Skip to main content

Innovative and Practical Breeding Tools for Improved Dairy Products from More Robust Dairy Cattle

Final Report Summary - ROBUSTMILK (Innovative and Practical Breeding Tools for Improved Dairy Products from More Robust Dairy Cattle)


Executive Summary:

Since the mid-nineties most leading EU dairy cattle breeding programs have expanded their breeding goals to include health and fertility traits in addition to milk production. The ROBUSTMILK project was designed to develop new practical technologies to allow breeders to re-focus their selection to include milk quality and dairy cow robustness and to evaluate the consequences of selection for these traits taking cognisance of various milk production systems. Six organisations were involved: Wageningen Livestock Research (The Netherlands), SRUC (Scotland), Teagasc Moorepark (Ireland), Gembloux Agricultural University (Belgium), SLU (Sweden) and Wageningen University (The Netherlands).

The project produced more than 39 peer-reviewed papers, 89 conference presentations and 45 other dissemination activities. The project was presented at 35 conferences and during the 2011 Interbull Meeting (Stavanger, Norway) two sessions were dedicated to animal robustness and milk quality. Also attached to the consortium meeting, at the 2012 EAAP meeting in Bratislava 18 papers were given by the RobustMilk team. Finally, a special issue of the Journal Advances in Animal Biosciences will published with 6 papers as final summary of the project.

The first step was to create an international database that includes unique and scarcely recorded phenotypic measurements ( feed intake, body condition scoring and detailed health and fertility recordings), as well as the genotypes. The first genome wide association analyses unravelling the genetics of energy balance, feed intake, fertility and milk quality were published within the project, but impact goes beyond the project. The database created the setting for several across continent genetics and genomics projects.

The RobustMilk team has been very active in developing new methods of accurately and routinely predicting milk quality at little or no extra cost using the infrared spectroscopy of milk machine used in milk laboratories. The utilisation of these equations has been initiated in the UK, Ireland and Belgium. Also ROBUSTMILK developed the first predictions for energy balance from mid-infrared spectra. This might become an important route for including energy balance in our breeding objectives.

RobustMilk focused on environmental sensitivity, and have developed models that can estimate if the offspring from some sires are more sensitive than the offspring from other sires. Furthermore, we have developed new traits using as opposed to the current lactation average SCC. These statistical tools have, potentially, huge implications for how we genetically evaluate SCC and could result in greater genetic gain for udder health.

Genomic selection is now considered to be the optimal method of genetic evaluation in international dairy cattle populations. Progress in genomic selection for some traits is hampered by access to sufficiently large datasets with phenotypes. ROBUSTMILK has answered the potential of genomically selecting for difficult to measure traits (e.g. feed intake) but has also proven that by combining information from multiple traits the accuracy obtained using genomic selection can be increased thereby increasing genetic gain.

Take home messages

• More robust cows can be selected to produce milk that is more healthy for humans using readily available national data and infrared spectroscopy of milk.
• Genomic and statistical tools will further aid selection for unique traits like feed intake and fertility,
• “In the age of the genotype, the phenotype is king” and therefore sharing phenotypes od scarcely recorded traits has proven to be a major step forward.

Project Context and Objectives:

Robust cows, healthier milk

Our European cows produce a lot of milk thanks to years of breeding and good craftsmanship of our farmers. However, our cows are considered less ‘robust’ now than they have been. Simultaneously, our aim is to produce milk that adds greater nutritional value to our diet. Thus the challenge for animal breeding is to deal with both the robustness of dairy cows and making their milk healthier for humans simultaneously.

What is ROBUSTMILK?

ROBUSTMILK is a project that has been funded through the EU Framework 7 Programme to join together six world-leading research organisations within EU that are actively working in dairy cattle breeding and have strong links with the dairy industry. It is called Innovative and Practical Breeding Tools for Improved Dairy Products from More Robust Dairy Cattle abbreviated to ROBUSTMILK for ease of use (and speed!).

Who is involved?

There are six organisations involved in the project and all have a strong background in dairy cattle breeding. They are all well respected in their own countries and each has a strong reputation for ensuring that research and innovation is disseminated quickly to industry. These organisations are:

• Animal Science Group, Lelystad (Netherlands)
• The Scottish Agricultural College (SAC Scotland)
• Teagasc Moorepark (Ireland)
• Gembloux Agricultural University (Belgium)
• Swedish University of Agricultural Science (Sweden)
• Wageningen University (Netherlands)

What will ROBUSTMILK do?

The objective of ROBUSTMILK is to develop new, useful and practical technologies to allow dairy farmers and the dairy industry to refocus their selection decisions to include additional traits such as milk quality and dairy cow robustness. It is of utmost importance that farmers can evaluate the consequences of selection for these novel and additional traits within their own milk production systems. Likewise, it is important that the inclusion of traits such as milk quality does not compromise health, fertility, 'robustness', or in other words the profitability of the cow. We seek the win-win situation where dairy cow milk is healthy for humans and producing it is also healthy for the cow.

The overall objective of ROBUSTMILK will be achieved by having five integrated work packages each having their own objective:

