· Dissertation zur Erlangung des Doktorgrades der Agrar- und Ernährungswissenschaftlichen...

Schriftenreihe des Instituts für Tierzucht und Tierhaltung der

Christian-Albrechts-Universität zu Kiel, Heft 233, 2019

©2019 Selbstverlag des Instituts für Tierzucht und Tierhaltung

der Christian-Albrechts-Universität zu Kiel

Olshausenstraße 40, 24098 Kiel

Schriftleitung: Prof. Dr. J. Krieter

ISSN: 0720-4272

Gedruckt mit Genehmigung des Dekans der Agrar- und Ernährungswissen-

schaftlichen Fakultät der Christian-Albrechts-Universität zu Kiel

Aus dem Institut für Tierzucht und Tierhaltung

der Agrar- und Ernährungswissenschaftlichen Fakultät


Die Dissertation wurde mit dankenswerter finanzieller Unterstützung aus Mitteln des

Bundesministeriums für Ernährung und Landwirtschaft und der H. Wilhelm Schaumann Stiftung

angefertigt.

GENETIC PARAMETERS AND GENOMIC EVALUATION OF FEED

INTAKE AND ENERGY BALANCE IN GERMAN DAIRY COWS

Dissertation

zur Erlangung des Doktorgrades

der Agrar- und Ernährungswissenschaftlichen Fakultät


vorgelegt von

M.Sc. agr.

IMKE HARDER

aus Bad Segeberg, Schleswig-Holstein

Kiel, 2019

Dekan: Prof. Dr. Dr. C. Henning

Erster Berichterstatter: Prof. Dr. G. Thaller

Zweiter Berichterstatter: Prof. Dr. Hermann Swalve

Tag der mündlichen Prüfung: 26. Juni 2019

Meiner Familie

Table of contents

General Introduction 1

Chapter I:

Lactation curves and model evaluation for feed intake and energy balance in dairy

cows 7

Chapter II:

Estimation of genetic parameters and breeding values for feed intake and

energy balance using pedigree relationships or single-step genomic evaluation

in Holstein-Friesian cows 33

Chapter III:

Zucht auf Futteraufnahme mit Hilfe der genomischen Selektion 63

General Discussion 83

Outlook and ongoing research 101

General Summary 105

Zusammenfassung 109

General Introduction

1


Milk production per cow has been distinctly improved over the past decades (Spurlock et al., 2012).

Nevertheless, the higher amount of milk has been accompanied by a higher incidence of health and

fertility problems (Banos et al., 2005). These antagonistic relationships between production, health

and fitness affected the efficiency of the cow. Thus, the focus on breeding programs shifted milk yield

towards a more balanced approach. So far, health and fertility traits become the focus of livestock

farming (Egger-Danner et al., 2015). Feed intake and energy balance are such novel traits and their

inclusion in the breeding goal is useful for various reasons.

On the one hand, feed intake is a trait of economic interests, because it accounts for the largest

proportion of operating costs (Li et al., 2018). An increase in milk production and a decrease in feed

intake increase feed efficiency, which is per se desirable. However, on the other hand, the negative

side effects, especially the health related factors, which come along with an increase in efficiency of

the cows, can most notably be observed in terms of a negative energy balance in early lactation

(Collard et al., 2000).

In particular, early lactation is a critical phase for dairy cows, because rapidly increased milk

production results in an increase in feed requirement. During the early lactation, the milk yield raised

by 50 %, whereas the feed intake just raised by 25 % and a lack between energy intake and energy

requirement occurs (Brade and Brade, 2016). To meet the energy requirements at the same time, body

reserves are mobilized and a reduction in body weight is the consequence (Prendiville et al., 2011).

The cow slips into an energy deficit. The lost body reserves need to be refilled during later lactation

and lead to a decrease in efficiency in that stage (Mäntysaari et al., 2012). If the resulting metabolic

situation cannot be met, it affects the constitution of the cow, which in turn cause metabolic stress

(Roche et al., 2009; Buttchereit et al., 2011). Furthermore, the loss of body reserves in that phase can

causes metabolic diseases, impair reproduction performance (Roche et al., 2007) and higher risk of

health problems (Leesen et al., 2014). The disease rate of high performing dairy cows is higher due

to a longer lasting energy deficit which is a consequence of the high amount of milk yield. Thus, the

aim is to minimize the gap between intake and requirement, while simultaneously increasing the feed

intake to cover the demands of the cows’ milk production, especially during the critical period at the

beginning of lactation (Liinamo et al., 2012).

A direct breeding on higher feed intake with a resulting improved energy balance failed due to the

extensive and high costs related to measuring feed intake, which is only available in research herds

(Berry et al., 2014; Tetens et al., 2014). The insufficient recording hindered accurate estimation of


2

genetic parameters and prevented the implementation of this trait in the breeding goal (Berry and

Crowley, 2013; Li et al., 2016).

With the implementation of genomic selection, breeding for hard to measure traits like feed intake

and feed efficiency are possible now (Pryce et al., 2012; Yao et al., 2017). To make best use of

genomic values, selection requires a large reference population of animals with both phenotypes and

genotypes. Furthermore, for traits that are more difficult to measure, a cow reference population is

the most effective approach (Chesnais et al., 2016). The success of genomic breeding values depends

on their reliability. The reliability of genomic prediction is a function of the number of individuals in

a reference population (Goddard and Hayes, 2009), the heritability of the trait (Daetwyler et al., 2010;

Goddard and Hayes, 2009) and the relationship between the evaluated animals and a training data set

(Habier et al., 2010; Pszczola et al., 2012; Pszczola et al., 2018).

To predict genomic estimated breeding values, the method single-step provides the best possibilities

to combine genotyped, phenotyped and not-phenotyped animals within a single evaluation.

Furthermore, it has the potential to yield more accurate and less biased genomic evaluations (Legarra

et al., 2014; Guarini et al., 2018).

For an initial assessment of genomic selection potential, the actuality of data and the selected pool of

cows, which are relevant to the candidates need to be permanently updated (VanRaden, 2008). To

create such a unique data set to collect many phenotypes for feed intake and energy balance and to

implement a genomic breeding value estimation, the project “optiKuh” was set emplace. It is a

collaboration of partners from universities, research institutes and business companies. The aim was

to improve livestock farming, especially through breeding for feed intake and metabolic stability,

considering feed efficiency and environmental impacts. Twelve research farms from across Germany

collaborated to establish management tools, innovations in livestock farming and animal welfare.

These involved farms measured individual feed intake by either the breed Holstein-Friesian or

Simmental cattle.

Given the challenges for an appropriate approach to integrate feed intake and energy balance in the

breeding goal, the thesis aims to estimate the genetic parameters to finally develop genomic breeding

values for both traits and to analyze the applicability of this reference population. For this purpose,

lactation curves, genetic analyses, genomic breeding value estimation and finally the discussion about

a balance breeding goal was conducted. Furthermore, a comparison of genomic estimated breeding

values and traditional breeding values were computed.


3

In Chapter 1, the data of the project optiKuh were described. Several fixed and random models were

tested to find the best model for the traits feed intake and energy balance for further genetic analyses.

Lactation curves and repeatabilities were estimated. These results were used for the evaluation of the

genetic parameters and the estimation of breeding values and are shown in Chapter 2. The focus was

set on the comparison between a pedigree-based data set and a combined data set, consisting of

genomic and pedigree-based data. For this purpose, the appropriate procedure single-step was used

to generate reliabilities for the trait feed intake and energy balance. The results were used to underline

the presumption, that feed intake can be integrated in the breeding goal of livestock farming.

In Chapter 3, an evaluation was conducted to analyze how to integrate the trait feed intake in the

practical livestock farming and how the results, achieved in Chapter 1 and Chapter 2, should be

classified under consideration of different approaches for the trait feed intake and its related traits.

References

Banos, G.; Coffey, M. P.; Brotherstone, S. Modeling Daily Energy Balance of Dairy Cows in the

First Three Lactations. J. Dairy Sci. 2005, 88 (6), 2226–2237. DOI: 10.3168/jds.S0022-

0302(05)72898-8.

Berry, D. P.; Coffey, M. P.; Pryce, J. E.; Haas, Y. de; Løvendahl, P.; Krattenmacher, N.; Crowley, J.

J.; Wang, Z.; Spurlock, D.; Weigel, K.; Macdonald, K.; Veerkamp, R. F. International genetic

evaluations for feed intake in dairy cattle through the collation of data from multiple sources. J.

Dairy Sci. 2014, 97 (6), 3894–3905. DOI: 10.3168/jds.2013-7548.

Berry, D. P.; Crowley, J. J. Cell Biology Symposium: genetics of feed efficiency in dairy and beef

cattle. Journal of animal science 2013, 91 (4), 1594–1613. DOI: 10.2527/jas.2012-5862.

Brade, W.; Brade, E. Die Futteraufnahme von hochleistenden Milchkühen im ersten Laktationsdrittel

aus der Sicht der Züchtung und des Tierwohles. 73 kB / Berichte über Landwirtschaft - Zeitschrift

für Agrarpolitik und Landwirtschaft, Band 94, Heft 2, August 2016 [Online] 2016.

Buttchereit, N.; Stamer, E.; Junge, W.; Thaller, G. Short communication: genetic relationships among

daily energy balance, feed intake, body condition score, and fat to protein ratio of milk in dairy

cows. J. Dairy Sci. 2011, 94 (3), 1586–1591. DOI: 10.3168/jds.2010-3396.

Chesnais, J. P.; Cooper, T. A.; Wiggans, G. R.; Sargolzaei, M.; Pryce, J. E.; Miglior, F. Using

genomics to enhance selection of novel traits in North American dairy cattle. J. Dairy Sci. 2016,

99 (3), 2413–2427. DOI: 10.3168/jds.2015-9970.


4

Collard, B. L.; Boettcher, P. J.; Dekkers, J.C.M.; Petitclerc, D.; Schaeffer, L. R. Relationships

Between Energy Balance and Health Traits of Dairy Cattle in Early Lactation. J. Dairy Sci. 2000,

83 (11), 2683–2690. DOI: 10.3168/jds.S0022-0302(00)75162-9.

Daetwyler, H. D.; Pong-Wong, R.; Villanueva, B.; Woolliams, J. A. The impact of genetic

architecture on genome-wide evaluation methods. Genetics 2010, 185 (3), 1021–1031. DOI:

10.1534/genetics.110.116855.

Egger-Danner, C.; Cole, J. B.; Pryce, J. E.; Gengler, N.; Heringstad, B.; Bradley, A.; Stock, K. F.

Invited review: Overview of new traits and phenotyping strategies in dairy cattle with a focus on

functional traits. Animal 2015, 9 (2), 191–207. DOI: 10.1017/S1751731114002614.

Goddard, M. E.; Hayes, B. J. Mapping genes for complex traits in domestic animals and their use in

breeding programmes. Nature reviews. Genetics 2009, 10 (6), 381–391. DOI: 10.1038/nrg2575.

Guarini, A. R.; Lourenco, D. A. L.; Brito, L. F.; Sargolzaei, M.; Baes, C. F.; Miglior, F.; Misztal, I.;

Schenkel, F. S. Comparison of genomic predictions for lowly heritable traits using multi-step and

single-step genomic best linear unbiased predictor in Holstein cattle. J. Dairy Sci. 2018, 101 (9),

8076–8086. DOI: 10.3168/jds.2017-14193.

Habier, D.; Tetens, J.; Seefried, F.-R.; Lichtner, P.; Thaller, G. The impact of genetic relationship

information on genomic breeding values in German Holstein cattle. Genetics, selection, evolution

: GSE 2010, 42, 5. DOI: 10.1186/1297-9686-42-5.

Leesen, R. von; Tetens, J.; Stamer, E.; Junge, W.; Thaller, G.; Krattenmacher, N. Effect of genetic

merit for energy balance on luteal activity and subsequent reproductive performance in

primiparous Holstein-Friesian cows. J. Dairy Sci. 2014, 97 (2), 1128–1138. DOI:

10.3168/jds.2013-7185.

Legarra, A.; Christensen, O. F.; Aguilar, I.; Misztal, I. Single Step, a general approach for genomic

selection. Livestock Science 2014, 166, 54–65. DOI: 10.1016/j.livsci.2014.04.029.

Li, B.; Fikse, W. F.; Lassen, J.; Lidauer, M. H.; Løvendahl, P.; Mäntysaari, P.; Berglund, B. Genetic

parameters for dry matter intake in primiparous Holstein, Nordic Red, and Jersey cows in the first

half of lactation. J. Dairy Sci. 2016, 99 (9), 7232–7239. DOI: 10.3168/jds.2015-10669.

Li, B.; Fikse, W. F.; Løvendahl, P.; Lassen, J.; Lidauer, M. H.; Mäntysaari, P.; Berglund, B. Genetic

heterogeneity of feed intake, energy-corrected milk, and body weight across lactation in


5

primiparous Holstein, Nordic Red, and Jersey cows. J. Dairy Sci. 2018, 101 (11), 10011–10021.

DOI: 10.3168/jds.2018-14611.

Liinamo, A.-E.; Mäntysaari, P.; Mäntysaari, E. A. Short communication: Genetic parameters for feed

intake, production, and extent of negative energy balance in Nordic Red dairy cattle. J. Dairy Sci.

2012, 95 (11), 6788–6794. DOI: 10.3168/jds.2012-5342.

Mäntysaari, P.; Liinamo, A.-E.; Mäntysaari, E. A. Energy efficiency and its relationship with milk,

body, and intake traits and energy status among primiparous Nordic Red dairy cattle. J. Dairy Sci.

2012, 95 (6), 3200–3211. DOI: 10.3168/jds.2011-4685.

Prendiville, R.; Pierce, K. M.; Delaby, L.; Buckley, F. Animal performance and production

efficiencies of Holstein-Friesian, Jersey and Jersey × Holstein-Friesian cows throughout lactation.

Livestock Science 2011, 138 (1-3), 25–33. DOI: 10.1016/j.livsci.2010.11.023.

Pryce, J. E.; Arias, J.; Bowman, P. J.; Davis, S. R.; Macdonald, K. A.; Waghorn, G. C.; Wales, W. J.;

Williams, Y. J.; Spelman, R. J.; Hayes, B. J. Accuracy of genomic predictions of residual feed

intake and 250-day body weight in growing heifers using 625,000 single nucleotide

polymorphism markers. J. Dairy Sci. 2012, 95 (4), 2108–2119. DOI: 10.3168/jds.2011-4628.

Pszczola, M.; Stock, K. F.; Mucha, E.; Sell-Kubiak, E. Genetic architecture of methane emissions

from dairy cows. WCGALP [Online] 2018.

Pszczola, M.; Strabel, T.; Mulder, H. A.; Calus, M. P. L. Reliability of direct genomic values for

animals with different relationships within and to the reference population. J. Dairy Sci. 2012, 95

(1), 389–400. DOI: 10.3168/jds.2011-4338.

Roche, J. R.; Berry, D. P.; Lee, J. M.; Macdonald, K. A.; Boston, R. C. Describing the body condition

score change between successive calvings: A novel strategy generalizable to diverse cohorts. J.

Dairy Sci. 2007, 90 (9), 4378–4396. DOI: 10.3168/jds.2006-729.

Roche, J. R.; Friggens, N. C.; Kay, J. K.; Fisher, M. W.; Stafford, K. J.; Berry, D. P. Invited review:

Body condition score and its association with dairy cow productivity, health, and welfare. J. Dairy

Sci. 2009, 92 (12), 5769–5801. DOI: 10.3168/jds.2009-2431.


6

Spurlock, D. M.; Dekkers, J. C. M.; Fernando, R.; Koltes, D. A.; Wolc, A. Genetic parameters for

energy balance, feed efficiency, and related traits in Holstein cattle. J. Dairy Sci. 2012, 95 (9),

5393–5402. DOI: 10.3168/jds.2012-5407.

Tetens, J.; Thaller, G.; Krattenmacher, N. Genetic and genomic dissection of dry matter intake at

different lactation stages in primiparous Holstein cows. J. Dairy Sci. 2014, 97 (1), 520–531. DOI:

10.3168/jds.2013-7301.

VanRaden, P. M. Efficient methods to compute genomic predictions. J. Dairy Sci. 2008, 91 (11),

4414–4423. DOI: 10.3168/jds.2007-0980.

Yao, C.; Los Campos, G. de; Vandehaar, M. J.; Spurlock, D. M.; Armentano, L. E.; Coffey, M.; Haas,

Y. de; Veerkamp, R. F.; Staples, C. R.; Connor, E. E.; Wang, Z.; Hanigan, M. D.; Tempelman, R.

J.; Weigel, K. A. Use of genotype × environment interaction model to accommodate genetic

heterogeneity for residual feed intake, dry matter intake, net energy in milk, and metabolic body

weight in dairy cattle. J. Dairy Sci. 2017, 100 (3), 2007–2016. DOI: 10.3168/jds.2016-11606.

Chapter I

7

Chapter I

Lactation curves and model evaluation for feed intake and energy balance in dairy cows

I. Harder1, E. Stamer², W. Junge1, G. Thaller1

1Institute of Animal Breeding and Husbandry, Christian-Albrechts-University, D-24098 Kiel

²TiDa Tier und Daten GmbH, D-24259 Westensee/Brux

Published in Journal of Dairy Science

DOI: https://doi.org/10.3168/jds.2018-15300

Chapter I

8

Abstract

Nowadays, a good health status of high performing dairy cows is essential for successful production.

Feed intake affects the metabolic stability of dairy cows and can be utilized as a measurement for

energy balance. By implementing feed intake and energy balance into the breeding goal, these traits

provides great potential for an improvement in the health of dairy cows by breeders.

In this study, fixed and random regression models were tested to establish appropriate models for a

further analysis of this approach. 1,374 Holstein-Friesian cows (HF) and 327 Simmental cows (SI)

from twelve German research farms participating in a collaboration called “optiKuh” were

phenotyped. Feed intake data recording was standardized across farms and energy balance was

calculated using phenotypic information on milk yield, milk ingredients, live weight, gestation stage

and feed intake. The phenotypic data-set comprises a total of 40,012 HF and 16,996 SI with average

weekly dry matter intakes of 21.8 ± 4.3 kg/d and 20.2 ± 3.6 kg/d respectively. Observations of days

in milk (DIM) 1 to 350 were used to evaluate the best fitting models and to estimate the repeatability

and correlations between cow effects at different stages for feed intake and energy balance. Four

parametric functions (Ali and Schaeffer and Legendre polynomials of second, third and fourth degree)

were compared to model the lactation curves. Based on the corrected Akaike information criterion

and the Bayesian information criterion, the goodness of fit was evaluated to choose the best fitting

model for the finest description of lactation curves for the traits energy balance and feed intake.

Legendre Polynomials fourth degree was the best fitting model for random regression models. In

contrast, Ali and Schaeffer was the best choice for fixed regression models. Feed intake and energy

balance acted as expected: The feed intake increased slowly at the beginning of the lactation and the

negative energy balance switched to a positive range around the 40th to 80th day of lactation. The

repeatabilities of both traits were quite similar and the repeatabilities for SI were the highest for both

traits. Additionally, correlations between cow effects were closest between early DIM. These results

emphasize, the possibility, that the unique “optiKuh” dataset can be used for further genetic analyses

to enable genomic selection for the trait feed intake or energy balance.

Introduction

Feed intake (FI) is a very important trait for high- performance dairy cows for various reasons. On

the one hand, feed costs represent up to 50% of dairy production costs and have consequently a major

impact on the economic success of livestock farming (Connor, 2015). On the other hand, it strongly

influences the health maintenance of dairy cows and therewith is an important factor for animal

welfare in dairy production (Boichard and Brochard, 2012). Milk yield (MY) is highly correlated to

Chapter I

9

FI. However, FI cannot compensate for the increased energy requirements especially in early lactation

stages (Ingvartsen and Andersen, 2000). Since an increase in FI is slower than an increase in milk

production, body reserves that were generated during gestation are utilized to compensate for the

resulting energy deficit (Coffey et al., 2002). This is a normal reproduction strategy of mammals.

However, since dairy breeding strategies focused for a long time only on increasing milk production,

they slip into a severe and lengthy intensified postpartum energy deficit that can be associated with

metabolic diseases (Haas et al., 2015). Cows with an extreme state of negative energy balance are

more susceptible to metabolic diseases like acidosis, ketosis and milk fever (Collard et al., 2000;

Spurlock et al., 2012; Leesen et al., 2014) and a decline in fertility (Butler, 2003). This indicates that

the deficit depends not only on the amount of milk produced, but also on FI (Buttchereit et al., 2010).

So, high FI at the beginning of lactation could counteract these problems and, therefore, FI should be

part of the dairy cow breeding goal. This is in contrast to the generally valid goal of an efficient cow,

which should eat less but produce more (Veerkamp et al., 2013). However, the improvement in feed

efficiency causes a worsening in energy balance and therewith a rise in cows’ health problems.

Therefore, it makes sense to breed on a high feed intake at the beginning of lactation to overcome the

difficult period although it causes problems with efficiency.

Therefore, reliable FI phenotypes have to be recorded, which presents practical challenges for dairy

farms. The measurement of FI is complicated, expensive and connected with a high number of

technical requirements (Coffey et al., 2004; Berry et al., 2007; Haas et al., 2012). In previous studies,

data-sets have consisted of data collected on several experimental farms, which have resulted in an

insufficient data-base ( Veerkamp et al., 2013; Berry et al., 2014;). To improve feed intake or energy

balance, respectively, genomic selection provides the most practical approach, because this eliminates

the need to collect phenotypes for each animal and achieves higher accuracies compared to traditional

selection (Connor, 2015; Yao et al., 2017).

In the “optiKuh” project a collaboration of twelve German research farms was established. All of the

research farms provided the technical equipment for a consistent measurement of unique FI

phenotypes. In this way a unique data set was created, providing the possibility to consider the traits

FI and energy balance (EB) in the breeding goal to support the health and fertility of dairy cows

(Veerkamp et al., 2000; Coffey et al., 2001; Coffey et al., 2002).

A little success could be generated in developing breeding values for FI to select for this trait (Berry

et al., 2014).

The traits FI and EB can be measured at different intervals (e.g. test day, test week…) which can be

evaluated using different models. These include repeatability models, multi trait models and random

Chapter I

10

regression models with different functions (Buttchereit et al., 2010; Spurlock et al., 2012; Li et al.,

2018; Uddin et al., 2018).

In the evaluation of the genetic merit of dairy cows, test-day models are used for repeated records of

production traits (Berry et al., 2014). One type of these models is the random regression model, which

is the model of choice to use the given benefits to account for individual differences in the shape of

lactation and environmental factors that affect cows at different stages of lactation (Jamrozik and

Schaeffer, 1997; Jensen, 2001; Banos et al., 2012). However, a fixed average lactation curve and a

random regression for the individual deviations is used to model a lactation curve of a cow in the

random regression model (Pool et al., 2000).

The objective of this study was to pool FI and EB data of twelve German dairy research farms. A

subsequent joint analysis of these traits was to establish a basis to consider FI in the breeding goal of

high-performance dairy cows to support health and fertility, especially at the beginning of lactation.

Therefore, the aim was to develop well suited statistical models for both variance component

estimation and breeding value estimation of the traits FI and EB.

Materials and Methods

The phenotype data for FI and EB originated from 1,374 HF and 327 SI cows. Individual FI was

recorded from December 2014 to March 2017 on twelve German research farms (Achselschwang,

Aulendorf, Braunschweig, Dummerstorf, Grub, Futterkamp, Hohenheim, Iden, Karkendamm,

Neumühle, Riswick, Triesdorf). The individual feed intake of each dairy cow was measured via

feeding troughs equipped with a weighing unit and automatic cow identification. The farms involved

in the project “optiKuh” were divided into A and B farms. In A farms, project-specific feeding trials

were performed with different roughage energy levels (6.5 MJ NEL/kg DM vs. 6.1 MJ NEL/kg DM)

and different amounts of concentrates (250 g/kg ECM vs. 150 g/kg ECM). NEL represents net energy

lactation defined as the amount of energy in a feed which is available for milk production and body

maintenance (Kirchgeßner, 2014). DM stands for dry matter, and ECM is the energy-corrected milk

yield containing 4% fat and 3.4% protein (according to (Kirchgeßner, 2014).

ECM (kg/d) = milk yield (kg/d) * ((1.05 + 0.38 * milk fat (%) + 0.21 * milk protein (%)) / 3.28)

B farm cows were fed with partly mixed, farm- specific rations (PMR) or totally mixed rations

(TMR). The only restriction on B farms for farm-specific, high-yield cow rations was a target value

for concentrations of 250 g/kg ECM. The realized feeding groups together with the duration of the

Chapter I

11

trials are displayed in Table 1. All cows were fed ad libitum. If the compilation of the components

changed, the rations were calculated again. If more than one analysis per component within ration

was carried out, the mean was taken. The frequency of sampling differed between research farms.

Every farm computed and allocated information for each day and ration (and types of concentrate).

Data was edited and summarized within farm. Due to the different frequencies of measurement (e.g.

daily FI and weekly milk ingredients) the data was computed within each calendar week as mean

values. Most of the research farms collected FI data during the dry period as well. Since the cows of

some of the research farms were housed separately during the first DIM, no FI data was available for

this period.

Table 1. Research farms with diets, feeding groups and duration of trials

Research farm Breed Diet *Feeding group Duration

6.1 / 150 6.1 / 250 6.5 / 150 6.5 / 250 (months)

A Braunschweig HF TMR X X 8

A Dummerstorf HF PMR X X X X 24

A Riswick (A) HF PMR X X X X 24

A Aulendorf SI TMR X X 24

A Grub SI PMR X X X X 24

A Triesdorf SI PMR X X 24

**Number of feeding groups – trial variants

B Futterkamp I HF TMR 2 – different amounts of essential oils 3

B Futterkamp II HF TMR 2 – different amounts of carbohydrates 3

B Futterkamp III HF TMR 2 – mycotxin binder 3

B Iden 2015 HF TMR 1 6

B Iden 2016 HF TMR 2 – different amounts of carbohydrates 6

B Karkendamm HF PMR 1 22

B Neumühle HF TMR 1 23

B Riswick I HF TMR 2 – different protein content 5

B Riswick II HF TMR 3 – different protein content 4

B Riswick III HF TMR 4 – different chop lengths of corn silage 6

B Hohenheim HF TMR 1 18

B Achselschwang SI PMR 2 – different concentrate levels 3

B Achselschwang SI PMR 4 – different amounts of grass and corn

silage and two concentrate levels

4

B Achselschwang SI PMR 2 – different chop lengths of corn silage 2

* 6.1 or 6.5 MJ NEL – with 150 or 250 g/kg ECM, respectively

** farm specific high yielding cow ration, target value for concentrates: 250g/kg ECM

Additionally, data was excluded, if a sample day had less than 22 h records after checking the feed

weighing technique. Furthermore, if the research farm fed PMR, total FI was only calculated if

mixture and concentrates were available. Calendar days with extreme mean herd weights were also

excluded by visual inspection. To use the records of the amount of milk, at least two milkings had to

Chapter I

12

be available. One research farm worked with an automatic milking system with 24 h data recording.

Thus, lactation for further analysis was defined from 1 to 350 days in milk. Observations outside the

range of ± four standard deviations (SD) of the mean value were excluded from further analyses. The

structure of the data set before raw data editing is summarized in Table 2. Four farms (Futterkamp,

Iden, Riswick (B) and Achselschwang) carried out more consecutive feeding experiments, e.g. 2-2-2

stands for three feeding experiments each with two groups. Additionally, Achselschwang included

the breed Brown-Swiss (BS) in their data set. Furthermore, two experimental farms (Braunschweig

and Dummerstorf) had no animals included in lactation 1. The number of animals was subsumed into

both breeds HF and SI: 557 HF and 125 SI for lactation 1, 509 for HF and 141 SI lactation 2, 370 HF

and 135 SI lactation 3, and 512 HF and 209 SI for lactation ≥ 4.

Table 2. Description of data-set before editing - number of cows, breeds, feeding groups, lactations,

days in milk (DIM) depending on farm No. No. DIM

Research farm cows Breed1 feeding groups Lactation² Mean Range

Braunschweig 64 HF 4 2-5, 9 67 0-171

Dummerstorf 30 HF 2 2-3 191 0-622

Futterkamp 179 HF 2-2-2³ 1-8 174 26-377

Hohenheim 51 HF 1 1-5, 7-8 59 0-464

Iden 188 HF 1-2³ 1-10 63 1-309

Karkendamm 341 HF 1 1-9 173 11-627

Neumühle 199 HF 1 1-11 95 1-405

Riswick (A) 83 HF 4 1-7 165 0-405

Riswick (B) 239 HF 2-3-4³ 1-10 86 0-468

Achselschwang 105 SI (BS) 2-4-2³ 1-6, 9 175 20-348

Aulendorf 59 SI 2 1-9 159 0-431

Grub 97 SI 4 1-8, 10 160 0-462

Triesdorf 66 SI 2 1-7 155 0-403 1HF = Holstein-Friesian; SI = Simmental; BS = Brown-Swiss.

²Range of lactation stages.

³Multiple numbers in no. of feeding groups = additional feeding experiments within feeding group.

The data was edited to ensure that the cows had at least one record per week: the resulting means for

FI, MY, and EB were 21.8 kg, 35.5 kg, and 3.20 MJ NEL for HF, and 20.2 kg, 27.4 kg, and 1.06 MJ

NEL for SI, respectively (Table 3). A comparison of the mean values of both breeds show higher

values for HF in both traits with + 2 kg FI and 7.01 MJ NEL, respectively (Table 4).

