Replace This with the Actual Title Using All Caps

UNDERSTANDING THE GENETICS UNDERLYING MASTITIS USING A MULTI-PRONGED APPROACH

A Dissertation Presented to the Faculty of the Graduate School of Cornell University

In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

by Asha Marie Miles December 2019

UNDERSTANDING THE GENETICS UNDERLYING MASTITIS USING A MULTI-PRONGED APPROACH

Asha Marie Miles, Ph. D. Cornell University 2019

This dissertation addresses deficiencies in the existing genetic characterization of mastitis due to granddaughter study designs and selection strategies based primarily on lactation average somatic cell score (SCS). Composite milk samples were collected across 6 sampling periods representing key lactation stages: 0-1 day in milk (DIM), 3- 5 DIM, 10-14 DIM, 50-60 DIM, 90-110 DIM, and 210-230 DIM. Cows were scored for front and rear teat length, width, end shape, and placement, fore udder attachment, udder cleft, udder depth, rear udder height, and rear udder width. Independent multivariable logistic regression models were used to generate odds ratios for elevated SCC (≥ 200,000 cells/ml) and farm-diagnosed clinical mastitis. Within our study cohort, loose fore udder attachment, flat teat ends, low rear udder height, and wide rear teats were associated with increased odds of mastitis. Principal component analysis was performed on these traits to create a single new phenotype describing mastitis susceptibility based on these high-risk phenotypes. Cows (N = 471) were genotyped on the Illumina BovineHD 777K SNP chip and considering all 14 traits of interest, a total of 56 genome-wide associations (GWA) were performed and 28 significantly associated quantitative trait loci (QTL) were identified. Special focus was given to the aforementioned mastitis risk traits, and candidate gene investigation revealed both immune function and cell proliferation related genes in the areas surrounding associated QTL, suggesting that selecting for mastitis resistant cows based on these traits would be an effective method for increasing mastitis resiliency in a herd. Tracking the progression of SCS during the study period identified extreme populations of cows that remained “chronic” (SCS ≥ 4) or “healthy” (SCS

< 4). Fixation indices were calculated and 2 SNPs identified that demonstrated moderate allelic differentiation of “healthy” from both “chronic” and “average” cows (FST ≥ 0.4). GWAs were performed for SCS at each sampled stage in lactation, area under the SCS curve, and median SCS, and each approach significantly associated unique QTL spanning the genome. This suggests that alternative methods to lactation average SCS must be employed to more efficiently select for mastitis-resistant cows.

BIOGRAPHICAL SKETCH

Asha Miles was born in Ithaca, NY, just before her family moved to California’s Central Valley. Growing up in a little town in rural San Joaquin County, her passion for agriculture sprouted when she joined the 4-H program in high school. She graduated high school in 2008 during which she spent three years in a Biotechnology Regional Occupation Program. She attended University of California at Davis and earned her Bachelor of Science in Biotechnology in 2012, where she completed an undergraduate honors research thesis in which she created a bioassay for endocrine disrupters in the glucocorticoid signaling pathway. Shortly thereafter, she completed a Master of Science in Animal Biology in 2015, examining the differential expression of antimicrobial peptides in the porcine small intestine after E. coli challenge and treatment with human lysozyme transgenic goat milk. Returning to her Ithaca roots, she enrolled in Cornell’s Animal Science doctoral program in 2015, investigating the genetics underlying mastitis in dairy cattle. Upon completion of her Ph.D., she will begin a postdoctoral scholar position at Pennsylvania State University exploring the role of the microbiome in food animal health and disease. As a biotechnologist, she believes the technological augmentation of agriculture is a critical tool. As an academic, she believes progressive change always begins with education. As a humanitarian, she believes that technology and education can only have meaning if brought to people who need them in an accessible manner. For these reasons, she is dedicated to agricultural research and the global dissemination of information to multiple facets of society, including scientists, regulators, students, and farmers.

To all the teachers, learners, and inspirers I’ve met along the way.

ACKNOWLEDGEMENTS

This work would not have been possible without the support of my Special Committee, the work of many graduate and undergraduate students associated with Cornell Animal Science, the patience of the Cornell Statistical Consulting Unit, research funding by the NIFA Animal Health Project #NYC-127898, personal funding by Investigative Biology in the College of Arts and Sciences, the participation of dairy farmers, and of course, their cows.

vii

TABLE OF CONTENTS

BIOGRAPHICAL SKETCH …………………………………………………………………v DEDICATION ……………………………………………………………………………... vi ACKNOWLEDGEMENTS ……………………………………………………………….. vii

CHAPTER 1: INTRODUCTION 1.A: Origins …………………………………………………………………………. 1 1.B: A critical issue for the dairy industry ………………………………………….. 2 1.C: Genetic evaluation of complex traits …………………………………………... 5 1.D: Mastitis and genomic selection ………………………………………………... 9 1.E: Current understanding and gaps in knowledge ……………………………….. 11 References …………………………………………………………………………. 14

CHAPTER 2: UDDER AND TEAT CONFORMATIONAL RISK FACTORS FOR ELEVATED SOMATIC CELL COUNT AND CLINICAL MASTITIS IN NEW YORK HOLSTEINS Title Page …………………………………………………………………………. 23

Abstract …………………………………………………………………………… 24 Introduction ……………………………………………………………………….. 26 Materials and Methods ……………………………………………………………. 27 Results …………………………………………………………………………….. 33 Discussion ………………………………………………………………………… 40 Acknowledgements ……………………………………………………………….. 45 References ………………………………………………………………………… 46

CHAPTER 3: GENOME WIDE ASSOCIATIONS FOR UDDER AND TEAT CONFORMATIONAL RISK FACTORS FOR MASTITIS IN HOLSTEIN COWS Title Page ………………………………………………………………………….. 50 Abstract ……………………………………………………………………………. 51 Introduction ………………………………………………………………………... 53 Materials and Methods …………………………………………………………….. 55 Results ……………………………………………………………………………... 57 Discussion …………………………………………………………………………. 64 Conclusions ………………………………………………………………………... 72 Declarations ………………………………………………………………………... 73

viii

References …………………………………………………………………………. 75

CHAPTER 4: TIME- AND POPULATION-DEPENDENT GENETIC PATTERNS UNDERLIE BOVINE MILK SOMATIC CELL COUNT Title Page ………………………………………………………………………….. 90 Abstract ……………………………………………………………………………. 91 Introduction ………………………………………………………………………... 93 Materials and Methods …………………………………………………………….. 94 Results ……………………………………………………………………………... 99 Discussion ………………………………………………………………..………. 105 Conclusions ………………………………………………………………………. 114 Acknowledgements ..……………………………………………………………... 115 References ..………………………………………………………………………. 116

CHAPTER 5: IMPLICATIONS FOR FUTURE RESEARCH AND CONCLUDING REMARKS

Concluding Remarks.……………………………………………………………... 139 References ..………………………………………………………………………. 144

LIST OF FIGURES

Figure 1.1. The “Iceberg Principle”………………………………………………………….. 4 Figure 2.1. The characterization of teat ends………………………………………………..32 Figure 3.1. Creation of a new “mastitis risk” trait…………………………………………..59 Figure 3.2. Manhattan plot for fore udder attachment………………………………………61 Figure 3.3. Manhattan plot for udder depth…………………………………………………62 Figure 3.4. Manhattan plot for udder height………………………………………………...63 Figure 3.5. Manhattan plot for rear teat end shape………………………………………….65 Figure 3.6. Manhattan plot for rear teat width………………………………………………66 Supplementary Figure 3.1. Quantile-Quantile plots………………………………………...86 Figure 4.1. Overview of SCS-based population stratification………………………………97 Figure 4.2. FST by marker comparison of populations……………………………………..101 Figure 4.3. FST by marker and linkage disequilibrium on BTA 4………………………….102 Figure 4.4. Manhattan plots for SCS by lactation stage……………………………………106 Supplementary Figure 4.1. Quantile-Quantile plots………………………………………..129 Figure 5.1. Gene ontology across all approaches…………………………………………..141

LIST OF TABLES

Table 2.1. Explanation of six selected sampling periods………………………………....…29 Table 2.2. A description of the cow demographics………………………………………….34 Table 2.3. Udder and teat trait frequencies……………………………………………… ….36 Table 2.4. Contrast odds ratios for elevated SCC model……………………………………38 Table 2.5. Contrast odds ratios for clinical mastitis model………………………………….39 Table 2.6. The Goodman-Kruskal Gamma statistic of association………………………….41 Table 2.7. Univariate analyses of variables of interest……………………………………... 42 Table 3.1. Udder and teat genome-wide association models………………………………. 58 Supplementary Table 3.1. Significantly associated QTL and candidate genes…………….. 87 Table 4.1. Descriptive statistics of each SCS phenotype………………………………….. 100 Table 4.2. Differentiating markers and their allele frequencies……………………………103 Table 4.3. SCS genome-wide association models………………………………………… 104 Supplementary Table 4.1. Differentiated QTL and candidate genes……………………… 130 Supplementary Table 4.2. Significantly associated QTL and candidate genes…………… 132

CHAPTER 1: INTRODUCTION

1.A: Origins

The term “mastitis” originates from the Greek mastos, meaning “breast”, and -itis, a suffix meaning “inflammation”. Inflammation etymology can be traced to the Latin inflammationem, which may be translated as “a kindling, a setting on fire”. Consequently, inflammation can be clinically described as heat, pain, redness, and swelling; symptoms which reflect the influence of inflammatory mediators on the infected tissue. Heat, redness, and swelling are accounted for by the increased permeability of blood vessels, subsequent leakage of fluid and proteins, and pain explained by the migration and action of immune response cells.

We define these effects as “clinical mastitis”, producing visible abnormalities in the milk and mammary gland. Mastitis can also present subclinically, where inflammatory effects are only detectable with diagnostic tests for microorganisms and/or the influx of inflammatory response cells (somatic cells) in the milk. This host response to threats on biological homeostasis is a protective function intended destroy the pathogen and return the mammary gland to normal function. The innate immune response predates metazoans, and adaptive immunity first appeared in jawed fish prior to the evolution of mammals; it is therefore likely that mastitis has been a prevalent issue since mammals first appeared in the late Triassic period, an estimated

200 million years ago (Novacek, 1997; Flajnik and Kasahara, 2010; Muthamilarasan and

Prasad, 2013).

Archaeozoological studies exploring mitochondrial DNA variation suggest that cattle were first domesticated approximately 10,500 years ago, and historical evidence suggests that humans began milking cows around 3000 years after that (Itan et al., 2009; Bollongino et al.,

2012). We know today that mastitis is propagated by both environmental and contagious pathogens, the latter of which is often spread during milking. It is likely that concerns with the management and control of mastitis began in Central Europe with early humans and the first dairy farmers. However, our ability to understand mastitis was limited until the advent and improvement of the light microscope in the 17th century, and consequently the earliest scientific report on bovine mastitis came in 1917, when Breed and Brew presented their microscopic method of grading dairies based on the quantity of bacteria in milk (Breed and Brew, 1917;

Wollman et al., 2015). Since then, research efforts have evolved from an initial focus on understanding the etiology of intramammary infections to the application of preventative measures including investigating genetic predisposition to mastitis and the possibility of breeding cows resistant to infection.

1.B: A critical issue for the dairy industry

Mastitis is internationally regarded as an economic problem with animal welfare and public health implications. While the mortality rate for clinical mastitis is very low, farms with mastitis problems suffer substantial economic loss due to factors such as milk production decrease, diagnostic tests, drug treatments, discarded milk, veterinary services, treatment labor, product quality decline, as well as culling (Seegers et al., 2003). Considering all of these factors, a recent study estimated the average cost of a case of clinical mastitis to be $444, suggesting that annual losses in the United States alone amount to nearly 2 billion dollars (Rollin et al.,

2015; iGEM, 2016). Secondary to lameness, mastitis is the leading antagonist of animal health and welfare, and improving cow recovery and reducing infection severity is considered to be

the fastest way to enhance cow welfare with regards to intramammary infection (Hillerton and

Berry, 2005). Furthermore, bacterial presence in milk poses a public health problem, as mastitis pathogens have been known to cause human disease and improper or lack of pasteurization techniques can result in the contamination of food products (Oliver et al., 2005). Additionally, the often low cure rates and potential antibiotic residues in milk lead to fears of increasing antimicrobial resistance and a call for methods of mastitis control other than antibiotic treatment

(Gomes and Henriques, 2016).

A major obstacle for mastitis control lies in the problem of disease detection. A nation- wide survey estimated clinical mastitis to have 25% prevalence, but with no mandate for participation in monthly milk testing there is no way to accurately measure the occurrence of subclinical mastitis in United States dairy herds (National Animal Health and Monitoring

System, 2016). This is known as the “Iceberg Principle”, where only a small proportion of information is visible while the bulk of the information about a phenomenon is unavailable or undetectable (Figure 1.1). In the case of mastitis, the problem extends below the subclinical mastitis tier to the entire population of milking cows, all of whom are susceptible to intramammary infection. In response to this high, population-wide risk for developing mastitis, stringent protocols for mastitis management and control have been implemented. Approaches have included reducing exposure to pathogens by prioritizing improvements to cow housing, milking hygiene, and minimizing teat lesions (Dodd, 1983; Bartlett et al., 1992). Additional emphasis has been given to milking machine effects and examining resistance of the teat canal itself to bacterial invasion, considering factors like over-milking, milk flow, vacuum level, liner slips, and pulsation (Thompson et al., 1978; Natzke et al., 1982; Baxter et al., 1992; Lacy-

Hulbert and Hillerton, 1995; Wieland et al., 2018). Research focus has also been given to the

Figure 1.1. The “Iceberg Principle”. Only a small percentage of mastitis cases are clinically detectable. Beneath those cases a much larger number of cows have subclinical mastitis, and beneath that there is an even larger population of susceptible cows who are at risk of developing mastitis.

potential dispersal of bacteria in the udder by considering factors like hand-stripping to remove foremilk and frequency of udder evacuations (Mein et al., 2004; Shoshani et al., 2017). Other external factors such as cow age, season, and climate also have been examined for their influence on mastitis occurrence (Fox et al., 1995; Bates and Dohoo, 2016; Zhang et al., 2016b).

While numerous studies have shown regulation of these factors to reduce the incidence of mastitis on farm, they do not address the effect of the individual cow on mastitis susceptibility.

Mastitis is well-established as a multi-faceted issue with an estimated 30% of mastitis problems attributable each to farm and milking management, 20% to the milking machine, and 20% to the cow herself (Mein et al., 2004). In 1994, milk somatic cell count (SCC) was added to national genomic evaluation systems specifically to address the high prevalence of mastitis in

United States dairy herds (Schutz, 1994).

1.C: Genetic evaluation of complex traits

Mastitis is an example of a complex trait, where the phenotype can be attributed to multiple genes, their interactions with each other, and the environment. The genetics underlying complex traits such as disease susceptibility are investigated using a quantitative genetics approach because, as compared to simple Mendelian inheritance, more advanced statistical methods are required to investigate traits which are both polygenic and influenced by environmental factors. Such traits have a complicated genotype to phenotype relationship, where a single genotype may produce a range of phenotypes based on the genotype by environment interaction. Furthermore, the wide phenotypic ranges for each genotype may overlap, presenting a challenge in distinguishing whether individuals differ due to genetic or environmental factors. This phenomenon is amplified by the polygenic nature of the trait,

resulting in a multitude of highly varied phenotypes and creating an obstacle to discovering the genetic regulation of the trait. Chromosomal regions responsible for a small proportion of the genetic variation in a complex trait are called quantitative trait loci (QTL) and their identification and mapping are key to understanding the genetics underlying complex traits.

The use of genetics in bovine production originated with blood-typing for parentage verification to ensure correct pedigree records, but advances in molecular methods allowed the commercial utilization of marker-based parentage testing by the turn of the 21st century

(Stormont, 1967; Dekkers, 2004). However, genetic progress attributable to marker-based selection was limited due to the difficulty in generating high quality data sets, as well as the realization that most economically important dairy cattle traits are controlled by many genes with small effects (Cole et al., 2009). This was remedied following the successful sequencing of the human genome and the initiation of the Bovine Genome Project in 2003, which together with additional low-depth sequencing of multiple breeds provided enough information for single nucleotide polymorphism (SNP) assay development (Wiggans et al., 2017).

Unfortunately, the 15,036 SNPs on the first genotyping panel released in 2005 were poorly distributed across the genome and not able to adequately detect linkage disequilibrium (LD), making it inappropriate for use in QTL mapping and genomic selection (Khatkar et al., 2007;

Hayes et al., 2013). In December 2007, the Illumina BovineSNP50 BeadChip offering 54,001

SNPs was released, followed three years later by the high-density Bovine HD chip which boasted 777,962 SNPs (Matukumalli et al., 2009; Illumina, 2010). Genotype imputation of data from low density SNP chips has been used to maximize genome coverage while minimizing costs, and is based on the principle that any two individuals may share genetic information derived from a distant common ancestor. Under this assumption, low density genotypes may

be compared to a large reference population with no missing data (e.g., whole genome sequences), and their complete genotypes inferred or “imputed” based on the haplotypes (sets of markers inherited together) observed in the reference population (Das et al., 2018). While genotype imputation can dramatically increase genome coverage, it introduces bias towards population averages and limits researcher ability to detect rare variants which do not occur frequently in the reference population (Marchini and Howie, 2010; Calus et al., 2014; Marete et al., 2018a).