• The creation of a common database across country partners that includes unique and scarcely recorded measurements for traits underlying individual dairy cow robustness and milk quality. These traits include measures such as feed intake, regular body condition scoring and detailed health and fertility recordings. These databases are held at each of the research partners involved and the first thing to do is to create a framework that enables bringing that data together to make it useable by this and future projects.
• To develop tools for easily and cheaply measuring dairy cow robustness (energy balance) and milk quality (lactoferrin and fatty acid composition). The tool pursed in ROBUSTMILK is mid- infrared spectrometry which is a methodology used to determine the fat, protein and lactose concentration of all milk samples from milk recorded herds internationally. Preliminary analyses indicated that equations could be developed using this methodology to predict milk fatty acid content; the objective of this task is to strengthen these calibration equations and evaluate whether they can also be used to predict dairy cow robustness.
• A robust cow maintains good milk quality (e.g. low levels of somatic cell count, SCC) over a wide range of environments and also throughout her life. In this work package we will develop statistical tools to select for both types of robustness, with special emphasis on SCC. Increased SCC is an indicator of both compromised udder health and lowered milk quality - therefore decreasing SCC is an example of a win-win situation.
• To develop genomic tools for selection for robustness and milk quality traits. The merged data from all of the partners will be accompanied by DNA from each of the animals so that new genomic technologies can be used to identify which DNA profiles are associated with 'good' milk and which with 'bad' milk and similarly can be used to differentiate between genes for high or low 'robustness'. Once identified, these markers can then be included in genetic evaluations thereby providing more power to individual farmers when making selection decisions. Such an objective cannot be achieved without strong and open collaboration among several research groups with access and willingness to share their individual cow records and DNA samples.
• Integrate and disseminate knowledge on the consequences of selection practices on robustness and milk quality. ROBUSTMILK has the potential to enhance the competitiveness of European agriculture through the production of higher quality dairy products and more sustainable dairy production systems.

ROBUSTMILK will contribute significantly towards the Knowledge Based Bio Economy objective of the EU, through a greater understanding of factors contributing to genetic variation and exploiting this variation in a sustainable manner in genetic improvement programmes. Research findings will be updated regularly at the ROBUSTMILK website (www.robustmilk.eu/)

What will be produced?

Primarily, knowledge and tools will be generated but of course when these are applied, through the strong links between this research group and industry, the ultimate beneficiaries will be EU consumers and farmers. For example, once we know the genetic profile of cows that have improved robustness and produce more healthy milk, then we can include this in selection programmes, farmers can choose better bulls to increase the profitability of their herd, and society will benefit from healthier food being produced by better cows. These cows will require less treatment for disease, will live longer and consequently dairy production from them will have less impact on the environment.

How will it benefit farmers and society?

Ultimately, when the research on milk quality is applied in a systematic manner within the food chain, healthier milk can be placed on supermarket shelves. At the same time we will know how to do this without compromising the health of the cow and so cows (and farmers) will benefit. Indeed, irrespective of the human health aspects of milk, this project will lead to cows that are more 'robust' which translates into cows that can produce large amounts of milk economically for the farmer, that can remain healthy in so doing and can live a long time without the need for veterinary intervention or drug usage. This benefits both farmers and society but above all, benefits cows and the environment, because herds with cows that live longer have an overall lower impact on the environment.

Project Results:

In this report a more popular summary of the research results is given. For further reading with the scientific context and references, we refer to the special issue of the Journal Advances in Animal Biosciences - Robust Cows, that will be published in 2013 with summary papers of the work done in this project:

R.F. Veerkamp, L. Kaal, Y. de Haas and J. D. Oldham, 2013. Breeding for robust cows and that produce healthier milk: RobustMilk. Advances in Animal Biosciences (In press)

Berry, D.P S. McParland, C. Bastin, E. Wall, N. Gengler, and H. Soyeurt, 2013. Phenotyping of robustness and milk quality. Advances in Animal Biosciences (In press)

Strandberg, E., M. Felleki, W.F. Fikse, J. Franzén, H.A. Mulder, L. Rönnegård, J.I Urioste,J Windig, 2013. Statistical tools to select for robustness and milk quality. Advances in Animal Biosciences (In press)

Bovenhuis, H, M. Visker and A. Lundén, 2013. Selection for milk fat and milk protein composition. Advances in Animal Biosciences. (In press)

Calus, M.P.L. D.P. Berry, G. Banos, Y. de Haas and R. F. Veerkamp, 2013. Genomic selection: the option for new robustness traits? Advances in Animal Biosciences (In press)

Coffey, M.P. S. McParland, C. Bastin, E. Wall, D.P. Berry and R.F. Veerkamp, 2013. Implementation in breeding programmes. Advances in Animal Biosciences (In press)

The results in this report is laid out to following the major objectives:

• create a common European data-base that includes unique and scarcely recorded phenotypic measurements for traits underlying robustness and milk quality, together with productivity records and fertility (WP1)
• develop phenotypic measurements tools for robustness (energy balance) and milk quality (lactoferrin and fatty acid composition) using mid-infrared spectrometry (WP2) WP4)
• create statistical tools to select for robustness and milk quality (udder health and SCC) taking into account complex biological backgrounds (WP3)
• develop genomics tools for selection for productivity, robustness (fertility, energy balance and udder health) and milk quality traits (lactoferrin and fatty acid composition) (WP4)

Common European data-base

Undertaking genetic studies of animal phenotypes presupposes accurate recording on sufficient numbers of animals to help dissect the genetics of the trait. The development of elaborate national monitoring schemes has facilitated routine population-wide on-farm accurate recording of several conventional traits, mostly associated with production (International Committee for Animal Recording, 2011). Certain indicators of functional traits, such as milk somatic cell count in dairy cattle, are included in these programmes. Nevertheless, several increasingly important traits associated with health, fitness and efficiency are currently not possible to routinely record in the commercial population. Understanding the genetic background of such traits depends then on data from experimental resource populations where animals are raised in controlled, closely monitored environments. Such populations, however, are usually of limited size. Combining data from different experimental herds would provide an expanded dataset that would allow a more rigorous genetic and genomic analysis of difficult- and expensive-to-record traits. The creation of a common database across country partners that includes unique and scarcely recorded measurements for traits underlying individual dairy cow robustness and milk quality is unique. These traits include measures such as feed intake, regular body condition scoring and detailed health and fertility recordings. These databases are held at each of the research partners involved and the first thing to do is to create a framework that enables bringing that data together to make it useable by this and future projects.