Chapter I

13

Table 3. Description of data-set after editing - mean and standard deviation of feed intake, energy

balance, milk yield, energy corrected milk (ECM) and energy intake (1 to 350 days in milk)

Feed intake

(kg DM/d)

Energy

balance

(MJ NEL/d)

Milk yield

(kg/d)

ECM

(kg/d)

Energy intake

(MJ NEL/d)

Research farm Mean SD Mean SD Mean SD Mean SD Mean SD

Braunschweig 21.1 3.89 -5.98 24.8 35.6 6.30 36.1 5.47 146 28.4

Dummerstorf 21.3 4.30 -3.81 30.6 33.7 8.27 33.8 7.39 144 29.2

Futterkamp 22.4 2.84 1.49 15.7 36.5 7.32 34.9 6.00 156 21.5

Hohenheim 21.8 3.80 -4.82 18.0 36.8 8.08 36.3 6.80 148 28.7

Iden 23.2 3.97 -10.1 21.4 42.0 7.95 40.9 6.65 163 28.7

Karkendamm 21.6 4.13 -2.40 29.3 37.1 8.42 35.6 7.29 155 29.7

Neumühle 20.7 3.72 -15.0 25.1 37.1 7.37 36.4 7.14 145 25.9

Riswick (A) 22.4 4.51 22.3 24.3 27.1 8.08 27.0 7.21 149 31.0

Riswick (B) 21.9 5.03 17.5 29.8 34.2 8.11 33.4 6.84 159 37.7

Achselschwang 22.7 3.38 4.31 18.6 33.0 8.00 32.6 7.13 154 24.6

Aulendorf 21.8 3.23 7.80 19.8 28.2 7.73 29.7 7.51 147 23.7

Grub 18.6 3.35 -0.31 18.9 23.6 7.28 24.9 6.94 124 24.6

Triesdorf 19.5 3.02 -5.58 15.0 28.9 7.48 29.9 6.88 133 22.5

Chapter I

14

Table 4. Descriptive statistics for weekly averages of feed intake, milk yield and energy balance as

well as energy relevant traits for both breeds

Trait

Holstein-Friesian

No.

cows

No.

lactations Obs. Mean SD

Feed intake (kg DM/d) 1,341 1,928 40,012 21.8 4.25

Milk yield (kg/d) 1,338 1,917 39,838 35.5 8.81

Energy balance (MJ NEL) 1,322 1,865 33,376 3.20 29.4

*Feed intake (kg DM/d) 22.3 4.04

*Weight (kg) 658 73.9

*ECM (kg/d) 34.3 7.71

*Milk yield (kg/d) 35.8 8.76

*Fat (%) 3.77 0.63

*Protein (%) 3.31 0.32

Trait

Simmental cattle

No.

cows

No.

lactations Obs. Mean SD

Feed intake (kg DM/d) 327 604 16,996 20.2 3.60

Milk yield (kg/d) 326 603 16,933 27.4 8.16

Energy balance (MJ NEL) 326 583 14,527 1.06 18.9

*Feed intake (kg DM/d) 20.4 3.41

*Weight (kg) 750 75.3

*ECM (kg/d) 28.3 7.58

*Milk yield (kg/d) 27.5 8.04

*Fat (%) 4.24 0.61

*Protein (%) 3.59 0.33 Obs. = number of weekly averages; * = traits for calculating energy balance

EB was calculated as the difference between energy intake (MJ NEL/d) and estimated energy

requirements for maintenance, milk yield, growth (only for primiparous cows), and gestation. These

terms were calculated by the following formulas predetermined by the project “optiKuh” according

to the German Society of Nutrition Physiology. Energy intake was computed by summing energy

amounts of eaten PMR and concentrates or TMR, respectively. Energy requirement for maintenance

was calculated according to Kirchgeßner (2014).

Maintenance costs (MJ NEL/d) = 0.293 * body weight 0.75

On some research farms cows the were weighed weekly. In other research herds the cows were

automatically weighed after every milking. Daily values were derived by averaging morning and

Chapter I

15

evening body weight (BW). The Energy requirement for milk yield was computed by multiplying

ECM with 3.28 (MJ NEL/kg).

The energy requirements for growth of primiparous cows (MJ NEL/d)= 0.007895 MJ NEL/kg*BW14

BW14 represents the mean weight of the first 14 DIM.

The Energy requirement during gestation (MJ NEL/d) = (0.044 * e0.0162 * t + udder deposition) / 0.29

where t represents the number of days after conception, and constants of udder deposition are 0.8 MJ

NEL/d (8th to 7th week ante partum), 1.1 MJ NEL/d (6th to 4th week ante partum), and 1.5 MJ NEL/d

(3rd week ante partum to expected calving date).

By comparing well established parametric functions 1) the best fitting function for the average

lactation curve was evaluated and chosen as the basis for 2) the evaluation of the best (co)variance

function to model cow-specific lactation curves. Afterwards, using the best fitting models lactation

curves were calculated, and both repeatabilities and correlations between cow effects of different

lactation stages were estimated.

To find the best function for fitting the fixed lactation curve within breed for the traits FI and EB, the

four parametric functions Ali and Schaeffer (Ali and Schaeffer, 1987) (AS) and the Legendre

Polynomials (Brotherstone et al., 2000) (LG) of 2nd to 4th degree were compared regarding their

goodness of fit within lactation (1 to 4; 1 to 6); see Table 6 and 7.

For the trait milk yield no model evaluation was conducted because the Ali and Schaeffer function is

well established as shown in many former analyses (Buttchereit et al., 2010; Stamer et al., 2011;

Melzer et al., 2017).

For the evaluation of the fixed lactation curve the following model was used.

yijk= µ + HTWi + LNOj + ∑ 𝑏𝑚𝑙=1 l * xijkl (d) + eijk

where yijk is the observation of FI or EB, µ is the overall mean, HTWi is the fixed effect of the ith

herd-test week (i= 1,…,847) for FI and EB, LNOj is the other fixed effect of the jth lactation class

(j=1,…,4), bl is set as the fixed jth regression coefficient of the lth function term of lactation day, x is

set as the function of the lactation curve (Table 5), and eijk stands for the random residual effect.

Chapter I

16

Table 5. Function terms of lactation day (d) for four models using the Ali and Schaeffer curve (AS)

and Legendre Polynomials of second (LG2), third (LG3) and fourth degree (LG4) to simulate the

lactation curve for days in milk from 1 - 350

Function terms

Function of

lactation curve Xijk0 xijk/1 Xijk/2 xijk/3 xijk/4

AS (d = 1 to 350) 1 d/350 (d/350)² ln(350/d) [ln(350/d)]²

LG2 1 sd1 0.5(3sd²-1) LG3 1 sd1 0.5(3sd²-1) 0.5(5sd³-3sd) LG4 1 sd1 0.5(3sd²-1) 0.5(5sd³-3sd) 0.125(35sd4-30sd²+3)

sd = standardized lactation day d

The effect of the feeding group was tested in preceding within the farm analyses by choosing the Ali

and Schaeffer to model the lactation curves. In the case of significant group differences, the herd test

week was extended by feeding group. Herd test weeks had to comprise at least three observations,

otherwise they were added to a neighboring test week. The fixed regression models were analyzed

using the SAS procedure MIXED and the Maximum Likelihood (ML) method (SAS, 2012). The

function with the best model fit was chosen as the fixed lactation curve for further analyses. As

already shown by Buttchereit et al. (2010), the described four functions (see Table 5) were chosen to

model animal dependent covariances between repeated measurements of the same cow in a second

step of model evaluation to design final random regression models for statistical analyses of FI, MY

and EB.

For the evaluation of the random deviations from the fixed lactation curve, the following model was

used.

yijk= µ + HTWi + LNOj + ∑ 𝑏𝑚𝑛=1 l * xijkl (d) + ∑ 𝑐𝑜𝑤𝑚

𝑛=1 km * xijklm (d) + ejkl

where cowkm is the mth random regression coefficient of the kth animal. The random regression models

were analyzed using the procedure MIXED in SAS (SAS, 2012) and the REML method. An

unstructured covariance matrix was used for the random cow effect.

The corrected Akaike information criterion (AICC) (Burnham and Anderson, 1998) and the Bayesian

information criterion (BIC) (Schwarz, 1978) were used as evaluation criteria of goodness of fit of the

differently analyzed models.

AICC = -2lnL + 2s (N* / (N* – s – 1))

BIC = -2lnL + s (ln N*)

Chapter I

17

In the case of the function of the curves for the general, fixed lactation curve (maximum likelihood

estimation), s is set as the sum of fixed (p) and random effects (q), lnL is the logarithm of the restricted

maximum likelihood function, N* is the number of observations (n). For the computation of the

goodness of fit for the function curve of the random cow effect (REML-estimation), it is set as

s = q and N* = n – p (= number of residual degrees of freedom).

Generally, the model which minimizes AICC or BIC should be applied, but if the values for AICC

and BIC are similar, the simpler model should be preferred (Littell, 2007). The calculated correlation

between the real observation and the predicted value was used is an additional criterion (Guo and

Swalve, 1995). Furthermore, the estimated error variance was considered.

Based on the final random regression model, repeatabilities for all DIM (1 to 350) were calculated

within breed by using the estimated regression coefficients. Also, cow- effect correlations, within

trait, were computed between twelve DIM points, separated by regular intervals of 30 days.

Results

The evaluation criteria for the traits FI and EB of the fixed and random regression models are shown

in Table 6 and Table 7 and illustrate the goodness of fit of different lactation curve functions,

represented as the difference to the best fitting model. Table 6 refers to the fixed regression model.

The criteria were computed within breed and trait. Additionally, a possible effect of the classification

into four or six lactation classes on evaluation criteria was tested by comparing results of both

goodness of fit criteria. The most suited models for the fixed lactation curve for both traits and breeds

seems to be the AS function. Additionally, the nested function within lactation accomplished better

results than the alternative with lactation class as the main effect. The simpler model with four classes

was chosen for further analyses due to the lack of clear differences between the models with four or

six lactation classes.

Table 7 refers to the random regression part of the model. For both HF and SI, the LG4 model shows

the lowest information criteria and also the lowest residual variances. There were no consistent results

for the correlations between observed and predicted trait values. Nevertheless, for these data, the best

suited statistical model seems to be the one which includes the fixed effects herd test week, lactation

class (four classes) and the AS function as fixed regression. The Legendre Polynomial of fourth

degree should be chosen as the random regression.

Chapter I

18

Table 6. Differences between the Bayesian information criterion (∆BIC) and corrected Akaike’s

information criterion (∆AICC), for EB and FI dependent on the fixed regression DIM functions Ali

and Schaeffer (AS) and the Legendre Polynomials of second (LG2), third (LG3) and fourth degree

(LG4) for the data-sets of the breeds Holstein-Friesian (HF) and Simmental cattle (SI) (DIM 1 to 350)

Feed intake

Breed Lactation Function Function Function within lactation ∆AICC ∆BIC ∆AICC ∆BIC

HF

1 to 4

AS 1739 1552 257 173

LG2 5529 5352 4364 4211

LG3 3332 3137 1897 1778

LG4 2183 1996 721 636

1 to 6

AS 1547 1377 0A 0B

LG2 5353 5166 4113 4011

LG3 3139 2961 1646 1595

LG4 1993 1823 438 438

SI

1 to 4

AS 366 213 64 0C

LG2 5580 5411 5362 5238

LG3 2391 2229 2108 2014

LG4 828 674 543 479

1 to 6

AS 300 161 0D 11

LG2 5517 5363 5301 5222

LG3 2333 2187 2053 2019

LG4 763 624 475 486

Absolute value: A209259; B216759; C81974; D74594

Energy balance

Breed Lactation Function Function Function within lactation ∆AICC ∆BIC ∆AICC ∆BIC

HF

1 to 4

AS 252 133 18 0A

LG2 1877 1743 1639 1554

LG3 620 494 393 341

LG4 276 158 44 25

1 to 6

AS 251 149 0B 65

LG2 1879 1761 1629 1594

LG3 620 510 385 400

LG4 275 174 18 82

SI

1 to 4

AS 483 388 7 0C

LG2 2230 2120 1839 1773

LG3 1015 912 582 545

LG4 503 408 35 28

1 to 6

AS 486 405 0D 66

LG2 2232 2136 1829 1807

LG3 1018 930 576 597

LG4 506 425 19 84

Absolute value: A307132; B300379; C123696; D117187

Chapter I

19

Table 7. Differences between the Bayesian information criterion (∆BIC) and corrected Akaike’s

information criterion (∆AICC), correlation between recorded and predicted value (R q,q̂) and the

residual variances (var e) for EB and FI dependent on the random regression DIM functions Ali and

Schaeffer (AS) and the Legendre Polynomials of second (LG2), third (LG3) and fourth degree (LG4)

for both breeds

Absolute values: A 180,491; B 180,574; C 275773; D 275856; E 58535; F 58596; G 102880; H 102940

The resulting parameter estimates of this model were used for the calculation of the fixed regression

curves within lactation class. The intercept estimate, lactation class effect and fixed regression

coefficients were used for this calculation. Additionally, herd test week was considered by its overall

mean effect, and random effect was set to its mean (0).

Figure 1 shows the average standardized error of the complete model (AS for fix lactation curve and

LG4 for random cow effect) which were plotted against 1 to 350 DIM. The almost homogenous

residual pattern underlined the chosen model. Exemplarily, figures for HF and both traits FI and EB

are illustrated.

Figure 1. Distribution of average standardized error in course of lactation of FI (left) and EB (right)

predicted by the final model (AS for the fixed lactation curve, LG4 for the random cow effect) from

DIM 1 to 350 for HF

Figures 2 and 3 illustrate lactation curves for the first and second lactations FI and EB and milk yield

(MY). The energy deficit in the early lactation is indicated as a hatched area. Cows hit peak milk

Feed intake Energy balance

Breed Function ∆AICC ∆BIC var e R q,q̂ ∆AICC ∆BIC var e R q,q̂

HF

AS 1182 1156 4.56 0.88 446 420 209.4 0.89

LG2 1642 1596 4.69 0.87 818 771 217.3 0.88

LG3 765 766 4.48 0.88 302 276 209.1 0.89

LG4 0A 0B 4.28 0.89 0C 0D 203.1 0.89

SI

AS 524 505 1.62 0.94 382 364 77.2 0.90

LG2 1027 993 1.72 0.94 747 714 81.4 0.89

LG3 342 323 1.61 0.94 185 167 75.7 0.90

LG4 0E 0F 1.54 0.95 0G 0H 73.3 0.91

-1.4

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0 50 100 150 200 250 300 350

av

era

ge s

tan

dard

ized

err

or

DIM

-1.4

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0 50 100 150 200 250 300 350

av

era

ge s

tan

dard

ized

err

or

DIM

Chapter I

20

production before they hit peak FI. Generally, difference between the first and second lactations is

evidently higher in HF compared to SI lactation curves of FI and MY are higher in second lactation,

while the EB curve is lower in the second lactation in both breeds. Both breeds show similar

trajectories of lactation in second lactation, but HF display a higher level overall. Cows show a more

pronounced energy deficit (hatched area, Figure 2) in the initial stage of lactation in first lactation,

but achieve a positive EB in second lactation around ten days later than in first lactation with a switch

at around DIM 50.

All in all, Figure 2 and Figure 3 show the central and well-known problem of negative EB. The

moment of highest MY (before 50 DIM) and highest FI (after 50 DIM) appear with a certain time,

which results in an energy deficit. In consequence, body reserves were used for milk production,

which in turn can cause metabolic stress and health problems. First lactating SI cows show a slightly

negative EB after 200 DIM.

Figure 2. Lactation curves for FI, MY and EB of Holstein-Friesian in first (left) and second (right)

lactation modeled with the AS function

Figure 3. Lactation curves for FI, MY and EB of Simmental cows in first (left) and second (right)

lactation modeled with the AS function

-50

-40

-30

-20

-10

0

10

20

30

40

50

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

En

ergy

bal

ance

(M

J N

EL

/d)

Fee

d in

take

(kg

DM

/d)

/ M

ilk y

ield

(kg/

d)

DIM

FI

MY

EB

-50

-40

-30

-20

-10

0

10

20

30

40

50

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

En

ergy

bal

ance

(M

J N

EL

/d)

Fee

d in

take

(kg

DM

/d)

/ M

ilk y

ield

(kg/

d)

DIM

FI

MY

EB

-50

-40

-30

-20

-10

0

10

20

30

40

50

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

En

ergy

bal

ance

(M

J N

EL

/d)

Fee

d in

take

(kg

DM

/d)

/ M

ilk y

ield

(kg/

d)

DIM

FI

MY

EB

-50

-40

-30

-20

-10

0

10

20

30

40

50

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

En

ergy

bal

ance

(M

J N

EL

/d)

Fee

d in

take

(kg

DM

/d)

/ M

ilk y

ield

(kg/

d)

DIM

FI

MY

EB

Chapter I

21

Repeatabilities for the traits FI, EB and MY for both breeds in the course of lactation (DIM 1 to 350)

are shown in Figure 4. The values ranged between 0.56 and 0.95 (HF) and 0.54 and 0.95 (SI). For

both breeds, the highest repeatability was observed at the beginning and the end of lactation, whereas

EB showed the poorest repeatability with values ranging between 0.49 to 0.85. In comparison

between the breeds, SI have a higher level of trajectories within traits.

Figure 4. Repeatability of feed intake, energy balance and milk yield for Holstein-Friesian (black)

and Simmental cow (grey)

Table 8 (HF) and Table 9 (SI) provide correlations between cow effects at different lactations stages

for FI (above the diagonal) and weekly EB (below the diagonal). Generally, the FI correlations are

higher than the EB correlations. Neighboring stages of lactation are more closely correlated than more

distant stages. This accounts for HF, SI and for both traits. Higher values were observed between

adjacent DIM in the middle and at the end of lactation and lower estimates were observed between

DIM records at the beginning of lactation. While correlations of FI at the beginning of lactation, the

relevant period in this study, varied from 0.41 to 0.95 for HF and for SI 0.30 to 0.96, respectively.

The correlations for EB varied from 0.42 to 0.96 for HF and from 0.26 to 0.93 for SI.

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0 50 100 150 200 250 300 350

Rep

eata

bili

ty

DIM

HF_FI

HF_EB

HF_MY

SI_FI

SI_EB

SI_MY

Chapter I

22

Table 8. Correlations between cow effects of different DIM points for FI (above the diagonal) and

EB (below the diagonal) in Holstein cows

DIM 10 40 70 100 130 160 190 220 250 280 310 340

10 0.76 0.53 0.41 0.33 0.29 0.26 0.22 0.17 0.13 0.1 0.06

40 0.81 0.93 0.80 0.65 0.52 0.47 0.46 0.45 0.45 0.42 0.16

70 0.55 0.93 0.95 0.82 0.97 0.62 0.57 0.54 0.53 0.52 0.28

100 0.42 0.78 0.96 0.96 0.86 0.77 0.67 0.59 0.54 0.54 0.39

130 0.36 0.64 0.83 0.96 0.97 0.89 0.77 0.64 0.56 0.55 0.45

160 0.31 0.50 0.67 0.85 0.97 0.97 0.87 0.73 0.63 0.60 0.46

190 0.26 0.40 0.56 0.74 0.89 0.97 0.96 0.86 0.77 0.70 0.41

220 0.18 0.33 0.47 0.62 0.76 0.87 0.96 0.97 0.90 0.81 0.36

250 0.08 0.27 0.40 0.50 0.6 0.71 0.84 0.96 0.98 0.88 0.33

280 -0.01 0.21 0.33 0.40 0.47 0.56 0.7 0.86 0.97 0.94 0.40

310 -0.06 0.13 0.26 0.33 0.39 0.47 0.59 0.73 0.84 0.93 0.67

340 -0.00 -0.01 0.08 0.21 0.31 0.37 0.39 0.39 0.40 0.50 0.77

Table 9. Correlations between cow effects of different DIM points for FI (above the diagonal) and

EB (below the diagonal) in Simmental cows

DIM 10 40 70 100 130 160 190 220 250 280 310 340

10 0.74 0.42 0.30 0.29 0.34 0.39 0.42 0.42 0.41 0.41 0.39

40 0.82 0.90 0.78 0.68 0.61 0.58 0.56 0.55 0.53 0.49 0.36

70 0.51 0.90 0.96 0.87 0.77 0.68 0.61 0.55 0.50 0.46 0.34

100 0.26 0.69 0.93 0.97 0.89 0.80 0.69 0.58 0.50 0.46 0.39

130 0.09 0.45 0.74 0.93 0.97 0.90 0.79 0.65 0.54 0.50 0.46

160 -0.00 0.23 0.51 0.77 0.94 0.97 0.88 0.74 0.62 0.57 0.52

190 -0.06 0.08 0.31 0.57 0.80 0.94 0.96 0.86 0.75 0.67 0.56

220 -0.10 -0.01 0.15 0.35 0.57 0.77 0.93 0.96 0.88 0.78 0.58

250 -0.14 -0.06 0.04 0.17 0.33 0.52 0.74 0.93 0.97 0.88 0.60

280 -0.20 -0.10 -0.02 0.05 0.15 0.31 0.53 0.78 0.95 0.95 0.68

310 -0.29 -0.20 -0.09 -0.01 0.08 0.21 0.38 0.6 0.78 0.92 0.86

340 -0.34 -0.32 -0.18 -0.03 0.11 0.20 0.26 0.31 0.23 0.35 0.80

Discussion

The data-set used for this study comprises of lactation data for days 1 to 350 of lactation. To eliminate

measurement errors, observations with deviating data from the range of ± 4 standard deviations were

excluded from analyses. Because the data were nearly exclusively recorded in feeding experiments,

pre-analyses were needed to check if the feeding groups were significant. In the case of significant

feeding groups, the herd test week effect was extended to a herd test week group effect. Thus, one

herd test week changed to two or four effect classes corresponding to the number of feeding groups.

The results emphasize the necessity of pre-analyses of the significance of the feeding group effect.

To investigate the goodness of fit for the different lactation models, it has to be considered, that each

model depends on the underlying data structure, such as duration of lactation and shape of lactation

Chapter I

23

curve (Melzer et al., 2017). Lactation curves generally synthesize the main aspects of the shape of

lactation and give good results when animal groups are homogenous and fitted to average lactation

patterns (Macciotta et al., 2015; Pulina et al., 2016;). In this study these circumstances were given

and good results, which complied with literature, could be generated.

Fixed and random regression were used to estimate not only lactation curves but also repeatabilities

and correlations between cow effects (Hüttmann et al., 2009). Random regression models permit the

use of incomplete lactations and the subsequent inclusion of a large amount of data from the same

animal (Bignardi et al., 2011). Pool et al. (2000) concluded that the trajectory of the curves are not

well predicted by random regression models unless the complete data which is relevant to the

trajectory were used.

LG could be used to model the trajectories of random animal genetic and permanent environmental

variations. In general, Legendre polynomials with minimal order should be used, because the

computation capacity of the application of a random regression model is limited. On the other hand,

the LG with higher orders are useful when conventional orders fail due to better convergence (Pool

et al., 2000) wherefore LG4 was chosen in this study as the best alternative. Additionally, Pool et al.

(2000) recommended the selection of LG4 for genetic and permanent environmental effect, because

this model is well equipped. In general, random regression models performed better than fixed

regression models. The fixed lactation curve for all cows was the same and differed only in height

(Schaeffer and Jamrozik, 2008). The AS model for fixed regression was most suitable for both FI and

EB, but the LG4 modeled the lactation curve almost as well as the AS model. Also for other data sets

and different studies, this model was observed to be the best suited one (e.g. (Buttchereit et al., 2010).

The nested variants performed better for both traits, because different courses of lactation justify the

nesting.

The FI, EB and MY lactation curves behave as expected: High energy deficits at the beginning of

lactation are due to a delay in the increase in FI in combination with fast increasing MY. Differences

between first lactating HF and SI cows can arise from different growth rates. For the calculation of

EB, 15% higher energy requirements for growth of primiparous cows are assumed. It seems the

assumption of 15% is too high for SI. So, breed -specific growth requirements might be necessary in

the future. The trajectories of the mean lactation curves are similar to the results in literature

(Hüttmann et al., 2009; Macciotta et al., 2015; Li et al., 2018). Clear differences during the course of

lactation between the traits can be observed. At first sight, the EB estimates show slightly different

patterns compared to the FI cause of measurements that are more reliable. The trajectory of the curves

of lactation of the current study were similar to the trajectory of least-square mean phenotypic values

Chapter I

24

by day of lactation reported by Coffey et al. (2001). Lactation curves for MY tend to have a peak

around 40 DIM, which is conforms to the studies of Hüttmann et al. (2009) and López et al. (2015).

Also, the results for FI are in accordance with Hüttmann et al. (2009). The energy balance curves for

“optiKuh” cows represented in this data-set are consistent with those reported in other studies (Vries

and Veerkamp, 2000; Coffey et al., 2001). Hüttmann et al. (2009) assumed that EB depends more on

milk yield than FI and (Buttchereit et al., 2012b) also assigned meanings to feed intake in this relation.

Additionally, Villa-Godoy et al. (1988) and Veerkamp et al. (2000) suggested that the nutrient intake

is a big influence factor on the metabolic situation. To analyze the relationships between these traits

within the course of lactation, the run of the curves points to a causal relationship between the three

traits. However, it must be considered that FI and MY are measured directly, whereas EB lactation

curves are modeled by several traits such as MY, milk ingredients, FI and BW.

Possibilities in calculating FI with auxiliary characteristics, as Egger-Danner et al. (2015) showed,

involves risks of inaccuracy and a higher rate of error. (Buttchereit et al., 2012b) described the

possibility to build the energy status with fat:protein ratio. Furthermore, the difference between the

first and second lactations within the breed show an enhanced FI intake, higher milk yield and

improved EB. Intriguingly, the EB curve reaches the positive area in the same time (around 40-80 d

post partum) in both lactations, but has a generally higher trajectory in second lactation. These

findings are in line with the analyses of Villa-Godoy et al. (1988), Vries and Veerkamp, (2000) and

Coffey et al. (2002). FI and EB stabilize around the same time. The lactation curves of SI population

are lower with a better persistency compared to the trajectory of HF population, caused by the fact of

a lower energy deficit and less compensatory scoring (Coffey et al., 2001).

High repeatability estimates suggest a precise and ideal measurement over the period (Buttchereit et

al., 2012b). Furthermore, high repeatabilities promise high heritability. Thereby, highly useful genetic

variation for breeding editing is possible under the assumption that the proportion of the permanent

environment is not dominated in the genetic model, which is split into permanent environmental effect

and an additive genetic effect of the cow.

Not surprisingly, high repeatabilities were found for FI (0.59 to 0.84 for HF and 0.64 to 0.89 for SI)

which confirms the results of the study by Tetens et al. (2014), who found very high values for DIM

up to 0.56. In contrast, the studies of Veerkamp and Thompson (1999b), Coffey et al. (2002) and

Berry et al. (2007) reported high residual variance for FI in early lactation. Explanations for different

levels of repeatabilities between SI and HF are inter alia different technique measurement, e.g. access

to feed of SI-farms is controlled by troughs. This means a more precise time allocation for visits of

the individual cow can be possible. One HF research farms has free access and thus faster changes

Chapter I

25

between cows. For milk, there is a broad consistency of the measurements between SI and HF

research farms.

The low correlations between cow effects for EB between early and late DIM are supported by the

results of Buttchereit et al. (2010) who observed even a negative phenotypic correlation. The

correlations between cow effects, referring to the HF data-set, are higher at the beginning of the

lactation based on EB, which goes is in line with the results of Buttchereit et al. (2010) and Reist et

al. (2002). The development of both traits is the same in the SI population with slightly lower values.

EB values decreased from early and late DIM. Under the assumption of correlations between cow

effects generated in this study, selection for higher FI at the beginning does not necessarily lead to

over -conditioning at the end of lactation.

Conclusions

This study evaluated, feed intake and energy balance during lactation and the estimation of the

parameters for the random regression model for each trait. The results of the lactation curves,

repeatability and animal correlations between DIM are consistent with those found in the literature.

The animal correlations between different DIM ranged between 0.74 and 0.96 for feed intake and

0.81 and 0.96 for energy balance, during the beginning of lactation. The data set of the “optiKuh”

project provides good possibilities to use the traits FI or EB for selective decisions and the foundation

for an estimation of genetic parameters and variance components such as heritabilities to generate

reliable and genomic estimated breeding values for FI and EB in German HF and SI populations. The

estimated correlations between cow effects of different stages suggest possibilities to select for higher

feed intake at the beginning of lactation to avoid high energy deficits associated with only small

changes in feed intake at the end of lactation.

The “optiKuh” data represent an essential basis for a national reference population. These data

provide possibilities for a national or even an international expansion to obtain reliable breeding

values for FI or EB.

Acknowledgements

The project was supported by funds of the Federal Ministry of Food and Agriculture (BMEL) based

on a decision of the Parliament of the Federal Republic of Germany via the Federal Office for

Agriculture and Food (BLE) under the innovation support program (Germany).

Chapter I

26

References

Ali, T., and L. R. Schaeffer. 1987. Accounting for covariance among test day milk yields in dairy

cows. Can. J. Anim. Sci. 67(3):637–644. https://doi.org/10.4141/cjas87-067.

Banos, G., M. P. Coffey, R. F. Veerkamp, D. P. Berry, and E. Wall. 2012. Merging and characterising

phenotypic data on conventional and rare traits from dairy cattle experimental resources in three

countries. Animal an international journal of animal bioscience 6(7):1040–1048.

https://doi.org/10.1017/S1751731111002655.

Berry, D. P., M. P. Coffey, J. E. Pryce, Y. de Haas, P. Løvendahl, N. Krattenmacher, J. J. Crowley,

Z. Wang, D. Spurlock, K. Weigel, K. Macdonald, and R. F. Veerkamp. 2014. International genetic


Dairy Sci. 97(6):3894–3905. https://doi.org/10.3168/jds.2013-7548.

Berry, D. P., B. Horan, M. O'Donovan, F. Buckley, E. Kennedy, M. McEvoy, and P. Dillon. 2007.

Genetics of grass dry matter intake, energy balance, and digestibility in grazing irish dairy cows.

J. Dairy Sci. 90(10):4835–4845. https://doi.org/10.3168/jds.2007-0116.

Bignardi, A. B., L. El Faro, R. A. A. Torres Júnior, V. L. Cardoso, P. F. Machado, and L. G.