Prior to the availability of high density genomic data complex traits were primarily analyzed via QTL mapping. This approach is based on the concept of linkage disequilibrium, which contrary to Mendel’s principle of independent assortment, suggests that alleles at different loci are linked and may be inherited together. This method assumes markers that are consistently associated with the inheritance of a particular trait must be linked with a QTL which contributes to that trait. Note that QTL mapping is traditionally performed within families or in controlled crosses of contrasting parents for the trait of interest, where only one recombination event is recorded in the population, and the F1 generation is assessed for correlation between the inheritance of a particular allele and a quantitative phenotype (Broman,

2001). This presents both a computing and financial problem when applied to large dairy cattle populations, considering high genotyping costs, attempts to increase resolution with advanced intercrosses, and the long generation intervals required for genotyping progeny: enter the genome-wide association study.

Unlike traditional linkage mapping, which assesses whether a disease and an allele show correlated transmission within a pedigree, genome-wide association studies (GWAS) associate markers and traits within a freely interbreeding population and do not consider familial

inheritance patterns (Lander and Schork, 1994). These GWAS were originally conducted as case-control studies comparing allelic frequencies in affected and unaffected populations; appropriately selecting a control group is a major challenge in contrast to linkage methods, which do not require one. While GWAS are high resolution and less labor intensive, they are sensitive to underlying population structure which can cause spurious associations (Li, 1969;

Platt et al., 2010). This is a particular problem in dairy cattle, in which desirable traits have been selected for over centuries using both phenotypic and genotypic data, meaning that positive associations must be cautiously interpreted. Observed positive associations may occur because the allele in question is actually causative or if that allele is in LD with the actual cause, but disturbingly, can also occur as an artifact of population admixture. In other words, if independent alleles and traits occur in high frequency within a certain population they may be falsely associated with one another. This problem is best ameliorated by conducting GWAS only in large, homogenous populations and by selecting control groups which are perfectly matched for ancestry (Lander and Schork, 1994). Because that is impossible in practice, numerous methods have been implemented to account for bias introduced by underlying population structure, including selection of internal controls by haplotype relative risk estimation and the inclusion of identity-by-state matrices in GWAS models (Falk and

Rubinstein, 1987; Kang et al., 2010).

Other challenges with the genetic analysis of complex traits lie in the determination of statistical significance. A GWAS represents a massive multiple testing problem, where depending on the genotype density researchers may be running millions of regressions for genetic variants against a single trait. To mitigate association inflation with false positives, a

Bonferroni multiple testing correction may be applied, or the less stringent False Discovery

Rate calculation may be performed, where the researcher sets a significance level accounting for a certain proportion of falsely associated QTL (Benjamini and Hochberg, 1995; Weller et al., 1998). Despite measures to minimize false positives and correct population structure, the only evidence that positive associations are “real” is repeatability in independent populations, thus GWAS validation is crucial. This presents a challenge in dairy cattle where the overuse of favorite sires have contributed to high inbreeding and replicated population structure in multiple data sets, suggesting that the most convincing validation is across breeds (Forutan et al., 2018).

To minimize these confounding issues, quality control measures should also be applied to all genetic data, including the filtering of highly related animals, poorly represented SNPs in the study population, the removal of rare variants which GWAS do not have the power to detect, and evaluation of Hardy-Weinberg equilibrium as a potential indicator of non-random influences on the population. Significantly associated markers are most compelling when linked with biologically relevant genes, which should be further validated through targeted sequencing and functional analyses at the transcriptional and translational level.

1.D: Mastitis and genomic selection

Reports on mastitis heritability, or the degree of variation in mastitis that can be explained by additive genetic differences, were made as early as 1944 and have ranged from

0.1 up to 0.88 (Gaunt et al., 1980; Shook and Schutz, 1994; Nash et al., 2000; Vallimont et al.,

2009; Ruegg, 2017). The lack of consensus for mastitis heritability estimates is due not only to high variation in management systems and environments, but the lack of standardized mastitis diagnosis and reporting. Consequently, genetic selection for mastitis resistance stagnated until the widespread adoption of Dairy Herd Improvement testing and milk SCC measures, which

though founded in 1905, still has only 46% nation-wide participation today (Voelker, 1981;

DairyOne, 2019). Nevertheless, significant effort has been made to determine genetic markers associated with mastitis susceptibility or resistance.

Current genetic evaluation for mastitis is based on indicator traits, where measures like somatic cell score (SCS) and udder composite traits are used as a proxy due to the time and financial complications of measuring clinical mastitis directly. However, the predictions for these traits have only minimally reduced mastitis incidence in U.S. dairy herds, likely because of the incomplete correlation of mastitis with these proxy traits, possible inaccuracy in farm diagnosis and recording, as well as the inability to standardize across farms (Wenz and Giebel,

2012; Vukasinovic et al., 2017). Furthermore, due to high genotyping costs these genetic predictions have been historically made through indirect associations of daughter production traits with bull genotypes. Daughter yield deviations (DYD) are averages of daughter performance adjusted for the effects of their dam, and because they are not regressed on their sire’s breeding values, are considered the most accurate and unbiased measure of daughter performance and typically used in national genomic evaluations (Szyda et al., 2008). These evaluations include a genomic prediction estimated for SNP effects on the traits of interest, information from traditional evaluations (based solely on performance and pedigree data), and a “subset evaluation” to determine what proportion of the traditional data was accounted for by genomics (VanRaden et al., 2009). Genomic evaluation accuracy is measured via reliabilities, a measure of the standard error of genomic breeding values, which are routinely evaluated to maximize genomic improvement efficiency (Liu et al., 2017).

Selection indices are based on these genomic evaluations, and calculated by the weighted regression of genomic prediction values for traits of interest. The most common

selection index used by United States producers in culling and breeding decisions is Net Merit, in which the only considerations of mastitis are the inclusion of SCS with a reliability of merely

75%, and an udder composite score with reliability of 80% (Wiggans et al., 2017). While 75% is much higher than the reliability for SCS using traditional evaluations and no incorporation of genomic information (35.3%), it is considerably lower than reliabilities for directly evaluated phenotypes like protein, fat, and milk yield (all 90%) (VanRaden and Cole, 2014). Diminished reliabilities in genomic evaluation for SCS relate back to the complex trait problem. Numerous human medical studies have struggled to identify the genetics underlying disease when armed only with a generalized clinical phenotype, a problem remedied by restricting case definitions to specific etiologies and comparing extremes in the distribution of continuous traits (Lander and Schork, 1994). Somatic cell score as a generalized proxy for mastitis is an inefficient phenotype, given the varied pathogenesis of mastitis and the non-pathogenic reasons for inflammation in the mammary gland, including injury, lactogenesis, and changes in metabolic state across a cow’s lactation (Bionaz et al., 2007; Graugnard et al., 2012; Akbar et al., 2015).

1.E: Current understanding and gaps in knowledge

Progress on the investigation of quantitative traits in cattle is documented in the Cattle

Quantitative Trait Locus Database (Cattle QTLdb) to facilitate meta-analyses and validation of association data (Hu et al., 2013). As of the most recent release in August 2019, a total of

2,045 QTL have been associated with the mastitis phenotypes of SCC, SCS, and clinical mastitis diagnosis. These QTL span all 29 autosomes and the X chromosome, and were generated via a combination of traditional linkage analysis, QTL mapping, and genome-wide association studies (CattleQTLdb, 2019). The complex, polygenic nature of mastitis is

reinforced by the significant association of many different regions across the genome, but also brings into question whether these genomic loci can be refined with more selective phenotyping.

Though udder and teat morphology are widely accepted to influence mastitis susceptibility, especially given their relation to complete milk evacuation and the effect of the milking machine, there is little consensus in the literature regarding their heritabilities and direct association to mastitis risk (Seykora and McDaniel, 1985). Furthermore, udder composite scores included in Net Merit estimations do not currently include teat characteristics, and teat width and end shape have never been used for genomic evaluation (Wiggans et al., 2017). Most extant research on the genetics of udder and teat type traits have relied on traditional evaluation methods and the indirect association of bull genotypes with the phenotypes of their daughters

(Rupp and Boichard, 1999; Nash et al., 2000; Chrystal et al., 2001; DeGroot et al., 2002).

Similarly, there is a long history of genetic evaluation of SCC, but these studies have relied on genotype imputation which may introduce error and bias allele frequencies towards the population average, as well as indirect genotype-phenotype associations, de-regressed breeding values, and pedigree-based methods to estimate disease allele inheritance (Zhang et al., 1998;

Cole et al., 2011a; Biffani et al., 2017; Fang et al., 2017). Furthermore, there is a need to refine the use of SCS as a mastitis indicator by removing phenotype bias introduced by non- pathogenic causes of inflammation, such as lactogenesis and uterine involution during early lactation (Bionaz et al., 2007; Graugnard et al., 2012).

This dissertation is aimed towards addressing these gaps in mastitis genetic investigation. Our objectives were to first assemble thorough cow phenotypes in a prospective cohort study that allows researcher control over exposure and disease assessment and

consideration of multiple indicator traits for mastitis. Secondly, our objective was to assemble high density genotypes on the same cows to determine a direct genotype to phenotype relationship. The major findings of this project are detailed in the following chapters.

REFERENCES

Akbar, H., Grala, T.M., Vailati Riboni, M., Cardoso, F.C., Verkerk, G., McGowan, J.,

Macdonald, K., Webster, J., Schutz, K., Meier, S., Matthews, L., Roche, J.R., Loor, J.J.,

2015. Body condition score at calving affects systemic and hepatic transcriptome

indicators of inflammation and nutrient metabolism in grazing dairy cows. Journal of

Dairy Science 98, 1019-1032.

Bartlett, P.C., Miller, G.Y., Lance, S.E., Hancock, D.D., Heider, L.E., 1992. Managerial risk

factors of intramammary infection with Streptococcus agalactiae in dairy herds in Ohio.

American journal of veterinary research 53, 1715-1721.

Bates, A.J., Dohoo, I., 2016. Risk factors for peri-parturient farmer diagnosed mastitis in New

Zealand dairy herds: findings from a retrospective cohort study. Prev Vet Med 127, 70-

76.

Baxter, J.D., Rogers, G.W., Spencer, S.B., Eberhart, R.J., 1992. The effect of milking machine

liner slip on new intramammary infections. J Dairy Sci 75, 1015-1018.

Benjamini, Y., Hochberg, Y., 1995. Controlling the False Discovery Rate: A Practical and

Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series

B (Methodological) 57, 289-300.

Biffani, S., Del Corvo, M., Capoferri, R., Pedretti, A., Luini, M., Williams, J.L., Pagnacco, G.,

Minvielle, F., Minozzi, G., 2017. An alternative experimental case–control design for

genetic association studies on bovine mastitis. animal 11, 574-579.

Bionaz, M., Trevisi, E., Calamari, L., Librandi, F., Ferrari, A., Bertoni, G., 2007. Plasma

Paraoxonase, Health, Inflammatory Conditions, and Liver Function in Transition Dairy

Cows. Journal of Dairy Science 90, 1740-1750.

Bollongino, R., Burger, J., Powell, A., Mashkour, M., Vigne, J.-D., Thomas, M.G., 2012.

Modern Taurine Cattle Descended from Small Number of Near-Eastern Founders.

Molecular Biology and Evolution 29, 2101-2104.

Breed, R.S., Brew, J.D., 1917. The Control of Public Milk Supplies by the Use of the

Microscopic Method1. Journal of Dairy Science 1, 259-271.

Broman, K.W., 2001. Review of statistical methods for QTL mapping in experimental crosses.

Lab animal 30, 44-52.

Calus, M.P., Bouwman, A.C., Hickey, J.M., Veerkamp, R.F., Mulder, H.A., 2014. Evaluation

of measures of correctness of genotype imputation in the context of genomic prediction:

a review of livestock applications. Animal 8, 1743-1753.

CattleQTLdb, 2019. QTL/associations for trait type Mastitis in the Cattle Genome.

Chrystal, M.A., Seykora, A.J., Hansen, L.B., Freeman, A.E., Kelley, D.H., Healey, M.H., 2001.

Heritability of Teat-End Shape and the Relationship of Teat-End Shape with Somatic

Cell Score for an Experimental Herd of Cows1,2. Journal of Dairy Science 84, 2549-

2554.

Cole, J.B., VanRaden, P.M., O'Connell, J.R., Van Tassell, C.P., Sonstegard, T.S., Schnabel,

R.D., Taylor, J.F., Wiggans, G.R., 2009. Distribution and location of genetic effects for

dairy traits. Journal of Dairy Science 92, 2931-2946.

Cole, J.B., Wiggans, G.R., Ma, L., Sonstegard, T.S., Lawlor, T.J., Crooker, B.A., Van Tassell,

C.P., Yang, J., Wang, S., Matukumalli, L.K., Da, Y., 2011. Genome-wide association

analysis of thirty one production, health, reproduction and body conformation traits in

contemporary U.S. Holstein cows. BMC Genomics 12, 408.

DairyOne, 2019. Dairy Herd Improvement Association: What is it? Who is it?

Das, S., Abecasis, G.R., Browning, B.L., 2018. Genotype Imputation from Large Reference

Panels. Annual review of genomics and human genetics 19, 73-96.

DeGroot, B.J., Keown, J.F., Van Vleck, L.D., Marotz, E.L., 2002. Genetic Parameters and

Responses of Linear Type, Yield Traits, and Somatic Cell Scores to Divergent Selection

for Predicted Transmitting Ability for Type in Holsteins1. Journal of Dairy Science 85,

1578-1585.

Dekkers, J.C., 2004. Commercial application of marker- and gene-assisted selection in

livestock: strategies and lessons. J Anim Sci 82 E-Suppl, E313-328.

Dodd, F.H., 1983. Mastitis--progress on control. J Dairy Sci 66, 1773-1780.

Falk, C.T., Rubinstein, P., 1987. Haplotype relative risks: an easy reliable way to construct a

proper control sample for risk calculations. Annals of human genetics 51, 227-233.

Fang, L., Sahana, G., Su, G., Yu, Y., Zhang, S., Lund, M.S., Sørensen, P., 2017. Integrating

Sequence-based GWAS and RNA-Seq Provides Novel Insights into the Genetic Basis

of Mastitis and Milk Production in Dairy Cattle. Sci Rep 7, 45560-45560.

Flajnik, M.F., Kasahara, M., 2010. Origin and evolution of the adaptive immune system:

genetic events and selective pressures. Nat Rev Genet 11, 47-59.

Forutan, M., Ansari Mahyari, S., Baes, C., Melzer, N., Schenkel, F.S., Sargolzaei, M., 2018.

Inbreeding and runs of homozygosity before and after genomic selection in North

American Holstein cattle. BMC Genomics 19, 98-98.

Fox, L.K., Chester, S.T., Hallberg, J.W., Nickerson, S.C., Pankey, J.W., Weaver, L.D., 1995.

Survey of intramammary infections in dairy heifers at breeding age and first parturition.

J Dairy Sci 78, 1619-1628.

Gaunt, S.N., Raffio, N., Kingsbury, E.T., Damon, R.A., Jr., Johnson, W.H., Mitchell, B.A.,

1980. Variation of lactoferrin and mastitis and their heritabilities. J Dairy Sci 63, 1874-

1880.

Gomes, F., Henriques, M., 2016. Control of Bovine Mastitis: Old and Recent Therapeutic

Approaches. Current microbiology 72, 377-382.

Graugnard, D.E., Bionaz, M., Trevisi, E., Moyes, K.M., Salak-Johnson, J.L., Wallace, R.L.,

Drackley, J.K., Bertoni, G., Loor, J.J., 2012. Blood immunometabolic indices and

polymorphonuclear neutrophil function in peripartum dairy cows are altered by level of

dietary energy prepartum1. Journal of Dairy Science 95, 1749-1758.

Hayes, B.J., Lewin, H.A., Goddard, M.E., 2013. The future of livestock breeding: genomic

selection for efficiency, reduced emissions intensity, and adaptation. Trends Genet 29,

206-214.

Hillerton, J.E., Berry, E.A., 2005. Treating mastitis in the cow--a tradition or an archaism.

Journal of applied microbiology 98, 1250-1255.

Hu, Z.-L., Park, C.A., Wu, X.-L., Reecy, J.M., 2013. Animal QTLdb: an improved database

tool for livestock animal QTL/association data dissemination in the post-genome era.

Nucleic acids research 41, D871-D879. iGEM, C., 2016. Market Analysis: Bovine Mastitis.

Illumina, 2010. Bovine HD Genotyping BeadChip Data Sheet., Illumina, San Diego, CA.

Itan, Y., Powell, A., Beaumont, M.A., Burger, J., Thomas, M.G., 2009. The Origins of Lactase

Persistence in Europe. PLOS Computational Biology 5, e1000491.

Kang, H.M., Sul, J.H., Service, S.K., Zaitlen, N.A., Kong, S.-y., Freimer, N.B., Sabatti, C.,

Eskin, E., 2010. Variance component model to account for sample structure in genome-

wide association studies. Nature Genetics 42, 348.

