computing, Vienna, Austria).

Live-Cell Imaging. Lifeact-RFP/Atg8-GFP–expressing Dictyostelium cells wereinfected with mCHERRY-expressing M. marinum and at 4 hpi transferred toa 35 mm μ-dish (Ibidi). At 6 hpi, nonfluorescent Dictyostelium cells were

added (ratio 10:1). The live cells were imaged at 25 °C using a Zeiss 510 Metaconfocal microscope equipped with a 63× Europlan apochromat oil immer-sion objective (N.A. 1.4) and the Zeiss confocal microscope software release3.2. Two different laser-lines were used (Argon-Laser 488 nm and HeNe-Laser 543 nm). The imaging was performed in a multitrack-mode with apinhole of 2 Airy units. Cells were imaged for 10 min with time intervals of6 s between each scan. Brightness and contrast adjustments to images, aswell as annotations, were performed using ImageJ.

Coinfection Assay. ABD-GFP expressing Dictyostelium cells were simulta-neously infected with both green fluorescent (GFP-expressing) M. marinumand red fluorescent (mCHERRY-expressing) M. smegmatis or M. marinumΔRD1 as described under Infection assay. The bacteria were grown sepa-rately in shaking culture and added to the adherent Dictyostelium cells atthe same time (defined as 0 hpi) with MOIs of 100 for M. marinum-GFP and30 for mCHERRY-expressing M. smegmatis or M. marinum ΔRD1. Phagocy-tosis and washing was performed as described (infection assay) and cellsprepared for immunofluorescence (as described) at 24 phi.

ACKNOWLEDGMENTS. We gratefully acknowledge Francisco Rivero for pro-viding Dictyostelium mutant racH− and Lalita Ramakrishnan for providingM. marinum and plasmids; and Ulrike Fröhlke and Carola Schneider for tech-nical help. This work was supported by The Leibniz Society, Deutsche For-schungsgemeinschaft (DFG; HA3474/3-1, M.H.), Ciencia sem fronteiras (CsF,BEX2607-13-1, postdoctoral fellowship to R.P.), Cancer Research-UK and TheRoyal Society Research Grant RG130655 (to J.S.K.), and a grant from the SwissNational Science Foundation (to the T.S. laboratory). This work would not havebeen possible without dictyBase (37).

1. Huang J, Brumell JH (2014) Bacteria-autophagy interplay: A battle for survival. NatRev Microbiol 12(2):101–114.

2. Starr T, et al. (2012) Selective subversion of autophagy complexes facilitates com-pletion of the Brucella intracellular cycle. Cell Host Microbe 11(1):33–45.

3. Delpeut S, Rudd PA, Labonté P, von Messling V (2012) Membrane fusion-mediatedautophagy induction enhances morbillivirus cell-to-cell spread. J Virol 86(16):8527–8535.

4. Collins CA, et al. (2009) Atg5-independent sequestration of ubiquitinated mycobac-teria. PLoS Pathog 5(5):e1000430.

5. Lerena MC, Colombo MI (2011) Mycobacterium marinum induces a marked LC3 re-cruitment to its containing phagosome that depends on a functional ESX-1 secretionsystem. Cell Microbiol 13(6):814–835.

6. Manzanillo PS, et al. (2013) The ubiquitin ligase parkin mediates resistance to in-tracellular pathogens. Nature 501(7468):512–516.

7. Castillo EF, et al. (2012) Autophagy protects against active tuberculosis by suppressingbacterial burden and inflammation. Proc Natl Acad Sci USA 109(46):E3168–E3176.

8. van der Wel N, et al. (2007) M. tuberculosis and M. leprae translocate from thephagolysosome to the cytosol in myeloid cells. Cell 129(7):1287–1298.

9. Hagedorn M, Soldati T (2007) Flotillin and RacH modulate the intracellular immunityof Dictyostelium to Mycobacterium marinum infection. Cell Microbiol 9(11):2716–2733.

10. Simeone R, et al. (2012) Phagosomal rupture by Mycobacterium tuberculosis resultsin toxicity and host cell death. PLoS Pathog 8(2):e1002507.

11. Houben D, et al. (2012) ESX-1-mediated translocation to the cytosol controls virulenceof mycobacteria. Cell Microbiol 14(8):1287–1298.

12. Brodin P, et al. (2006) Dissection of ESAT-6 system 1 of Mycobacterium tuberculosisand impact on immunogenicity and virulence. Infect Immun 74(1):88–98.