In the first study data from 4 different experimental resource dairy populations (1 herd in each of Scotland and Ireland, and 2 in the Netherlands) was pooled in a reference population for joint genetic and genomic analyses. Data included a total of 60,058 weekly records from 1,630 Holstein-Friesian cows across the 4 herds and included 7 traits: milk, fat and protein yield, milk somatic cell count, live weight, dry matter intake, and energy intake and balance. Missing records were predicted using random regression models, so that at the end there were 44 weekly records, corresponding to the typical 305-day lactation, for each cow. Data were subsequently merged and analysed with mixed linear models. Genetic variance and heritability estimates were greater (P<0.05) than zero for all traits except for average milk somatic cell count in weeks 16-44. Proportion of total phenotypic variance due to genotype by environment (sire by herd) interaction was not different (P>0.05) from zero. When estimable, the genetic correlation between herds for the same trait ranged from 0.85 to 0.99. Results suggested that merging experimental herd data into a single dataset is both feasible and sensible, despite potential differences in management and recording of the animals in the four herds. Merging experimental data will increase the precision of parameter estimates in a genetic analysis and augment the potential reference population in genome-wide association studies especially of difficult-to-record traits.

Subsequently the database was expanded using data from the other partners and new data from the same partners. At present a total of 4,473 animals have phenotypic data stored in the database, linking to 561,940 milk samples. Also we have genotyped these animals with the Illumina 50,000 SNP chip. Using the accompanied DNA from each of the animals new genomic technologies can be used to identify which DNA profiles are associated with ‘good’ milk and which with ‘bad’ milk and similarly can be used to differentiate between genes for high or low “robustness”.

Phenotypic measurements tools

Measuring milk quality

The groups in Belgium (GxABT-ULG), Edinburgh (SAC) and Ireland (Moorpark) have been very active in developing new methods of accurately and routinely predicting milk quality at little or no extra cost. The same machine in milk laboratories that determines fat and protein percentage in all milk samples (i.e. individual cows and bulk tank samples), also provides much more currently information that is currently unused. Earlier work had already suggested that this information can be used to predict fatty acid composition in the milk. Fatty acid composition is of interest if we want to improve the human health characteristics of the fat in the milk by altering the ratios of the different fatty acids. In ROBUSTMILK this methodology was further improved and validated in independent data the UK, Ireland and The Netherlands. Accuracy of prediction for particularly the saturated fat content was extremely high and these equations are already being applied in Belgium. The infrastructure is present for applying this technology in other countries including the UK.

Tools for easily and cheaply estimating dairy cow robustness (energy balance) and milk quality (lactoferrin and fatty acid composition) were developed. The tools pursued in ROBUSTMILK is mid-infrared spectrometry, which is a methodology used to determine the fat, protein and lactose concentration on all milk samples from milk recorded herds internationally. Preliminary analyses indicated that equations could be developed using this methodology to predict milk fatty acid content; the objective of this task is to strengthen these calibration equations and evaluate whether they can also be used to predict dairy cow robustness.

Prediction of milk quality

The addition of milk MIR and fatty acid information to the database of samples available from GxABT-ULG prior to the start of ROBUSTMILK resulted in a greater accuracy of predicting fatty acid content from milk using MIR. A total of 517 samples from the WP2 participants GxABT-ULG, SAC and Moorepark were used. Different breeds (Holstein-Friesian, Dual Purpose Belgian Blue, Normande, Red and White, Jersey, Norwegian Red, Montbeliarde) and crossbreds as well as different production systems (confinement and grazing) and stages of lactation were represented. Different pre-analysis treatment of data and statistical approaches were investigated. The accuracy of prediction after cross-validation for the individual fatty acids varied from 71% to 97%. However, the accuracy of predicting milk saturated fatty acid content was nearly 100%. The database was further added to and was externally validated against two separate datasets; the database used to develop the final MIR prediction equations consisted of 1,236 samples. The first validation dataset originated from different breeds in the Netherlands; the accuracy of predicting the saturated fatty acid content in the 190 samples was 99%. The second validation dataset consisted of 143 milk samples from a range of commercial Irish dairy cows; the accuracy of predicting the saturated fatty acid content was also 99%.

Lactoferrin content of 2,499 milk samples from Belgium (n=384), Ireland (n=1658), and Scotland (n=731) were determined as part of the ROBUSTMILK project. A total of 274 were used to externally validate the accuracy of predicting milk lactoferrin content from milk MIR; the equation was developed in the remaining samples. Different statistical approaches were evaluated to develop the more robust and accurate prediction equation. The accuracy of external prediction of milk fatty acid content using the alternative statistical approaches was up to 60%. Therefore, the accuracy of predicting milk lactoferrin content was not as high as the accuracy of predicting the milk content of some fatty acids. Using data available through collaboration with the University of Wisconsin, the marginal benefit of including lactoferrin in a prediction model for mastitis over and above a prediction model that already included somatic cell count was found to be small.

Prediction of animal robustness


Animal robustness was here defined as energy status which is known to influence animal health and reproduction. Data used in the study originated from research herds in Scotland and Ireland. Energy balance information was available from feed intake and energy sinks such as milk yield and composition and live-weight. Body energy status was also predicted by (change in) body weight and body condition score. The ability to predict energy intake from milk MIR was also assessed. In total, energy status information with corresponding milk MIR data was available from 564 lactations from Scotland and 338 lactations from Ireland. Several different combinations of calibration and validation datasets were generated and included combining data within country or across countries. Clear evidence existed, corroborating results from the development of MIR prediction equations for fatty acid, that the spectral variation observed in the validation dataset needed to be represented in the calibration dataset. For example, the accuracy of energy status prediction equations developed using Scottish data from cows in confinement was not high when applied to Moorepark cows fed grazed grass. The opposite was also true. When prediction equations were developed on the combined dataset and validated on a sub-selection of the combined dataset the accuracy of predicting energy status including energy intake was up to 80%. The predictive ability of energy intake was greater than either energy balance or energy content.

Genetics of predicted milk quality and robustness

A large quantity of milk MIR data was collected by the participants of WP2. Milk samples taken monthly (3 individual daily milkings) in SAC were sent to Moorepark where MIR was determined. MIR analysis of milk sample taken twice weekly at Moorepark were also undertaken and stored. During the ROBUSTMILK project 282,357 MIR records from Moorepark and 20,429 MIR records from SAC were generated. MIR spectral information from samples in Walloon animals was also routinely stored. At the end of the ROBUSTMILK project more than 2,305,000 MIR samples in Belgium were available. The large database of MIR records was required for estimating precise genetic parameters.

