Genomic and quantitative genetic analyses of female fertility and calving traits in German Holstein cattle using alternative random regression modelling approaches

Abstract

This thesis presents a comprehensive investigation of the genetic architecture of female fertility and calving traits in German Holstein cattle by integrating advanced quantitative genetic modeling and longitudinal genomic analyses. Due to their low heritability, strong environmental influence, and complex biological regulation, reproductive traits pose a long-standing challenge for dairy breeding programs. The work aims to improve the accuracy and biological relevance of genetic evaluations for traits such as non-return rate at 56 days (NRR56), calving-to-first service interval (CTFS), days open (DO), calving ease (CE), and stillbirth (SB). To achieve this, the thesis employs classical genetic models, random regression approaches, and longitudinal genome-wide association studies (GWAS), thereby combining statistical, genomic, and biological perspectives. Chapter 1 provides an extensive introduction to reproductive biology in dairy cattle and reviews key fertility and calving traits, along with their economic and welfare relevance. It discusses factors influencing reproductive performance, including management, behavior, health, and genetics, and outlines the limitations of traditional statistical methods that assume constant genetic effects across time. The chapter emphasizes the importance of longitudinal models and introduces the conceptual foundation of random regression models (RRM) and longitudinal GWAS. These approaches enable the modeling of heterogeneous variances across parities and capture time-dependent genetic effects. The chapter concludes by presenting the main objectives of the thesis: to estimate genetic parameters across reproductive traits using advanced modeling, to integrate genomic data into dynamic analyses, and to evaluate the biological function of identified genomic regions. Chapter 2 focuses on fertility traits and applies Multiple-trait models (MTM) and random regression models (RRM) to a large dataset comprising more than 592,000 fertility records. Genotypes from approximately 21,300 animals were integrated using a genomic relationship matrix. This chapter demonstrates that genetic variances and heritabilities for NRR56, CTFS, and DO generally increase with parity, particularly distinguishing heifers from cows. The RRM framework proved more biologically realistic, revealing parity-specific genetic patterns and declining genetic correlations as parity distance increased. Notably, correlations between heifer NRR56 and cow NRR56 were low (0.25-0.50), indicating distinct genetic expressions in early versus later reproductive cycles. The chapter shows that RRMs enable dynamic estimated breeding values (EBVs), which support more precise selection across the reproductive lifespan. Chapter 3 presents a longitudinal genome-wide association study (GWAS) for fertility traits, incorporating time-dependent single-nucleotide polymorphism (SNP) effects. Using repeated fertility measurements across six lactations, the study identifies significant genomic regions whose effects vary across reproductive stages. Circular Manhattan plots and quantile-quantile (QQ) plots illustrate both stage-specific SNP associations and overall model accuracy. Gene annotation and enrichment analysis reveal biological pathways relevant to reproduction, including hormonal regulation (for example, involving the gene CSMD1), cell adhesion (genes TMEM132C and DCHS2), and cell proliferation and oocyte (egg cell) development (gene CSNK1A1). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis further highlight the contribution of signaling mechanisms such as Hippo, Wnt, and gonadotropin-releasing hormone (GnRH) to fertility regulation. These findings underscore the importance of incorporating temporal genetic variation in genomic evaluations and demonstrate the power of longitudinal GWAS for identifying biologically meaningful candidate genes. Chapter 4 investigates calving traits, specifically calving ease (CE) and stillbirth (SB), using three modelling approaches: a maternal model with direct and maternal genetic effects, a multiple-trait model (MTM) treating each parity (calving event per cow) as a distinct trait, and a random regression model (RRM), which describes calving performance across multiple parities. Using nearly half a million calving records, the chapter shows that incorporating maternal genetic effects substantially improves model fit and biological interpretation. The RRM approach again provides smoother variances across parities and realistic covariance structures. Moderate heritabilities and strong genetic correlations across parities highlight the potential for genetic improvement of calving traits. This chapter further demonstrates that modeling CE and SB at the dam (mother cow) level, rather than attributing them solely to the calf, better reflects underlying physiology and leads to more accurate estimated breeding values (EBVs). Chapter 5 synthesizes the results of all studies and discusses their implications for dairy breeding. Building on these findings, the thesis concludes that reproductive traits exhibit dynamic genetic architecture and cannot be fully captured by static models. Notably, random regression models consistently outperform conventional approaches in describing time-varying variances and correlations. Furthermore, longitudinal GWAS complements quantitative models by identifying functional genes and pathways involved at different reproductive stages. Collectively, this research provides a foundation for implementing longitudinal modeling in national breeding programs, enabling more accurate selection for fertility, calving performance, animal welfare, and overall herd sustainability.