Albuquerque. 2011. Random regression models using different functions to model test-day milk

yield of Brazilian Holstein cows. Genetics and molecular research GMR 10(4):3565–3575.

https://doi.org/10.4238/2011.October.31.4.

Boichard, D., and M. Brochard. 2012. New phenotypes for new breeding goals in dairy cattle. Animal

an international journal of animal bioscience 6(4):544–550.

https://doi.org/10.1017/S1751731112000018.

Brotherstone, S., I. M. S. White, and K. Meyer. 2000. Genetic modelling of daily milk yield using

orthogonal polynomials and parametric curves. Anim. Sci. 70(03):407–415.

https://doi.org/10.1017/S1357729800051754.

Burnham, K. P., and D. R. Anderson. 1998. Model selection and inference: A practical information-

theoretic approach. Springer, New York NY u.a.

Butler, W. R. 2003. Energy balance relationships with follicular development, ovulation and fertility

in postpartum dairy cows. Livestock Production Science 83(2-3):211–218.

https://doi.org/10.1016/S0301-6226(03)00112-X.

Chapter I

27

Buttchereit, N., E. Stamer, W. Junge, and G. Thaller. 2010. Evaluation of five lactation curve models

fitted for fat:protein ratio of milk and daily energy balance. J. Dairy Sci. 93(4):1702–1712.

https://doi.org/10.3168/jds.2009-2198.

Buttchereit, N., E. Stamer, W. Junge, and G. Thaller. 2012. Modellevaluierung und Schätzung

genetischer Parameter für Energiebilanz und verwandte Merkmale bei Milchkühen.

Züchtungskunde: 84:74–89.

Coffey, M. P., G. C. Emmans, and S. Brotherstone. 2001. Genetic evaluation of dairy bulls for energy

balance traits using random regression. Animal Science 73(01):29–40.

https://doi.org/10.1017/S1357729800058021.

Coffey, M. P., G. Simm, and S. Brotherstone. 2002. Energy Balance Profiles for the First Three

Lactations of Dairy Cows Estimated Using Random Regression. J. Dairy Sci. 85(10):2669–2678.

https://doi.org/10.3168/jds.S0022-0302(02)74352-X.

Coffey, M. P., G. Simm, J. D. Oldham, W. G. Hill, and S. Brotherstone. 2004. Genotype and diet

effects on energy balance in the first three lactations of dairy cows. J. Dairy Sci. 87(12):4318–

4326. https://doi.org/10.3168/jds.S0022-0302(04)73577-8.

Collard, B. L., P. J. Boettcher, J.C.M. Dekkers, D. Petitclerc, and L. R. Schaeffer. 2000. Relationships

Between Energy Balance and Health Traits of Dairy Cattle in Early Lactation. J. Dairy Sci.

83(11):2683–2690. https://doi.org/10.3168/jds.S0022-0302(00)75162-9.

Connor, E. E. 2015. Invited review: improving feed efficiency in dairy production: challenges and

possibilities. Animal 9(3):395–408. https://doi.org/10.1017/S1751731114002997.

Egger-Danner, C., J. B. Cole, J. E. Pryce, N. Gengler, B. Heringstad, A. Bradley, and K. F. Stock.

2015. Invited review: Overview of new traits and phenotyping strategies in dairy cattle with a

focus on functional traits. Animal 9(2):191–207. https://doi.org/10.1017/S1751731114002614.

Guo, Z., and H. H. Swalve. 1995. Modelling of the lactation curve as a sub-model in the evaluation

of test day records. Proc. Interbull Mtg. Prague, International Bull Evaluation Service, Uppsala,

Schweden.

Haas, Y. de, M. P. L. Calus, R. F. Veerkamp, E. Wall, M. P. Coffey, H. D. Daetwyler, B. J. Hayes,

and J. E. Pryce. 2012. Improved accuracy of genomic prediction for dry matter intake of dairy

Chapter I

28

cattle from combined European and Australian data sets. J. Dairy Sci. 95(10):6103–6112.

https://doi.org/10.3168/jds.2011-5280.

Haas, Y. de, J. E. Pryce, M. P. L. Calus, E. Wall, D. P. Berry, P. Lovendahl, N. Krattenmacher, F.

Miglior, K. Weigel, D. Spurlock, K. A. Macdonald, B. Hulsegge, and R. F. Veerkamp. 2015.

Genomic prediction of dry matter intake in dairy cattle from an international data set consisting

of research herds in Europe, North America, and Australasia. J. Dairy Sci. 98(9):6522–6534.

https://doi.org/10.3168/jds.2014-9257.

Hüttmann, H., E. Stamer, W. Junge, G. Thaller, and E. Kalm. 2009. Analysis of feed intake and

energy balance of high-yielding first lactating Holstein cows with fixed and random regression

models. Animal 3(2):181–188. https://doi.org/10.1017/S175173110800325X.

Ingvartsen, K. L., and J. B. Andersen. 2000. Integration of metabolism and intake regulation: A

review focusing on periparturient animals. J. Dairy Sci. 83(7):1573–1597.

https://doi.org/10.3168/jds.S0022-0302(00)75029-6.

Jamrozik, J., and L. R. Schaeffer. 1997. Estimates of Genetic Parameters for a Test Day Model with

Random Regressions for Yield Traits of First Lactation Holsteins. J. Dairy Sci. (Journal of dairy

science) 80(4):762–770. https://doi.org/10.3168/jds.S0022-0302(97)75996-4.

Jensen, J. 2001. Genetic Evaluation of Dairy Cattle Using Test-Day Models. J. Dairy Sci. (Journal of

dairy science) 84(12):2803–2812. https://doi.org/10.3168/jds.S0022-0302(01)74736-4.

Kirchgeßner, M. 2014. Tierernährung. 14th ed. DLG-Verlag, Frankfurt am Main, Germany.

Leesen, R. von, J. Tetens, E. Stamer, W. Junge, G. Thaller, and N. Krattenmacher. 2014. Effect of

genetic merit for energy balance on luteal activity and subsequent reproductive performance in

primiparous Holstein-Friesian cows. J. Dairy Sci. 97(2):1128–1138.

https://doi.org/10.3168/jds.2013-7185.

Li, B., W. F. Fikse, P. Løvendahl, J. Lassen, M. H. Lidauer, P. Mäntysaari, and B. Berglund. 2018.

Genetic heterogeneity of feed intake, energy-corrected milk, and body weight across lactation in

primiparous Holstein, Nordic Red, and Jersey cows. J. Dairy Sci. 101(11):10011–10021.

https://doi.org/10.3168/jds.2018-14611.

Littell, R. C. 2007. SAS for mixed models. 2. ed., 2. print. SAS Press, Cary, NC.

Chapter I

29

López, S., J. France, N. E. Odongo, R. A. McBride, E. Kebreab, O. AlZahal, B. W. McBride, and J.

Dijkstra. 2015. On the analysis of Canadian Holstein dairy cow lactation curves using standard

growth functions. J. Dairy Sci. 98(4):2701–2712. https://doi.org/10.3168/jds.2014-8132.

Macciotta, N. P. P., G. Gaspa, L. Bomba, D. Vicario, C. Dimauro, M. Cellesi, and P. Ajmone-Marsan.

2015. Genome-wide association analysis in Italian Simmental cows for lactation curve traits using

a low-density (7K) SNP panel. J. Dairy Sci. 98(11):8175–8185. https://doi.org/10.3168/jds.2015-

9500.

Melzer, N., S. Trißl, and G. Nürnberg. 2017. Short communication: Estimating lactation curves for

highly inhomogeneous milk yield data of an F2population (Charolais × German Holstein). J.


Pool, M. H., L.L.G. Janss, and T.H.E. Meuwissen. 2000. Genetic Parameters of Legendre

Polynomials for First Parity Lactation Curves. J. Dairy Sci. 83(11):2640–2649.


Pulina, G., N. Macciotta, and A. Nudda. 2016. Milk composition and feeding in the Italian dairy

sheep. Italian Journal of Animal Science 4(sup1):5–14. https://doi.org/10.4081/ijas.2005.1s.5.

Reist, M., D. Erdin, D. von Euw, K. Tschuemperlin, H. Leuenberger, Y. Chilliard, H. M. Hammon,

C. Morel, C. Philipona, Y. Zbinden, N. Kuenzi, and J. W. Blum. 2002. Estimation of energy

balance at the individual and herd level using blood and milk traits in high-yielding dairy cows.

J. Dairy Sci. 85(12):3314–3327. https://doi.org/10.3168/jds.S0022-0302(02)74420-2.

SAS. 2012. SAS 9.2 TS Level 1 Mo., Cary, NC27513, USA. SAS Institute Inc.

Schaeffer, L. R., and J. Jamrozik. 2008. Random regression models: A longitudinal perspective.

Journal of animal breeding and genetics = Zeitschrift fur Tierzuchtung und Zuchtungsbiologie

125(3):145–146. https://doi.org/10.1111/j.1439-0388.2008.00748.x.

Schwarz, G. 1978. Estimating the Dimension of a Model. The Annals of Statistics 6(2):461–464.

https://doi.org/10.1214/aos/1176344136.

Spurlock, D. M., J. C. M. Dekkers, R. Fernando, D. A. Koltes, and A. Wolc. 2012. Genetic parameters

for energy balance, feed efficiency, and related traits in Holstein cattle. J. Dairy Sci. 95(9):5393–

5402. https://doi.org/10.3168/jds.2012-5407.

Chapter I

30

Stamer, E., W. Brade, and G. Thaller. 2011. Modellentwicklung und Schätzung genetischer

Parameter für den Harnstoffgehalt in der Milch erst- und zweitlaktierender Holstein-Kühe.

Züchtungskunde: 83:104–117.

Tetens, J., G. Thaller, and N. Krattenmacher. 2014. Genetic and genomic dissection of dry matter

intake at different lactation stages in primiparous Holstein cows. J. Dairy Sci. 97(1):520–531.

https://doi.org/10.3168/jds.2013-7301.

Uddin, M. E., T. Meuwissen, and R. F. Veerkamp. 2018. Adjusting for heterogeneity of experimental

data in genetic evaluation of dry matter intake in dairy cattle. Journal of animal breeding and

genetics = Zeitschrift fur Tierzuchtung und Zuchtungsbiologie 135(1):28–36.

https://doi.org/10.1111/jbg.12300.

Veerkamp, R. F., J. K. Oldenbroek, H. J. van der Gaast, and J.H.J. van der Werf. 2000. Genetic

Correlation Between Days Until Start of Luteal Activity and Milk Yield, Energy Balance, and

Live Weights. J. Dairy Sci. (Journal of dairy science) 83(3):577–583.


Veerkamp, R. F., J. E. Pryce, D. Spurlock, D. P. Berry, M. P. Coffey, P. Løvendahl, R. d. van Linde,

J. Bryant, F. Miglior, Z. Wang, M. Winters, N. Krattenmacher, N. Charfeddine, J. Pedersen, and

Y. de Haas. 2013. Selection on Feed Intake of Feed Efficiency: A Position Paper from gDMI

Breeding Goal Discussion. Interbull Bulletin(No. 47):15–22.

Veerkamp, R. F., and R. Thompson. 1999b. Multi-trait covariance functions to estimate genetic

correlations between milk yield, dry-matter intake and live weight during lactation. Metabolic

stress in dairy cows. J.D. Oldham, G.Simm, A.F. Groen, B.L. Nielsen, J.E. Pryce and

T.L.J.Lawrence, ed. British Society of Animal Science, Occasional publication No. 24:147–151.

Villa-Godoy, A., T. L. Hughes, R. S. Emery, L. T. Chapin, and R. L. Fogwell. 1988. Association

Between Energy Balance and Luteal Function in Lactating Dairy Cows. J. Dairy Sci. (Journal of

dairy science) 71(4):1063–1072. https://doi.org/10.3168/jds.S0022-0302(88)79653-8.

Vries, M. J. de, and R. F. Veerkamp. 2000. Energy Balance of Dairy Cattle in Relation to Milk

Production Variables and Fertility. J. Dairy Sci. (Journal of dairy science) 83(1):62–69.


Chapter I

31

Yao, C., G. de Los Campos, M. J. Vandehaar, D. M. Spurlock, L. E. Armentano, M. Coffey, Y. de

Haas, R. F. Veerkamp, C. R. Staples, E. E. Connor, Z. Wang, M. D. Hanigan, R. J. Tempelman,

and K. A. Weigel. 2017. Use of genotype × environment interaction model to accommodate

genetic heterogeneity for residual feed intake, dry matter intake, net energy in milk, and metabolic

body weight in dairy cattle. J. Dairy Sci. 100(3):2007–2016. https://doi.org/10.3168/jds.2016-

11606.

Chapter I

32

Chapter II

33

Chapter II

Estimation of genetic parameters and breeding values for feed intake and energy

balance using pedigree relationships or single-step genomic evaluation in Holstein-

Friesian cows

I. Harder1, E. Stamer², W. Junge1, G. Thaller1

1Institute of Animal Breeding and Husbandry, Christian-Albrechts-University, D-24098 Kiel

²TiDa Tier und Daten GmbH, D-24259 Westensee/Brux

Accepted in Journal of Dairy Science

Chapter II

34

Abstract

At the beginning of lactation, high-performing dairy cows often experience a severe energy deficit,

which in turn is associated with metabolic stress. Increasing feed intake (FI) or reducing the energy

deficit in this period could improve the metabolic stability and thus the health of the animals. Genomic

selection for the first time enables the inclusion of this hard to measure trait in breeding programs.

The objective of the current study was the estimation of genetic parameters and genomic breeding

values for FI and energy balance (EB).

For this purpose, 1,374 Holstein-Frisian (HF) dairy cows from eight German research farms were

phenotyped with standardized FI data protocols.

After data editing phenotypic data of HF comprised a total of 40,012 average weekly FI records with

a mean of 21.8 ± 4.3 kg/d. For EB 33,376 average weekly records were available with a mean of 3.20

± 29.4 MJ NEL/d. With the Illumina BovineSNP50 BeadChip 1,128 of phenotyped cows were

genotyped. Female candidates of 35 cows of the HF population were genotyped but not phenotyped.

Pedigree information contained sires and dams four generations back. The random regression animal

model included the fixed effects of herd test week alternatively herd group test week, parity and stage

of lactation modelled by the function according to Ali and Schaeffer. For both, the random permanent

environmental effect across lactations and the random additive genetic effect, third-order Legendre

polynomials were chosen. Additionally, a random permanent environmental cow effect within

lactation was included.

Analyses for heritabilities, genetic correlations between different lactation stages and breeding values

were estimated using both, pedigree relationships and single-step genomic evaluation, carried out

with the DMU software package, respectively. This allowed for comparison of conventional

reliabilities with genomic assisted reliabilities based on real data to evaluate the gain of genotyping.

Heritability estimates ranged between 0.12 and 0.50 for FI and 0.15 and 0.48 for EB and increased

towards the end of lactation. Genetic correlations were weak between early and late lactation with a

value of 0.05 for FI and negative with a value of -0.05 for EB. Reliabilities for genomic values of

cows for FI and EB ranged between 0.33 and 0.61 and 0.27 and 0.47, respectively. For the genotyped

cows without phenotypes, the inclusion of genomic relationship leads to the increase of the average

reliability of the breeding value for FI by nearly 9% and for EB by 4%. The results show the

possibility to combine pedigree, genotypes and phenotypes for increasing FI or EB to reduce health

and reproductive problems especially at the beginning of lactation. Nevertheless, the reference

population needs to be extended to reach higher breeding value reliabilities.

Chapter II

35

Introduction

Since a long time, cows were bred for high milk yield (MY), but the considerable increase in milk

production has been accompanied by a higher occurrence of health and fertility problems (Mäntysaari

et al., 2012). In this context, different aspects need to be considered. On the one hand, feed costs

constitute the major expense in dairy production for which reason it is economically important to

improve feed efficiency (Wallén et al., 2017). Because higher efficiency and increased milk

production is accompanied with a pronounced energy deficit, it can lead, on the other hand, to

metabolic stress and health problems (Mäntysaari et al., 2012; Spurlock et al., 2012).

Especially at the beginning of lactation, MY increases faster compared to feed intake (FI) (Spurlock

et al., 2012). When energy expenditures exceed intake, dairy cows experience energy deficits with

eventually high intensity and long duration ( Vries and Veerkamp, 2000; Leesen et al., 2014;).

Moreover, the pronounced negative energy balance (EB) is associated with metabolic diseases like

acidosis, ketosis and milk fever (Randhawa et al., 2014). Metabolic imbalances, reproductive

problems and other health issues are the limiting factors for reproductive life span (Rauw et al., 1998;

Heringstad et al., 2000; Reist et al., 2002) and producers must be aware of the potential correlated

changes that might influence fitness and welfare of dairy cows (Oltenacu and Broom, 2010). These

antagonistic aspects are well known and in this context, the traits FI or EB have been suggested as

selection traits to improve fitness (Spurlock et al., 2012).

Consequently, increased FI can enable an adequate energy supply resulting in improved health,

metabolic stability and sustainable mean of increasing farm profitability (Coffey et al., 2002).

So far, breeding for FI using conventional breeding values was hampered due to an insufficient data

basis caused by a difficult and costly-to-measure trait (Haas et al., 2015). Mainly research farms have

possibilities to collect individual FI data of cows.

The usefulness of an inclusion of FI and EB in breeding goals appropriately depends strongly on

genetic parameters (Tetens et al., 2014). Sound knowledge of the genetic architecture and the

reliability of the inheritance are essential. FI seems to be influenced by different genes at different

stages of lactation (Veerkamp and Koenen, 1999; Berry et al., 2007; Tetens et al., 2014). This means

that genetic parameters such as heritabilities and genetic correlations between different lactation

stages vary across lactation. Therefore, different sections of lactation need to be analyzed as different

but correlated traits.

Since the concept of genomic selection (Meuwissen et al., 2001), integration of the hard-to-measure

traits FI or EB into the breeding goal of dairy livestock populations seems to be more efficient.

Chapter II

36

Already in 2011, shortly after the implementation of genomic selection in dairy breeding, an

international project (gDMI) was initiated and realized to establish a sufficiently precise breeding

value estimation for the trait FI (Veerkamp et al., 2013; Berry et al., 2014; Haas et al., 2015).

In the current study, the single-step method is used for genomic evaluations (Aguilar et al., 2010;

Christensen and Lund, 2010). Via application of this method, genotypic, pedigree and phenotypic

information are combined and all available information can be used simultaneously (Legarra et al.,

2014). Single-step provides the best comprehensive option for genomic evaluation and is considered

to be the most effective approach to generate genomic breeding values with accurate results (Aguilar

et al., 2010; Christensen and Lund, 2010; Přibyl et al., 2015).

Achieving high accuracy is important because the reliability of predicted genomic breeding values

determine the benefit from genomic selection (Meuwissen et al., 2001; Edriss et al., 2013). Besides

the size of the reference population, it depends on the heritability and the genetic relationship

(Goddard and Hayes, 2009).

The aim of this study is the implementation of the single-step evaluation for the breed Holstein-

Friesian (HF) using data from eight dairy research farms located along Germany for the two traits FI

and EB.

These data are used to analyze heritabilities, genetic correlations between different lactation stages

and breeding values. FI and EB were compared between pedigree-based and single-step estimation

runs. Furthermore, realized breeding value reliabilities for cows and sires with or without phenotypic

information are compared.

Material and Methods

Phenotypic data

For the estimation of the genomic and conventional breeding values, 1,341 HF for FI and 1,322 for

EB with weekly averaged values were available (Table 1). Energy intake represents net energy

lactation (NEL) and was computed by summing energy amounts of eaten partly or total mixed rations

and concentrates. For the calculation of EB, energy intake, milk energy, maintenance costs, gestation

stage and body weight at the beginning of lactation as parameter of growth of first lactation cows

were used. Individual feed intake was recorded within twelve German research farms via feeding

troughs equipped with a weighing unit and automatic cow identification. Cows were fed ad libitum.

The target value for concentrates was predetermined by the project “optiKuh” with 250 g/kg ECM.

Data was edited and summarized within farm. Due to different frequencies of measurement (e.g. daily

FI and weekly milk ingredients) the traits were averaged within calendar week. The considered

Chapter II

37

lactation period was defined from 1st to 350th day. Observations outside of four standard deviations

were excluded from further analyses. The resulting means for FI, MY, and EB were 21.8 kg, 35.5 kg,

and 3.20 MJ NEL. Available pedigree information for cows is traced four generations back. Detailed

information are given in Harder et al. (2019).

Table 1. Descriptive statistics for weekly records of the traits feed intake, energy balance and milk

yield

Trait No.

cows

No.

parities

No.

obs Mean SD

Feed intake (kg DM) 1341 1928 40,012 21.8 4.25

Energy balance (MJ NEL) 1322 1865 33,376 3.20 29.4

Milk yield (kg) 1338 1917 39,838 35.5 8.81

Genotypic data

Genotypic data consisted of 1,895 cows, sires and sons of the sires in total, genotyped with the

Illumina Bovine SNP50 BeadChip (50K, Illumina Inc., San Diego, CA). Three data panels were used

for the genetic analyses. The main panel consists of 1,163 cows genotyped with the Illumina Bovine

SNP50 Bead Chip Version v2_C and v3_A1. Of these cows 1,128 were phenotyped. For another 35

HF no phenotypic information was available. In the second panel 491 sires and grandsires of the cows

were genotyped with chip versions A and v2_C. As well, 241 sons of these sires (fathers) were

genotyped with the versions A and v2 as well (Table 2). For final evaluation, a joint SNP set was

created containing all common SNPs of the three chip versions (n=49,184).

To ensure sufficient genotypic data quality SNP and individuals with call rates lower than 95% and

mean GenCall Score (GC-Score) lower than 0.6 (explanations e.g. Edriss, 2013) were excluded. After

quality control the final SNP data set contained 1,828 animals (1,163 cows, 448 sires, 217 grandsires)

and 45,373 common SNPs.

Chapter II

38

Table 2. Genotypic data – number of animals and SNPs of raw data and after quality control

Chipversion No of animals

SNPs Cows Sires/grandsires Sons of sires

Illumina BovineSNP50 BeadChip_A 54,001 - 345 55

Illumina BovineSNP50 BeadChip_v2_C 54,609 1026 146 186

Illumina BovineSNP50 BeadChip_v3_A1 53,218 137 - -

Total number of animals 1163 491 241

Animal callrate (≥ 0.95) 1163 448 217

No of SNPs

joint SNPs (version 1, 2 and 3) 49,184

GC-score ≥ 0.6 45,648

SNP callrate ≥ 0.95 45,373

Genotypes were coded as 0 for missing nucleobase information, 1 for Adenine, 2 for Cytosine, 3 for

Guanine and 4 for Thymine. Using the software package G-matrix (Madsen et al., 2013), the genomic

relationship matrix was calculated with the method VanRaden (VanRaden, 2008). A minor allele

frequency of 1% (representing the default justification of G-matrix) was chosen. The SAS procedure

INBREED was used to enable a comparison between the pedigree-based and genomic relationship

(SAS, 2012).

Statistical Model

Model evaluations for FI and EB were conducted in the previous study of Harder et al. (2019). Based

upon the results of this study, the general lactation curves were modeled within lactation number by

the function Ali and Schaeffer (AS) (Ali and Schaeffer, 1987). Random regression coefficients for

permanent cow effects and additive genetic effects were modeled by Legendre polynomials of 3rd

degree (LP3) (Brotherstone et al., 2000). Additionally a mean permanent environmental effect of the

cow within lactation was considered. A finer modeling of this effect by a Legendre polynomial could

not be realized due to missing convergence. Furthermore, the model includes the fixed effects

lactation number and herd test week. In case of significant feeding group effects, herd test week was

replaced by herd group test week, which had to contain at least two observations. This resulted in the

following linear random regression model for the estimation of genetic parameters:

Model evaluations for FI and EB were conducted in the previous study of Harder et al. (2019). Based

upon the results of this study, the general lactation curves were modeled within lactation number by

the function Ali and Schaeffer (AS) (Ali and Schaeffer, 1987). Random regression coefficients for

permanent cow effects and additive genetic effects were modeled by Legendre polynomials of 3rd

degree (LP3) (Brotherstone et al., 2000). Additionally a mean permanent environmental effect of the

cow within lactation was considered. A finer modeling of this effect by a Legendre polynomial could

Chapter II

39

not be realized due to missing convergence. Furthermore, the model includes the fixed effects

lactation number and herd test week. In case of significant feeding group effects, herd test week was

replaced by herd group test week, which had to contain at least two observations. This resulted in the

following linear random regression model for the estimation of genetic parameters:

yijklm = µ + HTWi + LNOj + ∑ 𝐶𝑗𝑛4𝑛=1 asjn + ∑ 𝑝𝑘𝑛

4𝑛=1 lpkn + pl + ∑ 𝑎𝑘𝑛

4𝑛=1 lpkn + eijklm

where yijklm is the weekly trait observation; µ is the overall intercept; HTWi is the fixed effect of

the ith herd (group) test week; LNOj is the fixed effect of the jth lactation class (j=1,…,4), class 4

includes the 4th and higher lactations; Cjn is the nth fixed regression coefficient within lactation class

j; asjn is the nth term of the Ali and Schaeffer function for DIM within lactation class j; pkn is the nth

random regression coefficient of the permanent environmental effect of the kth cow; lpkn is the nth

term of the third-order Legendre polynomial for DIM of cow k; pl is the random effect of the lth

combination between permanent environmental cow effect and lactation (1,…11); akn is the nth

random regression coefficient of the additive genetic effect of the kth cow; eijklm is the random

residual effect.

In matrix notation the RRM can be written as:

y = Xb + Z1p + Z2l + Z3a + e

where b = unknown parameters for fixed effects and fixed regression coefficients, p = random

regression coefficients for permanent environmental cow effects across lactations, l = vector of

permanent environmental cow effects within lactation, a = random regression coefficients for additive

genetic cow effects, e = vector of temporary environmental (= residual) effects, and X, Z1, Z2, and

Z3 are the corresponding incidence and covariates matrices.

The mixed model equations (MME) are represented as:

(

𝐗′𝐗 𝐗′𝐙𝟏 𝐗′𝐙𝟐 𝐗′𝐙𝟑𝐙𝟏′𝐗 𝐙𝟏

′ 𝐙𝟏 + 𝐈 𝛂𝟏 𝐙𝟏′ 𝐙𝟐 𝐙𝟏

′ 𝐙𝟑𝐙𝟐′𝐗 𝐙𝟐

′ 𝐙𝟏 𝐙𝟐′ 𝐙𝟐 + 𝐈α2 𝐙𝟐

′ 𝐙𝟑𝐙𝟑′𝐗 𝐙𝟑

′ 𝐙𝟏 𝐙𝟑′ 𝐙𝟐 𝐙𝟑

′ 𝐙𝟑 + 𝐇−𝟏 𝛂𝟑)

(

𝐛𝐩𝐥𝐚

) =

(

𝐗′𝐘𝐙𝟏′ 𝐘

𝐙𝟐′ 𝐘

𝐙𝟑′ 𝐘)

where H is the combined numerator and genomic relationship matrix; I is an identity matrix,

α1 = P-1𝑒2, α2 = 𝑒

2/𝑙2, α3 = G-1𝑒

2, G is the variance-covariance matrix of the additive genetic

random regression coefficients; P is the variance-covariance matrix of the permanent environmental

random regression coefficients across lactations; 𝑙2 is the permanent environmental variance within

lactation.

Variance components, heritabilities, genetic correlations between lactation stages and breeding values

for weekly FI and EB were estimated by two ways using the software DMU (Madsen et al., 2013).

Chapter II

40

First, a genetic evaluation was carried out using phenotypes and the pedigree relationship matrix.

Second, the single-step method was applied combining genotype, pedigree and phenotypic

information. The pedigree relationship matrix was augmented by genomic relationship. So, the

inverse of the resulting matrix H can be seen as a modification of regular pedigree relationships to

accommodate genomic relationships. Any formulation using inverse relationship A-1 can use H-1

instead (Legarra et al., 2009).

Based on the procedures outlined in Aguilar et al. (2010) and Christensen and Lund (2010) the inverse

of the matrix H can be written according to Legarra et al. (2014) as:

𝐇−1 = 𝐀−1 + [0 0𝟎 ((𝟏 −𝐰)𝐆 +𝐰 𝐀22

−𝟏 )−1 − 𝐀22−𝟏],

where A-1 is the inverse of the pedigree relationship matrix for all animals, G is the genomic

relationship matrix for genotyped animals, and A22-1 represents the inverse of the pedigree

relationship matrix for all genotyped animals. Previous studies have investigated the use of weight

(w) to avoid convergence problems and bias (Guarini et al., 2018). In this study, a weight of 5% for

both traits FI and EB was applied. Estimated (co) variances and LP3 coefficients were used for both

calculation of heritabilities for all considered DIM and calculation of genetic correlations between

selected lactation stages. Finally, reliabilities of mean breeding values were

computed as follows:

𝑟2 = 1 −PEV

𝜎𝑎2 ,

where PEV is the predicted error variance of the intercept for the estimated cow specific LP3 function

and σa2 is the additive genetic variance. The intercept represents the mean breeding value.

For this study, different data sets were prepared to estimate genetic parameters and breeding values

for FI and EB:

First data set (1) consists of all cows of the “optiKuh” training population only with pedigree

relationship (Data_1), (2) consists of all cows of the “optiKuh” training population with pedigree and

genomic relationship (Data_2), (3) augmented the data set with genotyped sires and dam sires

(Data_3) and (4) genotyped sons of the sires (Data_4) are included as well.

Results

To ensure high quality of a SNP, the signal strength of every SNP GenCall (GC)-score was measured.

The threshold (grey line in Figure 1) was set at 0.6, so every SNP with a mean GC-score lower than

0.6 was excluded from further analyses. Overall 7,453 SNPs (14.64%) lay below this threshold and

were discarded.