Khatkar, M.S., Zenger, K.R., Hobbs, M., Hawken, R.J., Cavanagh, J.A.L., Barris, W.,

McClintock, A.E., McClintock, S., Thomson, P.C., Tier, B., Nicholas, F.W., Raadsma,

H.W., 2007. A primary assembly of a bovine haplotype block map based on a 15,036-

single-nucleotide polymorphism panel genotyped in Holstein-Friesian cattle. Genetics

176, 763-772.

Lacy-Hulbert, S.J., Hillerton, J.E., 1995. Physical characteristics of the bovine teat canal and

their influence on susceptibility to streptococcal infection. The Journal of dairy research

62, 395-404.

Lander, E.S., Schork, N.J., 1994. Genetic dissection of complex traits. Science 265, 2037.

Li, C.C., 1969. Population subdivision with respect to multiple alleles. Annals of human

genetics 33, 23-29.

Liu, Z., VanRaden, P.M., Lidauer, M.H., Calus, M.P., Benhajali, H., Jorjani, H., Ducrocq, V.,

2017. Approximating Genomic Reliabilities for National Genomic Evaluation. Interbull

Bulletin Number 51, 75-85.

Marchini, J., Howie, B., 2010. Genotype imputation for genome-wide association studies. Nat

Rev Genet 11, 499-511.

Marete, A., Lund, M.S., Boichard, D., Ramayo-Caldas, Y., 2018. A system-based analysis of

the genetic determinism of udder conformation and health phenotypes across three

French dairy cattle breeds. PLOS ONE 13, e0199931.

Matukumalli, L.K., Lawley, C.T., Schnabel, R.D., Taylor, J.F., Allan, M.F., Heaton, M.P.,

O'Connell, J., Moore, S.S., Smith, T.P.L., Sonstegard, T.S., Van Tassell, C.P., 2009.

Development and Characterization of a High Density SNP Genotyping Assay for Cattle.

Plos One 4, 13.

Mein, G., Reinemann, D., Schuring, N., Ohnstad, I., 2004. Milking machines and mastitis risk:

a storm in a teatcup. In, National Mastitis Council Annual Meeting, Charlotte, NC.

Muthamilarasan, M., Prasad, M., 2013. Plant innate immunity: an updated insight into defense

mechanism. Journal of biosciences 38, 433-449.

Nash, D.L., Rogers, G.W., Cooper, J.B., Hargrove, G.L., Keown, J.F., Hansen, L.B., 2000.

Heritability of Clinical Mastitis Incidence and Relationships with Sire Transmitting

Abilities for Somatic Cell Score, Udder Type Traits, Productive Life, and Protein Yield.

Journal of Dairy Science 83, 2350-2360.

National Animal Health and Monitoring System, 2016. Dairy 2014: Milk quality, milking

procedures, and mastitis on U.S. dairies. United States Department of Agriculture, Fort

Collins, CO.

Natzke, R.P., Everett, R.W., Bray, D.R., 1982. Effect of overmilking on udder health. J Dairy

Sci 65, 117-125.

Novacek, M.J., 1997. Mammalian evolution: An early record bristling with evidence. Current

Biology 7, R489-R491.

Oliver, S., Jayarao, B., Almeida, R., 2005. Foodborne pathogens, mastitis, milk quality, and

dairy food safety.

Platt, A., Vilhjálmsson, B.J., Nordborg, M., 2010. Conditions Under Which Genome-Wide

Association Studies Will be Positively Misleading. Genetics 186, 1045-1052.

Rollin, E., Dhuyvetter, K.C., Overton, M.W., 2015. The cost of clinical mastitis in the first 30

days of lactation: An economic modeling tool. Prev Vet Med 122, 257-264.

Ruegg, P.L., 2017. A 100-Year Review: Mastitis detection, management, and prevention.

Journal of Dairy Science 100, 10381-10397.

Rupp, R., Boichard, D., 1999. Genetic parameters for clinical mastitis, somatic cell score,

production, udder type traits, and milking ease in first lactation Holsteins. J Dairy Sci

82, 2198-2204.

Schutz, M.M., 1994. Genetic Evaluation of Somatic Cell Scores for United States Dairy Cattle.

Journal of Dairy Science 77, 2113-2129.

Seegers, H., Fourichon, C., Beaudeau, F., 2003. Production effects related to mastitis and

mastitis economics in dairy cattle herds. Vet Res 34, 475-491.

Seykora, A.J., McDaniel, B.T., 1985. Udder and teat morphology related to mastitis resistance:

a review. J Dairy Sci 68, 2087-2093.

Shook, G.E., Schutz, M.M., 1994. Selection on somatic cell score to improve resistance to

mastitis in the United States. J Dairy Sci 77, 648-658.

Shoshani, E., Cohen, M., Doekes, J.J., 2017. Short- and long-term effects of high milking

frequency during the first 21 days of lactation on production and reproductive

performance in high-lactating cows. Animal 11, 91-100.

Stormont, C., 1967. CONTRIBUTION OF BLOOD TYPING TO DAIRY SCIENCE

PROGRESS. Journal of Dairy Science 50, 253-&.

Szyda, J., Ptak, E., Komisarek, J., Zarnecki, A., 2008. Practical application of daughter yield

deviations in dairy cattle breeding. Journal of applied genetics 49, 183-191.

Thompson, P.D., Schultze, W.D., Sauls, J.N., Arapis, S.C., 1978. Mastitis infection from abrupt

loss of milking vacuum. J Dairy Sci 61, 344-351.

Vallimont, J.E., Dechow, C.D., Sattler, C.G., Clay, J.S., 2009. Heritability estimates associated

with alternative definitions of mastitis and correlations with somatic cell score and yield.

J Dairy Sci 92, 3402-3410.

VanRaden, P.M., Cole, J.B., 2014. Net merit as a measure of lifetime profit: 2014 revision.

Animal Improvement Program Research Report NM$5 (10-14). Animal Genomics and

Improvement Laboratory, Agricultural Research Service, USDA, Beltsville, MD 20705.

VanRaden, P.M., Van Tassell, C.P., Wiggans, G.R., Sonstegard, T.S., Schnabel, R.D., Taylor,

J.F., Schenkel, F.S., 2009. Invited review: Reliability of genomic predictions for North

American Holstein bulls. Journal of Dairy Science 92, 16-24.

Voelker, D.E., 1981. Dairy Herd Improvement Associations. Journal of Dairy Science 64,

1269-1277.

Vukasinovic, N., Bacciu, N., Przybyla, C.A., Boddhireddy, P., DeNise, S.K., 2017.

Development of genetic and genomic evaluation for wellness traits in US Holstein cows.

J Dairy Sci 100, 428-438.

Weller, J.I., Song, J.Z., Heyen, D.W., Lewin, H.A., Ron, M., 1998. A new approach to the

problem of multiple comparisons in the genetic dissection of complex traits. Genetics

150, 1699-1706.

Wenz, J.R., Giebel, S.K., 2012. Retrospective evaluation of health event data recording on 50

dairies using Dairy Comp 305. Journal of Dairy Science 95, 4699-4706.

Wieland, M., Nydam, D.V., Alveby, N., Wood, P., Virkler, P.D., 2018. Short communication:

Teat-end shape and udder-level milking characteristics and their associations with

machine milking-induced changes in teat tissue condition. J Dairy Sci 101, 11447-

11454.

Wiggans, G.R., Cole, J.B., Hubbard, S.M., Sonstegard, T.S., 2017. Genomic Selection in Dairy

Cattle: The USDA Experience. Annual Review of Animal Biosciences 5, 309-327.

Wollman, A.J.M., Nudd, R., Hedlund, E.G., Leake, M.C., 2015. From Animaculum to single

molecules: 300 years of the light microscope. Open biology 5, 150019-150019.

Zhang, Q., Boichard, D., Hoeschele, I., Ernst, C., Eggen, A., Murkve, B., Pfister-Genskow, M.,

Witte, L.A., Grignola, F.E., Uimari, P., Thaller, G., Bishop, M.D., 1998. Mapping

quantitative trait loci for milk production and health of dairy cattle in a large outbred

pedigree. Genetics 149, 1959-1973.

Zhang, Z., Li, X.P., Yang, F., Luo, J.Y., Wang, X.R., Liu, L.H., Li, H.S., 2016. Influences of

season, parity, lactation, udder area, milk yield, and clinical symptoms on

intramammary infection in dairy cows. J Dairy Sci 99, 6484-6493.

CHAPTER 2: UDDER AND TEAT CONFORMATIONAL RISK FACTORS FOR

ELEVATED SOMATIC CELL COUNT AND CLINICAL MASTITIS IN NEW YORK

HOLSTEINS

Asha M. Milesa, Jessica A. A. McArtb, Francisco A. Leal Yepesa, Cassandra R. Stambuka, Paul

D. Virklerbc, and Heather J. Husona1 aDepartment of Animal Science, and bDepartment of Population Medicine and Diagnostic Sciences, and cQuality Milk Production Services, Animal Health Diagnostic Center, Cornell University,

Ithaca, NY 14853

1Corresponding author: Heather J. Huson, 201 Morrison Hall, 507 Tower Road, Ithaca, NY

14853. Phone: (607) 255-2289. E-mail: [email protected]

Preventative Veterinary Medicine 163 (2019) 7-13

ABSTRACT

Our primary objective was to identify udder and teat conformational risk factors associated with the occurrence of elevated somatic cell count (SCC) and clinical mastitis using a prospective cohort study design with careful assessment of exposure and disease outcomes. Mastitis prevalence was evaluated by parity across 6 sampling periods representing key physiological transitions during lactation: 0 to 1 day in milk (DIM), 3 to 5 DIM, 10 to 14 DIM, 50 to 60 DIM, 90 to 110 DIM, and 210 to 230 DIM. Cows were scored for front and rear teat length, width, end shape, and placement, fore udder attachment, udder cleft, udder depth, rear udder height, and rear udder width. Two independent multivariable logistic regression models were used to generate odds ratios (OR) for elevated SCC (≥

200,000 cells/ml) and farm-diagnosed clinical mastitis. We identified that loose fore udder attachment

(reference level: strong fore udder attachment, OR = 2.1, 95% confidence interval (CI) = 1.2 to 3.8) and flat teat end shape (reference level: round teat end shape, OR = 1.4, 95% CI = 1.1 to 1.9) increased the odds of an elevated SCC event, whereas a negative California Mastitis Test score at 0 to 1 DIM decreased the odds of an elevated SCC event (OR = 0.6, 95% CI = 0.4 to 0.8). Loose fore udder attachment (reference level: strong fore udder attachment, OR = 3.7, 95% CI = 1.3 to 10.7), flat teat end shape (reference level: round teat end shape, OR = 1.5, 95% CI = 1.0 to 2.4), low rear udder height

(reference level: intermediate rear udder height, OR = 2.8, 95% CI = 0.3 – 6.2), and increasing rear teat width (OR = 2.2, 95% CI = 1.2 – 4.4) heightened the odds of developing clinical mastitis. We identified that within our study cohort, loose fore udder attachment and flat teat ends had an important association with increased odds of both an elevated SCC event and clinical mastitis diagnosis. The identification of these udder and teat conformational risk factors for mastitis can provide farmers an effective and inexpensive tool to manage mastitis.

Key words: mastitis, udder conformation, teat conformation, logistic regression

INTRODUCTION

Mastitis is one of the most prevalent diseases among United States dairy herds, and consequently each year the dairy industry suffers substantial economic loss (Fetrow, 2015;

Rollin et al., 2015). Mastitis prevalence can be attributed to problems with herd and farm management, milking management, the milking machine, and the cow itself (Mein et al., 2004).

Many herd-level risk factors for mastitis such as season, climate, and management have already been identified (Fox et al., 1995; Cook et al., 2002; Bates and Dohoo, 2016; Zhang et al.,

2016b), but identifying udder and teat conformational risk factors can improve our strategies for on-farm mastitis management by identifying high-risk animals.

Anecdotal commentary on relationships among parity, udder size, and mastitis is not uncommon, and some work has been done to illustrate the relationships among teat dimensions and mastitis (Zwertvaegher et al., 2013; Guarin and Ruegg, 2016). However, there is little consensus in the literature regarding the relationship of elevated somatic cell count (SCC) and clinical mastitis to various udder and teat traits, or their respective heritabilities (Seykora and

McDaniel, 1985). For example, some researchers report no association between SCC and teat end shape (Chrystal et al., 2001; Guarin et al., 2017), contrary to previous research indicating that less pointed teat ends may be associated with fewer cases of mastitis (Hodgson et al., 1980).

These disparities might be explained by retrospective study designs which do not always allow researcher control over exposure and disease assessment, mitigation of selection and information bias, and the ability to establish a time sequence. Indeed, most studies assessed disease as lactation averages of somatic cell score instead of considering test day data individually across a lactation period (Rogers et al., 1998; Rupp and Boichard, 1999; Koeck et al., 2012). Many of these extant studies examining udder and teat morphology emphasize

genetic correlations, and the external validity of these studies is under further question given the use of genotyped bulls and the indirect phenotyping of their daughters.

Therefore, our aim was to conduct a prospective cohort study with careful and consistent assessment of exposure and disease to elucidate the relationships among udder and teat conformation and the occurrence of elevated SCC and clinical mastitis at any time during lactation. This work was devised as an intensive genetics study to improve genomic selection for mastitis-resistant cows by examining multiple mastitis related traits using direct genotyping of cows to address the inherent limitations of characterizing mastitis genetics using sire evaluations. Here the epidemiological analysis of our phenotypic data is presented.

MATERIALS AND METHODS

Study Design

We conducted a prospective cohort study on 2 commercial Holstein herds in New York

State from June 2015 to July 2016. The enrollment period lasted from June to December 2015, with follow-up on all cows completed by July 2016. Farm 1 milked roughly 1,000 cows twice daily in a 14 by 14 parallel parlor, and cows were housed in a free-stall barn bedded with sand and sawdust. Farm 2 milked roughly 3,800 cows thrice daily in a 100-stall rotary parlor, and cows were housed in free-stall barns bedded with anaerobically digested and separated manure solids with added quicklime. Both farms maintained extensive data records during our study period describing parity, pedigree, mating, reproduction, health traits, and productive yield. In addition, each farm participated in monthly DHI testing (including test day milk, protein, and fat yields, as well as individual cow SCC). Both farms worked with Cornell University Quality

Milk Production Services (QMPS) in maintaining ongoing mastitis control programs and

followed QMPS milking protocol recommendations. Any Holstein cow whose milk was collected within 24 hours of freshening was a candidate for inclusion in our study and remained in the study cohort for as long as the cow was part of the herd. Three quarter cows at time of calving were excluded. A convenience sample of 523 cows was taken equally across both farms.

Composite milk samples representing equal volumes from all 4 quarters were collected from each cow for SCC analysis, and body condition, udder, and teat conformation were scored prior to milking (Table 2.1). This study was approved by the Institutional Animal Care and Use

Committee under authorization reference number 2014-0121 on 2/20/2015.

Sample Size Justification

A sample size calculation was conducted with desired α = 0.05 and power of 80% using an open-source calculator (OpenEpi, http://www.openepi.com/SampleSize/SSCohort.htm) based on our primary conformational exposure of interest, flat teat end shape. We estimated an unequal ratio of unexposed (round teat ends) to exposed (flat teat ends) cows of 1.5:1, with only

5% of unexposed cows experiencing mastitis and 25% of exposed cows experiencing mastitis based on a recent nation-wide survey (National Animal Health and Monitoring System, 2016).

This yielded a minimum sample size of 479 cows with the Fleiss method of sample size calculation with continuity correction. We inflated the enrollment requirement an additional

10% to account for loss to follow up.

Data Collection

Two trained researchers diagnosed high or low inflammatory cell presence in the composite colostrum sample collected on 0 or 1 DIM using a California Mastitis Test (CMT).

