Haibo Lu

Milk yield and composition change during lactation and have been suggested as indicators e.g., for cow health and fertility. The changes in milk yield and composition were due to different metabolic pathways of milk synthesis. For example, milk synthesis is affected by negative energy balance in early lactation and pregnancy in late lactation. The differences in metabolic pathways might affect the genetic background of milk production traits during lactation. It is known that for milk production traits genetic variances change and genetic correlations differ from unity during lactation. For specific QTL underlying milk synthesis e.g., diacylglycerol O-acyltransferase 1 (DGAT1) K232A polymorphism, genetic effects change during lactation. However, most genome-wide association studies (GWAS) to identify genetic background of milk production traits do not account for the changes in genetic effects during lactation. These studies were based on accumulated records, e.g., 305 d milk yield or test-day records with constant genetic effects during lactation. Therefore, these GWAS might miss QTL whose effects change during lactation. The objective of this thesis was to unravel the changes in the genetic background of milk production traits during lactation. Accounting for these changes in the genetic background during lactation might contribute to the development of better indicators based on milk yield and composition.

Chapter 2 focused on the different approaches to detect QTL with changing effects during lactation. Four different GWAS approaches using a 50k SNP panel were performed based on 19,286 test-day milk protein content records: 1) separate GWAS for specific lactation stages; 2) GWAS for estimated Wilmink lactation curve parameters; 3) a GWAS using a repeatability model where SNP effects are assumed constant during lactation; and 4) a GWAS for genotype by lactation stage interaction using a repeatability model and accounting for changing genetic effects during lactation. Separate GWAS for specific lactation stages suggested that the detection power greatly differs between lactation stages and that genetic effects of some QTL change during lactation. GWAS for estimated Wilmink lactation curve parameters detected many chromosomal regions for Wilmink parameter a (protein content level), whereas 2 regions for Wilmink parameter b (decrease in protein content towards nadir), and no regions for Wilmink parameter c (increase in protein content after nadir). Twenty chromosomal regions were detected with effects on milk protein content, however, there was no evidence that their effects changed during lactation. For 5 chromosomal regions located on chromosomes 3, 9, 10, 14, and 27 there was significant evidence for genotype by lactation stage interaction and thus that their effects on milk protein content changed during lactation. Three of these 5 regions were only identified using a GWAS for genotype by lactation stage interaction. These results further elucidated the genetic background of milk protein content and demonstrated that GWAS for genotype by lactation stage interaction offers new possibilities to unravel the changes in the genetic background of milk composition.

Chapter 3 explicitly performed GWAS for genotype by lactation stage interaction for 7 other milk production traits, i.e., milk yield, lactose yield, lactose content, fat yield, fat content, protein yield, and somatic cell score (SCS) to screen the whole genome for QTL with changing effects during lactation. For this study 19,286 test-day records of 1,800 first-parity Dutch Holstein-Friesian cows were available that were genotyped using a 50k SNP panel. A total of 7 genomic regions with changing effects during lactation were detected in the GWAS for genotype by lactation stage interaction. Two regions on chromosomes 14 and 19 were also significant in the GWAS that assume constant genetic effects during lactation. Five regions on chromosomes 4, 10, 11, 16, and 23 were only significant in the GWAS for genotype by lactation stage interaction. The biological mechanisms that cause these changes in genetic effects are still unknown, but negative energy balance in early lactation and effects of pregnancy in late lactation may play a role. These findings increased our understanding of the genetic background of lactation and might contribute to the development of better management indicators based on milk composition.