Chapter II

41

Figure 1. Frequency distribution of the mean GC-score (quality control threshold at 0.6)

The pedigree check step for data quality control ensures the identification of every animal with an

incorrectly recorded pedigree, a misidentified cow, a wrong cow sample or a mistake in breed

belonging. In total, 14 animals were excluded. It allows the checking of full-sibs and half-sibs. Figure

2 illustrates the relatedness of cows, by the comparison of pedigree relationship coefficients and

marker relationship coefficients. Spots are expected to be located along the diagonal, in the area

around 0 (unrelated), 0.25 (half-sibs), 0.5 (full-sibs) and 1 (related to itself). Few outliers in the area

around 0 and around 0.3 may constitute inconsistencies in, firstly, the pedigree under assumption of

correct genomic coefficients or, secondly, blood samples which were assigned wrongly under the

assumption of correct pedigree-information. The negative marker relationships are due to the G-

matrix method which uses the allele frequencies in the base population, but these frequencies are

unknown (VanRaden, 2008). Therefore, the G-matrix was calculated by using the allele frequencies

of the available population, which resulted in a partly negative marker relationship.

0

1000

2000

3000

4000

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Fre

qu

ency

GC-score

Chapter II

42

Figure 2. Comparison of the relationship coefficients between pedigree relationship matrix and

genomic relationship matrix

Figure 3 depicts heritability estimates of FI and EB across lactation, distinguished between Data_1

and Data_2. Heritabilities of FI ranged from 0.12 to 0.50 with the lowest values in early lactation,

which increased slightly from lactation DIM 150 onward with the highest values found at the end of

lactation. Similar to this pattern, the heritability estimates for EB ranged from 0.15 to 0.48. In general,

estimated heritabilities of both traits were lower at the beginning of lactation but increased until the

end of the considered lactation. The heritability estimates derived from Data_2 were rather similar at

the beginning of lactation but had a lower trajectory so it shifted around 100 DIM and the

conventional heritability was higher at the end.

Chapter II

43

Figure 3. Heritabilities of feed intake (FI) and energy balance (EB) of Data_1 (consists of all cows of

the “optiKuh” training population only with pedigree relationship) and Data_2 (consists of all cows

of the “optiKuh” training population with pedigree and genomic relationship) across the first 350

DIM

Genetic correlations of FI and EB between twelve selected, equidistantly DIM are presented in Table

3 (Data_1) and Table 4 (Data_2). For FI values ranged from 0.05 to 1.00 in Data_1 and from 0.05 to

0.99 in Data_2. In general, estimated genetic correlations were slightly higher in Data_2 (Table 4).

EB genetic correlations were all positive and ranged from 0.00 to 0.97 in Data_1 whereas values in

Data_2 are positive until DIM 310 and then turn into negative range.

In general, genetic correlations between neighbored DIM are high and decrease with increasing DIM

intervals. Both traits show the same trajectory in the course of lactation. Estimates from early and

mid lactation are in middle range, whereas correlations between early and late lactation are low. One

value in Data_2 for EB turns even into negative.

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0 30 60 90 120 150 180 210 240 270 300 330

Heri

tab

ilit

y

DIM

EB pedigree-based

EB Single Step (C)

FI pedigree-based

FI Single Step (C)

Chapter II

44

Table 3. Genetic correlations of Data_1 (consists of all cows of the “optiKuh” training population

only with pedigree relationship) for FI (above the diagonal) and EB (below the diagonal) between

different DIM

DIM 10 40 70 100 130 160 190 220 250 280 310 340

10 0.80 0.52 0.35 0.24 0.17 0.13 0.10 0.09 0.08 0.07 0.05

40 0.82 0.93 0.82 0.70 0.56 0.44 0.35 0.32 0.35 0.42 0.49

70 0.49 0.90 0.97 0.89 0.77 0.63 0.53 0.50 0.53 0.62 0.70

100 0.26 0.75 0.96 0.97 0.89 0.78 0.69 0.65 0.68 0.75 0.80

130 0.12 0.61 0.87 0.97 0.97 0.90 0.83 0.80 0.81 0.84 0.84

160 0.04 0.47 0.74 0.88 0.97 0.98 0.94 0.91 0.91 0.90 0.83

190 0.00 0.34 0.59 0.75 0.88 0.97 0.99 0.98 0.96 0.92 0.80

220 0.00 0.25 0.46 0.63 0.78 0.90 0.98 1.00 0.98 0.92 0.78

250 0.02 0.22 0.40 0.55 0.70 0.84 0.94 0.99 0.99 0.94 0.80

280 0.04 0.26 0.43 0.56 0.69 0.80 0.89 0.94 0.98 0.97 0.87

310 0.06 0.35 0.52 0.62 0.69 0.74 0.77 0.80 0.86 0.95 0.96

340 0.05 0.42 0.60 0.65 0.66 0.63 0.59 0.58 0.63 0.77 0.93

Table 4. Genetic correlations of Data_2 (consists of all cows of the “optiKuh” training population

with pedigree and genomic relationship) for FI (above the diagonal) and EB (below the diagonal)

between different DIM

DIM 10 40 70 100 130 160 190 220 250 280 310 340

10 0.85 0.60 0.44 0.34 0.28 0.24 0.22 0.21 0.20 0.18 0.14

40 0.88 0.93 0.83 0.74 0.65 0.56 0.49 0.46 0.47 0.51 0.54

70 0.63 0.92 0.97 0.92 0.83 0.73 0.65 0.61 0.64 0.69 0.74

100 0.42 0.78 0.96 0.98 0.92 0.84 0.76 0.73 0.75 0.80 0.83

130 0.27 0.64 0.87 0.97 0.98 0.93 0.87 0.84 0.84 0.87 0.87

160 0.18 0.50 0.74 0.88 0.97 0.98 0.95 0.92 0.92 0.92 0.87

190 0.12 0.37 0.58 0.74 0.87 0.97 0.99 0.98 0.96 0.93 0.85

220 0.09 0.27 0.45 0.62 0.78 0.91 0.98 1.00 0.98 0.94 0.83

250 0.07 0.23 0.40 0.56 0.72 0.87 0.96 0.99 0.99 0.95 0.84

280 0.05 0.25 0.44 0.60 0.74 0.87 0.94 0.96 0.98 0.98 0.90

310 0.00 0.30 0.54 0.69 0.79 0.84 0.85 0.84 0.86 0.94 0.97

340 -0.05 0.34 0.60 0.72 0.74 0.70 0.62 0.56 0.58 0.71 0.91

The realized reliabilities for mean breeding values of the cows for FI are shown in Figure 4. The

reliabilities increase with rising numbers of FI weekly records. Up to 40 week averages, the reliability

for Data_2 was higher, in comparison to Data_1. This indicates, that in the current study the less data

is available, the higher benefit of using Single-step data with respect to reliabilities of breeding values.

However, the more data is available for the computations, the higher the reliabilities of the breeding

values with Data_1 is. Highest reliabilities for Data_1 ranged at 0.62 and for Data_2 at 0.61.

As expected, the differences between the four data sets within the single-step are marginal and added

genomic information of sires and sons of sires did not improve the reliabilities significantly in this

study.

Chapter II

45

For the 35 cows without phenotypic but genotypic information, single-step reliability of 0.23 could

be realized which is 0.08 higher compared to the pedigree-based reliability with 0.14.

Similar to FI, the reliabilities of EB breeding values are higher using single-step for cows with a low

number of up to 20 observations but this changes with an increasing number of observations per cow.

Figure 4. Realized reliabilities of breeding values for feed intake out of the four data sets – Data_1

(consists of all cows of the “optiKuh” training population only with pedigree relationship), Data_2

(consists of all cows of the “optiKuh” training population with pedigree and genomic relationship),

Data_3 (augmented the data set with genotyped sires and dam sires) and Data_4 (augmented the data

set with sons of the sires)

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0 1-10 11-20 21-30 31-40 41-50 51-60 61-70 71-80 81-90 91-99

35 53 378 240 149 110 40 42 69 29 18

Reliabilit

y o

f bre

edin

g v

alu

es

Number of mean week values / Number of cows

Data_1

Data_2

Data_3

Data_4

Chapter II

46

In general, the observed reliabilities for FI are slightly higher than for EB (max. value for FI 0.62 and

EB 0.60) (Figure 5).

Figure 5. Realized reliabilities of breeding values for energy balance out of the four data sets – Data_1

(consists of all cows of the “optiKuh” training population only with pedigree relationship), Data_2

(consists of all cows of the “optiKuh” training population with pedigree and genomic relationship),

Data_3 (augmented the data set with genotyped sires and dam sires) and Data_4 (augmented the data

set with sons of the sires)

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0 1-10 11-20 21-30 31-40 41-50 51-60 61-70 71-80 81-90 91-99

39 178 393 211 100 69 36 69 46 20 2

Reli

ab

ilit

y o

f b

reed

ing v

alu

es

Number of mean week values / Number of cows

Data_1

Data_2

Data_3

Data_4

Chapter II

47

For the 1,163 genotyped cows, FI breeding values of pedigree-based results (Data_1) are contrasted

with Single-step estimations (Data_2). The fitted linear regression shows a coefficient of

determination of R² = 0.86 (r = 0.93) (Figure 6).

Figure 6. Scatter plot between mean feed intake (FI) breeding values of Data_1 and Data_2 estimation

(1,163 genotyped cows)

R² = 0.8601

-6.0

-5.0

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

4.0

5.0

6.0

-5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0

Feed

in

tak

e (

kg D

M)

-si

ngle

-ste

p

Feed intake (kg DM) - pedigree-based

Chapter II

48

Figure 7 and 8 show reliabilities of the FI and EB breeding values for the sires. With increasing

number of daughters, the reliability gets higher. A sire with max. five daughters could generate

reliabilities of roughly 0.27 for FI and 0.21 for EB.

Figure 7. Realized reliabilities of the bull breeding values for feed intake out of the four data sets –

Data_1 (consists of all cows of the “optiKuh” training population only with pedigree relationship),

Data_2 (consists of all cows of the “optiKuh” training population with pedigree and genomic

relationship), Data_3 (augmented the data set with genotyped sires and dam sires) and Data_4

(augmented the data set with sons of the sires)

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

228 31 12 3 0 1

1-5 6-10 11-15 16-20 21-25 26-30

Rli

ab

ilit

y o

f b

reed

ing v

alu

es

Number of bulls / Number of daughters

Data_1

Data_2

Data_3

Data_4

Chapter II

49

Figure 8. Realized reliabilities of the bull breeding values for energy balance out of the four data sets

– Data_1 (consists of all cows of the “optiKuh” training population only with pedigree relationship),

Data_2 (consists of all cows of the “optiKuh” training population with pedigree and genomic

relationship), Data_3 (augmented the data set with genotyped sires and dam sires) and Data_4

(augmented the data set with sons of the sires)

An additional inclusion of genotyped sires, dam-sires and sons has led to higher reliabilities only for

sires with a maximum of ten daughters. Reliabilities increased up to 5.3% for FI and up to 3.8% for

EB. For sires with more than ten daughters, genotyped ancestors in the data set had no benefit.

Breeding values for FI and EB of all genotyped cows and the genotyped sires were contrasted (Figure

9 and Figure 10). The results provide an indication for the strength of the genetic relationship between

FI and EB. The fitted linear regression resulted in a correlation of 0.63 (R² = 0.40) for cows and 0.66

(R² = 0.43) for sires.

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

228 31 12 3 0 1

1-5 6-10 11-15 16-20 21-25 26-30

Reli

ab

ilit

y o

f b

reed

ing v

alu

es

Number of bulls / Number of daughters

Data_1

Data_2

Data_3

Data_4

Chapter II

50

Figure 9. Scatter plot between mean breeding values of feed intake and energy balance (Data_2, 1,163

genotyped cows)

Figure 10. Scatter plot between mean breeding values of feed intake and energy balance (Data_2, 275

genotyped sires)

R² = 0.4004

-30.00

-20.00

-10.00

0.00

10.00

20.00

30.00

-6.00 -5.00 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00

En

erg

y b

ala

nce (

MJ

NE

L)

-S

ingle

-ste

p

Feed intake (kg DM) - Single-step

R² = 0.4291

-30.00

-20.00

-10.00

0.00

10.00

20.00

30.00

-5.00 -4.00 -3.00 -2.00 -1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00

En

erg

y b

ala

nce (

MJ

NE

L)

-Sin

gle

-ste

p

Feed intake (kg DM) - Single-step

Chapter II

51

Discussion

Genomic selection enables an efficient and indispensable process (Haas et al., 2015) and thus it is

one solution to integrate hard to measure traits like FI and EB in the dairy cow breeding program to

improve their health and metabolic stability. Nevertheless, the success of genomic breeding

evaluation and selection is strongly dependent on the recording of the trait (Daetwyler et al., 2010).

Previous studies stated that only very few FI and EB data were available, which causes a lack of

reliable parameters due to the cost-intensive data collection (Vallimont et al., 2011; Veerkamp et al.,

2012).

The objective of this study was to estimate genomic breeding values for the traits FI and EB. In order

to evaluate the importance of genotyping, cows with pedigree relationship (Data_1) and cows with

pedigree and genomic relationship (Data_2) were compared and the method single-step was used to

apply this method on the “optiKuh” data set.

The project “optiKuh” provides a well-defined data set for FI and EB due to standardized

measurements from eight research farms, which represent the whole HF-population across Germany.

Within “optiKuh”, a homogenous data recording across herds characterized by widely similar feeding

and management conditions was established. Thereby, an adequate number of animals within a

relative brief time period of two years was achieved. Furthermore, entire lactations of several parities

were considered. This is in contrast to other studies, where FI was mostly recorded only until DIM

150 or 180 and not throughout the lactation (Buttchereit et al., 2011; Berry and Crowley, 2013;

Manzanilla Pech et al., 2014). Such an appropriate homogenous data set with a high number of

animals with a close relationship in the actual reference population is a precondition for getting

realistic accuracies of breeding values (Pszczola et al., 2012).

Previous results from the international FI project gDMI (Berry et al., 2007; Veerkamp et al., 2013;

Berry et al., 2014) were less satisfying. The connection between populations were weak underlined

by only few common sires or grandsires. Different parities and farming practices and as well as the

non-standardized rations might be reasons for relatively low reliabilities (0.14 to 0.29) and the benefit

of the joining data was smaller than expected. (Veerkamp et al., 2013; Haas et al., 2015).

For the use of genomic information, it is generally necessary to accomplish certain quality criteria. In

the current study, a threshold for the GC-score at 0.6 was chosen as a compromise between quantity

and quality of SNPs (Cunningham et al., 2008; Pszczola et al., 2018). This lies above 0.2, a threshold

Illumina Inc. (2005) and Yokoyama et al. (2010) recommend, because mean GC-scores below 0.2

usually report failed genotypes.

Chapter II

52

As expected, FI and EB reliabilities of pedigree-based and single-step data increase with higher

number of weekly records. Genomic reliabilities for FI, ranging between 0.33 and 0.61, are high

compared to the average reliabilities across validation sets in other studies, ranging between 0.04 and

0.20 (Pszczola et al., 2013; Pryce et al., 2014; Haas et al., 2015).

For cows with few phenotypic information, single-step estimation generated higher reliabilities for

both traits, whereas somewhat unexpected the highest reliabilities with max. 0.62 could be observed

in cows with only pedigree-based relationship Data_1, if the number of weekly records exceeded 40.

On contrary Pszczola et al. (2013) observed substantially higher genomic reliabilities (0.11)

compared to pedigree-based reliabilities (0.07) for FI. Also Manzanilla-Pech et al. (2017) found

similar results with max. 0.14 for genomic and max. 0.09 for pedigree-based reliabilities.

Possible explanations for the deviating results in this study could be the higher heritabilities found at

the end of lactation, so cows with an entire lactation profile have more observations and benefit from

these higher heritabilities in the pedigree-based alternative (see Figure 3).

In contrast, genomic estimated breeding values provide more accurate reliabilities for animals without

performance data (Muir, 2007; Zambrano et al., 2015). This supports the benefits of using genomic

calculations for selection in early age, especially for hard to measure traits. The generated reliabilities

for the 35 (FI) and 39 (EB) cows without phenotypic data lies at 0.23 for the trait FI and 0.17 for EB.

These values are above 8.6% higher for FI and 4.2% higher for EB, compared to pedigree-based

relationship Data_1.

In order to improve the connectedness among animals in the reference population to increase their

reliabilities of the breeding values, genotypic information of sires and dam-sires were included.

Improvements were only observed for reliabilities of FI and EB breeding values in males if sires had

low numbers of daughters, but not for female animals.

Genetic parameters for FI have already been available for a long time (van Arendonk et al., 1991;

Veerkamp, 1998) and the current results were generally in line with those of previous analyses.

In this study, heritabilities of both data sets slightly decrease at the beginning of lactation but increase

along DIM until the end of lactation. For the single-step, somewhat higher heritability values were

found at the beginning of lactation but this changed around DIM 50 when conventionally estimated

heritability values got higher. The very high heritability values at the end of lactation might be

overestimated, especially in case of pedigree-based alternative. This might be due to fewer

observations in total at the end of lactation and emphasizes the usefulness of genomic data.

In general, genomic data better reflects the realized relationship and helps to untangle the genetic and

the permanent environment.

Chapter II

53

In the current study, heritabilities for the whole lactation up to 350 DIM were estimated. The values

are in middle range, which is comparable to studies of Coffey et al. (2001), Liinamo et al. (2012) and

Li et al. (2018). Heritability values in other investigations, where just the first period of lactation

could be considered, were also in middle range (Hüttmann et al., 2009; Spurlock et al., 2012;

Krattenmacher et al., 2019).

Heritabilities as estimated in this study, range between 0.12 and 0.50 comparable to results of Li et

al. (2018) who reported values from 0.3 to 0.55. EB heritabilities were somewhat lower and ranged

between 0.15 and 0.48. This is in agreement with Berry et al. (2007). Nevertheless, heritabilities for

EB for the first 180 DIM reported by Krattenmacher et al. (2019) displayed a lower level and ranged

between 0.29 and 0.49 but also increase with later DIM. In contrast, Liinamo et al. (2012) found

highest values for heritabilities in early lactation (around 0.37) which declined later on. Nevertheless,

the moderate heritabilities for FI and EB are promising to implement genomic selection tools for both

traits.

In the present study, genetic correlations estimated for the traits between different lactation stages are

in line with other studies (Berry et al., 2007; Hüttmann et al., 2009). Our results strongly support, that

FI and EB in early estimates (20 DIM) are weakly correlated with estimates in late lactation (330

DIM) with values for FI at 0.23 (Data_2) and 0.08 (Data_1) and for EB at 0.08 and 0.18, respectively.

Only few studies had the possibility to use the entire lactation for analyses (Liinamo et al., 2012; Li

et al., 2018), but also report the same properties of genetic correlation across DIM.

For EB, Krattenmacher et al. (2019) found genetic correlations between early and mid lactation at

0.37. This is in contrast to the present results where values between DIM 10 and DIM 180 tend to be

quite lower with 0.03, which is in close agreement with results of Liinamo et al. (2012).

The pattern of genetic correlations across lactation indicates that FI, as well as EB, have a different

genetic background at different DIM. This is underlined by the results of the GWAS of Tetens et al.

(2014), which suggest that different genome regions affect the FI in course of lactation.

The correlation between FI and EB breeding values of 0.63 confirms, that FI and EB are different

traits but are correlated (Buttchereit et al., 2011; Tetens et al., 2014), due to the fact, that the

calculation of EB includes FI measurements.

To define a breeding goal to reduce energy deficit at the beginning of lactation, it has to be considered

which trait, FI or EB, should be used. Thereby it has to be noticed, that calculation of EB needs more

data than for FI like milk yield, milk ingredients, gestation stage, body weight. The genetic

correlations promise a possibility to increase FI or EB at the beginning of lactation together with a

minor increase of these traits at the end of lactation. So, in this lactation stage extreme high body

Chapter II

54

condition scores are unlikely. Especially at the beginning of lactation a high FI or a low energy deficit

is required to reduce risk of health problems that predominantly occur in this stage of lactation. A FI-

dependent management strategy could help to avoid over-conditioned cows by means of less or rather

demand-driven supply in the second half of lactation (Lin et al., 2013). A possible breeding strategy

is to choose sires with high breeding values for FI at the beginning and low breeding values at the

end of lactation.

Selection on higher FI or EB seems to be contrary to an efficient cow. Feed efficiency can be

characterized by gross efficiency (output/input relations) or by residual feed intake (RFI). Thereby,

the last named trait has been suggested as an important new trait since feed accounts for the largest

proportion of operating costs in dairy production (Connor, 2015; Manzanilla-Pech et al., 2016). RFI

is defined as the difference between energy intake and demand and is usually estimated as the

residuals from a model regressing FI on the various energy sinks and shows independence from the

independent variables (Berry and Crowley, 2013; Berry and Pryce, 2014). Then, efficient cows eat

less than predicted (= negative RFI). Due to nearly the same traits for calculation EB and RFI are

mathematically very similar (Savietto et al., 2014). So, selection for higher EB to realize less health

and reproductive disorders , especially in the first part of lactation, conflicts with feed efficiency

(positive RFI = inefficient). Therefore, feed efficiency traits should consider absolutely the health and

reproductive status of the cows as Veerkamp et al. (2013) recommended for life-time feed efficiency.

Before DMI or EB can be included into a breeding goal reliable genetic correlations between these

traits and health and reproductive traits have to be estimated to confirm the positive effects of reducing

the energy deficit. Also the genetic relationship between EB and RFI should be analyzed. For this

purpose a huge data set comprising DMI, EB, RFI and accurate recorded diagnoses and treatments is

necessary.

Independently of selection strategies envisaged levels of heritabilities, genetic correlations and the

reliabilities of breeding values for the most relevant traits provide sufficient preconditions for genetic

reduction of energy deficits at the beginning of lactation.

Conclusions

In the present study, the pedigree-based data set leads to higher reliabilities for animals with many

phenotypic records compared to the single-step alternative. In contrast, by combining effectively

phenotyped and genotyped animals using single-step genomic analytic procedures, relatively high

reliabilities for not-phenotyped but genotyped animals were achieved. In general, acceptable

Chapter II

55

estimated breeding values could be generated and promise a selection for high feed intake or energy

balance to define a new breeding goal even for only genotyped animals.

Nevertheless, for more accurate breeding values the reference population has to be extended by both

more phenotypes and more genotypes. Further steps, such as the use of ancillary traits, new options

for measuring phenotypes or establishing of economic weights, should be generated to finally

implement these traits in the breeding goal of high-performing dairy cows.

Acknowledgements

The project was supported by funds of the Federal Ministry of Food and Agriculture (BMEL) based

on a decision of the Parliament of the Federal Republic of Germany via the Federal Office for

Agriculture and Food (BLE) under the innovation support program.

References

Aguilar, I., I. Misztal, D. L. Johnson, A. Legarra, S. Tsuruta, and T. J. Lawlor. 2010. Hot topic: A

unified approach to utilize phenotypic, full pedigree, and genomic information for genetic

evaluation of Holstein final score. J. Dairy Sci. 93(2):743–752. https://doi.org/10.3168/jds.2009-

2730.

Ali, T., and L. R. Schaeffer. 1987. Accounting for covariance among test day milk yields in dairy

cows. Can. J. Anim. Sci. 67(3):637–644. https://doi.org/10.4141/cjas87-067.

Berry, D. P., Pryce, J. E. 2014. Feed Efficiency in Growing and Mature Animals. Proceedings, 10th

World Congress of Genetics Applied to Livestock Production.





Berry, D. P., and J. J. Crowley. 2013. Cell Biology Symposium: genetics of feed efficiency in dairy

and beef cattle. Journal of animal science 91(4):1594–1613. https://doi.org/10.2527/jas.2012-

5862.




Chapter II

56

Boichard, D., and M. Brochard. 2012. New phenotypes for new breeding goals in dairy cattle. Animal

an international journal of animal bioscience 6(4):544–550.

https://doi.org/10.1017/S1751731112000018.

Brotherstone, S., I. M. S. White, and K. Meyer. 2000. Genetic modelling of daily milk yield using

orthogonal polynomials and parametric curves. Anim. Sci. 70(03):407–415.

https://doi.org/10.1017/S1357729800051754.

Buttchereit, N., E. Stamer, W. Junge, and G. Thaller. 2011. Short communication: genetic

relationships among daily energy balance, feed intake, body condition score, and fat to protein

ratio of milk in dairy cows. J. Dairy Sci. 94(3):1586–1591. https://doi.org/10.3168/jds.2010-3396.

Christensen, O. F., and M. S. Lund. 2010. Genomic prediction when some animals are not genotyped.

Genetics, selection, evolution GSE 42:2. https://doi.org/10.1186/1297-9686-42-2.

Coffey, M. P., G. C. Emmans, and S. Brotherstone. 2001. Genetic evaluation of dairy bulls for energy

balance traits using random regression. Animal Science 73(01):29–40.

https://doi.org/10.1017/S1357729800058021.






Cunningham, J. M., T. A. Sellers, J. M. Schildkraut, Z. S. Fredericksen, R. A. Vierkant, L. E.

Kelemen, M. Gadre, C. M. Phelan, Y. Huang, J. G. Meyer, V. S. Pankratz, and E. L. Goode. 2008.

Performance of amplified DNA in an Illumina GoldenGate BeadArray assay. Cancer

epidemiology, biomarkers & prevention a publication of the American Association for Cancer

Research, cosponsored by the American Society of Preventive Oncology 17(7):1781–1789.

https://doi.org/10.1158/1055-9965.EPI-07-2849.

Daetwyler, H. D., R. Pong-Wong, B. Villanueva, and J. A. Woolliams. 2010. The impact of genetic

architecture on genome-wide evaluation methods. Genetics 185(3):1021–1031.

https://doi.org/10.1534/genetics.110.116855.

Chapter II

57

Edriss, V., Guldbrandtsen B., M. S. Lund, and G. Su. 2013. Criteria of GenCall score to edit marker

data and methods to handle missing markers have an influence on accuracy of genomic

predictions. Archiv Tierzucht 56:778–788.

Goddard, M. E., and B. J. Hayes. 2009. Mapping genes for complex traits in domestic animals and

their use in breeding programmes. Nature reviews. Genetics 10(6):381–391.

https://doi.org/10.1038/nrg2575.

Guarini, A. R., D. A. L. Lourenco, L. F. Brito, M. Sargolzaei, C. F. Baes, F. Miglior, I. Misztal, and

F. S. Schenkel. 2018. Comparison of genomic predictions for lowly heritable traits using multi-

step and single-step genomic best linear unbiased predictor in Holstein cattle. J. Dairy Sci.

101(9):8076–8086. https://doi.org/10.3168/jds.2017-14193.





https://doi.org/10.3168/jds.2014-9257.

Harder, I., E. Stamer, W. Junge, and G. Thaller. 2019. Lactation curves and model evaluation for feed

intake and energy balance in dairy cows. J. Dairy Sci. 102:7204-7216.

https://doi.org/10.3168/jds.2018-15300

Heringstad, B., G. Klemetsdal, and J. Ruane. 2000. Selection for mastitis resistance in dairy cattle: a

review with focus on the situation in the Nordic countries. Livestock Production Science(64: 95-

106).

Hüttmann, H., E. Stamer, W. Junge, G. Thaller, and E. Kalm. 2009. Analysis of feed intake and


models. Animal 3(2):181–188. https://doi.org/10.1017/S175173110800325X.

Illumina Inc. 2005. Illumina GenCall Data Analysis Software.

Krattenmacher, N., G. Thaller, and J. Tetens. 2019. Analysis of the genetic architecture of energy

balance and its major determinants dry matter intake and energy-corrected milk yield in

primiparous Holstein cows. J. Dairy Sci. https://doi.org/10.3168/jds.2018-15480.

Leesen, R. von, J. Tetens, E. Stamer, W. Junge, G. Thaller, and N. Krattenmacher. 2014. Effect of


Chapter II

58

primiparous Holstein-Friesian cows. J. Dairy Sci. 97(2):1128–1138.

https://doi.org/10.3168/jds.2013-7185.

Legarra, A., I. Aguilar, and I. Misztal. 2009. A relationship matrix including full pedigree and

genomic information. J. Dairy Sci. 92:465-4663. https://doi.org/10.3168/jds.2009-2061.

Legarra, A., O. F. Christensen, I. Aguilar, and I. Misztal. 2014. Single Step, a general approach for

genomic selection. Livestock Science 166:54–65. https://doi.org/10.1016/j.livsci.2014.04.029.




https://doi.org/10.3168/jds.2018-14611.

Liinamo, A.-E., P. Mäntysaari, and E. A. Mäntysaari. 2012. Short communication: Genetic

parameters for feed intake, production, and extent of negative energy balance in Nordic Red dairy

cattle. J. Dairy Sci. 95(11):6788–6794. https://doi.org/10.3168/jds.2012-5342.

Lin, Z., I. Macleod, and J. E. Pryce. 2013. Short communication: Estimation of genetic parameters

for residual feed intake and feeding behavior traits in dairy heifers. J. Dairy Sci. 96(4):2654–2656.

https://doi.org/10.3168/jds.2012-6134.

Madsen, P., J. Jensen, Labouriau R., L. G. Christensen, and G. Sahana. 2013. DMU - A Package for

Analyzing Multivariate Mixed Models in quantitative Genetics and Genomics.

Mäntysaari, P., A.-E. Liinamo, and E. A. Mäntysaari. 2012. Energy efficiency and its relationship

with milk, body, and intake traits and energy status among primiparous Nordic Red dairy cattle.


Manzanilla Pech, C. I. V., R. F. Veerkamp, M. P. L. Calus, R. Zom, A. van Knegsel, J. E. Pryce, and

Y. de Haas. 2014. Genetic parameters across lactation for feed intake, fat- and protein-corrected

milk, and liveweight in first-parity Holstein cattle. J. Dairy Sci. 97(9):5851–5862.

https://doi.org/10.3168/jds.2014-8165.

Manzanilla-Pech, C. I. V., R. F. Veerkamp, Y. de Haas, M. P. L. Calus, and J. ten Napel. 2017.