Cows were given a positive score if the CMT solution reacted to create distinct thickening of the sample, or a negative score if the sample remained liquid (Schalm and Noorlander, 1957).

ear

X X

Additional Additional

Teat Traits Teat

tive tive

Linear Linear

Traits

Descrip

X X X X X X

Body

Score

Condition

depth, width, height, cleft, fore udder attachment, and udder depth,attachment, width, cleft, height, fore

r r

X X X X X X

Milk

Sample

Health 2009) Health

pregnancy

positive energy balance; end end positiveenergy balance;

late lactation; mid lactation; late

Explanation of six selected sampling periods wherein the described data was collected on from cows Holstein 523 was collected data periods described sampling the wherein selected six of Explanation

Parturition;colostrum milk Baseline negative mastitis incidence; Peak balance energy Neutral/ period breeding voluntary of early pregnancy production; Peak Mid

DIM 1 to 0 DIM 5 to 3 DIM 14 to 10 DIM 60 to 50 DIM 110 to 90 DIM 230 to 210

assessed by 1 trained researchers (Holstein 2014) Association USA, researchers Inc. trained 1 by assessed

assessed by 4 trained researchers (Elanco Animal (Elanco researchers trained 4 by assessed

Table for scoring udde include traits Linear dairies. descriptive York New two and front r and shape end teat measuring and rear front scoring include traits teat Additional placement. teat rear and front length. width and teat a b

Specific assumptions regarding colostrum SCC were not made as there is evidence that established guidelines relating CMT scores to milk SCC are not applicable to colostrum

(Maunsell et al., 1998). Composite milk samples collected at 3 to 5, 10 to 14, 50 to 60, 90 to

110, and 210 to 230 DIM were outsourced to a milk laboratory, recognized and approved by the Association of Official Analytical Chemists Research Institute and FDA as an independent reference laboratory, for SCC by flow cytometry (FossomaticTM, Dairy One, Ithaca, NY). This same laboratory was used for DHI milk testing for both farms. All milk samples were collected in the parlor during milking with a minimum volume of 20 mL. Cases of elevated SCC were defined as SCC ≥ 200,000, as a composite milk sample from all 4 quarters with 200,000 cells/ml has been well-established as a threshold at which an intramammary infection is likely present

(McDermott et al., 1982; Dohoo and Leslie, 1991; Schepers et al., 1997). Cases of clinical mastitis were farm-diagnosed and did not require veterinary diagnosis. High SCC prevalence was calculated as new and existing cases over the total population at risk. Cows were given a body condition score (BCS) at each time point on a 5-point scoring system with 0.25 increments by 1 of 4 trained researchers (Ferguson et al., 1994; Elanco Animal Health, 2009). In addition, to identify cow-level risk factors, each cow was scored based on her udder and teat conformation by 1 trained researcher. Teats were assessed at 10 to 14 DIM and again at 90 to

110 DIM to evaluate consistency in researcher scoring. Teats were only evaluated on the right side of each cow to minimize interference in the milking routine. A difference in scores was observed in <3% of cows, and in those cows the second set of scores were used in analysis. At the termination of the study period, individual cow data (including test day SCC, parity, and health/treatment records) were collected from herd management software (DairyComp 305,

Valley Agricultural Software, Tulare, CA). Data on culled cows were included in analyses up until the point of culling.

Definitions

Udder conformation and teat placement traits were scored with 3 categorical levels adapted from the Holstein Association’s scoring for linear descriptive traits (Holstein

Association USA, 2014). Fore udder attachment was evaluated as extremely loose, of intermediate strength, or extremely strong. The udder cleft was scored as weak, intermediate, or extremely strong. The udder depth was determined by measuring the udder floor as extremely deep well below the hocks, of intermediate depth, or extremely high above the hocks. Rear udder height was assessed as extremely low, of intermediate height, or extremely high. Rear udder width was evaluated as narrow, of intermediate width, or extremely wide. Front and rear teat placement were scored as extremely wide placement on outside of quarter and pointing laterally, centrally placed on quarter and pointing straight down, or base of teats on extreme inside of quarter and pointing medially. In addition to these linear descriptive traits, front and rear right teat end shape were scored as flat, round, or pointed (Figure 2.1). Right teat width and length were also measured using a translucent measuring ruler with a scale unit of 1 cm.

Statistical Analysis

All statistical analyses were performed using R version 3.2.5 (R Core Team, 2016). An independent multiple logistic regression model was fit for each outcome, elevated SCC and clinical mastitis. Clinical mastitis outcomes were binary (0 = no event, 1 = at least one event during lactation) and varying days at risk accounted for by including lactation length as a covariate in the model. Elevated SCC outcomes were binary (0 = low, 1 = elevated), and modeled using repeated measure logistic regression including our sample SCC data and farm

The The characterization of teat ends as A) flat, B) round, or C) pointed. Arrows indicate hyperkeratosis at the teat

Figure Figure ends.

test day data, with cow as a random effect. Varying days at risk were accounted for by including lactation length as a covariate in the model.

All variables of interest were tested for association with elevated SCC or clinical mastitis using univariate analysis, and any variables with P < 0.2 were offered to the multivariable model. Because of the small proportion of observed pointed teat ends and the difficulty in distinguishing them from round teat ends in the presence of hyperkeratosis (Figure

1), pointed teat ends were grouped with round teat ends. The Goodman-Kruskal gamma statistic

(γ) was used to measure the strength of association among ordinal variables and Pearson’s product moment correlation (ρ) was used to measure the correlation among continuous variables (Goodman and Kruskal, 1954). Backward stepwise elimination was used to refine the model, all biologically plausible interactions were considered, variables were retained in the final model if P < 0.05, and final models evaluated for goodness of fit using a Likelihood Ratio

Test and by only considering models with less residual deviance than degrees of freedom. Any variables which caused more than 10% change in the estimates of one or more predictors were retained as confounders. Odds ratios were calculated to determine the odds of an elevated SCC event or developing clinical mastitis based on presence of the various risk factors measured for each cow. Both odds ratios and 95% confidence intervals (CI) were generated using the least squares means package (Lenth, 2016).

RESULTS

Enrollment demographics throughout the study period are described in Table 2.2.

Table 2.2. A description of the cow demographics (n = 523) across the sampling period Parity (n) 1 2 3+ Total Farm 1 85 78 92 255 Farm 2 68 100 100 268 Total Enrolled 153 178 192 523 Cows with complete linear descriptive traits 152 175 185 512 Cows with complete additional teat traits 152 176 181 509 Remaining at 3 to 5 DIM 153 177 190 520 Remaining at 10 to 14 DIM 152 177 185 514 Remaining at 50 to 60 DIM 152 171 173 496 Remaining at 90 to 110 DIM 149 167 165 481 Remaining at 210 to 230 DIM 140 157 137 434

Among the entire study cohort, the most common phenotypes were centrally placed teats pointing straight down, round teat-end shapes, loose fore udder attachment, intermediately defined udder cleft, intermediate udder depth, intermediate rear udder height, and intermediate rear udder width (Table 2.3). Front right teats had a median length of 5.0 cm with first and third quartiles 4.5 and 6.0 cm, respectively. Front right teats had a median width of 2.5 cm with first and third quartiles 2.5 and 3.0 cm, respectively. We observed a median rear right teat length of

4.0 cm with first and third quartiles 3.5 and 4.5 cm, respectively. Rear right teats had a median width of 2.5 cm with first and third quartiles 2.0 and 2.5 cm, respectively. From these statistics, udders tend to be deeper, wider, and have looser fore udder attachment with increasing parities, while there is no obvious age-related trend in teat conformation. At time of parturition, 3 to 5

DIM, and 210 to 230 DIM, the median BCS was 3.0 with first and third quartiles 3 and 3.25, respectively. At 10 to 14 DIM the median, first, and third quartiles were 3.0. At 50 to 60 and

90 to 110 DIM, the median BCS was 3.0 with first and third quartiles 2.75 and 3.0, respectively.

While cows followed a normal body condition curve across lactation, there was not enough variation for BCS to be informative and it was not significantly associated with high SCC or clinical mastitis. From Table 2.3, the majority of farm-culled cows were in their third or higher parity (59.5%). Forty-six percent of the farm-culled cows were removed from the herd for mastitis or chronically high SCC reasons.

At time of parturition, we observed a 24% prevalence of positive CMT scores. Positive

CMT prevalence was 42% within first parity animals, 15% within second parity animals, and

18% for cows in their third or higher parity. Cumulative incidences of 44.2% and 28.9% were observed for elevated SCC events and clinical mastitis, respectively, in close agreement with national reports (National Animal Health and Monitoring System, 2016).

Table 2.3. Udder and teat trait frequencies (%) measured for 523 Holstein cows, stratified by parity and animals culled by the farm prior to the end of the study period Parity 1 2 3+ Farm-culled Total n = 152 n = 175a/176b n = 185a/181b n = 74 n = 512a/509b Front teat placementa Lateral 0.7 2.8 7.0 2.7 3.7 Central 92.7 76.6 77.3 74.3 81.5 Medial 6.6 20.6 15.7 23.0 14.8 Front teat shapeb Flat 39.5 39.4 40.9 59.4 40.0 Round 55.2 48.6 49.2 33.8 50.9 Pointed 5.3 12.0 9.9 6.8 9.1 Fore udder attachmenta Loose 24.4 56.0 76.2 62.2 53.7 Intermediate 55.9 38.3 22.2 32.4 37.8 Strong 19.7 5.7 1.6 5.4 8.5 Rear teat placementa Lateral 20.4 1.7 6.5 2.7 3.4 Central 78.3 43.4 42.2 51.3 53.5 Medial 1.3 54.9 51.3 46.0 43.1 Rear teat shapeb Flat 38.8 43.5 45.3 63.5 42.9 Round 57.9 47.4 44.8 28.4 49.6 Pointed 3.3 9.1 9.9 8.1 7.5 Udder clefta Weak 20.4 5.1 9.2 20.3 11.2 Intermediate 66.4 65.8 61.1 54.1 64.4 Strong 13.2 29.1 29.7 25.6 24.4 Udder deptha Deep 0.7 13.7 57.3 36.5 25.2 Intermediate 28.9 58.9 40.5 47.3 43.5 High 70.4 27.4 2.2 16.2 31.3 Rear udder heighta Low 0.7 4.0 20.0 13.5 8.3 Intermediate 62.5 73.7 66.5 71.6 68.1 High 36.8 22.3 13.5 14.9 23.6 Rear udder widtha Narrow 27.0 6.3 5.9 10.8 12.4 Intermediate 72.3 86.8 75.7 77.0 78.3 Wide 0.7 6.9 18.4 12.2 9.3 Parity = 1 − − − 16.2 29.9 Parity = 2 − − − 24.3 34.4 Parity = 3+ − − − 59.5 35.7 aHolstein Association linear descriptive traits and associated sample sizes bAdditional teat conformational traits and associated sample sizes, measured only on right side

Final logistic regression models evaluating the odds of a case of elevated SCC or clinical mastitis are presented in Tables 2.4 and 2.5, respectively. Cows with loose fore udder attachment had 1.9 times greater odds of an elevated SCC event in an individual sample compared to cows with fore udder attachment of intermediate strength (P < 0.01), and 2.1 times the odds compared to cows with strong fore udder attachment (P = 0.03). Cows with flat right rear teat ends had 1.4 times greater odds of a high SCC event in an individual sample compared to those with round teat ends (P = 0.01). Cows in their first parity had 1.6 times greater odds of an elevated SCC event in a sample compared to those in their second parity (P = 0.04), and cows in their second parity had half the odds of an elevated SCC event in an individual sample than those in their third or higher parity (P < 0.01). There was no significant difference in odds between first parity cows and those in their third or higher parity, and compared to both, second parity cows had lesser odds of a high SCC event in a sample during their lactation. Cows with a negative CMT score just after parturition had 0.6 times lesser odds of a high SCC event in an individual sample during the rest of their lactation compared to cows with a positive CMT score

(P < 0.01).

In agreement with the elevated SCC model, cows with loose fore udder attachment had

2.2 times greater odds of a clinical mastitis event as those with attachments of intermediate strength (P < 0.01) and 3.7 times greater odds as those with strong attachments (P = 0.04).

Cows with flat right teat ends had 1.5 times the odds of a clinical mastitis event as those with round rear teat ends (P = 0.07). Cows with low rear udder height had 2.8 times the odds of developing clinical mastitis than those with intermediate rear udder height (P = 0.03). A cow’s odds of clinical mastitis increased 2.2 times for every 1 cm increase in right rear teat width (P

= 0.02).

Table 2.4. Contrast odds ratios (OR), confidence intervals (CI), and P-values reported for the final elevated SCC logistic regression model, reference levels are listed second in contrast column Variable Contrast OR 95% CI P-value Farm 1-2 0.4 0.3 – 0.5 <.01 Fore Udder Attachment loose-intermeda 1.9 1.4 – 2.6 <.01 loose-strong 2.1 1.2 – 3.8 0.03 intermeda-strong 1.1 0.7 – 2.0 0.89 Teat End Shape flat-round 1.4 1.1 – 1.9 0.01 Parity 1-2 1.6 1.1 – 2.3 0.04 1-3+ 0.9 0.6 – 1.3 0.76 2-3+ 0.5 0.4 – 0.8 <.01 CMT low-high 0.6 0.4 – 0.8 <.01 aintermediate

Table 2.5. Contrast odds ratios (OR), confidence intervals (CI), and P-values for the final clinical mastitis logistic regression model, reference levels are listed second in contrast column Variable Contrast OR 95% CI P-value Farm 1-2 0.1 0.1 – 0.2 <.01 Fore Udder Attachment loose-intermeda 2.2 1.3 – 3.6 <.01 loose-strong 3.7 1.3 – 10.7 0.04 intermeda-strong 1.7 0.6 – 5.0 0.61 Teat End Shape flat-round 1.5 1.0 – 2.4 0.07 Rear Udder Height low-intermeda 2.8 1.3 – 6.2 0.03 low-high 2.5 1.0 – 6.3 0.14 intermeda-high 0.9 0.5 – 1.7 0.92 Rear Teat Width (cm) NA 2.2 1.2 – 4.4 0.02 aintermediate

Variables of interest, including parity, CMT score, and the previously described udder and teat traits, were measured for strength and direction of association using a Goodman-

Kruskal gamma statistic (Table 2.6). Udder depth, rear udder height, udder width, and fore udder attachment were strongly associated (γ ≥ 0.5) with parity. Rear udder height, udder width, and fore udder attachment were strongly associated (γ ≥ 0.5) with udder depth. Rear udder height and udder width were also strongly associated (γ ≥ 0.5) with fore udder attachment. Front and rear right teat end shape demonstrated the highest degree of association (γ = 0.88). Front and rear teat placement were strongly associated with each other (γ ≥ 0.72), and rear teat placement was strongly associated with udder cleft (γ ≥ 0.6). Univariate inferential statistics relating each variable of interest to elevated SCC or clinical mastitis, even if they were not retained in thfinal models, are reported in Table 2.7.

DISCUSSION

Logistic regression models presented in this paper associated cow conformational factors with elevated SCC in an individual sample during lactation and CM. Flat teat ends have been associated with higher rate of milk flow as well as increased risk for milk leakage (Seykora and McDaniel, 1985; Klaas et al., 2005; Guarin and Ruegg, 2016), which may inform the higher odds of both elevated SCC and clinical mastitis in cows with flat right teat ends. It is likely that flat front teat ends were also problematic in cases of elevated SCC and clinical mastitis, though not retained in the final models due to high association with rear teat end shape. Farm-culled animals had a higher proportion of cows with flat teats compared to the other strata, but considering some of these animals were not culled for udder health reasons this does not necessarily indicate flat teats increase risk of culling.

Table 2.6. The absolute value of the Goodman-Kruskal Gamma statistic of association among categorical covariates considered in our models.

Udder

Parity FrontTeat Sha Teat Rear Shape Udder Depth Rear Height Udder Width FrontTeat Placement Teat Rear Placement Cleft Udder Front Teat Shape 0.05 Rear Teat Shape 0.12 0.88 Udder Depth 0.85 0.07 0.14 Rear Udder Height 0.50 0.01 0.04 0.56 Udder Width 0.61 0.01 0.14 0.70 0.24 Front Teat Placement 0.05 0.05 0.06 0.15 0.29 0.08 Rear Teat Placement 0.31 0.03 0.01 0.13 0.01 0.05 0.72 Udder Cleft 0.28 0.04 0.07 0.06 0.03 0.05 0.36 0.60 Fore Udder Attachment 0.61 0.10 0.01 0.77 0.52 0.61 0.24 0.18 0.05

Table 2.7. Univariate analysis of variables of interest against elevated SCC and clinical mastitis. Elevated SCC Clinical Mastitis Odds 95% CI P- Odds 95% Confidence P- Ratio value Ratio Interval value Parity 1 reference level 2 0.1 0.5 – 1.1 0.20 1.5 1.3 – 1.8 <.01 3+ 1.8 1.2 – 2.6 <.01 2.3 2.0 – 2.7 <.01 CMT high reference level low 0.4 0.2 – 0.5 <.01 0.8 0.5 – 1.3 0.41 Front Teat Length (cm) 1.0 0.8 – 1.1 0.73 1.1 1.0 – 1.2 <.01 Front Teat Width (cm) 1.2 0.9 – 1.5 0.12 2.3 2.0 – 2.6 <.01 Rear Teat Length (cm) 1.0 0.8 – 1.3 0.75 1.2 1.1 – 1.3 <.01 Rear Teat Width (cm) 1.9 1.2 – 2.9 <.01 2.5 2.2 – 2.9 <.01 Front Teat Shape flat reference level round 0.8 0.6 – 1.0 0.07 0.7 0.6 – 0.8 <.01 Rear Teat Shape Flat reference level Round 0.6 0.5 – 0.9 <.01 0.6 0.5 – 0.6 <.01 Udder Depth low reference level intermediate 0.4 0.3 – 0.6 <.01 0.3 0.3 – 0.4 <.01 high 0.4 0.2 – 0.5 <.01 0.2 0.2 – 0.3 <.01 Rear Udder Height low reference level intermediate 0.4 0.2 – 0.7 <.01 0.3 0.3 – 0.4 <.01 high 0.2 0.1 – 0.4 <.01 0.2 0.1 – 0.2 <.01 Front Teat Placement lateral reference level intermediate 0.6 0.3 – 1.4 0.22 0.6 0.5 – 0.8 <.01 medial 0.6 0.3 – 1.6 0.33 0.5 0.3 – 0.6 <.01 Rear Teat Placement lateral reference level intermediate 1.0 0.4 – 2.3 0.94 0.8 0.6 – 1.0 0.06 medial 1.2 0.5 – 2.9 0.65 0.8 0.6 – 1.1 0.20 Udder Width narrow reference level intermediate 1.4 0.9 – 2.3 0.17 2.2 1.8 – 2.6 <.01 wide 3.1 1.7 – 6.3 <.01 4.8 3.8 – 6.1 <.01 Fore Udder Attachment loose reference level intermediate 0.5 0.3 – 0.7 <.01 0.4 0.3 – 0.4 <.01 medial 0.4 0.2 – 0.7 <.01 0.2 0.2 – 0.3 <.01

The strong Goodman-Kruskal gamma associations (Table 2.6) make biological sense, suggesting that if udders are generally large in one dimension they will tend to also be large in other dimensions (e.g., if udders are wide, they are also deep). It also follows that the deepness of the udder cleft may inform the positioning and angle of rear teats. Furthermore, as cows age they tend to have lower, more pendulous udders. These associations are noteworthy as numerous previous studies suggested cows with higher, less pendulous udders have increased mastitis resistance (Seykora and McDaniel, 1985).