Chapter 4 further investigated the hypothesis that changes in genetic effects during late lactation might be related to pregnancy. Pregnancy is inseparable from the initiation of lactation and for maintaining the milk production cycle. Pregnancy affects milk production and therefore should be accounted for in the genetic evaluation. Furthermore, there might be genetic differences in pregnancy effects on milk composition. Therefore, phenotypic and genetic effects of pregnancy on milk production traits were estimated using 14,505 test-day records of 1,359 first-parity Dutch Holstein-Friesian cows with accurately estimated conception dates. Significant effects of pregnancy on all milk production traits were detected except SCS. The pregnancy effects on milk yield, lactose yield, protein yield, fat yield, and fat content were small during early gestation (< 150 d) and substantially increased in late gestation. The effects of pregnancy on milk protein yield were relatively stronger than those on fat yield. Interestingly, the effects of pregnancy on milk production traits differed for DGAT1 genotypes. Milk yield, lactose yield, protein yield, and fat yield of DGAT1 AA cows were more affected by pregnancy than that of DGAT1 KK cows. For example, the average cumulative effects of pregnancy on milk yield were -247 kg. However, effects of pregnancy were negligible for DGAT1 KK cows and were -443 kg for DGAT1 AA cows. Therefore, more evidence was identified to support the hypothesis that changing genetic effects of DGAT1 on milk yield, lactose yield, and fat content might be related to pregnancy. Besides, to relieve the impact of NEB on dairy cows in early lactation, alternative management strategies such as shortening or omitting the dry off period has been proposed to reduce peak milk production after calving. This management strategy results in additional milk yield during late gestation and decreases milk yield in the next lactation. It has been suggested that the suitability of this management strategy should be assessed with consideration of milk yield during the last 60 d in gestation, i.e., the conventional dry off period. Our study showed that milk yield during late gestation was affected by pregnancy and effects of pregnancy differ between cows with different DGAT1 genotypes. Although deciding which genotype is more suitable for shortening or omitting dry period length needs the milk production records in the next lactation, our analysis suggested that the suitability of cows for shortening or omitting the dry off period might depend upon their DGAT1 genotype. In addition, as pregnancy affects milk yield and composition, studies have been performed to investigate possibilities to predict pregnancy status based on milk infrared spectra. The current study showed that there were significant differences in pregnancy effects on milk production traits between cows with different DGAT1 genotypes. Therefore, accounting for genetic differences in pregnancy effects might improve the prediction of pregnancy by milk infrared spectra. Chapter 5 quantified phenotypic and genetic effects of season on milk production traits in the Netherlands based on 19,286 test-day milk production records of 1800 first-parity Holstein-Friesian cows that were genotyped using a 50k SNP panel. In the Netherlands and other countries, cows are grazed on pasture in summer whereas in winter cows are kept inside and fed silage. The different feeding regime during season might change the milk composition and affect the genetic background of milk production traits. The season effects were significant for all milk production traits. For example, milk fat yield and protein yield were lower in summer than in winter. The effects of season were largest for milk fat yield and fat content; smallest for milk yield, lactose yield, lactose content, and SCS; and intermediate for milk protein yield and protein content. Moreover, GWAS for genotype by season interaction were performed and 2 regions with major interaction signals on chromosomes 3 and 14 were identified. Chapter 6 is the general discussion. I first estimated the changes in genetic parameters of milk production traits during lactation for the data that were used in this thesis. Genetic variances and heritabilities of milk yield, lactose yield, lactose content, fat content, and protein content change during lactation. The changes in genetic parameters suggested that the genetic background of milk production traits might change during lactation and GWAS can be performed with consideration of changing genetic effects during lactation. Secondly, I discussed the approaches to model changing SNP effects during lactation especially in random regression models. GWAS based on random regression of polygenetic effects and fixed regression of SNP effects could investigate the changes in the genetic effects during lactation. However, a large dataset is needed to accurately estimate a large number of parameters; the selection of best Legendre polynomial order to fit each random and SNP effect in GWAS are computation demanding; and might estimate extreme values in the peripheries of lactation. Third, DGAT1 by lactation stage interaction on milk FA composition were estimated during d 63 to d 335 in lactation and no significant signals were identified. Therefore, no evidence in our dataset showed that the effects of DGAT1 on FA change during d 63 to d 335 in lactation.