Accuracies of breeding values for dry matter intake using nongenotyped animals and predictor

traits in different lactations. J. Dairy Sci. 100(11):9103–9114. https://doi.org/10.3168/jds.2017-

12741.

Chapter II

59

Manzanilla-Pech, C. I. V., R. F. Veerkamp, R. J. Tempelman, M. L. van Pelt, K. A. Weigel, M.

VandeHaar, T. J. Lawlor, D. M. Spurlock, L. E. Armentano, C. R. Staples, M. Hanigan, and Y.

de Haas. 2016. Genetic parameters between feed-intake-related traits and conformation in 2

separate dairy populations--the Netherlands and United States. J. Dairy Sci. 99(1):443–457.

https://doi.org/10.3168/jds.2015-9727.

Meuwissen, T.H.E., B. J. Hayes, and M. E. Goddard. 2001. Prediction of Total Genetic Value Using

Genome-Wide Dense Marker Maps. Genetics.

Misztal, I., A. Legarra, and I. Aguilar. 2009. Computing procedures for genetic evaluation including

phenotypic, full pedigree, and genomic information. J. Dairy Sci. 92(9):4648–4655.

https://doi.org/10.3168/jds.2009-2064.

Muir, W. M. 2007. Comparison of genomic and traditional BLUP-estimated breeding value accuracy

and selection response under alternative trait and genomic parameters. Journal of animal breeding

and genetics = Zeitschrift fur Tierzuchtung und Zuchtungsbiologie 124(6):342–355.

https://doi.org/10.1111/j.1439-0388.2007.00700.x.

Oltenacu, P. A., and D. M. Broom. 2010. The impact of genetic selection for increased milk yield on

the welfare of dairy cows. Anim. Welfare(19 (S)):39–49.

Přibyl, J., J. Bauer, V. Čermák, P. Pešek, J. Přibylová, J. Šplíchal, H. Vostrá-Vydrová, L. Vostrý, and

L. Zavadilová. 2015. Domestic estimated breeding values and genomic enhanced breeding values

of bulls in comparison with their foreign genomic enhanced breeding values. Animal 9(10):1635–

1642. https://doi.org/10.1017/S1751731115001044.

Pryce, J., and B. Hayes. 2012. A review of how dairy farmers can use and profit from genomic

technologies. Anim. Prod. Sci. 52(3):180. https://doi.org/10.1071/AN11172.

Pryce, J. E., W. J. Wales, Y. de Haas, R. F. Veerkamp, and B. J. Hayes. 2014. Genomic selection for

feed efficiency in dairy cattle. Animal 8(1):1–10. https://doi.org/10.1017/S1751731113001687.

Pszczola, M., K. F. Stock, E. Mucha, and E. Sell-Kubiak. 2018. Genetic architecture of methane

emissions from dairy cows. WCGALP.

Pszczola, M., T. Strabel, J. A. M. van Arendonk, and M. P. L. Calus. 2012. The impact of genotyping

different groups of animals on accuracy when moving from traditional to genomic selection. J.


Chapter II

60

Pszczola, M., R. F. Veerkamp, Y. de Haas, E. Wall, T. Strabel, and M. P. L. Calus. 2013. Effect of

predictor traits on accuracy of genomic breeding values for feed intake based on a limited cow

reference population. Animal 7(11):1759–1768. https://doi.org/10.1017/S175173111300150X.

Randhawa, S.N.S., R. Ranjan, R. Singh, and N. Chand. 2014. Diagnosis and Management of Negative

Energy Balance and Associated Production Diseases in Bovines. Intas Polivet(15 (2)):497–503.

Rauw, W.M., E. Kanis, E.N. Noordhuizen-Stassen, and F.J. Grommers. 1998. Undesirable side

effects of selection for high production efficiency in farm animals: A review. Livestock

Production Science 56(1):15–33. https://doi.org/10.1016/S0301-6226(98)00147-X.

Reist, M., D. Erdin, D. von Euw, K. Tschuemperlin, H. Leuenberger, Y. Chilliard, H. M. Hammon,

C. Morel, C. Philipona, Y. Zbinden, N. Kuenzi, and J. W. Blum. 2002. Estimation of energy

balance at the individual and herd level using blood and milk traits in high-yielding dairy cows.

J. Dairy Sci. 85(12):3314–3327. https://doi.org/10.3168/jds.S0022-0302(02)74420-2.

Savietto, D., Berry, D. P,. and Friggens, N. C. 2014. Towards an improved estimation of the biological

components of residual feed intake in growing cattle. J. Anim. Sci. 92: 467-476.

SAS. 2012. SAS 9.2 TS Level 1 Mo., Cary, NC27513, USA. SAS Institute Inc.



5402. https://doi.org/10.3168/jds.2012-5407.



https://doi.org/10.3168/jds.2013-7301.

Vallimont, J. E., C. D. Dechow, J. M. Daubert, M. W. Dekleva, J. W. Blum, C. M. Barlieb, W. Liu,

G. A. Varga, A. J. Heinrichs, and C. R. Baumrucker. 2011. Heritability of gross feed efficiency

and associations with yield, intake, residual intake, body weight, and body condition score in 11

commercial Pennsylvania tie stalls. J. Dairy Sci. 94(4):2108–2113.

https://doi.org/10.3168/jds.2010-3888.

van Arendonk, J.A.M., G. J. Nieuwhof, H. Vos, and S. Korver. 1991. Genetic aspects of feed intake

and efficiency in lactating dairy heifers. Livestock Production Science 29(4):263–275.

https://doi.org/10.1016/0301-6226(91)90103-W.

Chapter II

61

VanRaden, P. M. 2008. Efficient methods to compute genomic predictions. J. Dairy Sci.

91(11):4414–4423. https://doi.org/10.3168/jds.2007-0980.

Veerkamp, R. F. 1998. Selection for economic efficiency of dairy cattle using information on live

weight and feed intake: A review. J. Dairy Sci. 81(4):1109–1119.


Veerkamp, R. F., M. P. Coffey, D. P. Berry, Y. de Haas, E. Strandberg, H. Bovenhuis, M. P. L. Calus,

and E. Wall. 2012. Genome-wide associations for feed utilisation complex in primiparous

Holstein-Friesian dairy cows from experimental research herds in four European countries.

Animal an international journal of animal bioscience 6(11):1738–1749.

https://doi.org/10.1017/S1751731112001152.

Veerkamp, R. F., and E. P. C. Koenen. 1999. Genetics of food intake, live weight, condition score

and energy balance. British Society of Animal Science Occasional Publication No. 24:63–73.

Veerkamp, R. F., J. E. Pryce, D. Spurlock, D. P. Berry, M. P. Coffey, P. Løvendahl, R. d. van Linde,

J. Bryant, F. Miglior, Z. Wang, M. Winters, N. Krattenmacher, N. Charfeddine, J. Pedersen, and

Y. de Haas. 2013. Selection on Feed Intake of Feed Efficiency: A Position Paper from gDMI

Breeding Goal Discussion. Interbull Bulletin(No. 47):15–22.

Veerkamp, R. F., and R. Thompson. 1999. A covariance function for feed intake, live weight, and

milk yield estimated using a random regression model. J. Dairy Sci. 82(7):1565–1573.


Vries, M. J. de, and R. F. Veerkamp. 2000. Energy Balance of Dairy Cattle in Relation to Milk

Production Variables and Fertility. J. Dairy Sci. 83(1):62–69. https://doi.org/10.3168/jds.S0022-

0302(00)74856-9.

Wallén, S. E., M. Lillehammer, and T. H. E. Meuwissen. 2017. Strategies for implementing genomic

selection for feed efficiency in dairy cattle breeding schemes. J. Dairy Sci. 100(8):6327–6336.

https://doi.org/10.3168/jds.2016-11458.

Yokoyama, J. S., C. A. Erdman, and S. P. Hamilton. 2010. Array-based whole-genome survey of dog

saliva DNA yields high quality SNP data. PloS one 5(5):e10809.

https://doi.org/10.1371/journal.pone.0010809.

Chapter II

62

Zambrano, A. J., F. J. Rincón, H. Albeiro López, and Z. J. Julián Echeverri. 2015. Estimation and

comparison of conventional and genomic breeding values in Holstein cattle of Antioquia,

Colombia. Revista MVZ Córdoba 20(3):4739–4753.

Chapter III

63

Chapter III

Zucht auf Futteraufnahme mit Hilfe der genomischen Selektion

Imke Harder1, E. Stamer², W. Junge1, G. Thaller1

1Institut für Tierzucht und Tierhaltung, Christian-Albrechts-Universität zu Kiel, D-24098 Kiel;

2TiDa Tier und Daten GmbH, D-24259 Westensee/Brux;

Published in Züchtungskunde, 90, (6) S. 476-490, 2018, ISSN 0044-5401

Chapter III

64

Zusammenfassung

Das Merkmal Futteraufnahme hat erst kürzlich an Bedeutung gewonnen, da es wesentlich für die

leistungsgerechte Versorgung und die Gesunderhaltung der Hochleistungsmilchkühe ist. Auch wenn

ein Energiedefizit zu Laktationsbeginn als normale Reproduktionsstrategie gilt, so führt die hohe

genetische Veranlagung für Milchleistung zu einem deutlich stärker ausgeprägten Defizit. Durch eine

Erhöhung der aufgenommenen Futtermenge zu Laktationsbeginn kann das ausgeprägte

Energiedefizit verringert werden und verbessert dadurch die Stoffwechselstabilität der Milchkühe.

Mit dem Projekt optiKuh wurde ein einzigartiger Datensatz zur Messung der Futteraufnahme und

darauf aufbauend zur genomischen Selektion für die Futteraufnahme zu Laktationsbeginn geschaffen.

Hierfür standen zunächst Phänotypen von insgesamt 1.374 Holstein-Friesian Kühen mit 40.012

wöchentlichen Futteraufnahmedaten und einem Mittelwert von 21,8 ± 4,3 kg/Tag sowie von 327

Fleckvieh Kühen mit 16.996 wöchentlichen Futteraufnahmedaten und einem Mittelwert von 20,2 ±

3,6 kg/Tag zur Verfügung.

Für die Laktationskurven wurde ein Random Regression Modell verwendet, wobei für die fixe

Laktationskurve die Ali und Schaeffer Funktion und für den zufälligen permanenten Umwelteffekt

und den zufälligen additiv genetischen Effekt jeweils das Legendre Polynom 3. Grades angepasst

wurde.

Die genomische Zuchtwertschätzung konnte aktuell nur für die Rasse Holstein-Friesian durchgeführt

werden. Für die anschließende Schätzung der Parameter und die Varianzkomponenten- und

Zuchtwertschätzung wurde das Programm DMU genutzt.

Hierfür wurde der Datensatz von insgesamt 1.128 genotypisierten und 35 geno- aber nicht

phänotypisierten Kühen in zwei Teile untergliedert: zum einen in einen pedigreebasierten Ansatz und

zum anderen in einen kombinierten Ansatz aus Pedigree- und Genotypinformation („single-step“).

Die genomisch geschätzten Heritabilitäten bewegen sich im Laktationsverlauf in einem Bereich von

0,21 bis 0,47 und weisen zu Laktationsanfang ein höheres Niveau im Vergleich zu den

konventionellen Werten auf.

Mit der anschließenden Zuchtwertschätzung im Merkmal Futteraufnahme konnten durch die

Einbeziehung von genomischer Verwandtschaft hohe Sicherheiten erzielt werden. Für die 35 nicht

phänotypisierten Kühe konnte die Sicherheit um nahezu 10 % gesteigert werden gegenüber der

Variante ohne genomische Verwandtschaft, d.h. rein pedigreebasiert.

Schlüsselwörter: optiKuh, Futteraufnahme, Energiebilanz, genomische Zuchtwertschätzung

Chapter III

65

Summary

At the beginning of lactation, high performing dairy cows often experience a severe energy deficit,

which in turn is strongly associated with metabolic diseases. Increasing feed intake in this period

could improve the metabolic stability and thus the health of the animals. Genomic selection enables

for the first time the inclusion of this hard-to-measure trait in breeding programs.

For this purpose, in the project optiKuh 1.374 Holstein-Friesian (HF) dairy cows and 327 Simmental

cattle (SI) were phenotyped, where feed intake data recording was standardized across farms. After

data editing phenotypic data comprised a total of 40,0012 (HF) and 16,996 (SI) average weekly dry

matter intake records with a mean of 21.8 ± 4.3 kg/d for HF and 20.2 and ± 3.6 kg/d for SI.

For the subsequent breeding value estimation, data of SI could not be used due to a small data set

resulting in not converting estimation runs. 1,128 of HF phenotyped cows were genotyped and 35

animals were genotyped but not phenotyped. Variance components and breeding values were

estimated using both, pedigree relationships and single-step genomic evaluation, each carried out with

the DMU software package. With the underlying random regression model the fix Lactation stage

was modeled by the function of Ali and Schaeffer, and for both, the random permanent environmental

effect and the random additive genetic effect, third-order Legendre polynomials were chosen.

Heritability estimates ranged between 0.21 and 0.47 and increased towards the end of lactation. For

the genotyped cows with no phenotypic records, the inclusion of genomic relationship improves the

average reliability of the breeding value for feed intake by nearly 10 %.

Keywords: optiKuh, feed intake, energy balance, genomic breeding value estimation

Einleitung

Mit steigendem Interesse am Tierwohl in Bezug auf Gesundheit und Langlebigkeit der Tiere tritt das

Merkmal Futteraufnahme immer stärker in den Vordergrund und soll in Zukunft in das Zuchtziel der

Hochleistungsmilchkuh integriert werden.

In der Vergangenheit wurde über die Zucht auf höhere Milchleistung indirekt auch auf eine höhere

Futteraufnahme (FA) selektiert. Die Futteraufnahmekapazität stieg jedoch nicht im gleichen Umfang

wie die Milchleistung. Folglich führte dies zu einem immer stärker ausgeprägten Energiedefizit zu

Laktationsbeginn und ging häufig mit Stoffwechselkrankheiten wie Azidose, Ketose oder

Milchfieber einher (Coffey et al., 2001);(Leesen et al., 2014). Durch eine Steigerung der FA zu

Laktationsbeginn auf Basis züchterischer Selektionsentscheidungen könnte das starke Energiedefizit

ausgeglichen werden (Coffey et al., 2002). Aufgrund mangelnder Prüfkapazitäten war es bisher nicht

möglich, die FA mit Hilfe der konventionellen Zuchtwertschätzung direkt zu verbessern (Haas et al.,

Chapter III

66

2012). Mit Einführung der genomischen Selektion könnte nun aber erstmals das bisher ausschließlich

in Testherden erfassbare und dementsprechend im geringen Datenumfang vorliegende Merkmal

Futteraufnahme züchterisch bearbeitet werden (BOICHARD und BROCHARD, 2012; CALUS et al.,

2013).

Die vorliegende Untersuchung soll zeigen, inwieweit die Daten aus dem nationalen Verbundprojekt

„optiKuh“ in Verbindung mit der genomischen Selektion genutzt werden können, über die begrenzte

Anzahl der geprüften Tiere hinaus aussagekräftige Zuchtwerte für weitere Tiere ohne

Leistungsinformationen (= Kandidaten) zu schätzen. Hierfür wurden Varianzkomponenten und

Zuchtwerte unter Einbeziehung genomischer Markerinformationen für das Merkmal FA geschätzt.

Mit dem single-step Verfahren erfolgte eine gleichzeitige Nutzung von pedigreebasierter als auch

genomischer Verwandtschaft.

Material und Methoden

Für die Berechnung der Varianzkomponenten und der anschließenden Schätzung der genomischen

und konventionellen Zuchtwerte für die beiden Merkmale FA und Energiebilanz (EB) standen

insgesamt 1.374 Holstein-Friesian Kühe (HF) und 327 Fleckvieh Kühe (FV) aus dem Projekt

„optiKuh“ zur Verfügung. Die Phänotypen für FA wurden mit Hilfe von speziellen Wiegetrögen mit

automatischer Tiererkennung in einem Zeitraum zwischen Dezember 2014 bis März 2017 auf 12

Versuchsbetrieben aus ganz Deutschland kontinuierlich und individuell erhobenen. Die zentral

gespeicherten täglichen Futteraufnahmedaten wurden ihre Plausibilität hin geprüft (Tab. 1).

Der Zeitraum innerhalb einer Laktation wurde auf den 8. bis 350. Laktationstag (LTG) beschränkt;

außerhalb dieses Bereichs lagen nur wenige Beobachtungen vor. Nach eingehender

Plausibilitätskontrolle wurden die Rohwerte aufgrund unterschiedlicher Erfassungsfrequenzen der

Merkmale (z.B. FA täglich und Milchinhaltsstoffe wöchentlich) zu individuellen

Wochenmittelwerten aggregiert. Beobachtungen außerhalb des Bereichs ± 4 Standardabweichungen

wurden ausgeschlossen. Die EB wurde berechnet aus der Differenz zwischen Energieaufnahme (MJ

NEL/Tag) und Energiebedarf (MJ NEL/Tag). Die Energieaufnahme wird berechnet, in dem die

Futteraufnahme (TS/Tag) mit der Nettoenergielaktation (MJ/kg TS) multipliziert wird. Der

Energiebedarf wiederum errechnet sich aus dem Erhaltungsbedarf (0,293 * Lebendmasse0,75) und

dem Energiebedarf für Milchbildung (GfE, 2001). Resultierend daraus ergaben sich Mittelwerte für

FA, Milchmenge (Mkg) und EB mit 21,8 kg/Tag, 35,5 kg/Tag und 3,20 MJ NEL/Tag für HF und

20,2 kg/Tag, 27,4 kg/Tag und 1,06 MJ NEL/Tag für FV (Tab. 2). Ein Vergleich beider Rassen zeigt

Chapter III

67

jeweils höhere Werte für die Rasse HF in allen drei Merkmalen mit + 1,6 kg TS/Tag für FA, + 8,1

kg/Tag für Milchmenge und + 2,14 MJ NEL/Tag für EB.

Tab. 1. Rasse, Futtergruppen, Laktationsnummer und Laktationstag in Abhängigkeit vom Betrieb

Breed, feeding groups, parity, lactation day relative to research farm

Betrieb Kühe Rasse Fütterungs- Laktations- Laktationstag gruppen nummer Mittelwert Min.-Max.

Holstein-Friesian

(n=1.374)

Braunschweig 64 HF 4 2-5, 9 67 0-171

Dummerstorf 30 HF 2 2-3 191 0-622

Futterkamp 179 HF 2-2-2* 1-8 174 26-377

Hohenheim 51 HF 1 1-5, 7-8 59 0-464

Iden (2015) 109 HF 1 1-9 52 1-309

Iden (2016) 79 HF 2 1-7, 9-10 73 1-177

Karkendamm 341 HF 1 1-9 173 11-627

Neumühle 199 HF 1 1-11 95 1-405

Riswick (A) 83 HF 4 1-7 165 0-405

Riswick (B) 239 HF 2-3-4* 1-10 86 0-468

Fleckvieh

(n=327)

Aulendorf 59 FV 2 1-9 159 0-431

Grub 97 FV 4 1-8, 10 160 0-462

Achselschwang 105 FV (BV) 2-4-2* 1-6, 9 175 20-348

Triesdorf 66 FV 2 1-7 155 0-403

BV: ca. 35 % Braunviehversuchskühe

*aufeinander folgende Fütterungsversuche mit jeweils unterschiedlicher Anzahl an

Fütterungsgruppen

1.163 HF und 232 FV Kühe mit Phänotypen wurden mit dem Illumina BovineSNP50 Bead Chip

genotypisiert. Von 35 HF lag keine Phänotypisierung vor, sodass hier Vorhersagen ohne eigene

Leistung getroffen wurden. Nach einer gängigen Qualitätskontrolle (Callrate > 0,95, GC-Score > 0,6,

MAF) wurden ca. 43.500 SNP (HF) und 45.500 SNP (FV) verwendet. In Vorbereitung der

nachfolgenden Analysen erfolgten jeweils innerhalb Rasse die Plausibilisierung, Aufbereitung und

Gegenüberstellung der genomischen und pedigreebasierten Verwandtschaftsmatrizen. Ein Pedigree

mit vier Ahnengenerationen wurde für HF vom vit (Vereinigte Informationssysteme Tierhaltung e.V.)

und FV vom LKV Bayern (Landeskuratorium der Erzeugerringer für tierische Veredelung e.V.)

bereitgestellt. Eine Gegenüberstellung der pedigreebasierten bzw. erwarteten mit den genomischen

bzw. realisierten Verwandtschaftskoeffizienten ergab weitgehende Übereinstimmungen. Lediglich

im Bereich der Halbgeschwister und Vollgeschwister treten wenige deutliche Abweichungen auf.

Chapter III

68

Unter Annahme korrekter genomischer Koeffizienten können in diesem Fällen Abstammungsfehler

im Pedigree angenommen werden.

Auf Basis erstellter finaler Auswertungsdatensätze der einzelnen Betriebe fand die Entwicklung der

linearen Auswertungsmodelle statt. Das gewählte Random Regression Tiermodell beinhaltet die

fixen Effekte Herdentestwoche bzw. Herdentestwochengruppe und Laktationsnummer (1, 2, 3, ≥ 4).

Für die Modellierung der allgemeinen Laktationskurve wurden die vier parametrischen Funktionen

Ali und Schaeffer sowie die Legendre Polynome 2. bis 4. Grades hinsichtlich ihrer Anpassungsgüte

mit den Bewertungskriterien korrigiertes Akaike Informationskriterium (AICC; (Burnham and

Anderson, 1998) und Bayesian Informationskriterium (Schwarz, 1978) verglichen. Als geeignetste

fixe Laktationskurve erwies sich die Ali und Schaeffer Funktion(Ali and Schaeffer, 1987). Für den

zufälligen permanenten Umwelteffekt und den zufälligen additiv genetischen Tiereffekt wurde

jeweils das Legendre Polynom 3. Grades genutzt.

Varianzkomponenten und Zuchtwerte wurden sowohl auf Basis der pedigree-basierten (PED)

Verwandtschaft als auch auf der kombinierten, aus Pedigree und genomisch (PGK) bestehenden

Verwandtschaft mit Hilfe des „single-step“ Verfahrens geschätzt. Hierfür wurden die Programme

Gmatrix und DMU (Madsen et al., 2013) verwendet. Aufgrund der geringen Tierzahlen und damit

einhergehender fehlender Konvergenz der Schätzläufe konnten für die Rasse FV nur

Laktationsverläufe, jedoch keine Varianzkomponenten und damit keine Zuchtwerte geschätzt

werden.

Chapter III

69

Tab. 2. Deskriptive Statistik für die Wochenmittelwerte der Merkmale Futteraufnahme, Milchmenge

und Energiebilanz sowie für die energiebilanzrelevanten Merkmale für HF

Descriptive statistics for the weekly averages of the traits feed intake, milk yield and energy balance

as well as energy relevant traits for HF

Merkmal Holstein-Friesian

Kühe lnr n �̅� s

Futteraufnahme (kg TS/Tag) 1.341 1.928 40.012 21,8 4,25

Milchmenge (kg/Tag) 1.338 1.917 39.838 35,5 8,81

Energiebilanz (MJ NEL) 1.322 1.865 33.376 3,20 29,4

*Futteraufnahme (kg TS/Tag) 22,3 4,04

Gewicht (kg) 658 37,9

ECM (kg/Tag) 34,3 7,71

Milchmenge (kg/Tag) 35,8 8,76

Fett (%) 3,77 0,63

Eiweiß (%) 3,31 0,32

Merkmal Fleckvieh

Kühe lnr n �̅� s

Futteraufnahme (kg TS/Tag) 327 604 16.996 20,2 3,60

Milchmenge (kg/Tag) 326 603 16.933 27,4 8,16

Energiebilanz (MJ NEL) 326 583 14.527 1,06 18,9

*Futteraufnahme (kg TS/Tag) 20,4 3,41

Gewicht (kg) 750 75,3

ECM (kg/Tag) 28,3 7,58

Milchmenge (kg/Tag) 27,5 8,04

Fett (%) 4,24 0,61

Eiweiß (%) 3,59 0,33

lnr=Laktationsnummer; n=Anzahl Wochenmittelwerte; �̅� =Mittelwert; s=Standardabweichung

* Merkmale für die berechneten Energiebilanzen

Chapter III

70

Ergebnisse

Mit Hilfe von Random Regression Modellen und der Funktion Ali und Schaeffer wurden, unter

Berücksichtigung der Herdentestwoche und Laktationsnummer, Laktationskurven für die Merkmale

FA, Mkg und EB geschätzt.

Abb. 1. Laktationskurven für die drei Merkmale Futteraufnahme (FA), Milchmenge (Mkg) und

Energiebilanz (EB) in der ersten (links) und zweiten (rechts) Laktation für HF, modelliert mit der Ali

und Schaeffer Funktion vom 8. bis 350. Laktationstag

Lactation curves for the three traits feed intake (FA), milk yield (Mkg) and energy balance in first

(left) and second (right) parity for Holstein-Friesian (HF), modeled with Ali and Schaeffer function

for days in milk 8.to 350

Abb. 2. Laktationskurven für die drei Merkmale Futteraufnahme (FA), Milchmenge (Mkg) und

Energiebilanz (EB) in der ersten (links) und zweiten (rechts) Laktation für FV, modelliert mit der Ali

und Schaeffer Funktion vom 8. bis 350. Laktationstag

Lactation curves for the three traits feed intake (FA), milk yield (Mkg) and energy balance in first

(left) and second (right) parity for Simmental cattle (FV), modeled with Ali and Schaeffer function

for days in milk 8. to 350

Für beide Rassen ergeben sich für die Laktationskurven in etwa die gleichen Verläufe. Die zu Beginn

der Laktation schnell ansteigende Milchleistung in Verbindung mit einer etwas verzögerten Zunahme

der Futteraufnahme führt zu einer negativen Energiebilanz. Diese ist in dem schraffierten Bereich

dargestellt. Die Energiebilanz erreicht ihr Minimum etwa zwischen dem 5. und 10. LTG und schlägt

-50

-40

-30

-20

-10

0

10

20

30

40

50

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

En

ergie

bil

anz

(MJ

NE

L/T

ag)

Fu

tter

aufn

ahm

e (k

g T

S/T

ag)

/ M

ilch

men

ge

(kg/T

ag)

Laktationstag

FA

Mkg

EB

-50

-40

-30

-20

-10

0

10

20

30

40

50

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

En

ergie

bil

anz

(MJ

NE

L/T

ag)

Fu

tter

aufn

ahm

e (k

g T

S/T

ag)

/ M

ilch

men

ge

(kg/T

ag)

Laktationstag

FA

Mkg

EB

-50

-40

-30

-20

-10

0

10

20

30

40

50

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

En

ergie

bil

anz

(MJ

NE

L/T

ag)

Fu

tter

aufn

ahm

e (k

g T

S/T

ag)

/ M

ilch

men

ge

(kg/T

ag)

Laktationstag

FA

Mkg

EB

-50

-40

-30

-20

-10

0

10

20

30

40

50

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

En

ergie

bil

anz

(MJ

NE

L/T

ag)

Fu

tter

aufn

ahm

e (k

g T

M/k

g)

/ M

ilch

men

ge

(kg/k

g)

Laktationstag

FI

MY

EB

Chapter III

71

erst ab etwa dem 60. LTG in eine positive Bilanz um (Abb. 1 und 2). Ein Vergleich der Laktationen

innerhalb Rasse zeigt ein insgesamt angestiegenes Niveau der FA und der Mkg, sowie eine absolut

gesehen verbesserte EB.

Die Heritabilitäten für die FA liegen in einem moderaten Bereich zwischen 0,21 und 0,41. Im

Vergleich dazu liegen die Werte für EB etwas niedriger in einem Bereich von 0,18 und 0,38. Die

Heritabilitäten der Mkg sind wie erwartet, am höchsten und rangieren zwischen 0,24 und 0,73. Bei

einem Vergleich der Schätzwerte zwischen den beiden Verwandtschaftsalternativen sind die Werte

des PGK Datensatzes im Bereich des Laktationsanfangs höher als die der PED Werte. Dieses

Verhältnis kehrt sich ab dem 100. Laktationstag um (Abb. 3). Insgesamt steigen die geschätzten

Heritabilitäten ab dem 150. Laktationstag an, sodass die höchsten Werte zum Ende der Laktation

vorliegen.

Abb. 3. Vergleich der geschätzten Heritabilitäten der unterschiedlichen Verwandtschaftsalternativen

(pedigree-basiert und kombiniert) für die drei Merkmale Futteraufnahme (FA), Milchmenge (Mkg)

und Energiebilanz (EB) beim HF im Laktationsverlauf (8. bis 350. Laktationstag)

Comparison between estimated heritabilities in course of lactation (days in milk 8. to 350.) for the

three traits feed intake (FI), milk yield (Mkg) and energy balance (EB) according to relationship

computation for HF

Die genetischen Korrelationen zwischen aufeinander folgenden Laktationsabschnitten zeigen zu

Beginn der Laktation (40. bis 160. Laktationstag) eine hohe Korrelation mit Schätzwerten für die FA

zwischen 0,65 und 0,98, sowie 0,50 und 0,97 für die EB. Ein ähnlicher Zusammenhang besteht

zwischen den Werten am Ende der Laktation. Hingegen sind die frühen (LTG 10) mit den späten

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0 50 100 150 200 250 300 350

Her

itab

ilit

ät

Laktationstag

FA kombiniert FA pedigree-basiert EB kombiniert

EB pedigree-basiert Mkg kombiniert Mkg pedigree-basiert

Chapter III

72

Laktationsabschnitten (LTG 340) sehr niedrig korreliert und betragen 0,14 für FA und -0,05 für EB

(Tab. 3).