While univariate inferential statistics are less comprehensive than a multivariable model, they allow characterization of the relationship of each variable of interest to elevated

SCC or clinical mastitis even if they were not retained in the final models (Table 2.7). Though not retained in the final elevated SCC model, udder depth, rear udder height, and udder width were significantly associated with repeated high SCC events. It is likely they were not retained in the final models due to their high association with fore udder attachment (γ = 0.77, 0.52, and

0.61, respectively) though individually they likely contribute to increased odds of elevated

SCC. Indeed, past studies have identified similar correlations between deeper udder floors and lower rear udder height with high somatic cell score (Young et al., 1960). Farm was a significant covariate in these models, reaffirming that farm size and management style plays a significant role in mastitis prevalence and risk. The results presented in this study are most reflective of farms with similar size, climate, and management practices.

Most variables of interest appear independently related to odds of clinical mastitis.

Similarly to the elevated SCC model, udder depth and width were likely not retained in the model due to their strong correlation with fore udder attachment (Table 2.6). Front and rear right teat length have a strong Pearson’s correlation coefficient (ρ = 0.69), and rear right teat

width has a moderately strong correlation with rear right teat length (ρ = 0.42), which likely explains why neither teat length trait was retained in the final clinical mastitis model. There is little consensus in the literature regarding the relationship of teat size to mastitis, with some studies reporting a positive correlation between teat diameter and mastitis incidence (Hickman,

1964) and others reporting no significant mastitis association or heritability (Seykora and

McDaniel, 1985). In addition, these univariate analyses suggest that wide right front teats were related to increased odds of clinical mastitis, and laterally pointing front teats may have influenced odds of clinical mastitis. The association between front teat placement and rear udder height (γ = 0.29), and a similar correlation between front and rear teat width (ρ = .38), may explain their individual association with clinical mastitis but statistical insignificance in the multivariate model. It is also possible that the small proportion of cows with laterally pointing teats reduced the power such that a statistically significant association of teat placement with clinical mastitis could not be made in the multivariable models.

CONCLUSIONS

Our study associated loose fore udder attachment and flat teat ends with increased odds of both elevated SCC and clinical mastitis diagnosis. Parity and CMT scores were related to a cow’s odds of experiencing a high SCC event in an individual sample during the course of her lactation, and rear udder height and rear teat width were related to odds of developing clinical mastitis. There is little consensus in the literature regarding the impact of udder and teat conformation on mastitis, and this prospective cohort study with precise measurement of exposure and disease outcomes aided in the clarification of those relationships. A notable aspect of this study was the use of repeated measure logistic regression to associate type traits with

elevated SCC events – as opposed to previous studies which examined lactation averages, markedly reducing the udder health information considered. In addition, monitoring these traits can be used as a low-cost management tool to recognize animals that may have increased risk of elevated SCC and clinical mastitis. In particular, if farmers are concerned with elevated SCC or high prevalence of clinical mastitis in their herd, utilizing the CMT at time of calving and noting fore udder attachment as well as teat end shape may indicate high risk animals within their herd. This could allow farmers to increase monitoring frequency of these high risk animals, mitigating the time and financial burden of adjusting management practices for the entire herd.

In addition, the most marked risk factors of loose fore udder attachment and flat teat ends may be effective criteria to include in culling protocols or inform mating strategies by selecting bulls with strong evaluations for udder and teat morphology without sacrificing milk yield.

ACKNOWLEDGEMENTS

This work was made possible through funding by the NIFA Animal Health Project

#NYC-127898, the invaluable participation of commercial dairy farms, and the work of many

Cornell University graduate and undergraduate students associated with the Huson lab. Special thanks to Stephen Parry of the Cornell Statistical Consulting Unit. Preliminary results were presented as an abstract at the ADSA Annual Meeting in Pittsburgh, PA on June 25-28, 2017 as well as the 57th Annual National Mastitis Council Meeting in Tucson, AZ on January 30-

February 2, 2018.

REFERENCES

Bates, A.J., Dohoo, I., 2016. Risk factors for peri-parturient farmer diagnosed mastitis in New

Zealand dairy herds: findings from a retrospective cohort study. Prev Vet Med 127, 70-

76.

Chrystal, M.A., Seykora, A.J., Hansen, L.B., Freeman, A.E., Kelley, D.H., Healey, M.H., 2001.

Heritability of Teat-End Shape and the Relationship of Teat-End Shape with Somatic

Cell Score for an Experimental Herd of Cows. J. Dairy Sci. 84 (11): 2549-2554.

Cook, N.B., Bennett, T.B., Emery, K.M., Nordlund, K.V., 2002. Monitoring nonlactating cow

intramammary infection dynamics using DHI somatic cell count data. J Dairy Sci 85,

1119-1126.

Dohoo, I.R., Leslie, K.E., 1991. Evaluation of changes in somatic cell counts as indicators of

new intramammary infections. Preventive Veterinary Medicine 10, 225-237.

Elanco Animal Health, 2009. The 5-point body condition scoring system.

https://www.elanco.us/pdfs/ai10752-body-condition-score-insert.pdf (accessed

February 2 2015).

Ferguson, J.D., Galligan, D.T., Thomsen, N., 1994. Principal descriptors of body condition

score in Holstein cows. J Dairy Sci 77, 2695-2703.

Fetrow, J., 2015. Annual losses due to mastitis above a desirable baseline level.

https://www.ansci.umn.edu/extension-outreach/dairy-diagnostics-toolbox (accessed

August 25 2016).

Fox, L.K., Chester, S.T., Hallberg, J.W., Nickerson, S.C., Pankey, J.W., Weaver, L.D., 1995.

Survey of intramammary infections in dairy heifers at breeding age and first parturition.

J Dairy Sci 78, 1619-1628.

Goodman, L.A., Kruskal, W.H., 1954. Measures of association for cross classifications. Journal

of the American Statistical Association 49, 732-764.

Guarin, J.F., Ruegg, P.L., 2016. Short communication: Pre- and postmilking anatomical

characteristics of teats and their associations with risk of clinical mastitis in dairy cows.

J Dairy Sci 99, 8323-8329.

Guarin, J.F., Paixao, M.G., Ruegg, P.L., 2017. Association of anatomical characteristics of teats

with quarter level somatic cell count. J. Dairy Sci. 100(1): 643-652.

Hickman, C.G., 1964. Teat Shape and Size in Relation to Production Characteristics and

Mastitis in Dairy Cattle¹. Journal of Dairy Science 47, 777-782.

Hodgson, A.S., Murdock, F.R., Hillers, J.K., 1980. Effect of teat-end shape on milking rate. J.

Dairy Sci. 63(Suppl. 1):147.

Holstein Association USA, I., 2014. Linear Descriptive Traits.

http://holsteinusa.com/pdf/print_material/linear_traits.pdf (accessed August 3 2014).

Klaas, I.C., Enevoldsen, C., Ersboll, A.K., Tolle, U., 2005. Cow-related risk factors for milk

leakage. J Dairy Sci 88, 128-136.

Koeck, A., Miglior, F., Kelton, D.F., Schenkel, F.S., 2012. Short communication: Genetic

parameters for mastitis and its predictors in Canadian Holsteins. J Dairy Sci 95, 7363-

7366.

Lenth, R.V., 2016. Least-Squares Means: The R Package lsmeans. 2016 69, 33.

Maunsell, F.P., Morin, D.E., Constable, P.D., Hurley, W.L., McCoy, G.C., Kakoma, I.,

Isaacson, R.E., 1998. Effects of mastitis on the volume and composition of colostrum produced

by Holstein cows. J Dairy Sci 81, 1291-1299.

McDermott, M.P., Erb, H.N., Natzke, R.P., 1982. Predictability by somatic cell counts related

to prevalence of intrammary infection within herds. J Dairy Sci 65, 1535-1539.

Mein, G., Reinemann, D., Schuring, N., Ohnstad, I., 2004. Milking machines and mastitis risk:

a storm in a teatcup, In: National Mastitis Council Annual Meeting, Charlotte, NC.

National Animal Health and Monitoring System, 2016. Dairy 2014: Milk quality, milking

procedures, and mastitis on U.S. dairies. United States Department of Agriculture, Fort

Collins, CO.

R Core Team, 2016. R: A language and environment for statistical computing. R Foundation

for Statistical Computing, Vienna, Austria.

Rogers, G.W., Banos, G., Nielsen, U.S., Philipsson, J., 1998. Genetic correlations among

somatic cell scores, productive life, and type traits from the United States and udder

health measures from Denmark and Sweden. J Dairy Sci 81, 1445-1453.

Rollin, E., Dhuyvetter, K.C., Overton, M.W., 2015. The cost of clinical mastitis in the first 30

days of lactation: An economic modeling tool. Prev Vet Med 122, 257-264.

Rupp, R., Boichard, D., 1999. Genetic parameters for clinical mastitis, somatic cell score,

production, udder type traits, and milking ease in first lactation Holsteins. J Dairy Sci

82, 2198-2204.

Schalm, O.W., Noorlander, D.O., 1957. Experiments and observations leading to development

of the California mastitis test. J Am Vet Med Assoc 130, 199-204.

Schepers, A.J., Lam, T.J., Schukken, Y.H., Wilmink, J.B., Hanekamp, W.J., 1997. Estimation

of variance components for somatic cell counts to determine thresholds for uninfected

quarters. J Dairy Sci 80, 1833-1840.

Seykora, A.J., McDaniel, B.T., 1985. Udder and teat morphology related to mastitis resistance:

a review. J Dairy Sci 68, 2087-2093.

Young, C.W., Legates, J.E., Lecce, J.G., 1960. Genetic and Phenotypic Relationships between

Clinical Mastitis, Laboratory Criteria, and Udder Height1. Journal of Dairy Science 43,

54-62.

Zhang, Z., Li, X.P., Yang, F., Luo, J.Y., Wang, X.R., Liu, L.H., Li, H.S., 2016. Influences of

season, parity, lactation, udder area, milk yield, and clinical symptoms on

intramammary infection in dairy cows. J Dairy Sci 99, 6484-6493.

Zwertvaegher, I., De Vliegher, S., Verbist, B., Van Nuffel, A., Baert, J., Van Weyenberg, S.,

2013. Short communication: Associations between teat dimensions and milking-

induced changes in teat dimensions and quarter milk somatic cell counts in dairy cows.

J Dairy Sci 96, 1075-1080.

CHAPTER 3: GENOME WIDE ASSOCIATIONS FOR UDDER AND TEAT

CONFORMATIONAL RISK FACTORS FOR MASTITIS IN HOLSTEIN COWS

Asha M. Milesa, Christian J. Posbergha, and Heather J. Husona1 aDepartment of Animal Science, Cornell University, Ithaca, NY 14853

1Corresponding author: Heather J. Huson, 201 Morrison Hall, 507 Tower Road, Ithaca, NY

14853. Phone: (607) 255-2289. E-mail: [email protected]

In Review: BMC Genomics, October 2019

ABSTRACT

The objective of our study was to conduct high-density genome-wide association studies of dairy cow udder and teat conformation with direct phenotyping. We identified and compared quantitative trait loci (QTL) for a novel composite mastitis risk trait and considered environmental impact of milking by comparing primiparous cows only. Cows (N = 471) were genotyped on the Illumina BovineHD 777K beadchip and scored for front and rear teat length, width, end shape, and placement, fore udder attachment, udder cleft, udder depth, rear udder height, and rear udder width. Principal component analysis was performed on fore udder attachment, rear teat end shape, rear teat width, and rear udder height, to create a single new phenotype describing mastitis susceptibility based on these high-risk traits. Over all 14 traits of interest, a total of 56 genome-wide associations were performed and 28 significantly associated

(Bonferroni multiple testing correction < 0.05) QTL were identified. The linkage disequilibrium

(LD) block surrounding the associated QTL or a 1 Mb window in the absence of LD was interrogated for candidate genes, resulting in the identification of genes with functions related to both cell proliferation and immune signaling, including ZNF683, DHX9, CUX1, TNNT1, and

SPRY1. We assessed a primiparous only subset of cows (n = 144) to account for the possibility that the genetic variance component of the phenotype is greater for cows who have had less exposure to the environment, and observed different associated QTL and inheritance patterns for udder depth in primiparous cows compared to the total cohort. Special focus was given to the aforementioned mastitis risk traits, and candidate gene investigation revealed both immune function and cell proliferation related genes in the areas surrounding significantly associated

QTL, suggesting that selecting for mastitis resistant cows based on these traits would be an effective method for increasing mastitis resiliency in a herd.

Key words: genome wide association, mastitis, principal component analysis, udder conformation, teat conformation

INTRODUCTION

Mastitis, a condition characterized by inflamed mammary tissue and udder gland, is the costliest disease facing U.S. dairy producers, accounting for an estimated $2 billion in annual losses and 11% of total milk lost according to a recent market analysis (iGEM, 2016). While mastitis has been historically considered a management problem, genetic correlations among milk yield, mastitis susceptibility, and udder morphology encouraged selection for udder and teat type traits as early as the 1950s (O’Bleness et al., 1960). In 2009, udder composite values were incorporated into official national genomic evaluation systems to account for the influence of cow conformation on health traits (Wiggans et al., 2011; Interbull, 2013). However, the lack of standardized reporting in the United States has led to a deficit in health-related phenotypes, hindering genetic improvement in U.S. dairy herds, and there is little consensus in the literature regarding mastitis and udder and teat trait relationships or their respective heritabilities

(Seykora and McDaniel, 1985; Parker Gaddis et al., 2014). Most extant research has relied on pedigree information to calculate relationship matrices for estimation of heritability, genetic correlation, and variance, given the strong correlations between health, production, and conformation traits (Seykora and McDaniel, 1986; Rupp and Boichard, 1999; Chrystal et al.,

2001; DeGroot et al., 2002). Focus has been also given to sire transmitting abilities, and while these studies comprehensively evaluate udder morphology, teat length, and teat placement, other teat characteristics such as width and shape are neglected (Nash et al., 2000; DeGroot et al., 2002).

As genotyping technologies have become increasingly cost-effective more molecular- based studies are emerging. A large study utilized publicly available whole genome sequences to impute medium and high-density single nucleotide polymorphism (SNP) data up to over 20

million sequence variants to identify quantitative trait loci (QTL) for a number of udder and teat type traits in Fleckvieh cattle (Pausch et al., 2016). While Pausch et al. had extremely high density genetic data, they relied on bull genotypes and the indirect phenotyping of their daughters, raising external validity concerns. Currently, only one study has directly associated udder traits with cow genotypes, relating beef cow teat length, teat diameter, and a composite

“udder support” score to genomic data from the low-density Bovine Illumina50k chip (Tolleson et al., 2017). Large-scale studies using genotype imputation from low density SNP arrays may maximize genome coverage and make large sample sizes more practical, but they introduce bias towards population averages and limit researcher ability to detect rare variants (Marete et al., 2018a).

A high-density genome wide association (GWA) study of dairy cow udder and teat conformation with direct phenotyping and no reliance on genotype imputation has yet to be conducted. Thus, we previously associated udder and teat conformational traits with mastitis in a prospective cohort study, where flat teat ends and loose fore udder attachment significantly increased odds of both elevated SCC and clinical mastitis, and low rear udder height and wider rear teats increased odds of clinical mastitis diagnosis alone (Miles et al., 2019). We posited that these risk factors may be effective criteria to include in culling protocols or inform mating strategies by selecting bulls with strong evaluations for udder and teat morphology without sacrificing milk yield. The purpose of this current study was to use GWA approaches to identify

SNP markers associated with udder and teat type traits, with special focus on these four risk traits for incorporation into genomic selection marker panels for mastitis-resistant cows.

MATERIALS AND METHODS

Phenotyping. We conducted a prospective cohort study of 523 Holstein cows on 2 commercial herds in New York State involving direct udder and teat phenotype determination as previously described (Miles et al., 2019). Udder and teat traits included fore udder attachment, udder cleft, udder depth, rear udder height, rear udder width, as well as front and rear teat placement, end shape, length, and width. Front and rear teat length and width were scored quantitatively, while the remainder were scored on 3 categorical levels according to the U.S. Holstein Association’s linear descriptive traits (Holstein Association USA, 2014). Phenotypes were first considered quantitatively to assess continuous variation in the trait, but also on a case-control basis comparing extremes in morphology (Nazari-Ghadikolaei et al., 2018).