MIR fatty acid prediction equations were applied to all data from Walloon, SAC and Moorepark; the MIR fatty acid prediction equations were generated using representative samples from each population. Therefore, fatty acid predictions were available on all milk samples. MIR spectral information from Walloon animals were sent to Moorepark and the MIR energy status prediction equations were applied to spectra represented in the energy status calibration dataset. This was to ensure no decline in accuracy of prediction from extrapolation beyond the parameter space represented in the calibration dataset.

Genetic analyses of milk fatty acids on Walloon data revealed that that de novo synthesized FA were under stronger genetic control than FA originating from the diet and from body fat mobilization. Heritability estimates on first-parity Holstein cows for saturated short- and medium-chain individual FA ranged from 0.35 for C4:0 to 0.44 for C8:0, whereas those for monounsaturated long-chain individual FA were lower (around 0.18). Moreover, de novo synthesized FA were more heritable in mid to late lactation. The heritability of saturated fatty acid content varied from 0.25 to 0.50. In general the genetic correlations between the milk fatty acid content and milk yield were negative across lactation which is similar to the negative correlation generally observed between milk yield and total milk fat composition.

Furthermore, the genetic correlations between milk fatty acid composition and days open, a measure of fertility, changed throughout lactation. The genetic correlations with days open for unsaturated fatty acids, monounsaturated fatty acids, long chain fatty acids, C18:0, and C18:1 cis-9 were positive in early lactation but negative after 100 days in milk. For the other fatty acids, genetic correlations with days open were negative across the whole lactation. At 5 days in milk, the genetic correlation between days open and C18:1 cis-9 was 0.39 while the genetic correlations between days open and the fatty acids C6:0 to C16:0 ranged from -0.37 to -0.23. The results suggest an opportunity to use fatty acid content in milk as an indicator trait to supplement the prediction of genetic merit for fertility.

The heritability of MIR predicted milk lactoferrin content from first-parity Walloon Holstein cows varied from 0.20 to 0.45 across lactation and was greatest in mid to late lactation.

The heritability of the MIR predicted energy status traits estimated using the research data at SAC and Moorepark varied from 0.13 to 0.20. The genetic correlation between the MIR predicted and gold standard measures for energy balance, energy content and energy intake was 0.59 0.50 and 0.76 respectively suggesting they are genetically similar traits and the accurate phenotypic prediction of energy balance is also reflected in accurate genetic prediction. The heritability of the MIR predicted energy balance using the Walloon data varied from 0.17 to 0.56 across lactation and parities.

All genetic analyses results suggest that the predicted traits are heritable and sufficient genetic variation is present to justify their inclusion in national breeding schemes. Prior to widespread implementation however, more MIR calibration data for energy status would be beneficial.

Statistical tools for environmental sensitivity and genotype by environment interaction

Cows may respond differently to differences in the macro-environmental system. For example, SCC of one cow may be low across production systems, while another cow may tend to have a high SCC in one production system but low in another. If this different response to a change in environment has a genetic basis this is called genotype by environment interaction (GxE). Genotype by environment interaction can be analyzed by using a multi-trait model in which a trait measured in different environments is considered as separate traits. This method is used when the environment is naturally classified into discrete categories (e.g. production systems, climatic zones). When the environmental descriptor is of continuous nature (e.g. temperature, herd size), one can either categorize it and use the multi-trait method or analyze it using a reaction norm model, in which the trait is considered to be a function of the environmental descriptor. Each of these methods has been used in several studies before, but we combined them in a study of SCC in Irish dairy cattle. We had production system as a categorical environmental descriptor (either spring calving or year-round calving) and herd average milk production as a continuous environmental descriptor and analyzed SCC across environments with a bivariate linear reaction norm model. In this model, the genetic correlation between systems was around 0.83 which was lower than if the reaction norm on milk production was not included (0.91). However, within the same production system, there was practically no GxE between herds with low or high milk production.

Micro-environmental Sensitivity or Genetic Heterogeneity of Residual Variance

Micro-environmental sensitivity or genetic heterogeneity of residual variance has been much less studied than macro-environmental sensitivity. In a review, Hill and Mulder (2010) showed that several studies, mainly in body weight traits or litter size, have detected genetic variation in residual variance. Most recent studies have used a Bayesian approach with Markov Chain Monte Carlo sampling algorithms for estimation of variance components. However, this method has computational limitations and for larger datasets the computation time becomes a severe restriction. Therefore, a new method for estimating genetic variance in residual variance was developed using double hierarchical generalized linear models (DHGLM). This approach is based on iterating between two sets of mixed model equations, one on the level of the observations and the other on the level of residual variances. This approach was validated both using simulated data and by application on the dataset on litter size of pigs, previously analysed by others. One particular problem that needed to be addressed was the estimation of the (genetic) correlation between the breeding value (BV) for the level of the trait (the ordinary BV) and the BV for the residual variance. The DHGLM theory was therefore extended and approximated by a bivariate linear mixed model.

The approach developed makes it possible to estimate genetic variation in residual variance also in large data sets. We applied it to a relatively large dairy cattle data with 180 000 Holstein cows having about 1.6 million test-day records. Estimation of variance components took less than a week on a normal Linux server. We found genetic variation in residual variance in both milk yield and SCC: a change of one genetic standard deviation of the breeding values for residual variance would result in a change in the residual variance of about 20%, for both traits. Genetic variation in residual variance was also found in a population of Swedish Red dairy cattle and estimated variance components were very similar to those estimatedin the Swedish Holstein population. Breeding values for residual variance were subsequently used for a genome-wide association study and signals on several chromosome were found.