Tab. 3. Genetische Korrelationen zwischen ausgewählten Laktationstagen für die Merkmale

Futteraufnahme und Energiebilanz (Variante kombiniert (PGK)) beim HF

Genetic correlations between lactation day for the traits feed intake and energy balance (version

combined (PGK)) for HF

FA

EB

LTG 10 40 70 100 130 160 190 220 250 280 310 340

10 0,85 0,60 0,44 0,34 0,28 0,24 0,22 0,21 0,20 0,18 0,14

40 0,88 0,93 0,83 0,74 0,65 0,56 0,49 0,46 0,47 0,51 0,54

70 0,63 0,92 0,97 0,92 0,83 0,73 0,65 0,61 0,64 0,69 0,74

100 0,42 0,78 0,96 0,98 0,92 0,84 0,76 0,73 0,75 0,80 0,83

130 0,27 0,64 0,87 0,97 0,98 0,93 0,87 0,84 0,84 0,87 0,87

160 0,18 0,50 0,74 0,88 0,97 0,98 0,95 0,92 0,92 0,92 0,87

190 0,12 0,37 0,58 0,74 0,87 0,97 0,99 0,98 0,96 0,93 0,85

220 0,09 0,27 0,45 0,62 0,78 0,91 0,98 1,00 0,98 0,94 0,83

250 0,07 0,23 0,40 0,56 0,72 0,87 0,96 0,99 0,99 0,95 0,84

280 0,05 0,25 0,44 0,60 0,74 0,87 0,94 0,96 0,98 0,98 0,90

310 0,00 0,30 0,54 0,69 0,79 0,84 0,85 0,84 0,86 0,94 0,97

340 -0,05 0,34 0,60 0,72 0,74 0,70 0,62 0,56 0,58 0,71 0,91

Die realisierten Sicherheiten für die Kuhzuchtwerte im Merkmal Futteraufnahme steigen mit der

Anzahl der Eigenleistungen (= Wochenmittelwerte). Dabei ergeben sich höhere Sicherheiten für die

kombinierte Variante (PGK) im Bereich bis 40 Eigenleistungen; bei einer höheren Anzahl an

Eigenleistungen sind die Sicherheiten der pedigree-basierten Variante höher (Abb. 4). Insgesamt

rangieren die genomischen Genauigkeiten zwischen 0,26 und 0,63. Für die 35 Kandidatentiere, Tiere

mit einer Geno- aber keiner Phänotypisierung, konnten Sicherheiten von 0,26 % realisiert werden.

Allein auf PED basierter Zuchtwertschätzung ergibt sich eine ca. 10 % niedrigere Sicherheit der

Kandidatenzuchtwerte. Eine Integration genotypisierter Vatertiere (Kuhväter und Väter der

Kuhmütter) ergab keinen Anstieg der Sicherheiten der Kuhzuchtwerte.

Chapter III

73

Abb. 4. Realisierte Sicherheiten der Kuhzuchtwerte für das Merkmal Futteraufnahme der Rasse HF

in Abhängigkeit von der Anzahl der Eigenleistungen (Wochenmittelwerte)

Realized reliabilities of the female breeding values for the trait feed intake and the breed HF relative

to the number of weekly averages

Zusätzlich wurden für die Väter der Kühe Sicherheiten ermittelt, die in einem Bereich von 0,3 (1

Tochter) bis 0,8 (35 Töchter) lagen. Mit steigender Anzahl an Töchtern pro Bulle steigen ebenfalls

die Sicherheiten an, wobei nur wenige Bullen eine hohe Töchteranzahl aufwiesen. Die Sicherheiten

blieben in der PGK Zuchtwertschätzung unverändert, da hier für die Bullen noch keine

Genotypisierungsergebnisse vorlagen. Dabei wiesen 99,44 % der Väter weniger als 21 Töchter auf.

Die zusätzliche Integration genotypisierter Väter und Muttersväter (kombiniert K, V-MV) brachte

höhere Sicherheiten für Bullen mit wenig Töchtern im Vergleich zu dem PGK-Datensatz (Abb. 5).

Diskussion

Generell wird die züchterische Bearbeitung des Merkmals FA mit konventionellen Zuchtwerten

durch eine aufwändige und teure Phänotypisierung erschwert. Dadurch, dass Futteraufnahmedaten

vorwiegend auf Versuchsbetrieben gemessen werden, liegen nur sehr geringe Tierzahlen vor und sehr

wenige Tiere werden leistungsgeprüft (Berry and Crowley, 2013; Pryce et al., 2015). Seit die

genomische Selektion erfolgreichen etabliert wurde, bietet dies neue Ansätze für die Zucht auf FA

(Boichard and Brochard, 2012; Meuwissen et al., 2001). Für eine effiziente Bearbeitung des

0,00

0,10

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

0 1-10 11-20 21-30 31-40 41-50 51-60 61-70 71-80 81-90 91-99

35 53 378 240 149 110 40 42 69 29 18

Sic

her

hei

t d

er Z

uch

twer

te

Anzahl der Wochenmittelwerte / Anzahl der Kühe

pedigree-basiert

genomisch u. pedigreebasiert

Chapter III

74

Abb. 5. Realisierte Sicherheiten der Bullenzuchtwerte für das Merkmal Futteraufnahme in

Abhängigkeit von der Anzahl der Töchter und der berücksichtigten Verwandtschaftsalternativen

(ausschließlich Pedigree: pedigree-basiert; Pedigree aller Tiere und Genotypisierung der Kühe:

kombiniert (K); Pedigree aller Tiere und Genotypisierungen der Kühe und Kuhväter, sowie der Väter

der Kuhmütter: kombiniert (K, V-MV)

Realized reliabilities of the male breeding values of HF for the trait feed intake relative to the

number of daughters and consideration of relationship computations (only Pedigree: pedigree-

based; Pedigree of all animals and genotyped cows: kombiniert (K);Pedigree of all animals,

genotyped cows and sires of the cows, as well as sires of cow dams: kombiniert (K, V-MV)

Merkmals müssen jedoch umfangreiche Lernstichproben vorliegen, um eine adäquate genomische

Zuchtwertschätzung durchführen zu können (Haas et al., 2015).

Erste Ansätze ergaben sich auf diesem Gebiet durch das im Jahr 2011 gestartete internationale Projekt

global Dry Matter Initiative (gDMI) mit dem Ziel einer Entwicklung eines „SNP-Keys“

(Schätzformel) zur Schätzung eines genomischen Zuchtwerts für FA. Bereits dieses Projekt konnte

zeigen, dass eine Erstellung einer gemeinsamen Lernstichprobe trotz erheblicher Unterschiede in der

Merkmalserfassung (keine kontinuierliche Datenerfassung, teilweise historische Daten), Fütterung

(Teilmischrationen, Grasfütterung etc.), Management, Tierstrukturen (inklusive Färsen, geringe

genetische Verknüpfung der Bullen) und Genotypisierungen (kein einheitlicher Chip bzw. imputierte

Daten) und damit ein sehr heterogener Datensatz zu durchaus vielversprechenden Ergebnissen führt

(Veerkamp et al., 2013; Haas et al., 2015). Darauf aufbauend wurde das nationale Projekt „optiKuh“,

mit der Zucht auf FA als eine der Zielsetzungen, durchgeführt. Basierend auf einer standardisierten

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

273 38 14 5 0 1

1-5 6-10 11-15 16-20 21-25 26-30

Sic

herh

eit

der

Zu

ch

twert

e

Anzahl der Bullen / Anzahl der Töchter

pedigree-basiert kombiniert (K) kombiniert (K, V-MV)

Chapter III

75

Datenerfassung über einen Zeitraum von zwei Jahren, ähnlichen Tierstrukturen (ausschließlich

Milchkühe der Rasse HF und FV), vergleichbares Management und einer kontinuierlichen

Datenerfassung wurden hervorragende Grundlagen zur Abschätzung der Sicherheiten für die Zucht

auf FA geschaffen.

Damit war es möglich, umfassende und belastbare Schätzungen für Varianzkomponenten und

Zuchtwerte über den Laktationsverlauf zu berechnen, um schlussendlich Tiere mit einer hohen

Futteraufnahme zu Laktationsbeginn zu identifizieren.

Diese genetischen Fundierungen wurden über den Laktationsverlauf durch eine Berechnung der

Laktationskurve für die drei Merkmale FA, Mkg und EB abgegriffen. Die Ergebnisse verhielten sich

wie erwartet und ähneln sich mit denen im der Literatur (Banos et al., 2005; Hüttmann et al., 2009).

Auch die Heritabilitätschätzung wurde im Laktationsverlauf abgeleitet. Die Ergebnisse liegen im

mittleren Bereich und stimmen mit den Literaturwerten überein (Berry et al., 2007; Tetens et al.,

2014). Zu Beginn sind die geschätzten Heritabilitäten niedriger als zum Ende der Laktation. Zum

einen ist der Varianzanteil des permanenten Umwelteffekts zu Beginn der Laktation höher. Zum

anderen nimmt die Anzahl der Kühe und Beobachtungen gegen Ende der Laktation, bedingt durch

Fütterungsversuche mit definierten Laktationsstadien, Abgänge und Trockenstehzeiten, ab. Den

Heritabilitätsberechnungen liegen die jeweils innerhalb der Verwandtschaftsinformation geschätzten

Varianzkomponenten zugrunde. Die höheren Heritabilitäten am Beginn und die niedrigeren

Heritabilitäten am Ende der Laktation, jeweils für die Variante der kombinierten Verwandtschaft, d.h.

der realisierten Verwandtschaft, deuten bei einer rein pedigree-basierten Verwandtschaft auf eine

Unterschätzung der Heritabilität zu Beginn und eine Überschätzung am Ende hin.

Für die Berechnung der genomisch basierten Zuchtwerte wurde die „single-step“ Methode gewählt

(Aguilar et al., 2010; Christensen and Lund, 2010). Mit dieser Methode lassen sich in nur einem

Schritt pedigree-basierte und genomische Daten miteinander kombinieren (Legarra et al., 2009;

Christensen and Lund, 2010) wodurch sich die Genauigkeit der Berechnung erhöht (Přibyl et al.,

2015). Integral können somit Zuchtwerte für Kandidaten berechnet und eine Vorselektion

durchgeführt werden (Masuda et al., 2016).

Die Sicherheiten für FA sind tendenziell niedriger als für routinemäßig erfasste Produktionsmerkmale

(Haas et al., 2015). Verglichen mit den Sicherheiten für FA liegen die Werte aus „optiKuh“ deutlich

höher als die Ergebnisse aus der gDMI-Studie von HAAS et al. (Haas et al., 2015), in der die Werte

in einem Bereich zwischen 0,05 und 0,28 lagen. Resultierend aus den Veränderungen der Heritabilität

zwischen dem PKG gegenüber dem PED Datensatz ergeben sich für die weiblichen Kandidaten um

ca. 10 % gesteigerte Sicherheiten. Die Sicherheit von 26 % entspricht einer Genauigkeit von 0,51 und

Chapter III

76

liegt auch hier im Vergleich zu gDMI ebenfalls höher (Haas et al., 2015). Diese Ergebnisse lassen

sich durch die homogenere Datenstruktur in dem Projekt „optiKuh“ erklären.

Durch die in dieser Studie gefundenen geringen genetischen Korrelationen wird deutlich, dass die

Futteraufnahme zu Beginn und zum Ende der Laktation unterschiedliche Merkmale präsentieren.

Auch (Manzanilla Pech et al., 2014) und (Veerkamp et al., 2015) fanden schwache genetische

Korrelationen zwischen der FA in der frühen, mittleren und späten Laktation. Dies führt ebenfalls zu

der Annahme, dass mehrere Gene an der Merkmalsausprägung beteiligt sind (Shonka et al., 2015).

Diese gezeigten Ergebnisse führen zu der Schlussfolgerung, dass eine züchterische Erhöhung der FA

in den ersten 50 bis 60 Tagen der Laktation nur eine geringe Erhöhung am Ende der Laktation nach

sich zieht. Eine Erhöhung der FA in der zweiten Laktationsphase ist unerwünscht, da die Tiere nicht

überkonditioniert in die Trockenstehphase starten sollten. Ein möglicher Lösungsansatz besteht in

einer Anpassung des Fütterungsmanagements. So könnte im Rahmen einer Zweiphasenfütterung zum

Ende der Laktation eine energetische Verdünnung der Ration – unter Berücksichtigung der

Körperkondition (BCS-Note) – erfolgen, einhergehend mit einer Reduzierung der Futterkosten in

dieser Phase. Alternativ kommt eine züchterische Beeinflussung der Futteraufnahme am Ende der

Laktation in Betracht. Hohe Zuchtwerte im ersten Laktationsabschnitt und niedrige Zuchtwerte am

Ende der Laktation schließen sich nicht aus (siehe genetische Korrelationen).

Die bisherigen Untersuchungen zeigen, dass die FA in einem engen Zusammenhang mit

Leistungsmerkmalen steht (Manzanilla Pech et al., 2014). So ermittelte (Veerkamp, 1998) eine hohe

genetische Beziehung zwischen Mkg und FA von 0,46 bis 0,65. Auch zwischen FA und EB besteht

ein enger Zusammenhang (Veerkamp et al., 2000). Das bedeutet, dass mit steigender FA die Mkg

und die EB ansteigen. Die genetische Korrelation von 0,37 zwischen FA und BCS ist dies etwas

niedriger und lässt darauf schließen, dass eine sinkende FA zu einem sinkenden BCS führt und

negative Auswirkungen auf die Gesundheit und die Fruchtbarkeit der Milchkuh haben kann. Eine

falsche Selektion zu Laktationsanfang kann fatale Konsequenzen für die Gesundheit der Tiere haben,

weshalb die bedarfsgerechte Versorgung stets im Vordergrund stehen sollte (Tetens et al., 2014).

Auch nach Brade und Brade (2016) ist eine ausreichende Energie- und Nährstoffversorgung für

hochlaktierende Milchkühe über die FA wichtig. Dadurch, dass heutige Hochleistungskühe deutliche

Energiedefizite zu Laktationsanfang aufweisen (Leesen et al., 2014) und die Energie aus ihren

Reserven nutzen müssen, kann dieser Sachverhalt durchaus der limitierende Faktor für die

Nutzungsdauer und damit die Wirtschaftlichkeit der Milchproduktion sein (Rauw et al., 1998). Wird

nun züchterisch die FA zu Laktationsbeginn erhöht, kann die Hochleistungsmilchkuh mehr Energie

für die Milchproduktion aus dem Futter gewinnen und muss weniger aus den Körperreserven

Chapter III

77

mobilisieren, die während der Trächtigkeit aufgebaut wurden (Coffey et al., 2004). Zusätzlich stellt

sich noch die Frage, ob mit einer Erhöhung der FA das Energiedefizit verbessert wird oder aber die

zusätzliche Energie für die Milchleistung oder eine erhöhte Lebendmassezunahme genutzt wird. Es

ist bekannt, dass zwischen FA und Lebendmasse ein hoher positiver Zusammenhang mit einer

Korrelation von bis zu 0,84 besteht (Manafiazar et al., 2016; Vallimont et al., 2011). Um das

Größenwachstum bzw. die Lebendmassezunahme der Tiere zu überprüfen, sollte in das Zuchtziel

zusätzlich zur FA eine Kontroll-bzw. Korrekturgröße, wie z.B. die Lebendmasse, eingeführt werden

(Veerkamp et al., 2013). Wenn korrelierte Hilfsmerkmale mit in das Zuchtziel integriert werden, kann

durch die indirekte Selektion eine Verbesserung der Zielmerkmale erreichen werden (Schüler et al.,

2001). Auch wenn die Zucht auf FA zunächst mit höheren Futterkosten zu Laktationsbeginn

verbunden ist, könnten Einsparungen in Form von sinkenden Tierarztkosten durch gesündere Tiere

mit einer verbesserten EB und einer daraus resultierenden Erhöhung der Lebensdauer verbunden mit

einer geringeren Remontierungsrate stattfinden; langfristig sind dadurch monetäre Vorteile zu

erwarten. Aber auch die kurzfristige monetäre Betrachtung in Form von Futtereffizienz sollte nicht

außeracht gelassen werden. Auf eine Erhöhung der Futtereffizienz wurde bereits indirekt über die

Zucht auf eine höhere Mkg gezüchtet, denn mit steigender Mkg wird ein immer größerer Anteil des

Futters für die Laktation und relativ weniger Futter für die Erhaltung genutzt (Veerkamp et al., 2013).

Dies lässt sich durch die hohe genetische Korrelation zwischen der Brutto-Futtereffizienz, welche

definiert ist als Verhältnis von Milchleistung (kg/Tag) zu Futteraufwand (kg TS/Tag und als gutes

Maß für die Berechnung der Futtereffizienz gilt (Connor, 2015), erklären (VALLIMONT ET AL., 2011).

Die stark negative Korrelation zwischen der Brutto-Futtereffizienz und FA wird hingegen als

problematisch eingestuft (Berry et al., 2003). Dies wiederum führt zu der Annahme, dass eine

verbesserte Brutto-Futtereffizienz ein ausgeprägteres Energiedefizit hervorruft.

Diese langfristigen Auswirkungen der Erhöhung der FA stehen der Futtereffizienz – und damit der

kurzfristigen Betrachtung gegenüber. Futtereffizientere Tiere erbringen die gleiche

Produktionsleistung bei verminderter Futteraufnahme (Connor, 2015). Direkte Vorteile finden sich

in der Einsparung von Futterkosten und Schonung natürlicher Ressourcen wie Wasser und Land und

fördern somit die ökonomische und ökologische Nachhaltigkeit der Milchproduktion (DILLON et

al., 2008). Dieser Konflikt zwischen langfristiger und kurzfristiger Betrachtung erschwert die

Zuchtzieldefinition. Die am bisher am geeignetste Möglichkeit scheint ein balanciertes Zuchtziel zu

sein: Eine Erhöhung der FA zu Laktationsbeginn und eine Zucht auf effizientere Kühe in der zweiten

Laktationshälfte, da sich die Phase des Energiedefizits nicht mehr im kritischen Bereich befindet

(Veerkamp et al., 2013).

Chapter III

78

Dadurch, dass die direkte Futteraufnahmemessung sehr teuer ist, gibt es einige Alternativen, um die

FA zu messen bzw. zu schätzen. Zum einen besteht die Möglichkeit, die FA über Hilfsmerkmale

abzuleiten. Die Lebendmasse ist ein guter Prädikator, da die FA zu 70 % aus der Lebendmasse und

der Mkg geschätzt werden kann. Während die Milchleistung leicht zu erfassen ist, verfügen nur

wenige Praxisbetriebe über eine zuverlässige Wiegetechnik (Zom et al., 2012) was diese

Messmethode erschwert. Eine weitere Möglichkeit ist die Schätzung der FA aus Exterieurmerkmalen

(Egger-Danner et al., 2015). (Manzanilla-Pech et al., 2016) konnte Vorhersagegenauigkeiten von 0,43

erreichen, diese Varianten sind aber, aufgrund der indirekten Ableitung, mit hohen Unsicherheiten

behaftet.

Schlussfolgerung

Das Verbundprojekt ermöglicht erstmals eine direkte Selektion auf Futteraufnahme. Insgesamt

konnte gezeigt werden, dass mit einer weitestgehend standardisierten Datenerfassung und mit Hilfe

des „single-step“ Verfahrens vielversprechend hohe Sicherheiten für genomische Zuchtwerte für

Futteraufnahme generiert werden konnten.

Eine umfassende Bewertung aller Aspekte der Futteraufnahme ermöglicht eine Prüfung, in wie fern

das Merkmal Futteraufnahme mit genomischen Zuchtwerten bestmöglich in das Zuchtziel der

Hochleistungsmilchkuh integriert werden kann. Ein möglicher Lösungsansatz wäre ein balanciertes

Zuchtziel mit einer höheren Futteraufnahme zu Laktationsbeginn und einer adäquaten Effizienz im

Verlauf der Laktation bei gleichzeitiger Berücksichtigung der Gesundheit und einer höheren

Stoffwechselstabilität der Tiere. Dafür sollten unbedingt die genetischen Beziehungen zu allen

anderen wichtigen Zuchtzielmerkmalen geschätzt werden, möglichst an einem umfangreicheren

Datenmaterial.

Um in Zukunft noch höhere Sicherheiten im Merkmal Futteraufnahme generieren zu können, muss

die Anzahl an Geno- und Phänotypisierungen erhöht werden und eine kontinuierliche Erfassung der

Daten für eine stetige Aktualisierung der Lernstichprobe sichergestellt werden.

Mit dem Verbundprojekt „optiKuh“ und seiner Fortführung im kürzlich begonnenen BLE-Projekt

„eMissionCow“ wird in Deutschland eine solide Grundlage für die Zucht auf FA geschaffen.

Danksagung

Die Förderung des Vorhabens erfolgte aus Mitteln des Bundesministeriums für Ernährung und

Landwirtschaft (BMEL) aufgrund eines Beschlusses des deutschen Bundestages. Die

Chapter III

79

Projektträgerschaft erfolgte über die Bundesanstalt für Landwirtschaft und Ernährung (BLE) im

Rahmen des Programms zur Innovationsförderung.

Diese Arbeit wurde dankenswerterweise durch ein Anschlussstipendium der H. Wilhelm Schaumann

Stiftung gefördert.

Literatur

Aguilar, I., I. Misztal, D.L. Johnson, A. Legarra, S. Tsuruta and T.J. Lawlor, (2010): Hot topic. A

unified approach to utilize phenotypic, full pedigree, and genomic information for genetic

evaluation of Holstein final score. J. Dairy Sci. 93 (2), 743–752.

Ali, T.E. and L.R. Schaeffer, (1987): Accounting for covariances among test day milk yields in dairy

cows. Can. J. Anim. Sci. 67 (3), pp. 637–644.

Banos, G., M.P. Coffey and S. Brotherstone, (2005): Modeling Daily Energy Balance of Dairy Cows

in the First Three Lactations. J. Dairy Sci. 88 (6), 2226–2237.

Brade, W. and E. Brade, (2016): Die Futteraufnahme von hochleistenden Milchkühen im ersten

Laktationsdrittel aus Sicht der Züchtung und des Tierwohles. Berichte über die Landwirtschaft,

94(2)

Berry, D.P., F. Buckley, P. Dillon, R.D. Evans, M. Rath and R.F. Veerkamp, (2003): Genetic

Parameters for Body Condition Score, Body Weight, Milk Yield, and Fertility Estimated Using

Random Regression Models. J. Dairy Sci. 86 (11), 3704–3717.

Berry, D.P. and J.J. Crowley, (2013): Cell Biology Symposium: genetics of feed efficiency in dairy

and beef cattle. J. Anim. Sci. 91 (4), pp. 1594–1613.

Berry, D. P., B. Horan, M. O'Donovan, F. Buckley, E. Kennedy, M. McEvoy and P. Dillon, (2007):


J. Dairy Sci. 90 (10), 4835–4845.

Boichard, D. and M. Brochard, (2012): New phenotypes for new breeding goals in dairy cattle.

Animal : an international journal of animal bioscience 6 (4), 544–550.

Burnham, K., P. Anderson and D. Raymond, (1998): Model selection and inference. A practical

information-theoretic approach. New York NY u.a.: Springer.

Calus M.P.L., Y. de haas, M. Pszczola and R.F. Veeerkamp: Predicted accuracy of and response to

genomic selection for new traits in dairy cattle. Animal 7 (2), 183 - 191.

Christensen, O., M.S. Lund, (2010): Genomic prediction when some animals are not genotyped.

Genetics, Selection, Evolution: GSE 42, p. 2.

Chapter III

80

Coffey, M.P., G.C. Emmans and S. Brotherstone, (2001): Genetic evaluation of dairy bulls for energy

balance traits using random regression. Anim. Sci. 73 (01), 29–40.

Coffey, M. P., G. Simm and S. Brotherstone, (2002): Energy Balance Profiles for the First Three

Lactations of Dairy Cows Estimated Using Random Regression. J. Dairy Sci. 85 (10), pp. 2669–

2678.

Coffey, M.P., G. Simm, J.D. Oldham, W.G. Hill and S. Brotherstone, (2004): Genotype and diet

effects on energy balance in the first three lactations of dairy cows. J. Dairy Sci. 87 (12), pp. 4318–

4326.

Connor, E. E., (2015): Invited review: improving feed efficiency in dairy production: challenges and

possibilities. Animal 9 (3), 395–408.

Dillon, P.A.T., T. Hennessy, L. Shalloo, F. Thorne and B. Horan, (2008): Future outlook for the Irish

dairy industry. A study of international competitiveness, influence of international trade reform

and requirement for change. Int. J Dairy Tech. 61 (1), 16–29.

Egger-Danner, C., J.B. Cole, J.E. Pryce, N. Gengler, B. Heringstad, A. Bradley and K.F. Stock,

(2015): Invited review. Overview of new traits and phenotyping strategies in dairy cattle with a

focus on functional traits. Animal 9 (2), 191–207.

GfE, (2001): Ausschuss für Bedarfsnormen der Gesellschaft für Ernährungsphysiologie. Energie- und

Nährstoffbedarf landwirtschaftlicher Nutztiere, Nr. 8: Empfehlungen zur Energie- und

Nährstoffversorgung der Milchkühe und Aufzuchtrinder. Frankfurt a. Main: DLG-Verlag.

Haas, Y. de, M.P.L. Calus, R.F. Veerkamp, E. Wall, M.P. Coffey and H.D. Daetwyler, et al., (2012):

Improved accuracy of genomic prediction for dry matter intake of dairy cattle from combined

European and Australian data sets. J. Dairy Sci., 95 (10), 6103–6112.

Haas, Y. de, J.E. Pryce, M.P.L. Calus, E. Wall, D.P. Berry, P. Lovendahl, et al., (2015): Genomic

prediction of dry matter intake in dairy cattle from an international data set consisting of research

herds in Europe, North America, and Australasia. J. Dairy Sci., 98 (9), 6522–6534.

Hüttmann, H., E. Stamer, W. Junge, G. Thaller and E. Kalm, (2009): Analysis of feed intake and


models. Animal 3 (2), pp. 181–188.

Leesen, R. von, J. Tetens, E. Stamer, W. Junge, G. Thaller and N. Krattenmacher, (2014): Effect of


primiparous Holstein-Friesian cows. J. Dairy Sci. 97 (2), 1128–1138.

Legarra, A., I. Aguilar and I. Misztal, (2009): A relationship matrix including full pedigree and

genomic information. J. Dairy Sci. 92 (9), 4656–4663.

Chapter III

81

Madsen, P., J. Jensen, R. Labouriau, L.G. Christensen and G. Sahana, (2013): DMU - A Package for


Manafiazar, G., L. Goonewardene, F. Miglior, D.H. Crews, J.A. Basarab, E. Okine and Z. Wang,

(2016): Genetic and phenotypic correlations among feed efficiency, production and selected

conformation traits in dairy cows. Animal 10 (3), 381–389.

Manzanilla Pech, C.I.V., R.F. Veerkamp, M.P.L. Calus, R. Zom, A. van Knegsel, J.E. Pryce and Y.

de Haas, (2014): Genetic parameters across lactation for feed intake, fat- and protein-corrected

milk, and liveweight in first-parity Holstein cattle. J. Dairy Sci. 97 (9), 5851–5862.

Manzanilla-Pech, C.I.V., R.F. Veerkamp, R.J. Tempelman, M.L. van Pelt, K.A. Weigel, M.

VandeHaar, et al., (2016): Genetic parameters between feed-intake-related traits and

conformation in 2 separate dairy populations--the Netherlands and United States. J. Dairy Sci. 99

(1), 443–457.

Masuda, Y., I. Misztal, S. Tsuruta, A. Legarra, I. Aguilar, D.A.L. Lourenco, et al. (2016):

Implementation of genomic recursions in single-step genomic best linear unbiased predictor for

US Holsteins with a large number of genotyped animals. J. Dairy Sci. 99 (3), 1968–1974.

Přibyl, J., J. Bauer, V. Čermák, P. Pešek, J. Přibylová, J. Šplíchal, et al. (2015): Domestic estimated

breeding values and genomic enhanced breeding values of bulls in comparison with their foreign

genomic enhanced breeding values. Animal 9 (10), 1635–1642.

Pryce, J. E., O. Gonzalez-Recio, G. Nieuwhof, W.J. Wales, M.P. Coffey, B.J. Hayes and M.E.

Goddard, (2015): Hot topic. Definition and implementation of a breeding value for feed efficiency

in dairy cows. J. Dairy Sci. 98 (10), 7340–7350.

Rauw, W.M., E. Kanis, E.N. Noordhuizen-Stassen and F.J. Grommers, (1998): Undesirable side

effects of selection for high production efficiency in farm animals. A review. Livest. Prod. Sci.

56 (1), 15–33.

Schüler, L., H. Swalve und K.-U. Götz, (2001): Grundlagen der quantitativen Genetik. 94 Tabellen.

Stuttgart (Hohenheim): Ulmer (UTB für Wissenschaft Agrarwissenschaft, 2183).

Schwarz, G., (1978): Estimating the Dimension of a Model. The Annals of Statistics 6 (2), 461–464.

Shonka, B.N., S. Tao, G.E. Dahl and D.M. Spurlock, (2015): Genetic regulation of prepartum dry

matter intake in Holstein cows. J. Dairy Sci. 98 (11), 8195–8200.

Tetens, J., G. Thaller and N. Krattenmacher, (2014): Genetic and genomic dissection of dry matter

intake at different lactation stages in primiparous Holstein cows. J. Dairy Sci. 97 (1), 520–531.

Vallimont, J.E., C.D. Dechow, J.M. Daubert, M.W. Dekleva, J.W. Blum, C.M. Barlieb, et al., (2011):

Heritability of gross feed efficiency and associations with yield, intake, residual intake, body

Chapter III

82

weight, and body condition score in 11 commercial Pennsylvania tie stalls. J. Dairy Sci. 94 (4),

2108–2113.

Veerkamp, R. F., (1998): Selection for economic efficiency of dairy cattle using information on live

weight and feed intake. A review. J. Dairy Sci. 81 (4), 1109–1119.