Genotyping and Quality Control. A whole blood sample from each cow was taken via the coccygeal vessel, collected in 10 mL K2EDTA anticoagulant vacutainers, and stored at 4ºC or

-20ºC until DNA extraction. Genomic DNA was extracted from whole blood according to the

Gentra Puregene Blood Kit protocol (Gentra Systems, Inc. Minneapolis, MN, USA) using laboratory-made buffers. In sum, 471 cows were submitted to GeneSeek (Neogen Genomics,

Lincoln, NE) for SNP genotyping on the Illumina BovineHD 777K beadchip (Illumina, Inc.,

San Diego, CA). Quality control filtering was applied to all genotypes via Golden Helix SNP

& Variation Suite software v8.3.4 (Golden Helix, Bozeman, MT). Genotypes were retained if

SNP and individual call rate ≥ 0.9, minor allele frequency ≥ 0.05, and allele number ≤ 2.

Identity-by-descent (IBD) estimates were calculated for all pairs of individuals based on available genotype data and individuals with IBD estimate ≥ 0.9 were removed.

Principal Component Analysis. Principal Component Analysis (PCA) of phenotype data was performed using RStudio version 3.2.5 (R Core Team, 2016) to identify components that may

describe multiple udder and teat traits. Four PCAs were performed: teat traits only (front and rear teat placement, end shape, length, and width), udder traits only (fore udder attachment, udder cleft, udder depth, rear udder height, and rear udder width), teat and udder traits together, and “risk” traits only (rear teat end shape, rear teat width, fore udder attachment, and rear udder height).

Genome-Wide Association. Efficient Mixed Model Linear analysis (EMMAX) models were employed in Golden Helix SVS allowing the inclusion of the IBD matrix to correct for any population structure among this cohort of cows (Kang et al., 2010). Additive, dominant, and recessive inheritance models were considered along with the variables of farm, parity, batch, rear teat length, udder depth, and rear udder width as potential covariates given their potential correlation and confounding effects on traits of interest. To address the possibility that the genetic variance component of the phenotype is greater for cows who have had less exposure to the environment, a primiparous-only subset of the larger cohort of cows was also evaluated

(n = 144). P-values were adjusted for multiple testing using a Bonferroni correction and False

Discovery Rate (FDR).

Model Selection and Candidate Gene Investigation. Quantile–quantile (QQ) plots of the log10

(Expected P-values) and log10 (Observed P-values) and a genomic inflation factor lambda

(median observed P-value divided by the median expected P-value) were used to assess goodness of fit. A stringent multiple testing correction was applied and only regions with SNPs passing Bonferroni correction (adjusted P-value < 0.05) were interrogated for candidate genes.

Any gene in linkage disequilibrium (LD) with an associated marker, or in a 500 kb upstream or downstream range of said marker if LD was not present, was identified using the NCBI

RefSeq Database (O'Leary et al., 2016). Gene functions were investigated using the GeneCards

database (Stelzer et al., 2016). All genome coordinates given use the most recent ARS UCD

1.2 bovine genome assembly.

RESULTS

Principal Component Analysis. Principal component (PC) analysis was used to create novel composite traits for udder conformation, and mastitis risk factors. Udder traits-only PC1 accounted for 41% of the variance in phenotype and described over-all udder size, loading towards deeper, lower, and wider rear udders, as well as looser fore udder attachment. Risk traits-only PC1 accounted for 35% of phenotypic variance, loading in the direction of the low risk phenotypes of thinner rear teats, tighter fore udders, and higher rear udders (Figure 3.1a).

No other PCs were informative or considered for GWA.

Genome-wide association. After all quality assurance measures were applied, 458 cows with

581,663 SNPs remained for analysis. Phenotypes were first considered quantitatively to assess continuous variation in the trait, then on a case-control basis comparing extremes in morphology. Primiparous cows (n = 144) were only evaluated linearly to preserve the sample size and due to low representation of many udder and teat morphologies in first parity cows

(Miles et al., 2019). Of the 56 GWA studies performed, 13 had significantly associated SNPs, minimally passing a multiple testing correction threshold of false discovery rate (FDR) < 0.05 and no evidence of genomic inflation (Table 3.1). Quantile-Quantile plots are presented in

Supplementary Figure 3.1. All QTL positions and candidate genes residing within an LD block or 1 Mb surrounding associated SNPs are reported in Supplementary Table 3.1. No significant associations (FDR < 0.05) with udder PC1, front teat end shape, rear teat placement, front teat placement, or udder cleft were identified in our total cohort of cows (N = 458).

Table 3.1. Models selected. Summary of final models including sample size (N), model type, inheritance patterns, quality control measures, trait heritabilities, and total number of significantly associated SNPs. Trait N Model Inheritance Pseudo- Pseudo- FDR2 Bonferroni2 Type Lambda1 heritability Front Teat 458 Linear Recessive 1.00 0.33 5 2 Length Front Teat 458 Case- Recessive 1.01 0.06 4 4 Width control3 Fore Udder 288 Case- Additive 1.00 0.45 4 3 Attachment control4 Risk PC1 458 Linear Dominant 1.01 0.07 2 1 Rear Teat 458 Case- Additive 1.00 0.42 1 1 Length control5 Rear Teat 227 Case- Recessive 1.02 0.50 1 1 End Shape control6 Rear Teat 458 Case- Recessive 1.02 0.03 3 1 Width control3 Rear Teat 458 Linear Recessive 0.98 0.18 50 23 Width Udder Depth 265 Case- Dominant 0.99 0.99 1 1 control7 Udder 458 Case- Recessive 1.00 0.02 129 32 Height control8 Udder Width 458 Case- Dominant 1.00 0.30 1 1 control9 Front Teat 144 Linear Recessive 1.02 0.38 10 7 Placement10 Udder 144 Linear Recessive 1.05 0.62 10 0 Depth10 1genomic inflation factor for model quality control, Quantile-Quantile plots in Supplementary Figure 1 2number of SNP associations passing either Bonferroni or False Discovery Rate (FDR) multiple testing corrections at P < 0.05 3narrow versus wide teats, split by median measurement 4loose versus tight fore udders, excluding all cows with intermediate fore udder attachment 5short versus long rear teats, split by median measurement 6flat versus pointed rear teat ends, excluding all cows with round rear teat end shape 7deep versus high udder depth, excluding all cows with intermediate udder depth 8low versus all other udder heights (intermediate and high rear udders combined) 9wide versus all other udder widths (intermediate and narrow udders combined) 10primiparous cow subset

In the primiparous cow subset (n = 144), significant associations (FDR < 0.05) were identified for front teat placement and udder depth only.

Risk PC1. A linear GWAS with dominant inheritance and covariates of farm, parity, rear teat length, udder depth, and udder width significantly associated 1 SNP on Bos taurus autosome

(BTA) 15 (Bonferroni < 0.05; Figure 3.1b).

Udder traits. An additive mixed model with no covariates comparing loose versus tight fore udder attachment significantly associated 1 SNP on BTA 2 and 2 SNPs on the X chromosome passing Bonferroni correction and 1 additional SNP on BTA 2 passing FDR (Figure 3.2).

Significant associations were made for udder depth in both the total cohort and primiparous subset populations of cows. In a case-control comparison of deep versus high udders among all cows, a dominant mixed model with parity and udder width as fixed effects significantly associated 1 SNP on BTA 5 passing both FDR and Bonferroni correction (Figure 3.3a). In a linear scoring of udder depth among the subset of primiparous cows only, a recessive mixed model significantly associated 10 SNPs on BTA 17 (FDR < 0.05; Figure 3.3b). In a case-control comparison of low udder heights to all others (intermediate and high udders combined), a recessive mixed model with covariates udder depth, udder width, and fore udder attachment identified 129 SNPs which passed FDR on 20 chromosomes spanning the genome, and 32 SNPs passing Bonferroni correction on BTAs 6, 14, 15, 18, and 22 (Figure 3.4). In a case-control comparison of wide versus all other widths (intermediate and narrow combined), a dominant mixed model with no covariates significantly associated 1 SNP passing both FDR and

Bonferroni correction on BTA 15.

Teat Traits. In a quantitative representation of front teat length, a recessive mixed model with farm, parity, front teat width, and rear teat length and width covariates significantly associated

control GWA for fore udder attachment, udder attachment, fore GWA for control

values by chromosome in a a in case chromosome by values

log10 log10

–

udders to tight fore udders (n = 288). The black line represents the Bonferroni Bonferroni the black represents multipleline udders uddersThe 288). (n tight to testing fore =

. Manhattan plot Manhattan . showing

Figure 3.2 Figure fore loose comparing threshold. correction

5 SNPs passing FDR on BTAs 9, 10, and 12, and 2 SNPs passing Bonferroni correction on

BTA 10. The most appropriate model for a case-control comparison of wide versus narrow front teats was a recessive mixed model with farm, parity, and front teat length included as fixed effects, yielding 4 SNPs passing Bonferroni correction on BTA 23. A case-control comparison of short versus long rear teats significantly associated (Bonferroni < 0.05) 1 SNP on BTA 2 in an additive mixed model with no covariates. In a case-control comparison of flat versus pointed rear teat ends, a recessive mixed model significantly associated (Bonferroni < 0.05) 1 SNP on

BTA 26 (Figure 3.5). In an examination of rear teat width, two recessive mixed models with no covariates yielded significantly associated SNPs. In the first model, a case-control comparison of narrow versus wide rear teats yielded 3 SNPs passing FDR on BTAs 18, 25, and 28, and 1

SNP passing Bonferroni correction on BTA 25 (Figure 3.6a). Rear teat width was represented quantitatively in the second model and 50 SNPs were significantly associated (FDR < 0.05) on

BTAs 1, 4, 5, 10, 11, 15, 16, 18, 19, 25, 26, 27, and the X chromosome. A total of 23 SNPs passed Bonferroni correction on BTAs 10, 11, 16, 18, 19, and 25 (Figure 3.6b). Examining primiparous cows only, a linear representation of front teat placement with a recessive mixed model yielded 10 SNPs passing FDR on BTA 9.

DISCUSSION

We performed genome-wide association studies to associate QTL with udder and teat conformation traits, with special focus on previously identified mastitis-risk traits of fore udder attachment, rear teat end shape, rear teat width, and udder height (Miles et al., 2019). These studies provide insight into the genetic regulation of teat and udder confirmation and mastitis susceptibility as well as novel traits and specific markers that may be used in genetic selection.

A comprehensive list of candidate genes may be found in Supplementary Table 3.1; here we discuss those most biologically relevant to mastitis risk associated traits.

Our goal in performing PCA on these risk traits was to identify a single measure which may describe mastitis risk based on udder and teat conformation. Risk PC1 described this risk

(excepting rear teat end shape), and GWA for this new composite measure significantly associated novel QTL (15:7287030-7311314) not identified by individual assessment of those risk traits (Figure 3.1). Candidate genes identified in this region were related to both cell division (Centrosomal Protein 126 (CEP126)) and immune cell progenitor differentiation

(Angiopoietin Like 5 (ANGPTL5)), suggesting that Risk PC1 does indeed reflect both mastitis and udder and teat morphology (Drake et al., 2011; Bonavita et al., 2014).

In this study cohort, we previously associated loose fore udder attachment with high odds of elevated milk somatic cell count and clinical mastitis diagnosis, making fore udder attachment among these cows a relevant criterion on which to base culling and management decisions for mastitis control (Miles et al., 2019). A case-control GWA for extremes in fore udder attachment (loose versus tight) identified a number of genes related to both immune function and cell proliferation near associated markers at 2:126359098-126364670 (Figure 3.2).

Wiskott-Aldrich Syndrome Protein Family 2 (WASF2) is a cytoplasmic protein implicated in cell migration, phagocytosis, and immune synapse formation (Thrasher and Burns, 2010).

Similarly, Nuclear Distribution C (NUDC) is critical to cytokinesis, Stratifin (SFN) may regulate cell cycle progression, and Keratinocyte Differentiation Factor 1 (KDF1) serves as an essential regulator of epidermis formation (Ghahary et al., 2004; Lee et al., 2013; Zhang et al.,

2016a). In contrast, this same genomic region houses genes related to immune function, including Ficolin 3 (FCN3) which is an essential component of the lectin complement pathway,

Ribosomal Protein S6 Kinase A1 (RPS6KA1) which has been implicated in activated Toll Like

Receptor 4 signaling, High Mobility Group Nucleosomal Binding Domain 2 (HMGN2) which is known to have antimicrobial activity against bacteria, viruses, and fungi, as well as Zinc

Finger 683 (ZNF683), a tissue-resident T-cell transcription regulator (Moller et al., 1994; Vieira

Braga et al., 2015; Plovsing et al., 2016; Tian et al., 2018). The variety in function among candidate genes associated with fore udder attachment, relating to both the physical trait and immune function, reinforce the role of this udder type trait as a major risk factor for mastitis.

Similarly, we previously associated rear teat end shape with increased odds of both elevated somatic cell count and clinical mastitis in this cohort of cows, and in this study investigated the potential genetic regulation of this trait (Miles et al., 2019). Upon examining rear teat end shape (Figure 3.5), a similar candidate gene pattern emerged with gene functions related to both cell division (explaining teat morphology variation) and immune response regulation (reinforcing teat end shape as an appropriate indicator of mastitis risk). Significantly associated SNP BovineHD2600014687 positioned at 26:50630351 resides within Kinase Non-

Catalytic C-Lobe Domain Containing 1 (KNDC1), which has been implicated in the regulation of cellular senescence and cell cycle progression (Zhang et al., 2014). In addition, Adhesion G

Protein-Coupled Receptor A1 (ADGRA1) was identified within a 1 Mb window of this SNP and belongs to a family of receptors known to regulate immune signaling (Knapp and Wolfrum,

2016). A pattern of candidate gene function emerged like that of fore udder attachment, reinforcing that it may be appropriate to select for mastitis-resilient cows based on udder and teat conformational traits.

In this cohort of cows, we previously associated increasing rear teat width with increased odds of clinical mastitis (Miles et al., 2019). We first assessed this trait quantitatively

to account for continuous variation in rear teat width, and a linear GWA for rear teat width also revealed candidate genes related to both cell division and immune function at a number of different QTL spanning the genome (Figure 3.6). Near a significantly associated SNP at BTA

11:104129366, we identified Caspase Recruitment Domain-Containing Protein 9 (CARD9), a key modulator of immune response related to TLR and NOD2 signaling pathways as well as

NF-kB activation (Bertin et al., 2000; Lamas et al., 2018). Furthermore, genes related to cell differentiation and survival were identified including Notch Receptor 1 (NOTCH1), a highly conserved protein with an extracellular domain containing many epidermal growth factor repeats and whose signaling is heavily involved in cell fate specification, and Epidermal

Growth Factor- Like 7 (EGFL7) which is involved in Notch binding (Siebel and Lendahl,

2017). In addition, a Myomaker Myoblast Fusion Factor (MYMK) resides in this region and has been associated with muscle hypertrophy, making it a strong candidate for impacting variation in teat morphology (Si et al., 2019). Genes were also investigated in various regions on BTA

16, including Centrosomal Protein 350 (CEP350) which plays a critical role in microtubule binding and spindle integrity during cell replication (Yan et al., 2006). This region is also home to Laminin Subunit Gamma 1 and 2 (LAMC1/2), which are thought to regulate cell organization into tissues, potentially contributing to variation in teat width (Nakad et al., 2017). In regards to immune function, nearby genes were investigated including Major Histocompatibility

Complex Class I-Related (MR1) which is critical to adaptive immune response, Ribonuclease

L (RNASEL) is involved in interferon regulation, and DExH-Box Helicase 9 (DHX9), which has been found to control TLR-stimulated immune responses (Squire et al., 1994; Yamaguchi et al., 1998; Dempsey et al., 2018). A significantly associated QTL at BTA 19:29058547-

29063744 is located within the Growth Arrest Specific 7 (GAS7) gene, which while previously

understood to influence neuron differentiation, was recently found to be abundantly expressed in murine alveolar macrophages, though its exact roles in immune responses are still unknown

(Xu et al., 2017). Furthermore, HIC ZBTB Transcriptional Repressor (HIC1) and Tyrosine 3-

Monooxygenase Activation Protein (YWHAE) have both been associated with the regulation of cell proliferation, and in the case of HIC1 its function as a transcriptional regulator has been specifically tied to immune homeostasis (Leal et al., 2016; Burrows et al., 2017). A significantly associated QTL on BTA 25:40126743-40190566 resides within the Sidekick Cell Adhesion

Molecule 1 (SDK1) gene, an adhesion molecule isoform in the immunoglobulin superfamily primarily known for synapse formation in the retina, though Sidekicks in general are expressed in many different tissue types (Yamagata and Sanes, 2008). A significantly associated SNP at

25:35208040 was interrogated for candidate genes in the surrounding area, which included Cut

Like Homeobox 1 (CUX1), known for its role in morphogenesis as well as regulation of antigen presenting cells (Kuhnemuth et al., 2015; Xu et al., 2018). Myosin Light Chain 10 (MYL10), implicated in immune cell transmigration, and Tripartite Motif Containing 56 (TRIM56), an ubiquitin-ligase with a role in antiviral innate immunity, were also found in this region (Oltz et al., 1992; Chen et al., 2018). A case-control GWA study was also performed to identify genes which may potentially drive extremes in morphology, and significantly associated a single SNP at BTA 25:38568564. Candidate genes near this SNP included Polypeptide N-

Acetylgalactosaminyltransferase 2 (GALNT2) which is believed to be involved in O-linked glycosylation of the immunoglobulin A1 hinge region and PiggyBac Transposable Element

Derived 5 (PGBD5), a transposase suspected to mediate genomic rearrangements (Iwasaki et al., 2003; Henssen et al., 2015). The large number of QTL identified spanning the genome suggests rear teat width may be highly polygenic.