The same approach was also applied to milk yield, SCS, and Saturated Fatty Acid (SFA), Unsaturated Fatty Acids (UFA) and C18:1 cis-9 contents in milk in Walloon Holstein cows. For milk and SCC the genetic variance in residual variance was a bit smaller than in Swedish Holsteins with genetic standard deviations of 17 and 16% respectively. For SFA, UFA and C18:1 cis-9 genetic standard deviations were 12%.

Simultaneous estimation of micro- and macro-environmental sensitivity

In one part of our project we tested, on simulated data, whether it was possible to estimate genetic variance in both micro- and macro-environmental sensitivity in the same analysis. The DHGLM was extended with a linear reaction norm to account for micro- and macro-environmental sensitivity.. A data set with true genetic variation in micro-, or macro-environmental sensitivity or both was analyzed with models including either or both of these components, and the models were compared using Akaike Information Criterion (AIC) or likelihood ratio test. In 95-100% or 97-100% of the cases the correct statistical model (corresponding to the true) was chosen, for AIC and likelihood ratio test, respectively. We also found that it was necessary to have at least 100 bulls with at least 100 daughters each to estimate all genetic parameters with sufficient precision. The estimate of the genetic variance in micro-environmental sensitivity had lower precision than the genetic variance of macro-environmental sensitivity. Furthermore, the genetic correlations between micro-environmental sensitivity with macro-environmental sensitivity or with the intercept of the reaction norm were estimated with low precision (SD across replicates > 0.15) even with designs with 100 bulls each with 100 daughters. In designs with only 20 daughters per bull, a majority of the replicates failed to converge to positive definite variance-covariance matrices.

Statistical tools to improve milk quality – Somatic Cell Count

In order to lower the SCC in milk it is necessary to decrease the incidence of both clinical (CM) and subclinical (SCM) mastitis. This can be done both by using observations on actual mastitis cases and by using SCC, being an indicator of both SCM and CM. Apart from the studies above on genetic variation in residual variance in SCC, which could be used to select for lower SCC (lower variation is also associated with lower average levels), we have studied how to decrease SCC in various ways.

Definition of alternative SCC traits

One option is to try to improve the actual trait used in the genetic evaluation. In our case we want a trait based on SCC that is more closely related to CM. Currently, most commonly an average of the (log of) SCC over a certain period in the lactation is used (e.g. first 50 or 150 days) or a test-day model is used and the EBV for the average level (or maybe a sum of SCC over a certain period based on the EBVs) is used for selection purposes. We first studied some alternative SCC-traits that could be defined using test-day SCC from weekly observations in an experimental herd and then tested if these would also be useful with only monthly observations. Standard deviation of SCC within the lactation (SCCSD) and a discrete indicator of at least one test-day with SCC above 500 000 cells/ml (TD>500) were those most strongly associated with CM in both the weekly and monthly data sets. In the weekly data set also number of days of the widest SCC peak was selected and in the monthly data set, number of SCC peaks (NPeak) and average number of days per SCC peak (AveDays) were chosen. Heritabilities were in general between 10 and 16% (except for AveDays: 5%). Heritabilities were at least as high in the monthly data set, which indicated that these traits would be potentially useful also in a field data set.

To test this, these traits were also studied in a large field data set consisting of about 178 000, 116 000 and 64 000 lactation records in the first three lactations, respectively. Heritability was estimated at 0.12-0.17 for SCCSD, TD>500 and AveDays, but lower, 0.06-0.10 for NPeak and TD41-80 (an indicator of at least one test-day with SCC between 41 000 and 80 000 cells/ml). All traits, except TD41-80, were highly positively genetically correlated to CM (0.67-0.82) and even more so with SCM (0.94-0.99) in all three lactations. TD41-80 showed a negative (as expected) correlation from -0.22 to -0.50 with CM and from -0.48 to -0.85 with SCM. All alternative traits, except TD41-80, were strongly correlated with each other within lactation but TD41-80 had a negative correlation of around 0.4-0.5 with other traits in first lactation.

Epidemiological reaction norms

One problem with selecting for better udder health is that the traits of interest have rather low heritabilities, especially so for CM. One possible reason for the low heritabilities is that not all animals have been exposed to the pathogens causing the disease. Consequently, in a typical study analysing the genetics of (sub)clinical mastitis using national recording schemes, healthy animals are a mixture of non-resistant animals that have been exposed and resistant animals that may or may not have been in contact with pathogens. One way to account for this is to quantify the exposure probability and estimate heritabilities in environments differing in exposure probability. We did this using a reaction norm model, where the environmental descriptor was a measure of prevalence based on what percentage of cows in a herd had either CM or elevated SCC, and the trait was CM or SCM as traits. We found that including information on herd prevalence in the model explained more variation in CM and SCM. We also found a somewhat curvilinear concave relation between heritability and prevalence, but no strong evidence for the hypothesis by Bishop and Woolliams (2010) of decreasing heritability with decreasing prevalence and exposure.

Transition probability model

The SCC in milk associated with mastitis could be decreased both by having cows that do not get mastitis and by having cows that recover quickly from mastitis. Previously, genetic evaluation and selection has aimed at decreasing the first factor, but in one part of our project we have aimed at including both aspects in our genetic evaluation. In this approach, we estimated breeding values for the probability of transition from a healthy to a diseased state as well as for the reverse probability. The approach takes into account the recovery process and repeated cases. In a scenario with about 16% of lactations having at least one mastitis case, and with 150 daughters per bull, the accuracy of estimated breeding values (EBVs) (correlation between true and estimated breeding values) was about 0.75 for the transition from healthy to diseased, and about 0.4 in the opposite direction. The lower accuracy for the EBV from diseased to healthy is expected because it is based on severely reduced data consisting only of daughters that have had mastitis. However, we managed to deliver as accurate EBVs of mastitis liability as some previously used methods using data of confirmed cases of mastitis, even though we estimated mastitis based on SCC only.,. In addition, the approach generated valuable information about the recovery process that could be included in the genetic evaluation of mastitis.