Veerkamp, R.F., Y. Haas, J.E. de; Pryce, M.P. Coffey, D. Spurlock, M.J. Vandehaar, (2015):

Guidelines to measure individual feed intake of dairy cows for genomic and genetic evaluations,

pp. 1–9.

Veerkamp, R.F., J.K. Oldenbroek, H.J. van der Gaast and J.H.J. van der Werf, (2000): Genetic

Correlation Between Days Until Start of Luteal Activity and Milk Yield, Energy Balance, and

Live Weights. J. Dairy Sci. 83 (3), 577–583.

Veerkamp, R.F., J.E. Pryce, D. Spurlock, D.P. Berry, M.P. Coffey, P. Løvendahl, et al., (2013):

Selection on Feed Intake of Feed Efficiency: A Position Paper from gDMI Breeding Goal

Discussion. Interbull Bulletin (No. 47), 15–22.

Zom, R.L.G., G. André and A.M. van Vuuren, (2012): Development of a model for the prediction of

feed intake by dairy cows 2. Evaluation of prediction accuracy. Livest. Sci. 143 (1), 58–69.

General Discussion

83

General Discussion

The aim of the thesis was to analyze feed intake and energy balance data from the project “optiKuh”.

The beginning of lactation is the most critical period for dairy cows where the majority of metabolic

diseases occur. Thus, it should be verified how feed intake and energy balance should be implemented

in the breeding goal and if so the health status can be improved. Therefore, a reference population

with a particular focus on heritability, genetic parameters and breeding value estimation with the use

of the method single-step was generated to underpin future genetic selection for Holstein-Friesian

cows. In addition, for the breed Simmental Cattle some interesting results not mentioned in the papers

are discussed in detail below. Furthermore, it should be clarified, whether and how large genomic

selection could offer advantages in comparison to conventional calculated reliabilities.

Metabolic issues and definition of the breeding goal

Traditionally, breeding programs for dairy cows focus on production traits. Genetic selection and

improved management methods have significantly increased milk production, resulting in increased

feed costs per feed unit, but also improved feed efficiency (Shonka et al., 2015). A defined general

breeding goal is an efficient cow that is well-conditioned, has a high health status with high milk yield

and a balanced feed intake.

On the one hand, feed accounts for a significant proportion of 50% of production costs and is the

major expense in dairy farming (Mäntysaari et al., 2012; European Commission, 2013; Connor,

2015).

On the other hand, especially the first weeks of lactation are the critical time period for the health and

metabolic stability of dairy cows (Drackley, 1999). Problematically, a high efficiency in combination

with an increased milk yield is associated with higher mobilization of body reserves and metabolic

diseases. The issue of the energy deficit, especially at the beginning of the lactation period, is

discussed in Chapter I.

Thus, feed intake is a major trait, which should be modified (Li et al., 2018), especially in the initial

phase of lactation. Already Persaud et al. (1991) indicate that breeding for higher feed intake in early

lactation seems to be a promising strategy to improve the energy balance in that stage and thereby

reduces the catabolism of body reserves. Even today, Brade and Brade (2016) still claim a high feed

intake that is essential to improve the health status. In addition, the significant impact on dairy

General Discussion

84

production, profitability and the economy as a long-term objective requires new perspectives

(Connor, 2015; Puillet et al, 2016).

To answer the question of the appropriate breeding goal for high performing dairy cows, a balance

between efficiency and health should be reached to prevent undesirable side effects (Brade and Brade,

2016). This can best be achieved through breeding strategies that select for important traits and their

connection to profitability and not just individual traits (Berry and Crowley, 2013; Pryce and Haas

2017). Eventually it could be useful to define different breeding goals for lactation sections and not

for the entire lactation.

Energy balance, feed intake and milk yield

In dairy cattle breeding, the slower increase of feed intake, in comparison to milk yield at the

beginning of lactation has been widely recognized. Li et al. (2018) reported the same development of

lactation trajectory for feed intake and energy balance as found in this study. After calving, feed intake

and milk yield increase with a different degree. Milk yield increased sharply to peak at days in milk

50 in both breeds and feed intake reaches the peak at days in milk 80 (Chapter I, Figure 2). This leads

to a lack of energy supply for milk production as well as increased body reserve mobilization in early

lactation (Banos et al., 2012) (see Chapter I).

The duration of the energy deficit of primiparous cows is not that long and not that intensive compared

to multiparous cows (Table 2). This is also shown in the study of Buttchereit (2011). Furthermore,

the energy deficit in higher lactations was endured longer but has a flatter trajectory. On average, the

cows return to positive energy balance between 53 and 77 days in milk, depending on lactation

number. Coffey et al. (2002) showed quite similar results with return to positive energy balance

around days in milk 72 to 95. Moreover, the rate of return to positive energy balance might be a useful

indicator of resumption of reproductive activity (Coffey et al., 2002). In addition, the time when the

energy balances from negative to positive could be used as a threshold for analyses regarding

metabolic stability and selection strategies.

As explained in Chapter I, the negative energy balance is longer lasting in Simmental Cattle but not

that pronounced. This might be due to the influence of age of first calving, on the length of the dry

period or the genetic architecture (Emmans, 1994). In this study, age of first calving for Holstein-

Friesian range between 24 and 26 months and for Simmental Cattle between 25 and 28 months,

respectively.

General Discussion

85

Table 1: Average days in milk when energy balance change from negative to positive for Holstein-

Friesian and Simmental Cattle

Genetic properties of feed intake and energy balance

Only for the „optiKuh“ reference Holstein-Friesian population heritabilities could be computed and

range between 0.12 and 0.50 (feed intake) and 0.15 and 0.48 (energy balance) for pedigree-based data

and 0.21 and 0.45 (feed intake) and 0.17 and 0.28 (energy balance) for single-step data. The values

are lower at the beginning of lactation but increased towards the end of lactation. An explanation

could be that at the end of lactation, decreasing number of animals were available. Similar results

were also found in other studies (Liinamo et al., 2012; Li et al., 2018). Also e.g. Spurlock et al. (2012)

or Krattenmacher et al. (2019) had same tendencies whereas only the first 180 DIM were considered.

The estimated entire-lactation genetic correlation between early and later lactation stages were found

to be far away from unity (0.14 for feed intake and 0.00 for energy balance), whereas the

corresponding correlation between cow effects was 0.06 for feed intake and 0.00 for energy balance.

For the breed Simmental Cattle, only the correlation between cow effects could be computed and

range at 0.39 for feed intake and -0.34 for energy balance. Genetic correlations with adjacent days in

milk are phenotypically and genetically high (See Chapter II). In general, in most of the studies the

genetic correlations in course of lactation are quite similar (Berry et al., 2007; Liinamo et al., 2012;

Manzanilla Pech et al., 2014). These results indicate that feed intake is controlled by different genes

(Tetens et al., 2014; Shonka et al., 2015). Consequently, observed weak and sometimes negative

correlations between early and later lactation stages create poor predictive ability and could not be

used as strong predictors for feed intake of days in milk in early lactation and vice versa (Li et al.,

2018). Furthermore, the genetic heterogeneity of feed intake across lactation needs to be carefully

considered in any recording, selection strategies and in the design of breeding goals.

Genomic selection

To date, genomic selection has led to significant changes in the dairy industry (Weller, 2016). For

example, breeding values can now be estimated for animals without phenotypic information for

difficult to measure traits, e.g. feed intake whose recording is either too expensive to be carried out

with large numbers of cows or where phenotypes are not easily accessible (Goddard and Hayes,

Lactation change

1 2 3 ≥ 4

Holstein-Friesian 56 57 65 75

Simmental Cattle 53 66 77 75

General Discussion

86

2009). In that case, benefits of genomic selection are greatest. Among others, these arguments justify

the use of genomic selection and the advantage over traditional best linear unbiased prediction (Muir,

2007) and is subject to the present investigation.

For feed intake and energy balance, the limited data size remains a challenge for genomic evaluation.

Thus, a female reference population can be a solution. Further advantages of the use of genotyped

females are direct modeling of record phenotypes (Macciotta et al., 2015).

Particularly, the requirements concerning the number of animals in the reference population is high

in order to achieve high reliabilities (Goddard and Hayes, 2009; Gonzalez-Recio et al., 2014).

Generally, it is difficult to generate a big reference population, wherefore the international

collaboration project gDMI (global Dry Matter Initiative) has been established (Berry et al., 2014; de

Haas et al., 2014; Pryce et al., 2014; Krattenmacher et al., 2019). The project gDMI is based on very

different production systems and measurements of feed intake with partly inadequate genetic

linkages. The use of a common international reference population is limited. With the pendant

“optiKuh”, national phenotype data were brought together. Results for reliabilities could be improved

with more homogenous data, compared to gDMI (Berry et al., 2014; Haas et al., 2015).

Quality criteria of genetic data for Holstein-Friesian and Simmental Cattle

Currently, genetic studies on feed intake or energy balance in other dairy cattle breeds than Holstein-

Friesian are rare and based on very small data sets (Søndergaard et al., 2002; Liinamo et al., 2012).

One possibility to overcome this problem of a too small test group in reference population, is a

combined data sets. In the present study, the attempt was to integrate Simmental Cattle in the analyses.

Unfortunately, the size of the population was too small to reach convergence for estimation of genetic

parameters and variance components. Nevertheless, initial analyses could be realized.

262 Simmental Cattle cows in the project “optiKuh” were genotyped with the Illumina BovineSNP50

v2 BeadChip 50K ( (Illumina Inc., San Diego, CA). Data was edited in the same way as for Holstein-

Friesian (Chapter II) and 49,345 SNPs were available.

To ensure sufficient genotypic data quality, SNPs with call rates lower than 95% were discarded. For

Gencall-Scores (GC-Scores) a threshold (grey line in Figure 1) was set at 0.6, so every SNP with a

mean GC-score lower than 0.6 was excluded from further analyses. For SI 3,775 SNPs (7.65%) lay

below this threshold and were discarded. It might have been possible to set the threshold at 0.2 in this

study to avoid the loss of information whilst also achieving a good quality ((Illumina Inc., 2005;

General Discussion

87

Edriss et al 2013). After quality control the final SNP data set contained 262 animals and 45.570

SNPs for FI.

Figure 1. Frequency distribution of the mean GC-score for the breed Simmental Cattle (quality control

threshold at 0.6)

To gain additional information, the minor allele frequency was calculated for both breeds, Holstein-

Friesian and Simmental Cattle. The frequency of the minor allele in a sequence variation in a

population is the second most frequent allele value. Markers with a low minor allele frequency, have

low heterozygosity and are therefore less informative. The distribution of the minor allele frequency

is given in Figure 2. The minor allele frequency of Holstein-Friesian is distributed as expected,

whereas the high frequency around zero of Simmental Cattel can be explained due to a small data set.

The minor allele frequency in the current study was calculated in the program G-matrix (Madsen et

al., 2013).

Figure 2. Distribution of Minor Allele Frequencies of the breed Holstein-Friesian (left) and

Simmental Cattle (right)

0

1000

2000

3000

4000

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Fre

qu

ency

GC-score

0

200

400

600

800

1000

1200

1400

0 0.1 0.2 0.3 0.4 0.5

Fre

qu

ency

MAF

0

200

400

600

800

1000

1200

1400

0 0.1 0.2 0.3 0.4 0.5

Fre

qu

ency

MAF

General Discussion

88

Analyses for the pedigree check for data quality control due to relationship coefficients between

pedigree-based and single-step data are shown in Chapter II, Figure 2. Equivalent analyses for the

breed Simmental Cattle are presented below. The relationship coefficients before data editing are

illustrated on the left hand side. The augmentation of data set for Simmental Cattle by another breed

was not expected. These data include 40 animals belonging to the breed Brown Swiss from one

research farm. On the right hand side, coefficients of relatedness are shown after exclusion of Brown

Swiss. It seems that these Brown Swiss are closer related within breed than Simmental Cattle due to

the higher level of relationship.

Figure 3. Comparison of the relationship coefficients between pedigree relationship matrix and

genomic relationship matrix for Simmental Cattle – before (left) and after (right) exclusion of 40

cows with a false breed classification (Simmental Cattle instead of Brown Swiss)

Reliabilities of breeding values

There is a growing interest for genomic breeding values with an acceptable reliability for the

examined traits (Pryce and Berry, 2014). In general, accuracy of genomic predictions is critical for

the expected genetic gains resulting from genomic selection. Several factors are decisive for the

reliability of breeding values which include but are not limited to heritability, the size of the reference

population and the genetic relationship between the animals and the available information (Goddard

and Hayes, 2009; Gonzalez-Recio et al., 2014; Dehnavi et al., 2018).

A study of Calus et al. (2013) showed that a reference population with 2000 cows could already gain

success in breeding selection for new, not yet in the breeding goal integrated traits. With up to 1,500

cows, “optiKuh” could be seen as an initial project with which first adequate reliabilities could be

realized.

General Discussion

89

To increase the reliability of breeding values, Shabalina et al. (2017) reported, that a close relationship

between the animals is required. In the current project, the average relationship between the eight

Holstein-Friesian research farms was 6%, whereby Hohenheim and Dummerstorf have a lower

relationship and in contrast, the relationship between Karkendamm and Neumühle lies above the

average. The results for Holstein-Friesian farms in “optiKuh” are higher, in comparison to a study of

Haas et al. (2015), in which the calculated relationship between the international farms ranged

between 2 and 4%. For example, between the research farms Riswick and Karkendamm 20 common

sires of totally 218 sires and 28 common grand sires of 118 grand sires of all “optiKuh”- Holstein-

Friesian cows, were identified. This is a small number of common used sires, which can still be

interpreted as a weak genetic linkage. These results were underlined by the results of the reliabilities

of breeding values for bulls for the trait feed intake. Only two bulls have 26 or more daughters, but

the most sires have five or less daughters. As Figure 10 in Chapter II shows, the more common sires

were used, the higher the reliabilities were. A possibility to increase reliabilities would be a tighter

link between research farms.

The analyses for the single-step data set were done with the procedure “single-step”, a method

described in Chapter II, where pedigree and genomic data is combined. This was carried out with the

program DMU (Derivative-free approach to Multivariate analysis by Restricted Maximum

Likelihood (REML)) (Madsen et al., 2013). This program, beside e.g. the program BLUPF90 family

(Misztal et al., 2002), can directly run single-step evaluations (Legarra et al., 2014).

In this study, breeding value reliabilities lead to higher results for the pedigree-based alternative but

only when weekly records exceed 40. Consequently, single-step estimation generated higher

reliabilities for both traits with less phenotypic information. These results are in contrary to the

literature, where the genomic alternative give always higher values for breeding reliabilities (Pszczola

et al. 2013; Manzanilla-Pech et al. 2017). One possible explanation might be the higher pedigree-

based heritability at the end of lactation. Increasingly, this occurs by cows with many records.

Additionally, Manzanilla Pech et al. (2014) reported, that recording of the trait in mid or late lactation

gave higher accuracy, because heritability increased towards the end of lactation.

With the genomic breeding value estimation and the establishment of an “optiKuh” reference

population, reliabilities for cows, which are not phenotyped but genotyped, could be generated.

In the current study, reliability of 26% for female candidates was achieved. This value is 10% higher

compared to only pedigree-based values. For energy balance, values of 19.1% were analyzed and are

6% higher compared to pedigree-based alternative. These results could be generated due to the strong

genetic relationship between the candidates and the cows in the reference population (Habier et al.,

General Discussion

90

2010; Pszczola et al., 2012). Genomic relationships among animals in the reference population might

be less important than those for the evaluated male or female candidates with no phenotypic

observations. This higher connectedness reduces bias and thus improves the genetic evaluation

(Pszczola et al., 2012). This leads to the assumption that candidates with a high genetic connection to

the corresponding reference population leads to higher reliabilities, compared to candidates with a

lower genetic relationship.

Selection strategy

To select cows with high feed intake at the beginning of lactation, the pattern of heritabilities and

genetic correlations must be taken into account for decision-making processes. Heritabilities for

estimating the breeding values are low in this period, compared to the end of lactation. This influences

the reliability of breeding values negatively.

Furthermore, the low genetic correlations between beginning and end of lactation give information

about how selection on high feed intake at the beginning of lactation has an impact on feed intake at

the end of lactation. When using breeding values at the onset of lactation, it should be considered,

how feed intake developed in the course of lactation. A possible strategy would be the selection of

cows with high values at the beginning and mid or negative breeding values at the end of lactation

(see Figure 4). These cows would have a higher feed intake at the beginning of lactation resulting in

an improve of health but additionally a balance feed intake during the course of lactation.

Therefore, it is important to use the values of the entire lactation. In this study, the first 350 days in

milk were used, although Figure 1 in Chapter I leads to the assumption that values should only be

considered until the 300 days in milk to avoid over estimation of heritabilities, genetic parameters

and breeding values.

Daily breeding values in the course of lactation provide some extra knowledge about possible

breeding strategies for increasing feed intake or reducing energy deficit at the beginning of lactation.

Thus, for all cows with at least ten weekly averages, resulting in 1,226 cows (from 1,342), values

were estimated. Furthermore, average daily breeding values were calculated for the first 75 (BV75)

lactation days and for the lactation days between 250 and 325 (BV325). BV75 was chosen according

to the exhibited negative energy balance with an average duration of the first 75 days. The correlations

between the breeding values at the beginning and at the end of lactation are positive with r = 0.59 for

feed intake and r = 0.31 for energy balance. These values are higher compared to the calculated

genetic correlations between the 35 days in milk and the 285 days in milk of lactation for feed intake

(0.44) and energy balance (0.23).

General Discussion

91

According to BV75 the animals were grouped into high, moderate and low daily breeding values

(light blue area). Class boundaries were set at > 1 (high), 1 to -1 (moderate) and < -1 (low) kg for

feed intake and > 7 (high), 7 to -7 (moderate) and < -7 (low) MJ NEL for energy balance (Table 2 and

Table 3). Furthermore, animals were grouped considering both, breeding values at the beginning and

at the end (marked in light grey).

Table 2. Grouped animals by average breeding values for feed intake (FI) in total (light blue box) and

cross overlapping between breeding values of first 75 lactation days and lactation days between 250

and 325

FI 250 to 325 DIM

(mean daily breeding value)

high

(> 1 kg)

moderate

(1 to -1 kg)

low

(< -1 kg)

No. of cows

(mean daily

breeding value)

186

(1.96 kg)

628

(-0.08 kg)

412

(-2.08 kg)

FI 1 to 75 DIM

(mean daily

breeding value)

high

(> 1 kg)

84

(1.38 kg) 45 35 4

moderate

(1 to -1 kg)

1036

(-0.00 kg) 137 573 326

low

(< -1 kg)

106

(-1.43 kg) 4 20 82

General Discussion

92

Table 3. Grouped animals by average breeding values for energy balance (EB) in total (light blue

box) and cross overlapping between breeding values of first 75 lactation days and lactation days

between 250 and 325

EB 250 to 325 DIM

(mean daily breeding value)

high

(> 7 MJ

NEL)

moderate

(-7 to 7 MJ

NEL)

low

(< -7 MJ

NEL)

No. of cows

(mean daily

breeding value)

95

(10.64 MJ

NEL)

1011

(-0.14 MJ

NEL)

120

(-10.67 MJ

NEL)

EB 1 to 75 DIM

(mean daily

breeding value)

high

(> 7 MJ NEL)

120

(10.97 MJ NEL) 31 86 3

moderate

(7 to -7 MJ NEL)

1007

(0.04 MJ NEL) 60 851 96

low

(< -7 MJ NEL)

99

(-9.60 MJ NEL) 4 74 21

General Discussion

93

Lactation curves of animals grouped of high/high, high/moderate, high/low and high/low average

daily breeding values were illustrated in Figure 4. Therefore, ten cows with highest values of each

column of the first row from Table 2 (feed intake) and Table 2 (energy balance) were chosen due to

the important time period, the beginning of lactation. Feed intake and energy balance curves show

nearly the same pattern in each section. Cows with high/high breeding values (a) and (d) have high

breeding values during the entire lactation.

Figure 4. Lactation curves of daily breeding values for feed intake (left (a-c)) and energy balance

(right (d – f)) for selected cows according to the mean lactation day 1-75 and mean lactation day 250

- 325

Animals with high feed intake and energy balance breeding values at the beginning of lactation and

a mid or negative breeding values at the end of lactation are grouped into high/moderate and

low/moderate. The breeding values dropped in course of lactation, whereas in group high/low the

decreasing is more intense. These animals, which are grouped into high/moderate and high/low, might

-8

-6

-4

-2

0

2

4

6

8

0 50 100 150 200 250 300 350

Gen

om

ic b

reed

ing

valu

e fo

r fe

ed in

take

(kg

DM

)

high (d1-75) / moderate (d250-325)

n=10

(b)

-8

-6

-4

-2

0

2

4

6

8

0 50 100 150 200 250 300 350

Gen

om

ic b

reed

ing

valu

e fo

r fe

ed in

take

(kg

DM

)

DIM

high (d1-75) / low (d250-325)(c)

-40

-30

-20

-10

0

10

20

30

40

0 50 100 150 200 250 300 350

Gen

om

ic b

reed

ing

valu

e fo

r en

ergy

bal

ance

(M

J N

EL

)

average breeding values for energy balance(d) high (d1-75) / high (d250-325)

n=10

-40

-30

-20

-10

0

10

20

30

40

0 50 100 150 200 250 300 350

Gen

om

ic b

reed

ing

valu

e fo

r en

ergy

bal

ance

(M

J N

EL

)

high (d1-75) / moderate (d250-325)

n=10

(e)

-40

-30

-20

-10

0

10

20

30

40

0 50 100 150 200 250 300 350

Gen

om

ic b

reed

ing

valu

e fo

r en

ergy

bal

ance

(M

J N

EL

)

DIM

high (d1-75) / low (d250-325)

n=3

(f)

-8

-6

-4

-2

0

2

4

6

8

0 50 100 150 200 250 300 350

Gen

om

ic b

reed

ing

valu

e fo

r fe

ed in

take

(kg

DM

)

average breeding values for feed intake

high (d1-75) / high (d250-325)

n=10

(a)

General Discussion

94

be interesting for the breeding on feed intake. Over conditioning could be avoided and cows can start

well-conditioned into the dry period. Furthermore, a breed on high values at the beginning seems to

be possible. The daily breeding values and the genetic correlations varying within the course of

lactation thus confirm the results of current thesis, that different traits are responsible for feed intake

and energy balance at the beginning and at the end of lactation.

In conclusion, the illustrations show the possibility to breed for high feed intake or energy balance at

the beginning of lactation without over supply at the end of lactation. Thus, metabolic stability

combined with no overly conditioned at the beginning of the dry period seems to be reachable. As a

quintessence, the whole lactation should be considered in case of selection strategies.

Relationship of daily breeding values of feed intake and energy balance to disease rate

Cows in a high energy deficit during early lactation are more likely to have higher incidences of

diseases such as metabolic disorders (Collard et al., 2000). Several significant phenotypic associations

between health traits and feed intake or energy balance have been reported in the literature, but little

information exists about their genetic relationship. For this purpose, average breeding values of the

first 75 days in milk for feed intake and energy balance were calculated depending on disease rates.

Diagnosis of diseases were recorded within the project “optiKuh”. Diseases were clustered into

metabolic and general disease. The aggregation of diagnosis to disease categories were necessary,

because of the limited amount of data. In addition, metabolic problems were most prevalent at the

beginning of lactation when the physiological requirements of the cows are high. “Metabolic

diseases” include e.g. acidosis or ketosis, wheares “other diseases” contain mastitis, claw and leg

diseases, reproduction disorders, respiatory disorders, digestive disorders and other illnesses.

Disease codes were generated in an analogoues manner for both categories. Each observation (day)

was allocated a code, “1” if the cow showed a disease and “0” otherwise. In case of recorded

metabolic diseases, the day of first diagnosis plus one day of mean treatment duration and the

following eight days as recovery time were coded with “1”. For the “general disease rate” code “1”

was chosen if the cow show a disease, first diagnosis or treatment duration, generally. Disease rate is

defined as the frequency of occurrence of disease within in the first 75 days in milk.

Table 4 and Table 5 show the disease rates of “metabolic diseases” and “general diseases” dependent

on breeding value class (high, moderate and low) due to feed intake and energy balance. Groups were

chosen to investigate the cows with extreme breeding values. Animals with high daily breeding values

for both traits show a lower disease rate for general – and metabolic disorders. Conversly, this means

that lower breeding values and increasing lactation number lead to an increasing disease rate.

General Discussion

95

The investigation of (Buttchereit et al., 2012) show that incidences for metablic disorders are high in

the first lactation. This is in agreement with Drackley (1999), who reported that metabolic disease

increased during this period. On contrary, Collard et al. (2000) indicate higher disease rates in

multiparous cows. This is consistent with the current study, in which it ould shown, that with

increasing lactation number the incidences of disorders also increased.

Furthermore, the differences between high and moderate categories of breeding values are lower

compared to moderate and low breeding value categories. Lower breeding values might have higher

impact on disease rate than high and moderate breeding values. This is observable in both, breeding

values for feed intake and energy balance.

As a conclusion, the first results of associations between average daily breeding values in the first

third of lactation suggest that general- and metabolic diseases could be reduced by selection for high

feed intake and energy balance.

Table 4. Metabolic (left) and general (right) disease rates of Holstein-Friesian cows dependent on

breeding value class for feed intake (days in milk 1 to 75) and lactation number (lno 1 to lno 4)

Breeding value class

for feed intake

Metabolic disease rate (%) general disease rate (%)

lno1 lno2 lno3 lno4 lno1 lno2 lno3 lno4

high (> 1 kg) 0.00 0.43 0.88 1.32 3.46 4.18 6.87 7.24

moderate (-1 to 1 kg) 0.92 1.12 3.40 6.02 5.92 6.29 9.41 12.8

Low (< -1 kg) 1.88 6.61 7.73 8.91 8.83 14.6 16.6 19.5

Number of animals in lactation number 1 (lno1): in total = 329; high = 34; moderate = 280; low = 15




General Discussion

96

Table 5. Metabolic (left) and general (right) disease rates of Holstein-Friesian cows dependent on

breeding value class for energy balance (days in milk 1 to 75) and lactation number (lno 1 to lno 4)

Breeding value class

for energy balance

metabolic disease rate (%) general disease rate (%)

lno1 lno2 lno3 lno4 lno1 lno2 lno3 lno4

high (> 7 MJ NEL) 1.02 0.64 1.98 2.22 4.29 4.54 7.05 11.7

moderate

(7 to -7 MJ NEL) 0.79 1.26 3.00 3.22 5.90 6.57 9.12 12.8

low (< -7 MJ NEL) 0.92 4.00 7.74 7.68 7.76 10.2 17.7 15.8





References

Banos, G., M. P. Coffey, R. F. Veerkamp, D. P. Berry, and E. Wall. 2012. Merging and characterising

phenotypic data on conventional and rare traits from dairy cattle experimental resources in three

countries. Animal an international journal of animal bioscience 6(7):1040–1048.

https://doi.org/10.1017/S1751731111002655.





Berry, D. P., and J. J. Crowley. 2013. Cell Biology Symposium: genetics of feed efficiency in dairy

and beef cattle. Journal of animal science 91(4):1594–1613. https://doi.org/10.2527/jas.2012-

5862.




Brade, W., and E. Brade. 2016. Die Futteraufnahme von hochleistenden Milchkühen im ersten

Laktationsdrittel aus der Sicht der Züchtung und des Tierwohles. 73 kB / Berichte über

General Discussion

97

Landwirtschaft - Zeitschrift für Agrarpolitik und Landwirtschaft, Band 94, Heft 2, August 2016.

https://doi.org/10.12767/BUEL.V94I2.111.G274.

Buttchereit, N. 2011. Model evaluation and estimation of genetic parameters for energy balance and

related traits in dairy cows. Schriftenreihe des Instituts für Tierzucht und Tierhaltung der

Christian-Albrechts-Universität zu Kiel 180. Selbstverl. des Inst. für Tierzucht und Tierhaltung

der Christian-Albrechts-Univ. zu Kiel, Kiel.

Buttchereit, N., E. Stamer, W. Junge, and G. Thaller. 2012. Genetic parameters for energy balance,

fat /protein ratio, body condition score and disease traits in German Holstein cows. Journal of

animal breeding and genetics = Zeitschrift fur Tierzuchtung und Zuchtungsbiologie 129(4):280–

288. https://doi.org/10.1111/j.1439-0388.2011.00976.x.

Calus, M. P. L., Y. de Haas, and R. F. Veerkamp. 2013. Combining cow and bull reference

populations to increase accuracy of genomic prediction and genome-wide association studies. J.





Collard, B. L., P. J. Boettcher, J.C.M. Dekkers, D. Petitclerc, and L. R. Schaeffer. 2000. Relationships

Between Energy Balance and Health Traits of Dairy Cattle in Early Lactation. J. Dairy Sci.

83(11):2683–2690. https://doi.org/10.3168/jds.S0022-0302(00)75162-9.



Dehnavi, E., S. A. Mahyari, F. S. Schenkel, and M. Sargolzaei. 2018. The effect of using cow genomic

information on accuracy and bias of genomic breeding values in a simulated Holstein dairy cattle

population. J. Dairy Sci. 101(6):5166–5176. https://doi.org/10.3168/jds.2017-12999.

Drackley, J. K. 1999. Biology of Dairy Cows During the Transition Period: The Final Frontier?

Journal of dairy science 82(11):2259–2273. https://doi.org/10.3168/jds.S0022-0302(99)75474-3.

Emmans, G. C. 1994. Effective energy: A concept of energy utilization applied across species. The

British journal of nutrition 71(6):801–821.

European Commission. 2013. Analysis of Milk Margins.

General Discussion

98

Goddard, M. E., and B. J. Hayes. 2009. Mapping genes for complex traits in domestic animals and

their use in breeding programmes. Nature reviews. Genetics 10(6):381–391.

https://doi.org/10.1038/nrg2575.

Gonzalez-Recio, O., M. P. Coffey, and J. E. Pryce. 2014. On the value of the phenotypes in the

genomic era. J. Dairy Sci. 97(12):7905–7915. https://doi.org/10.3168/jds.2014-8125.