We previously associated low rear udder height with increased odds of clinical mastitis

(Miles et al., 2019), and a case-control comparison of low udder height versus all other udder height types also identified several significantly associated QTL spanning the genome (Figure

4). A significantly associated SNP at 14:27024015 lay within the Aspartate Beta-Hydroxylase

(ASPH) gene, known for hydroxylating epidermal growth factors and contributing to dysmorphic features (Patel et al., 2014). Furthermore, a significantly associated QTL on BTA

18 was near Troponin T1, Slow Skeletal Type (TNNT1), a sarcomere regulatory complex associated with muscle weakness, which could feasibly contribute to weak rear udder attachments and consequently lower udder height (Fox et al., 2018). In addition, NLR Family

Pyrin Domain Containing 2 (NLRP2), a member of the NLR family known to be regulators of immune responses, was in this same region (Jones et al., 2016). While we considered low rear udder height as a risk factor for clinical mastitis, the majority of candidate genes identified in this area seem more related to the physical trait itself rather than the indirect trait of “mastitis- risk”, making rear udder height a less compelling trait by which to base breeding and culling decisions for mastitis control.

We hypothesized that the genotype by environment interaction may be greater in multiparous cows who have had greater mastitis exposure and mechanical manipulation of the udder and teats via milking, and thus evaluated a primiparous-only subset of cows (n = 144) for each trait. Only linear GWAs assessing continuous variation in the trait resulted in significant associations; the lack of success with the case-control approach is likely explained by few risk phenotypes (< 20%) observed in primiparous cows. In these linear GWA, QTL were only significantly associated with two traits (front teat placement and udder depth) likely due to low power from a smaller sample size. For udder depth, different inheritance patterns and QTL were

identified for each population (Figure 3.3), which is likely explained by farm culling resulting in different genotypic frequencies among primiparous and multiparous cows. The total cohort

GWA associated a single SNP at BTA 5:113268242 located within Transcription Factor 20

(TCF20) which is primarily associated with human neurodevelopmental disorders (Schafgen et al., 2016). The primiparous only subset significantly associated one QTL on BTA 17. Antisense genes Nudix Hydrolase 6 (NUDT6) and Fibroblast Growth Factor 2 (FGF2) were in linkage with this QTL, and are hypothesized to regulate cell proliferation (Asa et al., 2001). Also in linkage with this QTL lay Sprouty RTK Signaling Antagonist 1 (SPRY1), which has been shown to influence mammary epithelial morphogenesis during post-natal development by negatively regulating epidermal growth factor signaling in the murine mammary gland (Koledova et al.,

2016). We hypothesize that this QTL was masked in our analysis of the total cohort due to the differing genotypic frequencies in multiparous versus primiparous populations, suggesting to truly elucidate the genetic mechanism underlying these morphological characteristics primiparous populations with minimal exposure to selective pressures must be evaluated.

CONCLUSIONS

This is the first study to use high density SNP chip data with direct phenotyping and no reliance on imputation to investigate the genetic mechanisms regulating bovine udder and teat conformation. Potential biases due to indirect phenotyping and genotype imputation are mitigated by assessing high density genotypes, researcher-controlled phenotypes, and disease risk all within the same population of cows. For many traits, we found significantly associated

QTL spanning the genome, suggesting that udder and teat morphologies are complex traits with many causal variants with small effect sizes. A noteworthy aspect of this study was the

examination of a primiparous only subset of cows, which demonstrated different inheritance patterns and associated QTL than the total multiparous cohort. In particular, the objective of this study was to identify genetic mechanisms which may underlie our previously identified mastitis risk factors of loose fore udder attachment, flat rear teat ends, wide rear teats, and low udder height in this cohort of cows. In the case of these risk factors, candidate genes surrounding significantly associated QTL were identified with functions relating to both the immune response and cell proliferation and tissue morphology, suggesting these traits are indeed representative of mastitis susceptibility.

DECLARATIONS

Ethics approval and consent to participate: Farm owner signed consent was obtained prior to sampling and this study was approved under Cornell University IAUC Protocol #2014-0121.

Consent for publication. Not applicable.

Availability of data and materials. The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests. The authors declare that they have no competing interests.

Funding. This work was made possible through funding by the NIFA Animal Health Project

#NYC-127898.

Authors’ contributions. AM performed on-farm data collection, DNA extraction, sample preparation for genotyping, statistical and genetic analyses, interpretation, and manuscript preparation. CP aided in on-farm data collection, DNA extraction, and interpretation of genetic analyses. HH provided funding and study design. All authors read and approved the final manuscript.

Acknowledgements. We thanks commercial dairy farms for their invaluable participation, and many Cornell University graduate and undergraduate students associated with the Huson lab for their assistance with on farm sampling and lab work.

REFERENCES

Asa, S.L., Ramyar, L., Murphy, P.R., Li, A.W., Ezzat, S., 2001. The endogenous fibroblast

growth factor-2 antisense gene product regulates pituitary cell growth and hormone

production. Molecular endocrinology (Baltimore, Md.) 15, 589-599.

Bates, A.J., Dohoo, I., 2016. Risk factors for peri-parturient farmer diagnosed mastitis in New

Zealand dairy herds: findings from a retrospective cohort study. Prev Vet Med 127, 70-

76.

Bertin, J., Guo, Y., Wang, L., Srinivasula, S.M., Jacobson, M.D., Poyet, J.L., Merriam, S., Du,

M.Q., Dyer, M.J., Robison, K.E., DiStefano, P.S., Alnemri, E.S., 2000. CARD9 is a

novel caspase recruitment domain-containing protein that interacts with BCL10/CLAP

and activates NF-kappa B. J Biol Chem 275, 41082-41086.

Bonavita, R., Walas, D., Brown, A.K., Luini, A., Stephens, D.J., Colanzi, A., 2014. Cep126 is

required for pericentriolar satellite localisation to the centrosome and for primary cilium

formation. Biology of the cell 106, 254-267.

Burrows, K., Antignano, F., Bramhall, M., Chenery, A., Scheer, S., Korinek, V., Underhill,

T.M., Zaph, C., 2017. The transcriptional repressor HIC1 regulates intestinal immune

homeostasis. Mucosal immunology 10, 1518-1528.

Chen, Y., Zhao, J., Li, D., Hao, J., He, P., Wang, H., Zhang, M., 2018. TRIM56 Suppresses

Multiple Myeloma Progression by Activating TLR3/TRIF Signaling. Yonsei medical

journal 59, 43-50.

Chrystal, M.A., Seykora, A.J., Hansen, L.B., Freeman, A.E., Kelley, D.H., Healey, M.H., 2001.

Heritability of Teat-End Shape and the Relationship of Teat-End Shape with Somatic

Cell Score for an Experimental Herd of Cows1,2. Journal of Dairy Science 84, 2549-

2554.

Cook, N.B., Bennett, T.B., Emery, K.M., Nordlund, K.V., 2002. Monitoring nonlactating cow

intramammary infection dynamics using DHI somatic cell count data. J Dairy Sci 85,

1119-1126.

DeGroot, B.J., Keown, J.F., Van Vleck, L.D., Marotz, E.L., 2002. Genetic Parameters and

Responses of Linear Type, Yield Traits, and Somatic Cell Scores to Divergent Selection

for Predicted Transmitting Ability for Type in Holsteins1. Journal of Dairy Science 85,

1578-1585.

Dempsey, A., Keating, S.E., Carty, M., Bowie, A.G., 2018. Poxviral protein E3-altered

cytokine production reveals that DExD/H-box helicase 9 controls Toll-like receptor-

stimulated immune responses. J Biol Chem 293, 14989-15001.

Dohoo, I.R., Leslie, K.E., 1991. Evaluation of changes in somatic cell counts as indicators of

new intramammary infections. Preventive Veterinary Medicine 10, 225-237.

Drake, A.C., Khoury, M., Leskov, I., Iliopoulou, B.P., Fragoso, M., Lodish, H., Chen, J., 2011.

Human CD34+ CD133+ hematopoietic stem cells cultured with growth factors

including Angptl5 efficiently engraft adult NOD-SCID Il2rgamma-/- (NSG) mice.

PLoS One 6, e18382.

Elanco Animal Health, 2009. The 5-point body condition scoring system.

Ferguson, J.D., Galligan, D.T., Thomsen, N., 1994. Principal descriptors of body condition

score in Holstein cows. J Dairy Sci 77, 2695-2703.

Fetrow, J., 2015. Annual losses due to mastitis above a desirable baseline level. University of

Minnesota.

Fox, L.K., Chester, S.T., Hallberg, J.W., Nickerson, S.C., Pankey, J.W., Weaver, L.D., 1995.

Survey of intramammary infections in dairy heifers at breeding age and first parturition.

J Dairy Sci 78, 1619-1628.

Fox, M.D., Carson, V.J., Feng, H.Z., Lawlor, M.W., Gray, J.T., Brigatti, K.W., Jin, J.P., Strauss,

K.A., 2018. TNNT1 nemaline myopathy: natural history and therapeutic frontier. Hum

Mol Genet 27, 3272-3282.

Ghahary, A., Karimi-Busheri, F., Marcoux, Y., Li, Y., Tredget, E.E., Taghi Kilani, R., Li, L.,

Zheng, J., Karami, A., Keller, B.O., Weinfeld, M., 2004. Keratinocyte-releasable

stratifin functions as a potent collagenase-stimulating factor in fibroblasts. J Invest

Dermatol 122, 1188-1197.

Goodman, L.A., Kruskal, W.H., 1954. Measures of association for cross classifications. Journal

of the American Statistical Association 49, 732-764.

Guarin, J.F., Ruegg, P.L., 2016. Short communication: Pre- and postmilking anatomical

characteristics of teats and their associations with risk of clinical mastitis in dairy cows.

J Dairy Sci 99, 8323-8329.

Henssen, A.G., Henaff, E., Jiang, E., Eisenberg, A.R., Carson, J.R., Villasante, C.M., Ray, M.,

Still, E., Burns, M., Gandara, J., Feschotte, C., Mason, C.E., Kentsis, A., 2015. Genomic

DNA transposition induced by human PGBD5. eLife 4.

Hickman, C.G., 1964. Teat Shape and Size in Relation to Production Characteristics and

Mastitis in Dairy Cattle¹. Journal of Dairy Science 47, 777-782.

Holstein Association USA, I., 2014. Linear Descriptive Traits. iGEM, C., 2016. Market Analysis: Bovine Mastitis.

Interbull, 2013. Description of National Genetic Evaluation Systems: Conformation.

Iwasaki, H., Zhang, Y., Tachibana, K., Gotoh, M., Kikuchi, N., Kwon, Y.D., Togayachi, A.,

Kudo, T., Kubota, T., Narimatsu, H., 2003. Initiation of O-glycan synthesis in IgA1

hinge region is determined by a single enzyme, UDP-N-acetyl-alpha-D-

galactosamine:polypeptide N-acetylgalactosaminyltransferase 2. J Biol Chem 278,

5613-5621.

Jones, J.D., Vance, R.E., Dangl, J.L., 2016. Intracellular innate immune surveillance devices in

plants and animals. Science 354.

Kang, H.M., Sul, J.H., Service, S.K., Zaitlen, N.A., Kong, S.-y., Freimer, N.B., Sabatti, C.,

Eskin, E., 2010. Variance component model to account for sample structure in genome-

wide association studies. Nature Genetics 42, 348.

Klaas, I.C., Enevoldsen, C., Ersboll, A.K., Tolle, U., 2005. Cow-related risk factors for milk

leakage. J Dairy Sci 88, 128-136.

Knapp, B., Wolfrum, U., 2016. Adhesion GPCR-Related Protein Networks. Handbook of

experimental pharmacology 234, 147-178.

Koeck, A., Miglior, F., Kelton, D.F., Schenkel, F.S., 2012. Short communication: Genetic

parameters for mastitis and its predictors in Canadian Holsteins. J Dairy Sci 95, 7363-

7366.

Koledova, Z., Zhang, X., Streuli, C., Clarke, R.B., Klein, O.D., Werb, Z., Lu, P., 2016. SPRY1

regulates mammary epithelial morphogenesis by modulating EGFR-dependent stromal

paracrine signaling and ECM remodeling. Proceedings of the National Academy of

Sciences of the United States of America 113, E5731-5740.

Kuhnemuth, B., Muhlberg, L., Schipper, M., Griesmann, H., Neesse, A., Milosevic, N.,

Wissniowski, T., Buchholz, M., Gress, T.M., Michl, P., 2015. CUX1 modulates

polarization of tumor-associated macrophages by antagonizing NF-kappaB signaling.

Oncogene 34, 177-187.

Lamas, B., Michel, M.L., Waldschmitt, N., Pham, H.P., Zacharioudaki, V., Dupraz, L., Delacre,

M., Natividad, J.M., Costa, G.D., Planchais, J., Sovran, B., Bridonneau, C., Six, A.,

Langella, P., Richard, M.L., Chamaillard, M., Sokol, H., 2018. Card9 mediates

susceptibility to intestinal pathogens through microbiota modulation and control of

bacterial virulence. Gut 67, 1836-1844.

Leal, M.F., Ribeiro, H.F., Rey, J.A., Pinto, G.R., Smith, M.C., Moreira-Nunes, C.A.,

Assumpcao, P.P., Lamarao, L.M., Calcagno, D.Q., Montenegro, R.C., Burbano, R.R.,

2016. YWHAE silencing induces cell proliferation, invasion and migration through the

up-regulation of CDC25B and MYC in gastric cancer cells: new insights about YWHAE

role in the tumor development and metastasis process. Oncotarget 7, 85393-85410.

Lee, S., Kong, Y., Weatherbee, S.D., 2013. Forward genetics identifies Kdf1/1810019J16Rik

as an essential regulator of the proliferation-differentiation decision in epidermal

progenitor cells. Developmental biology 383, 201-213.

Lenth, R.V., 2016. Least-Squares Means: The R Package lsmeans. 2016 69, 33.

Marete, A., Lund, M.S., Boichard, D., Ramayo-Caldas, Y., 2018. A system-based analysis of

the genetic determinism of udder conformation and health phenotypes across three

French dairy cattle breeds. PLOS ONE 13, e0199931.

Maunsell, F.P., Morin, D.E., Constable, P.D., Hurley, W.L., McCoy, G.C., Kakoma, I.,

Isaacson, R.E., 1998. Effects of mastitis on the volume and composition of colostrum

produced by Holstein cows. J Dairy Sci 81, 1291-1299.

McDermott, M.P., Erb, H.N., Natzke, R.P., 1982. Predictability by somatic cell counts related

to prevalence of intrammary infection within herds. J Dairy Sci 65, 1535-1539.

Mein, G., Reinemann, D., Schuring, N., Ohnstad, I., 2004. Milking machines and mastitis risk:

a storm in a teatcup. In, National Mastitis Council Annual Meeting, Charlotte, NC.

Miles, A.M., McArt, J.A.A., Leal Yepes, F.A., Stambuk, C.R., Virkler, P.D., Huson, H.J., 2019.

Udder and teat conformational risk factors for elevated somatic cell count and clinical

mastitis in New York Holsteins. Preventive Veterinary Medicine 163, 7-13.

Moller, D.E., Xia, C.H., Tang, W., Zhu, A.X., Jakubowski, M., 1994. Human rsk isoforms:

cloning and characterization of tissue-specific expression. Am J Physiol 266, C351-359.

Nakad, B., Fares, F., Azzam, N., Feiner, B., Zilberlicht, A., Abramov, Y., 2017. Estrogen

receptor and laminin genetic polymorphism among women with pelvic organ prolapse.

Taiwanese journal of obstetrics & gynecology 56, 750-754.

Nash, D.L., Rogers, G.W., Cooper, J.B., Hargrove, G.L., Keown, J.F., Hansen, L.B., 2000.

Heritability of Clinical Mastitis Incidence and Relationships with Sire Transmitting

Abilities for Somatic Cell Score, Udder Type Traits, Productive Life, and Protein Yield.

Journal of Dairy Science 83, 2350-2360.

National Animal Health and Monitoring System, 2016. Dairy 2014: Milk quality, milking

procedures, and mastitis on U.S. dairies. United States Department of Agriculture, Fort

Collins, CO.

Nazari-Ghadikolaei, A., Mehrabani-Yeganeh, H., Miarei-Aashtiani, S.R., Staiger, E.A.,

Rashidi, A., Huson, H.J., 2018. Genome-Wide Association Studies Identify Candidate

Genes for Coat Color and Mohair Traits in the Iranian Markhoz Goat. Frontiers in

genetics 9, 105-105.