Genomic Tools

Genomic information is expected to make an important contribution to selection for traits which are difficult to improve by means of traditional selection. These might be traits which are difficult or expensive to measure and therefore routine recordings are not available, or traits which are recorded routinely but the phenotypes are not very accurate in a sense that the genetics is masked by disturbing environmental factors. However, before genomic information can make a contribution to genetic improvement of these traits either genes need to be identified that affect these traits or a reference population for setting up genomic prediction equations is needed (genomic selection). Therefore, progress in genomic selection is hampered by access to sufficiently large datasets to estimate the optimal DNA profile for a given trait (SNP key), but also by the lack of (statistical-) tools to utilise the DNA information with the scarce records. ROBUSTMILK brought the data together in a database, developed the required tools, and answered the potential of genomically selecting for difficult to measure traits.

Combined reference populations

The objective of the RobustMilk project was to combine data from individual partners recognising that each individual partner had too small a reference population to generate highly accurate genomic predictions. This work was expanded by collaborating with an Australian project with feed intake records. It was clearly shown that the multi-trait model achieved the highest gain in accuracy across all countries, and that all countries benefitted when all countries were included together in the prediction model. In a few situations when only two countries were included, however, realized accuracies in fact were lower compared to a scenario where only data from the country itself was used. These results indicate that generally there is an expected benefit of combining data, albeit that in each specific case compatibility of datasets needs to be investigated.

When a reference population needs to be established for difficult to measure traits, the most likely cost-effective strategy is to generate genotypes and phenotypes on cows, instead of genotyping bulls and phenotyping large paternal half-sib progeny groups. Based on cow reference populations consisting of 600 to 2,000 cows with single phenotypic measurements, the accuracy of direct genomic values (DGV) was established within RobustMilk for traits with heritabilities ranging from 0.2 to 0.5 and was reported to be in the range, which closely resembles the theoretical expected range of 0.29 to 0.63 based on a deterministic formula.

An important question was which animals should be included in the optimal construction of a reference population. The degree of relationship between reference populations and selection candidates affects the prediction accuracy and it was shown that strong relationships among animals in the reference population in fact have a negative effect on the average accuracy of genomic predictions in selection candidates. As a consequence, the optimal reference population design maximises the relationships between the reference population and the evaluated animals, while minimising the relationship among animals in the reference population. Both from the perspective of relationship to the evaluated animals and from the perspective of sampling extreme phenotypes, it is likely that adding ‘unique’ animals to the reference population leads to higher increases in accuracy compared to adding animals at random.

Genome wide association study for milk quality (SCC)

We analysed lactation-average SCC and test-day SCC standard deviation which were calculated for each cow based on test-day records. The standard deviation of test-day SCC aims to capture fluctuations in SCC associated with infection. The number of test-day records per cow on the experimental herds ranged between ten and 52 with an average of 31 test-days. These numbers are much higher as compared to routinely collected data on commercial herds where SCC usually are recorded monthly. Monthly SCC recordings have lower probability of detecting infections of short duration than intensive recording as practiced on experimental herds

Two SNP were found to be significantly associated with lactation-average SCC and one SNP with the standard deviation of SCC. One SNP on BTA18 was associated with both lactation-average SCC and the standard deviation of SCC. A SNP on BTA4 was uniquely associated with the lactation-average SCC. In total 98 cows in the data set had a case of clinical mastitis during their first lactation. The SNP with significant effects on the lactation-average SCC or the standard deviation of SCC were not significantly associated with clinical mastitis. Relatively few significantly associated SNP were detected, suggesting that both somatic cell count traits are controlled by many loci, each with a relatively small effect. Different SNP contributing to lactation-average SCC and the standard deviation of SCC could reflect differences in genetic regulation of both traits where lactation-average SCC may refer to base-line somatic cell count during lactation and standard deviation of SCC may reflect immune reactivity to infection.

Genome wide association study (GWAS) for fertility

Traditional fertility measures based on calving and insemination dates are heavily affected by decisions made by the farmer and therefore are not very accurate indicators for fertility. The interval from calving to first luteal activity (CLA) based on hormonal profiles has been suggested as a more accurate and therefore preferable measure for fertility. However, routine measures of CLA based on hormonal profiles are usually not available for reasons of cost and labour but were available on data collected on experimental herds. The GWAS results clearly

The traditional fertility traits were days from calving to first observed heat, days from calving to first service, calving interval, number of services, and pregnancy rate to first service. Besides traditional fertility traits information on the interval from calving to first luteal activity (CLA) based on hormonal profiles was available. CLA was defined as the number of days from calving to the first occurrence of two consecutive test-day records with a milk progesterone concentration of ≥3 ng/ml.

For traditional fertility traits (days from calving to first observed heat, days from calving to first service, calving interval, number of services, and pregnancy rate to first service) only weakly significant SNP were detected. Much stronger evidence for the presence of QTL was detected for CLA: a SNP on BTA 2 explained 0.51% of the genetic variance in CLA while a SNP on BTA 21 explained 0.35% of the genetic variance in CLA.

This work illustrateed the benefits of having an accurately measured phenotype and the potential benefits of physiological traits whose genetic control may be less complex than high level phenotypes.

Genome wide association study (GWAS) for feed utilisation

The importance of feed utilisation has received renewed interest in order to reduce the environmental impact of dairy production, but also because the direct link to energy balance of dairy cows. However, in practical dairy cattle breeding there is no direct selection for feed intake or feed efficiency. This is a consequence of the practical difficulties of measuring individual feed intake in dairy cows. The detection of genes regulating feed intake or the development of genomic selection tools could bring selection for feed intake based on genomic information within reach.

Several traits related to feed utilization were analysed yield, live weight, body condition score and dry matter intake. The averages for the predicted values for week 3 – 15 in lactation were used in the GWAS.

The GWAS for feed utilization traits revealed several significant chromosomal regions. Highly significant effects were found on BTA4 for body condition scores, on BTA7, BTA13 and BTA18 for live weight and on BTA27 for dry matter intake. The percentage of the total genetic variance of most significant SNPs ranged from 0.5 to 0.6% indicating that the magnitudes of the effects were relatively small.