Haas, Y. de, M. P. L. Calus, R. F. Veerkamp, E. Wall, M. P. Coffey, H. D. Daetwyler, B. J. Hayes,

and J. E. Pryce. 2012. Improved accuracy of genomic prediction for dry matter intake of dairy

cattle from combined European and Australian data sets. J. Dairy Sci. 95(10):6103–6112.

https://doi.org/10.3168/jds.2011-5280.





https://doi.org/10.3168/jds.2014-9257.

Habier, D., J. Tetens, F.-R. Seefried, P. Lichtner, and G. Thaller. 2010. The impact of genetic

relationship information on genomic breeding values in German Holstein cattle. Genetics,

selection, evolution GSE 42:5. https://doi.org/10.1186/1297-9686-42-5.

Illumina Inc. 2005. Illumina GenCall Data Analysis Software.

Krattenmacher, N., G. Thaller, and J. Tetens. 2019. Analysis of the genetic architecture of energy

balance and its major determinants dry matter intake and energy-corrected milk yield in

primiparous Holstein cows. J. Dairy Sci. https://doi.org/10.3168/jds.2018-15480.

Legarra, A., O. F. Christensen, I. Aguilar, and I. Misztal. 2014. Single Step, a general approach for

genomic selection. Livestock Science 166:54–65. https://doi.org/10.1016/j.livsci.2014.04.029.




https://doi.org/10.3168/jds.2018-14611.




General Discussion

99

Macciotta, N. P. P., G. Gaspa, L. Bomba, D. Vicario, C. Dimauro, M. Cellesi, and P. Ajmone-Marsan.

2015. Genome-wide association analysis in Italian Simmental cows for lactation curve traits using

a low-density (7K) SNP panel. J. Dairy Sci. 98(11):8175–8185. https://doi.org/10.3168/jds.2015-

9500.

Madsen, P., J. Jensen, Labouriau R., L. G. Christensen, and G. Sahana. 2013. DMU - A Package for


Mäntysaari, P., A.-E. Liinamo, and E. A. Mäntysaari. 2012. Energy efficiency and its relationship

with milk, body, and intake traits and energy status among primiparous Nordic Red dairy cattle.





https://doi.org/10.3168/jds.2014-8165.

Manzanilla-Pech, C. I. V., R. F. Veerkamp, Y. de Haas, M. P. L. Calus, and J. ten Napel. 2017.

Accuracies of breeding values for dry matter intake using nongenotyped animals and predictor

traits in different lactations. J. Dairy Sci. 100(11):9103–9114. https://doi.org/10.3168/jds.2017-

12741.

Misztal, I., S. Tsuruta, T. Strabel, B. Auvray, T. Druet, and D. H. Lee. 2002. BLUPF90 and related

programs (BGF90). http://nce.ads.uga.edu/~ignacy.

Muir, W. M. 2007. Comparison of genomic and traditional BLUP-estimated breeding value accuracy

and selection response under alternative trait and genomic parameters. Journal of animal breeding

and genetics = Zeitschrift fur Tierzuchtung und Zuchtungsbiologie 124(6):342–355.

https://doi.org/10.1111/j.1439-0388.2007.00700.x.

Persaud, P., G. Simm, and W. G. Hlll. 1991. Genetic and phenotypic parameters for yield, food intake

and efficiency of dairy cows fed ad libitum 1. Estimates for ‘total’ lactation measures and their

relationship with live-weight traits. Anim. Prod. 52(03):435–444.

https://doi.org/10.1017/S0003356100013003.

Pryce, J. E., and D. P. Berry. 2014. Implementation of genomic breeding values for novel traits such

as feed efficiency through female nucleus reference populations. ICAR.

General Discussion

100

Pszczola, M., T. Strabel, H. A. Mulder, and M. P. L. Calus. 2012. Reliability of direct genomic values

for animals with different relationships within and to the reference population. J. Dairy Sci.

95(1):389–400. https://doi.org/10.3168/jds.2011-4338.

Pszczola, M., R. F. Veerkamp, Y. de Haas, E. Wall, T. Strabel, and M. P. L. Calus. 2013. Effect of

predictor traits on accuracy of genomic breeding values for feed intake based on a limited cow

reference population. Animal 7(11):1759–1768. https://doi.org/10.1017/S175173111300150X.

Shabalina, T., E. C. G. Pimentel, C. Edel, L. Plieschke, R. Emmerling, and K.-U. Götz. 2017. Short

communication: The role of genotypes from animals without phenotypes in single-step genomic

evaluations. J. Dairy Sci. 100(10):8277–8281. https://doi.org/10.3168/jds.2017-12734.

Shonka, B. N., S. Tao, G. E. Dahl, and D. M. Spurlock. 2015. Genetic regulation of prepartum dry

matter intake in Holstein cows. J. Dairy Sci. 98(11):8195–8200. https://doi.org/10.3168/jds.2015-

9675.

Søndergaard, E., M. K. Sørensen, I. L. Mao, and J. Jensen. 2002. Genetic parameters of production,

feed intake, body weight, body composition, and udder health in lactating dairy cows. Livestock

Production Science 77(1):23–34. https://doi.org/10.1016/S0301-6226(02)00023-4.



5402. https://doi.org/10.3168/jds.2012-5407.



https://doi.org/10.3168/jds.2013-7301.

Weller, J. I. 2016. Genomic selection in animals. John Wiley & Sons Inc, Hoboken, NJ.

Outlook and ongoing research

101


The project “optiKuh” is a first step to evaluate genetic properties of feed intake and energy balance

to integrate these in the breeding goal of high performing dairy cows. With this unique data set,

satisfactory results for both traits could be evaluated. With the above data set it was possible to

estimate heritabilities, genetic correlations, variance components and finally, breeding value

reliabilities.

However, further analyses are essential to support the findings of the current study and to approve

the present results in the literature. Additionally, possible side effects of selection need to be identified

and excluded, if necessary.

For instance, genetic correlations to other breeding goals and health traits needs to be analyzed. In

the literature, feed intake has been recently reported to have positive genetic correlations with energy

balance, body condition score and conformation traits (Vallimont et al., 2011; Liinamo et al., 2012;

Manzanilla-Pech et al., 2016). The data of the project “optiKuh” can be used for further investigations

about proportionalities between feed intake, energy balance and milk yield.

To date, only little information about reliable correlations between those relevant traits and health

have been evaluated. The project provides metabolic parameters to assess the metabolic status of the

cow. Thereby it would be possible to determine the nearly exact physiology status in course of

lactation of the high performing dairy cows. Specific statements according to the actual energy status

would be feasible. Furthermore, diseases were documented within the project “optiKuh”. Thus,

appropriate data is available, to investigate relationships between health, energy balance and feed

intake.

Initial results could be achieved and associations between daily breeding values for feed intake,

energy balance and health could be investigated. Furthermore, high breeding values for feed intake

and energy balance can be associated with low disease rate and vice versa.

The data set provides information to conduct a multivariate computation basically, but for practice

application, an extension of the reference population would be necessary. Furthermore, the genetic

correlations for feed intake and energy balance between consecutive lactations should be investigated.

Moreover, the aspect should be taken into account that increased feed intake does not necessarily

have a positive effect e.g. on the health or milk production of dairy cows, but could have possibly

results in increased, extreme body condition score or unwanted size growth.

Genetic correlation between feed intake and body weight was found to be positive (Buttchereit, 2011)

and in the recent years, animals have increased in size. This means that the size of the animals should


102

be integrated as an auxiliary trait in the selection strategy or, alternatively, the body conditions score

should be taken into account as a correction variable. Nevertheless, “optiKuh” has limitations in the

data set, because only in few research farms, the body condition score was determined.

With the current reference population, adequate reliabilities for the traits feed intake and energy

balance could be achieved. To create higher reliabilities for the trait feed intake, the reference

population needs to be enlarged. This applies especially for the breed Simmental Cattle.

In the current study, only congruent SNPs of all three data sets were used and may cause a loss of

information. A possibility to decrease the loss of information and to increase the accuracy of breeding

values is to impute the genotypes (Pryce et al., 2014). Nevertheless, the potential and the influence

of the results of imputing could not be pre-estimate but a gain might be possible.

With the method single-step, high reliabilities could be estimated for animals which were involved in

the current project but had no phenotypic information. This was probably due to a genetically low

distance to the cows in the “optiKuh” reference population. Conversely, it is presumed that possible

candidates, which are genetically apart, would have lower reliabilities. This needs to be analyzed in

further investigations.

Besides the trait feed intake, further traits in this context are of interest. For example, residual feed

intake is a focus in many studies. It is defined as the difference between animal’s actual and estimated

feed intake, based on energy requirements for production and maintenance (Potts et al., 2017). This

concept is quite similar to residual energy intake calculations, but is derived on the basis of energy

intake of the feed instead of feed intake. Cows with a low residual feed intake are considered as more

efficient animals. However, feed efficiency is important, especially from the economic point of view.

Thus, these traits need to be analyzed and should be set in the connection of the trait feed intake.

The final question should be answered, which trait should be implemented in the breeding goal.

A transmission of the system in a routine farm should be started in future, with the aim to use all

prospective data recording of research farms in connection with genotyping for required estimations.

All in all, to gain further knowledge about the traits feed intake and energy balance, the reference

population needs to be expanded with more genotypes and phenotypes.

Therefore, the follow up project “eMissionCow” has been funded to increase the number of cows in

the reference population and to use the data from “optiKuh” for ongoing research.

References

Buttchereit, N. 2011. Model evaluation and estimation of genetic parameters for energy balance and

related traits in dairy cows. Schriftenreihe des Instituts für Tierzucht und Tierhaltung der


103

Christian-Albrechts-Universität zu Kiel 180. Selbstverl. des Inst. für Tierzucht und Tierhaltung

der Christian-Albrechts-Univ. zu Kiel, Kiel.







https://doi.org/10.3168/jds.2014-8165.

Manzanilla-Pech, C. I. V., R. F. Veerkamp, R. J. Tempelman, M. L. van Pelt, K. A. Weigel, M.

VandeHaar, T. J. Lawlor, D. M. Spurlock, L. E. Armentano, C. R. Staples, M. Hanigan, and Y.

de Haas. 2016. Genetic parameters between feed-intake-related traits and conformation in 2

separate dairy populations--the Netherlands and United States. J. Dairy Sci. 99(1):443–457.

https://doi.org/10.3168/jds.2015-9727.

Potts, S. B., J. P. Boerman, A. L. Lock, M. S. Allen, and M. J. Vandehaar. 2017. Relationship between

residual feed intake and digestibility for lactating Holstein cows fed high and low starch diets. J.


Pryce, J. E., J. Johnston, B. J. Hayes, G. Sahana, K. A. Weigel, S. McParland, D. Spurlock, N.

Krattenmacher, R. J. Spelman, E. Wall, and M. P. L. Calus. 2014. Imputation of genotypes from

low density (50,000 markers) to high density (700,000 markers) of cows from research herds in

Europe, North America, and Australasia using 2 reference populations. J. Dairy Sci. 97(3):1799–

1811. https://doi.org/10.3168/jds.2013-7368.

Vallimont, J. E., C. D. Dechow, J. M. Daubert, M. W. Dekleva, J. W. Blum, C. M. Barlieb, W. Liu,

G. A. Varga, A. J. Heinrichs, and C. R. Baumrucker. 2011. Heritability of gross feed efficiency

and associations with yield, intake, residual intake, body weight, and body condition score in 11

commercial Pennsylvania tie stalls. J. Dairy Sci. 94(4):2108–2113.

https://doi.org/10.3168/jds.2010-3888.


104

General Summary

105

General Summary

The inclusion of feed intake as a breeding trait in dairy cows is useful for various reasons. On the one

hand, feeding has a large share of costs for livestock production. On the other hand, feed intake has

an impact on energy balance at the beginning of lactation and thus it is of vital importance for health

and fertility of dairy cows. At the beginning of lactation feed intake increases slower, compared to

the fast increasing milk yield. Body reserves are used for the high milk production. This results in an

energy deficit, thus the risk for metabolic diseases increase.

With upcoming interest in animal welfare with regard to health and longevity of the animals, this trait

gets in focus in breeding. If feed intake could be increased, especially at the beginning of lactation,

the negative energy balance could be balanced to improve the metabolic stability of the cows.

Aim of the present study was, to evaluate genomic breeding values to integrate them in long- term

objective in the breeding goal of high performing dairy cows.

In chapter I the data for the both traits feed intake and energy balance of the project “optiKuh” were

descriptively prepared to be used for further analyses. To date, breeding measures were difficult due

to expensive data recording and are currently only implemented on few research farms.

With the project “optiKuh” a unique data set was created to measure feed intake and to use the data

for a genomic selection, subsequently.

Weekly records were chosen due to different measurement frequencies (e.g. daily FI and weekly milk

ingredients) of the traits and observations outside the range of ± 4 standard deviations are excluded

from further analyses.

After data editing, 1,341 Holstein-Friesian with 40,012 weekly records and a mean feed intake of

21.8 ± 4.3 kg/d are used. For the trait energy balance 1.322 cows and 33.376 weekly records were

available with a mean of 3,2 ± 29,4 MJ NEL.

The breed Simmental Cattle provided totally 327 cows with 16,996 weekly averages and a mean feed

intake of 20.2 ± 3,6 kg/d. For the energy balance 326 animals with 14,527 records and a mean of 1.1

± 18,9 MJ NEL were available.

Different models of lactation were tested and the most suitable was chosen based on the convergence

criteria AICC and BIC. The resulting random regression animal model included the fixed effects of

herd test week alternatively herd group test, parity and days in milk from 5 to 350. Lactation curves

were modeled by the function of Ali and Schaeffer, and for both the random permanent environmental

General Summary

106

effect and the random additive genetic effect, third-order Legendre polynomials were chosen.

Repeatability was high, ranging between 0.6 and 0.8.

In Chapter II the preparations of the analyses within breed were carried out to plausible, edit, and

contrast the genomic and pedigree-based relationship matrices. This was done due to already chosen

linear evaluation models for the final data sets.

For the estimation of the genomic and the conventional breeding values for the traits, feed intake and

energy balance, 1,341 Holstein-Friesian cows were available.

The direct and precise animal individual feed intake data, as well as the derived energy balance traits

created the basis of the estimation of variance components and breeding value estimation.

Simultaneously, genomic marker information were included.

Furthermore, 1,128 Holstein-Friesian and 232 Simmental Cattle with a phenotype profile were

additionally genotyped with the Illumina BovineSNP50 Bead Chip. 35 other Holstein-Friesian could

be genotyped but not phenotyped.

After a quality control of the genotyping results, 43,455 SNPs (HF) were used. With the program

DMU analyses of genetic parameters were realized. For two different data sets, first the pedigree-

based data set and second the combined data set, containing both pedigree and genotype information

(“single-step”), variance components and breeding values were estimated.

The computation of the pedigree-based relationship coefficients were done within breed under

consideration of four ancestor generations. Based on the limited number of animals the equation

system did not converge and breeding values for breed Simmental Cattle could not be estimated.

Based on homogenous feed intake data and sufficient animal numbers, a meaningful reference

population for the computation of variance components and breeding values could be used. The

heritabilities in course of lactation for feed intake range between 0.12 and 0.50. In comparison, the

energy balance heritabilities are lower with values, ranging between 0.15 and 0.48.

The genetic correlations of following lactation segments between the beginning and the end of

lactation showed low correlations for both traits with values at 0.05 and -0.05, respectively. It could

be assumed, that different traits are participated in gene expression. To select on these traits, these

circumstances should be considered.

Totally, the genomic reliabilities range between 0.33 and 0.61. The higher reliabilities are due to more

weekly records. For the 35 Holstein-Friesian female candidates, animals with a geno- but not

phenotyped information, reliabilities of 23% (feed intake) and 17% (energy balance) can be realized.

General Summary

107

Towards the variant without genomic information, the reliability increased at nearly 10% (feed

intake) and 6% (energy balance), respectively.

In chapter III the results for feed intake and energy balance are summarized and discussed in the

context of breeding goals. The strong relation between the traits feed intake to milk yield, energy

balance, health and fertility is an important parameter for breeding on healthy and long-living cows

and therefore the well-being of the animals. Therefore, feed intake should be increased at the

beginning of lactation. With an improved energy balance due to a higher feed intake, healthier cows

can be breed.

The closely related feed efficiency, which is the subject of many studies due to the economic interest,

should be regarded critically, especially at the beginning of lactation. Thus, it would be appropriate,

if the cows were breed on a high feed intake at the beginning of lactation whereas the efficiency

should be the focus in the middle and end of lactation and the cow can start the dry period in a good

balanced manner.

General Summary

108

Zusammenfassung

109

Zusammenfassung

Die Berücksichtigung der Futteraufnahme als Zuchtzielmerkmal bei Milchkühen ist aus

verschiedenen Gründen sinnvoll. Zum einen ist das Futter zu einem großen Anteil an den Kosten in

der Milchproduktion verantwortlich und zum anderen wirkt sich die Futteraufnahme direkt auf die

Energiebilanz in der Frühlaktation aus und ist somit von entscheidender Bedeutung für die

Gesundheit und Fruchtbarkeit. Denn zu Beginn der Laktation steigt die Futteraufnahme im Vergleich

zur rasch ansteigenden Milchleistung vergleichsweise langsam an. Dies kann teils erhebliche

Energiedefizite zur Folge haben. Körperreserven, die während eines Laktationsverlaufs aufgebaut

wurden, werden durch Einschmelzen von Körperfett zusätzlich für die Milchproduktion verwendet.

Dadurch kommt es zu einem steigenden Risiko für Stoffwechselerkrankungen, wie zum Beispiel

Azidose, Ketose oder Milchfieber. Aufgrund des wachsenden Interesses am Tierwohl in Bezug auf

Gesundheit und Langlebigkeit tritt die Futteraufnahme als Merkmal immer stärker in den

Vordergrund. Durch eine Erhöhung der Futtermenge, insbesondere zu Laktationsbeginn, kann das

Energiedefizit vermindert werden um die Stoffwechselstabilität der Hochleistungsmilchkühe zu

verbessern. Ziel der vorliegenden Arbeit war es, für das Merkmal Futteraufnahme genomische

Zuchtwerte zu entwickeln, um diese langfristig in das Zuchtziel der Hochleistungskühe zu

integrieren.

In Kapitel I wurden zunächst die Daten für die beiden zu untersuchenden Merkmale Futteraufnahme

und Energiebilanz aus dem „optiKuh“-Projekt deskriptiv aufbereitet, damit diese für weitere

Analysen genutzt werden können. Bisher standen züchterischen Maßnahmen die teure und

aufwendige Datenerfassung im Weg, wodurch eine routinemäßige Erfassung der

Grundfutteraufnahme lediglich auf wenigen Testbetrieben möglich war.

Mit dem Projekt „optiKuh“ wurde ein einzigartiger Datensatz zur Messung der Futteraufnahme und

darauf aufbauend zur genomischen Selektion für diese Merkmale geschaffen.

Aufgrund unterschiedlicher Erfassungsfrequenzen der Merkmale (zum Beispiel tägliche Messung der

Futteraufnahme aber nur wöchentliche Messung der Milchinhaltsstoffe) wurden Wochenmittelwerte

als Grundlage für die Berechnungen gewählt. Beobachtungen außerhalb des Bereichs ± 4

Standardabweichungen wurden von weiteren Auswertungen ausgeschlossen. Nach der

Datenbereinigung standen für die Rasse Holstein-Friesian 1.341 Kühe zur Verfügung. Für die

Futteraufnahme konnten insgesamt 40.012 wöchentlichen Daten mit einer mittleren Futteraufnahme

von 21,8 ± 4,3 kg/Tag verwendet werden. Die Energiebilanz wies 1.322 Kühe mit 33.376

Zusammenfassung

110

wöchentlichen Daten und einer mittleren Energiebilanz von 3,2 ± 29,4 MJ NEL auf. Für die Rasse

Fleckvieh standen insgesamt 327 Kühe mit 16.996 Wochenmittelwerten und einer mittleren

Futteraufnahme von 20,2 ± 3,6 kg/Tag zur Verfügung. Die Energiebilanz konnte von 326 Tieren mit

14.527 Werten genutzt werden, welche eine mittlere Energiebilanz von 1,1 ± 18,9 MJ NEL

aufwiesen.

Es wurden verschiedene Laktationsmodellierungen geprüft und anhand von den Konvergenzkriterien

AICC und BIC das am besten geeignete ausgesucht. Das gewählte Random Regression Tiermodell

beinhaltete die fixen Effekte Herdentestwoche bzw. Herdentestwochengruppe und

Laktationsnummer, sowie für die Modellierung der allgemeinen Laktationskurve die Funktion von

Ali und Schaeffer. Für den zufälligen permanenten Umwelteffekt und den zufälligen additiv

genetischen Tiereffekt wurden jeweils das Legendre Polynom 3. Grades gewählt. Die ermittelten

Wiederholbarkeiten waren hoch und lagen in einem Bereich zwischen 0,6 und 0,8.

In Kapitel II erfolgten in Vorbereitung auf diese Analysen jeweils innerhalb der Rasse die

Plausibilisierung, Aufbereitung und Gegenüberstellung der genomischen und pedigree-basierten

Verwandtschaftsmatrizen. Dies fand auf Basis bereits erstellter finaler Auswertungsdatensätze und

der Entwicklung linearer Modelle statt.

Zur Schätzung der genomischen und konventionellen Zuchtwerte der beiden Merkmale

Futteraufnahme und Energiebilanz standen insgesamt 1.163 Holstein Friesian Kühe aus dem Projekt

„optiKuh“ zur Verfügung.

Die direkt und damit genau erfassten tierindividuellen Futteraufnahmen sowie die abgeleiteten

Energiebilanzen bildeten die Grundlage für eine Varianzkomponenten - und Zuchtwertschätzung -

unter gleichzeitiger Einbeziehung genomischer Markerinformationen.

Außerdem wurden 1.128 Holstein-Friesian und 232 Fleckvieh Kühe mit einem phänotypischen Profil

zusätzlich mit dem Illumina BovineSNP50 Bead Chip genotypisiert. 35 weitere Holstein-Friesian-

Kühe konnten lediglich genotypisiert aber nicht phänotypisiert werden.

Nach einer gängigen Qualitätskontrolle der Genotypisierungsergebnisse konnten insgesamt 43.455

SNPs (Holstein-Friesian) genutzt werden. Mithilfe des Programms DMU wurde eine Analyse der

genetischen Parameter durchgeführt. Die Varianzkomponenten- und Zuchtwertschätzungen erfolgten

anhand zweier Datensätze – zum einen ein pedigree-basierter Datensatz und zum anderen ein

kombinierter Datensatz aus Pedigree- und Genotypinformation („single-step“).

Zusammenfassung

111

Die Berechnung der pedigree-basierten Verwandtschaftskoeffizienten erfolgte innerhalb Rasse unter

Berücksichtigung von vier Ahnengeneration. Aufgrund geringer Tierzahlen für die Rasse Fleckvieh

konvergierte das Gleichungssystem nicht und es konnten keine Zuchtwerte geschätzt werden.

Mit Hilfe einheitlicher Futteraufnahmemessungen und ausreichenden Tieranzahlen konnte eine

aussagekräftige Lernstichprobe zur Berechnung von Varianzkomponenten und Zuchtwerten erstellt

werden. Die Heritabilitäten über den Laktationsverlauf für die Rasse Holstein-Friesian für die

Futteraufnahme lagen in einem Bereich zwischen 0,12 und 0,50, während die Werte für die

Energiebilanz im Vergleich etwas niedriger zwischen 0,15 und 0,48 lagen.

Die genetischen Korrelationen zu Beginn der Laktation im Vergleich zum Ende der Laktation zeigen

niedrige Korrelationen von 0,05 für Futteraufnahme und -0,05 für die Energiebilanz.

Insgesamt rangierten die genomischen Genauigkeiten zwischen 0,33 und 0,61, wobei die höheren

Genauigkeiten aufgrund von mehr Beobachtungen zustande kamen.

Mit der anschließend durchgeführten Zuchtwertschätzung für weibliche Kandidaten können für die

nicht phänotypisierten 35 Tiere durch die Einbeziehung von genomischen Verwandtschaft

Sicherheiten von 23% (Futteraufnahme) und 17% (Energiebilanz) erzielt werden. Gegenüber der

Variante ohne genomische Verwandtschaft, d.h. rein pedigree-basiert, erhöht sich die Sicherheit

damit um nahezu 8,6% (Futteraufnahme) bzw. 4,2% (Energiebilanz).

In Kapitel III erfolgte eine Zusammenstellung der Ergebnisse und eine Diskussion über das Zuchtziel.

Durch die enge Beziehung zwischen dem Merkmal Futteraufnahme zu Milchleistung, Energiebilanz,

Gesundheit und Fruchtbarkeit, ist es ein bedeutsamer Parameter für die Zucht auf langlebige und

robuste Kühe und damit vorteilhaft für das Wohlbefinden der Hochleistungstiere. Aus diesem Grund

sollte die Futteraufnahme zu Beginn der Laktation erhöht werden um die Energiebilanz und damit

die Gesundheit der Hochleistungsmilchkühe zu verbessern. Die damit in engem Zusammenhang

stehende Futtereffizienz, die aufgrund des wirtschaftlichen Interesses Gegenstand vieler

Untersuchungen ist, ist insbesondere zu Beginn der Laktation kritisch zu betrachten. Sinnvoll wäre

es, wenn zu Laktationsbeginn auf eine hohe Futteraufnahme gezüchtet wird und im Laufe der

Laktation die Effizienz in den Vordergrund rückt, sodass die Milchkuh mit einer guten Bilanz in die

erneute Laktation startet.

Zusammenfassung

112

113

Danksagung

An dieser Stelle möchte ich mich bei allen bedanken, die maßgeblich an dem Gelingen dieser Arbeit

beteiligt waren.

Meinem Doktorvater Herrn Prof. Thaller möchte mich für die Überlassung des interessanten Themas,

die wissenschaftliche Betreuung und die Möglichkeiten, mein Thema auf Tagungen präsentieren zu

können, bedanken.

Außerdem bedanke ich mich für die Förderung des Vorhabens, welches aus den Mitteln des

Bundesministeriums für Ernährung und Landwirtschaft (BMEL) aufgrund eine Beschlusses des

deutschen Bundestages erfolgte. Die Projektträgerschaft erfolgte über die Bundesanstalt für

Landwirtschaft und Ernährung (BLE) im Rahmen des Programms zur Innovationsförderung.

Zusätzlich danke ich der H. Wilhelm Schaumann Stiftung für die finanzielle Unterstützung in Form

eines Stipendiums.

Ganz besonders bedanken möchte ich mich bei Herrn Dr. Stamer für eine tolle Betreuung und eine

konstante, herzliche, geduldige und motivierende Unterstützung, die wesentlich zu dem Gelingen

dieser Arbeit beigetragen haben.

Außerdem danke ich Frau Dr. Krattenmacher für den tollen Start in die Promotion und eine tolle

Unterstützung in Form von Korrekturlesen oder sehr hilfreichen Anregungen.

Herrn Dr. Wolfgang Junge, Hans-Otto, allen anderen Kollegen und Hiwis danke ich für die gute und

freundliche Zusammenarbeit während der Probennahme auf dem Versuchsgut Karkendamm.

Ich danke allen Kollegen für die schöne Zeit im Institut, die netten Ausflüge und die lustigen

feierabendlichen Aktivitäten.

Mein besonderer Dank geht an Lilian Gehrke, die mir über die Zeit sehr ans Herz gewachsen ist und

ohne die ich mich nicht jeden Tag so sehr aufs Büro gefreut hätte. Die täglichen Gespräche und

psychologischen Tipps haben für sehr viel Spaß am Arbeitsplatz gesorgt. Aber auch die privaten

Feierabendstunden werden mir sehr fehlen.

114

Darüber hinaus danke ich all meinen Freunden außerhalb des Instituts, die mich während der

gesamten Zeit unterstützt haben.

Aber vor allem möchte ich meiner Familie danken, ohne die diese Promotion niemals hätte stattfinden

können. Ich danke euch dafür, dass ihr immer für mich da seid und mich einfach in jeder Lebenslage

unterstützt. Und dir kleine Schwester danke ich besonders, dass du mich jeden Tag, mitgefiebert,

mitgelacht und mich teilweise ausgehalten hast. Ich möchte euch danken für die vielen aufbauenden

Worte und die schöne Zeit, die ihr mir dadurch ermöglich habt. Ich bin froh, dass ich euch habe.

115

Lebenslauf

Name: Imke Harder

Geburtsdatum: 16. November 1988

Geburtsort: Bad Segeberg

Familienstand: ledig

Staatsangehörigkeit: deutsch

Schulische Ausbildung

1995 – 1996 Grundschule in Schleswig

1996 – 1999 Grundschule in Bad Segeberg

1999 – 2008 Städtisches Gymnasium in Bad Segeberg

Abschluss: Abitur

Studium

2008 – 2015 Studium der Agrarwissenschaften an der Christian-Albrechts-Universität

zu Kiel,

Fachrichtung Nutztierwissenschaften

Abschluss: Master of Science

Praktika

Januar 2013

– März 2013 Auslandsaufenthalt in Kanada, Vancouver Island

Aushilfskraft auf Farmen

Berufliche Tätigkeit

116

2014 – 2015 Wissenschaftliche Hilfskraft am Institut für Tierernährung und

Stoffwechselphysiologie der Christian-Albrechts-Universität zu Kiel

seit April 2015 Wissenschaftliche Mitarbeiterin am Institut für Tierzucht und

Tierhaltung der Christian-Albrechts-Universität zu Kiel

bei Prof. Dr. Georg Thaller

Projekt „optiKuh“

· Dissertation zur Erlangung des Doktorgrades der Agrar- und Ernährungswissenschaftlichen...

Documents

Transcript of · Dissertation zur Erlangung des Doktorgrades der Agrar- und Ernährungswissenschaftlichen...