13. Abdallah AM, et al. (2007) Type VII secretion—mycobacteria show the way. Nat RevMicrobiol 5(11):883–891.

14. Simeone R, Bottai D, Brosch R (2009) ESX/type VII secretion systems and their role inhost-pathogen interaction. Curr Opin Microbiol 12(1):4–10.

15. Gutierrez MG, et al. (2004) Autophagy is a defense mechanism inhibiting BCG andMycobacterium tuberculosis survival in infected macrophages. Cell 119(6):753–766.

16. Saunders BM, Britton WJ (2007) Life and death in the granuloma: Immunopathologyof tuberculosis. Immunol Cell Biol 85(2):103–111.

17. Hagedorn M, Rohde KH, Russell DG, Soldati T (2009) Infection by tubercular myco-bacteria is spread by nonlytic ejection from their amoeba hosts. Science 323(5922):1729–1733.

18. Kolonko M, et al. (2014) WASH-driven actin polymerization is required for efficientmycobacterial phagosome maturation arrest. Cell Microbiol 16(2):232–246.

19. Ichimura Y, et al. (2000) A ubiquitin-like system mediates protein lipidation. Nature408(6811):488–492.

20. Johansen T, Lamark T (2011) Selective autophagy mediated by autophagic adapterproteins. Autophagy 7(3):279–296.

21. Otto GP, Wu MY, Kazgan N, Anderson OR, Kessin RH (2004) Dictyostelium macro-autophagy mutants vary in the severity of their developmental defects. J Biol Chem279(15):15621–15629.

22. Calvo-Garrido J, et al. (2010) Autophagy in Dictyostelium: Genes and pathways, celldeath and infection. Autophagy 6(6):686–701.

23. Polson HE, et al. (2010) Mammalian Atg18 (WIPI2) localizes to omegasome-anchoredphagophores and positively regulates LC3 lipidation. Autophagy 6(4):506–522.

24. King JS, Veltman DM, Insall RH (2011) The induction of autophagy by mechanicalstress. Autophagy 7(12):1490–1499.

25. Carlsson F, Joshi SA, Rangell L, Brown EJ (2009) Polar localization of virulence-relatedEsx-1 secretion in mycobacteria. PLoS Pathog 5(1):e1000285.

26. Florey O, Kim SE, Sandoval CP, Haynes CM, Overholtzer M (2011) Autophagy ma-chinery mediates macroendocytic processing and entotic cell death by targetingsingle membranes. Nat Cell Biol 13(11):1335–1343.

27. Thurston TL, Wandel MP, von Muhlinen N, Foeglein A, Randow F (2012) Galectin 8targets damaged vesicles for autophagy to defend cells against bacterial invasion.Nature 482(7385):414–418.

28. Thurston TL, Ryzhakov G, Bloor S, von Muhlinen N, Randow F (2009) The TBK1adaptor and autophagy receptor NDP52 restricts the proliferation of ubiquitin-coatedbacteria. Nat Immunol 10(11):1215–1221.

29. Wild P, et al. (2011) Phosphorylation of the autophagy receptor optineurin restrictsSalmonella growth. Science 333(6039):228–233.

30. Reddy A, Caler EV, Andrews NW (2001) Plasma membrane repair is mediated byCa(2+)-regulated exocytosis of lysosomes. Cell 106(2):157–169.

31. Jimenez AJ, et al. (2014) ESCRT machinery is required for plasma membrane repair.Science 343(6174):1247136.

32. Jackson WT, et al. (2005) Subversion of cellular autophagosomal machinery by RNAviruses. PLoS Biol 3(5):e156.

33. Robinson SM, et al. (2014) Coxsackievirus B exits the host cell in shed microvesiclesdisplaying autophagosomal markers. PLoS Pathog 10(4):e1004045.

34. King JS, et al. (2013) WASH is required for lysosomal recycling and efficient auto-phagic and phagocytic digestion. Mol Biol Cell 24(17):2714–2726.

35. Caroll P, et al. (2010) Sensitive detection of gene expression in mycobacteria underreplicating and non-replicating conditions using optimized far-red reporters. PLoSONE 5(3):e9823.

36. Hagedorn M, Neuhaus EM, Soldati T (2006) Optimized fixation and immunofluores-cence staining methods for Dictyostelium cells. Methods Mol Biol 346:327–338.

37. Fey P, et al. (2013) One stop shop for everything Dictyostelium: dictyBase and thedicty stock center. Methods Mol Biol 983:59–92.

E692 | www.pnas.org/cgi/doi/10.1073/pnas.1423318112 Gerstenmaier et al.

Dow

nloa

ded

by g

uest

on

Apr

il 4,

202

0