The success of genome-wide association studies is a function of, amongst others, the heritability of the trait under investigation and the number of phenotypic records on that trait. For traits which are expensive and difficult to measure the number of available records is a limiting factor. The present study is one of the first studies to show the potential of combining detailed phenotypic and genotypic data from research herds located in different countries, thereby increasing the power of the study.

Multi-trait genomic prediction – combining robustness and milk yield

One strategy to improve genomic selection for scarcely recorded traits is to use less-expensive predictor traits in multi-trait prediction models proved to be successful in traditional breeding programs to increase the accuracy of prediction. As part of the RobustMilk project, we optimised a multitrait genomic selection model and performed a simulation study to investigate this combined model. This research indeed supported that multi-trait genomic prediction can lead to a considerable increase in genomic prediction accuracy. Within the RobustMilk project, we also investigated the additional benefit of exploiting a multi-trait GBLUP-type model to predict genomic breeding values for dry matter intake, using measurements for milk yield and live weight as indicator traits. This study showed that indeed the indicator traits improved the accuracy of prediction for dry matter intake, but also indicated that the accuracy was similar to a multi-trait pedigree based model. This was most likely the result of moderate to strong genetic correlations between the predicted trait and the indicator traits, confirming the previous simulation study showing the added benefit of genomic information in multi-trait models.

Another strategy to improve genomic selection for scarcely recorded was to include phenotypic data from past experiments but from animals having no DNA available. However these can be used in the analysis by merging pedigree and genomic relationships, and as established within the RobustMilk project by estimating genetic parameters for feed utilisation within the project, the combined datasets gave higher precision of the estimated genetic correlation and heritability, in comparison to using either the pedigree or genomic data alone.

Furthermore, within the RobustMilk project, a Bayesian model was developed that can use data from two traits that are each measured on a separate group of animals. This model was applied to a scenario where the one group of animals was a cow reference population and the other group of animals was a bull reference population. Results showed that accuracies of genomic prediction for the trait measured on the cows benefitted from exploiting the additional information on the bull trait and helped to reduce potential bias in predicted breeding values. Additionally, using the cow and bull data combined, resulted in increased power to detect QTL using the Bayesian model. Similar results were found when using this model in a genome-wide association study for milk fatty acid content as part of the RobustMilk project. In addition, evidence for two QTL related to progesterone levels, detected using the single-trait model, improved when this bivariate Bayesian model was applied using information on correlated fertility traits

Predicted accuracy and response to genomic selection, using a multi-trait approach with one trait recorded on cows and the other on bulls, was investigated using deterministic simulation within the RobustMilk project. Scenarios were considered where phenotypes for a new trait were available on up to 10,000 cows, while a bull reference population of up to 20,000 animals was available for an overall index, which had a genetic correlation of -0.5 0.0 or 0.5 to the new trait. When the new trait had a negative correlation with the overall index, achieving favourable genetic progress for the new trait was only possible in the extreme situation where the new trait had a moderate heritability (0.30) and when it was economically as important as the overall index. In all other scenarios with a negative correlation between the new trait and the overall index, the cow reference population was not sufficient to achieve favourable genetic progress for the new trait. When the new trait had an economic value that was at least 20% of the economic value for the overall index, using the cow reference population did diminish the negative genetic response for the new trait caused by selection for the overall index. In all scenarios where the genetic correlation between the new trait and the overall index was zero or positive, considerable genetic response could be achieved using a reference population of up to 10,000 cows, the rate of response achievable depending on the heritability and the economic value of the new trait.

Integration milk quality and robustness

Among others, new breeding tools for milk quality traits (i.e. fatty acids profile in milk) and robustness traits (i.e. indicators of energy balance status of dairy cows) have been developed through the ROBUSTMILK project. Even if these new traits currently have no (or undefined) monetary value, it is suggested that better overall economic efficiency could be achieved if they are included in the breeding goals. However, before doing this, the consequences of shifting emphasis from the current traits in the breeding objective to these new traits should be investigated particularly given that genomics is likely to lead to accelerated rates of gain. Together with existing information from the different countries, and using selection index theory, consequences of selection for new robustness and milk quality traits were assessed. A total of 6 scenarios of selection (e.g. inclusion of new robustness and milk quality traits) were then defined. New traits considered were body condition score (BCS) for robustness and two indices representing the relative part of milk fat that is unsaturated or monounsaturated for milk quality. The other groups of traits considered were: production, udder health, longevity, fertility, conformation, robustness, and milk quality. Results indicated that currently used total merit indexes lead to favourable genetic gains for production but also for most of the other traits as a balanced emphasis is given to both production and non-production traits represented by an increasing number of functional traits. Adding new robustness and milk quality traits with reasonable weights in the breeding goal did not affect greatly relative genetic gain of existing traits.


Potential Impact:

The impact in this report is laid out following the major objectives:

• create a common European data-base that includes unique and scarcely recorded phenotypic measurements for traits underlying robustness and milk quality, together with productivity records and fertility (WP1)
• develop phenotypic measurements tools for robustness (energy balance) and milk quality (lactoferrin and fatty acid composition) using mid-infrared spectrometry (WP2) WP4)
• create statistical tools to select for robustness and milk quality (udder health and SCC) taking into account complex biological backgrounds (WP3)
• develop genomics tools for selection for productivity, robustness (fertility, energy balance and udder health) and milk quality traits (lactoferrin and fatty acid composition) (WP4)

Common data-base

The RobustMilk project has been instrumental in realising that also in animal breeding it holds that “In the age of the genotype, the phenotype is king”. Setting up the shared database, where research organisations share data. We developed the database and protocols, but also an acceptable “way of working”, that also is promoted by the G20 that sharing genetic and genomic data in the agricultural setting is essential.

Concrete the RobustMilk database has resulted in further collaborations between and between Australia and Europe (de Haas et al., 2012). This has now been expanded to the gDMI intiative. Other initiatives are underway to pool data for feed intake (Veerkamp et al., 2012a) and methane emission (de Haas et al., 2011).

Phenotypic measurements tools

An agreement has been drafted and signed between the ROBUSTMILK partners who contributed to the development of the milk fatty acid prediction equations. This agreement allows the prediction equations to be freely used by the member countries in their national milk recording and breeding programs. Research on genetic evaluations of predicted milk fatty acid is at an advanced stage in the Walloon region of Belgium. Ireland and the UK have finalised agreements with the manufacturers of milk infrared spectroscopy machines to routinely generate and store MIR data. In all countries, the fatty acid content of milk samples will be provided back to farmers and this information will also contribute to national genetic (and subsequently genomic) evaluations. This information can be used by a) farmers to monitor the milk fatty acid composition of their herd, b) breeding companies to identify elite germplasm for milk fatty acid composition, and c) milk processors to incentivise farmers to producer higher quality milk and potentially market products of superior milk quality.

Improving milk quality, especially the milk fatty acid content has implications for human health because of the known associations between saturated fat intake and human health including cardiovascular disease. Therefore providing a tool for milk processors to incentivise the production of milk with better human health characteristics has huge societal benefits. More importantly, however, ROBUSTMILK provides the tools for farmers to breed for improved milk quality thereby achieving the goals set out by the milk processor. Generating higher value milk using the tools developed in ROBUSTMILK will result in greater profitability since there is no additional operational cost of acquiring the predicted phenotypes because MIR is part of routine milk testing, both of individual cows and bulk tank samples. Also, breeding is cumulative and permanent and therefore the benefits of selection for improved milk quality remain within the herd.

The predictive ability of milk lactoferrin content from milk MIR is currently too low for national implementation although ability to categorise milk into high, medium or low lactoferrin content is being evaluated.

There is growing societal pressure for welfare and environmentally friendly animal production systems. Robust animals are partly defined as animals with less severe and shorter durations of negative energy balance and as a result, are expected to be less welfare compromised. Also, one of the biggest contributors to increased environmental footprint is animal wastage through impaired fertility. For example, modelling research from the UK suggests that if dairy cow fertility in the UK national herd could be restored to 1995 levels from 2003 levels then herd methane emissions could be reduced by 10 to 11% while ammonia emissions could be reduced by 9% under a milk quota environment; the respective reductions were 21 to 24% and 17% if ideal fertility levels were achieved. A reduction of 4 to 5% in herd methane emissions was expected in the UK if fertility levels were restored to 1995 levels from 2003 levels where no milk quota existed. Energy balance is known to be associated with animal health and fertility and therefore managing or breeding for more favourable energy balance profiles will have superior health and fertility which is more environmentally friendly. The MIR prediction tools provided through ROBUSTMILK for predicting energy balance, which was subsequently showed to exhibit heritable genetic variation, will therefore result in more welfare friendly and environmentally friendly dairy production systems.

Statistical tools

New genetic evaluation methods for SCC and mastitis may lead to more efficient selection, which in turn could lead to better milk quality and udder health. The genetic evaluation models developed for decreasing residual variation can be expected to decrease not only the average level of SCC, but also the variation in SCC, which could mean that fewer cows exceed the limit that give penalty in the payment for milk. This would result in better economy for the farmers as well as better milk quality for the processing plant. The method developed to also estimate the genetic ability to recover from mastitis could lead to not only fewer but also shorter periods of disease, which means lower amount of discarded milk due to mastitis.

Genomics tools

In this WP, methods have been developed to combine different sources of information in a genomic prediction model. For instance, information from different traits can be combined in one analysis, to improve accuracy of estimated SNP effects for traits for which few records are available. This extremely important for traits that important from an environmental point of view, such as feed efficiency and methane emission. In fact, the methods developed will be utilized in the gDMI (global Dry Matter Intake) initiative, that aims to predict genomic breeding values for dry matter intake.

Furthermore, the analyses within WP4 have generated wider knowledge on the genetic architecture of several robustness related traits. One of the clear conclusions, is that most robustness related traits, such as mastitis, feed efficiency and fertility, are mainly affected by a large number of genes, each having a small impact on these traits. Only few loci have been detected that have a clearly above average impact on any of those traits.

Main dissemination activities

Dissemination of results was done in different directions.

First, towards stakeholders through the internet, layman publications, farmers conferences/seminars/discussion groups, and intervention in the popular press. Second towards scientific community through publications, scientific conferences, workshops, and a satellite conference linked to INTERBULL on selection for robustness and milk quality.

A total of 35 papers in popular press and presentations in farmers conferences/seminars/discussion groups were completed during the project. People from the ROBUSTMILK team attended up to 35 conferences or workshops and more than 89 abstracts of conferences were listed. Finally, more than 39 peer-reviewed papers were written on the ROBUSTMILK research and results.

Papers published in local journal for animal breeding and animal production stakeholders as well as scientific papers and presentations of the results in scientific meetings were all published on the ROBUSTMILK website (www.robustmilk.eu).

For the dairy cattle breeding industry during the 2011 Interbull Meeting (organized in Stavanger, Norway, from August 26 to 28, in connection with the 2011 EAAP meeting), two sessions were dedicated to animal robustness and milk quality and were organized in cooperation with the ROBUSTMILK project. A total of 155 persons registered for the meeting. Out of total of 52 presentations, 17 were related to the robustness and milk quality and 9 were given by ROBUSTMILK research groups and were directly derived from research done inside the project. Also, a workshop with the industry about utilisation of the ROBUSTMILK outcomes took places during the meeting.

Attached to the final consortium meeting at the EAAP in Bratislava (2012) RobustMilk was present with 19 contributions to present the finding to scientific (and industry) community. Purposely in different session to reach a wider audience.

As a final summary of the outcomes of the project a special issue of the Journal Advances in Animal Biosciences - Robust Cows will be published in 2013.

List of Websites:

www.robustmilk.eu