O'Leary, N.A., Wright, M.W., Brister, J.R., Ciufo, S., Haddad, D., McVeigh, R., Rajput, B.,

Robbertse, B., Smith-White, B., Ako-Adjei, D., Astashyn, A., Badretdin, A., Bao, Y.,

Blinkova, O., Brover, V., Chetvernin, V., Choi, J., Cox, E., Ermolaeva, O., Farrell,

C.M., Goldfarb, T., Gupta, T., Haft, D., Hatcher, E., Hlavina, W., Joardar, V.S., Kodali,

V.K., Li, W., Maglott, D., Masterson, P., McGarvey, K.M., Murphy, M.R., O'Neill, K.,

Pujar, S., Rangwala, S.H., Rausch, D., Riddick, L.D., Schoch, C., Shkeda, A., Storz,

S.S., Sun, H., Thibaud-Nissen, F., Tolstoy, I., Tully, R.E., Vatsan, A.R., Wallin, C.,

Webb, D., Wu, W., Landrum, M.J., Kimchi, A., Tatusova, T., DiCuccio, M., Kitts, P.,

Murphy, T.D., Pruitt, K.D., 2016. Reference sequence (RefSeq) database at NCBI:

current status, taxonomic expansion, and functional annotation. Nucleic Acids Res 44,

D733-745.

O’Bleness, G.V., van Vleck, L.D., Henderson, C.R., 1960. Heritabilities of Some Type

Appraisal Traits and Their Genetic and Phenotypic Correlations with Production.

Journal of Dairy Science 43, 1490-1498.

Oltz, E.M., Yancopoulos, G.D., Morrow, M.A., Rolink, A., Lee, G., Wong, F., Kaplan, K.,

Gillis, S., Melchers, F., Alt, F.W., 1992. A novel regulatory myosin light chain gene

distinguishes pre-B cell subsets and is IL-7 inducible. Embo j 11, 2759-2767.

Parker Gaddis, K.L., Cole, J.B., Clay, J.S., Maltecca, C., 2014. Genomic selection for producer-

recorded health event data in US dairy cattle. Journal of Dairy Science 97, 3190-3199.

Patel, N., Khan, A.O., Mansour, A., Mohamed, J.Y., Al-Assiri, A., Haddad, R., Jia, X., Xiong,

Y., Megarbane, A., Traboulsi, E.I., Alkuraya, F.S., 2014. Mutations in ASPH cause

facial dysmorphism, lens dislocation, anterior-segment abnormalities, and spontaneous

filtering blebs, or Traboulsi syndrome. Am J Hum Genet 94, 755-759.

Pausch, H., Emmerling, R., Schwarzenbacher, H., Fries, R., 2016. A multi-trait meta-analysis

with imputed sequence variants reveals twelve QTL for mammary gland morphology

in Fleckvieh cattle. Genetics Selection Evolution 48, 14.

Plovsing, R.R., Berg, R.M., Munthe-Fog, L., Konge, L., Iversen, M., Moller, K., Garred, P.,

2016. Alveolar recruitment of ficolin-3 in response to acute pulmonary inflammation in

humans. Immunobiology 221, 690-697.

R Core Team, 2016. R: A language and environment for statistical computing. R Foundation

for Statistical Computing, Vienna, Austria.

Rogers, G.W., Banos, G., Nielsen, U.S., Philipsson, J., 1998. Genetic correlations among

somatic cell scores, productive life, and type traits from the United States and udder

health measures from Denmark and Sweden. J Dairy Sci 81, 1445-1453.

Rollin, E., Dhuyvetter, K.C., Overton, M.W., 2015. The cost of clinical mastitis in the first 30

days of lactation: An economic modeling tool. Prev Vet Med 122, 257-264.

Rupp, R., Boichard, D., 1999. Genetic parameters for clinical mastitis, somatic cell score,

production, udder type traits, and milking ease in first lactation Holsteins. J Dairy Sci

82, 2198-2204.

Schafgen, J., Cremer, K., Becker, J., Wieland, T., Zink, A.M., Kim, S., Windheuser, I.C.,

Kreiss, M., Aretz, S., Strom, T.M., Wieczorek, D., Engels, H., 2016. De novo nonsense

and frameshift variants of TCF20 in individuals with intellectual disability and postnatal

overgrowth. European journal of human genetics : EJHG 24, 1739-1745.

Schalm, O.W., Noorlander, D.O., 1957. Experiments and observations leading to development

of the California mastitis test. J Am Vet Med Assoc 130, 199-204.

Schepers, A.J., Lam, T.J., Schukken, Y.H., Wilmink, J.B., Hanekamp, W.J., 1997. Estimation

of variance components for somatic cell counts to determine thresholds for uninfected

quarters. J Dairy Sci 80, 1833-1840.

Seykora, A.J., McDaniel, B.T., 1985. Udder and teat morphology related to mastitis resistance:

a review. J Dairy Sci 68, 2087-2093.

Seykora, A.J., McDaniel, B.T., 1986. Genetics Statistics and Relationships of Teat and Udder

Traits, Somatic Cell Counts, and Milk Production1. Journal of Dairy Science 69, 2395-

2407.

Si, Y., Wen, H., Du, S., 2019. Genetic Mutations in jamb, jamc, and myomaker Revealed

Different Roles on Myoblast Fusion and Muscle Growth. Marine biotechnology (New

York, N.Y.) 21, 111-123.

Siebel, C., Lendahl, U., 2017. Notch Signaling in Development, Tissue Homeostasis, and

Disease. Physiological reviews 97, 1235-1294.

Squire, J., Zhou, A., Hassel, B.A., Nie, H., Silverman, R.H., 1994. Localization of the

interferon-induced, 2-5A-dependent RNase gene (RNS4) to human chromosome 1q25.

Genomics 19, 174-175.

Stelzer, G., Rosen, N., Plaschkes, I., Zimmerman, S., Twik, M., Fishilevich, S., Stein, T.I.,

Nudel, R., Lieder, I., Mazor, Y., Kaplan, S., Dahary, D., Warshawsky, D., Guan-Golan,

Y., Kohn, A., Rappaport, N., Safran, M., Lancet, D., 2016. The GeneCards Suite: From

Gene Data Mining to Disease Genome Sequence Analyses. Current Protocols in

Bioinformatics 54, 1.30.31-31.30.33.

Thrasher, A.J., Burns, S.O., 2010. WASP: a key immunological multitasker. Nature reviews.

Immunology 10, 182-192.

Tian, H., Miao, J., Zhang, F., Xiong, F., Zhu, F., Li, J., Wang, X., Chen, S., Chen, J., Huang,

N., Wang, Y., 2018. Non-histone nuclear protein HMGN2 differently regulates the

urothelium barrier function by altering expression of antimicrobial peptides and tight

junction protein genes in UPEC J96-infected bladder epithelial cell monolayer. Acta

biochimica Polonica 65, 93-100.

Tolleson, M.W., Gill, C.A., Herring, A.D., Riggs, P.K., Sawyer, J.E., Sanders, J.O., Riley, D.G.,

2017. Association of udder traits with single nucleotide polymorphisms in crossbred

Bos indicus–Bos taurus cows1,2. Journal of Animal Science 95, 2399-2407.

Vieira Braga, F.A., Hertoghs, K.M., Kragten, N.A., Doody, G.M., Barnes, N.A., Remmerswaal,

E.B., Hsiao, C.C., Moerland, P.D., Wouters, D., Derks, I.A., van Stijn, A., Demkes, M.,

Hamann, J., Eldering, E., Nolte, M.A., Tooze, R.M., ten Berge, I.J., van Gisbergen,

K.P., van Lier, R.A., 2015. Blimp-1 homolog Hobit identifies effector-type

lymphocytes in humans. European journal of immunology 45, 2945-2958.

Wiggans, G.R., VanRaden, P.M., Cooper, T.A., 2011. The genomic evaluation system in the

United States: Past, present, future. Journal of Dairy Science 94, 3202-3211.

Xu, H., He, J.H., Xu, S.J., Xie, S.J., Ma, L.M., Zhang, Y., Zhou, H., Qu, L.H., 2018. A group

of tissue-specific microRNAs contribute to the silencing of CUX1 in different cell

lineages during development. Journal of cellular biochemistry 119, 6238-6248.

Xu, Q., Liu, X., Wang, X., Hua, Y., Wang, X., Chen, J., Li, J., Wang, Y., Stoeger, T., Chen, S.,

Huang, N., 2017. Growth arrest-specific protein 7 regulates the murine M1 alveolar

macrophage polarization. Immunologic research 65, 1065-1073.

Yamagata, M., Sanes, J.R., 2008. Dscam and Sidekick proteins direct lamina-specific synaptic

connections in vertebrate retina. Nature 451, 465-469.

Yamaguchi, H., Kurosawa, Y., Hashimoto, K., 1998. Expanded genomic organization of

conserved mammalian MHC class I-related genes, human MR1 and its murine ortholog.

Biochem Biophys Res Commun 250, 558-564.

Yan, X., Habedanck, R., Nigg, E.A., 2006. A complex of two centrosomal proteins, CAP350

and FOP, cooperates with EB1 in microtubule anchoring. Molecular biology of the cell

17, 634-644.

Young, C.W., Legates, J.E., Lecce, J.G., 1960. Genetic and Phenotypic Relationships between

Clinical Mastitis, Laboratory Criteria, and Udder Height1. Journal of Dairy Science 43,

54-62.

Zhang, C., Zhang, W., Lu, Y., Yan, X., Yan, X., Zhu, X., Liu, W., Yang, Y., Zhou, T., 2016a.

NudC regulates actin dynamics and ciliogenesis by stabilizing cofilin 1. Cell research

26, 239-253.

Zhang, C., Zhen, Y.Z., Lin, Y.J., Liu, J., Wei, J., Xu, R., Hu, G., 2014. KNDC1 knockdown

protects human umbilical vein endothelial cells from senescence. Molecular medicine

reports 10, 82-88.

Zhang, Z., Li, X.P., Yang, F., Luo, J.Y., Wang, X.R., Liu, L.H., Li, H.S., 2016b. Influences of

season, parity, lactation, udder area, milk yield, and clinical symptoms on

intramammary infection in dairy cows. J Dairy Sci 99, 6484-6493.

Zwertvaegher, I., De Vliegher, S., Verbist, B., Van Nuffel, A., Baert, J., Van Weyenberg, S.,

2013. Short communication: Associations between teat dimensions and milking-

induced changes in teat dimensions and quarter milk somatic cell counts in dairy cows.

J Dairy Sci 96, 1075-1080.

Supplementary Figure 3.1. Quantile-Quantile plots showing the –log10(Expected P Values) plotted against the – log10(Observed P Values) for each final model A) Front Teat Length, B) Front Teat Width, C) Fore Udder Attachment, D) Risk PC1, E) Rear Teat Length, F) Rear Teat End Shape, G) Rear Teat Width (case-control), H) Rear Teat Width (linear), I) Udder Depth, J) Udder Height, K) Udder Width, L) Front Teat Placement (primiparous cows only), M) Udder Depth (primiparous cows only. The black line represents X = Y.

Supplementary Table 3.1. Significantly associated QTL (Bonferroni < 0.05) and potential candidate genes for all models Trait Chromosome1 Position (bp)1 Candidate Genes2 Front Teat 50107685- 10 FOXB1, LOC112448464 Length 50117321 PLA2G7, ANKRD66, TRD6, Front Teat Width 23 19946762 SLC25A27 20566460- 23 TNFRSF21, CD2AP, ADGRF2 20576922 WASF2, GPR3, CD164L2, FCN3, MAP3K6, SYTL1, TMEM222, WDTC1, SLC9A1, FAM46B, TRNP1, Fore Udder 126354670- KDF1, NUDC, NR0B2, GPN2, SFN, 2 Attachment 126359098 ZDHHC18, PIGV, LOC112443408, ARID1A, RPS6KA1, HMGN2, LIN28A, ZNF683, CRYBG2, UBXN11, CEP85 PHEX, LOC112445297, SMS, 121461599- LOC112445297, MBTPS2, SMPX, X 121477381 KLHL34, TRNAC-GCA, LOC112445299, CNKSR2 7287030- Risk PC1 15 CEP126, ANGPTL5, TRPC6 7311314 Rear Teat SCG2, AP1S3, WDFY1, MRPL44, 2 112245780 Length SERPINE2, FAM124B, CUL3 KNDC1, SYCE1, TCERG1L, CYP2E1, LOC528422, LOC112444506, ADGRA1, LOC100850437, Rear Teat End 26 50630351 LOC101906198, LOC112444503, Shape CFAP46, LOC112444502, LOC112444507, NKX6-2, INPP5A, LOC112444505 Rear Teat 18 18781078 BRD7, ADCY7 Width3 25 38568564 LOC618554, LOC101906717 LOC112444763, GALNT2, PGBD5, 28 2835028 LOC112444731, LOC100337323, LOC112444759, LOC614741 Rear Teat CTDSPL2, EIF3J, SPG11, 10 Width4 LOC112448593, PATL2 LOC112448855, LHX3, QSOX2, LOC787891, GPSM1, 11 104129366 LOC101902280, DNLZ, CARD9, SNAPC4, ENTR1, PMPCA, INPP5E, SEC16A, NOTCH1, LOC112448856,

EGFL7, MIR126, LOC101902839, LOC101902895, FAM69B, LOC112448928, LOC107132967, LOC100848307, LOC112448857, ABO, LOC112448956, SURF6, LOC11244890, MED22, RPL7A, LOC100139115, LOC100112448907, LOC100112448908, LOC100112448904, LOC100112448903, LOC100112448905, SURF2, SURF4, STKLD1, LOC107132968, REXO4, ADAMTS13, CACFD1, SLC2A6, LOC112448858, TRNAC-GCA, MYMK, ADAMTSL2, FAM163B, DBH, SARDH, VAV2 CEP350, QSOX1, LOC112441858, LHX4, ACBD6, MIR669, XPR1, 61802991- 16 TRNAC-ACA, LOC107133256, 62196774 KIAA1614, STX6, MR1, IER5, CACNA1E, LOC104974498, ZNF648, LOC101905162, GLUL, TEDDM1, RGSL1, RNASEL, RGS16, RGS8, 16 63823597 LOC101905664, NPL, DHX9, SHCBP1L, LAMC1, LAMC2, NMNAT2 LOC112442233, CBLN1, 18 17655467 C18H16orf78, ZNF423, TRNAG-CCC 18 20537778 LOC516179, TOX3 MIR2899, LOC112442482, TRNAG- 18 42468232 CCC, LOC617301, ZNF507, DPY19L3 VPS53, MIR2336, RFLNB, C19H17orf97, TRNAG-UCC, RPH3AL, LOC104975006, DOC2B, LOC112442619, YWHAE, TRNAE- UUC, CRK, MYO1C, INPP5K, 19 22640468 PITPNA, SLC43A2, SCARF1, RILP, PRPF8, TLCD2, MIR22, WDR81, SMYD4, SERPINF1, RPA1, RTN4RL1, LOC112442621, DPH1, OVCA1, MIR132, MIR212, HIC1, SMG6, LOC112442776 23978522- RAP1GAP2, OR1D5, LOC101906737, 19 23997890 LOC618593, OR1G1, LOC540082,

LOC532238, LOC522582, LOC112442765, LOC520835, LOC59525, LOC59526 29058547- 19 GLP2R, RCVRN, GAS7 29063744 CUX1, TRNAW-CCA, MIR2388, LOC112444316, LOC112444338, MYL10, COL26A1, LOC104970468, IFT22, FIS1, PLOD3, LOC618076, 25 35208040 NAT16, VGF, AP1S1, LOC101902751, SERPINE1, TRIM56, LOC101902689, LOC101909082, LOC107131854 40126743- 25 SDK1 40190566 Udder Depth 5 113268242 TCF20, LOC104976976 102964124- Udder Height 6 LOC100298890 102982437 14 27024015 CLVS1, ASPH 15545765- 15 AMOTL1, LOC112441606 15782913 TNNT1, PPP1R12C, LOC112442386, LOC112442387, EPS8L1, RDH13, 62273143- 18 LOC100848752, GP6, NLRP2, 62481417 LOC100336589, LOC100852077, LOC112442414 22 46733454 CACNA2D3, LOC112443534 PRDM11, LOC101906676, SYT13, LOC107133190, LOC112441655, CHST1, LOC104974324, Udder Width 15 75722222 LOC107133191, SLC35C1, CRY2, MAPK8IP1, C15H11orf94, PEX16, LARGE2, PHF21A, CREB3L1 Front Teat 58002055- 9 LOC101902249, LOC112448054 Placement5 58079933 34476230- SPRY1, SPATA5, LOC112442097, Udder Depth5,6 17 34552407 NUDT6, FGF2 1Positions based on ARS-UCD 1.2 2Genes in LD with significantly associated SNP, or +/- 500 kb of associated SNP in case of no LD, bolded text indicates associated SNP is within the gene 3case-control GWA of narrow versus wide rear teats divided at the median value 4linear GWA of quantitative rear teat width scores 5primiparous only subset of cows (N = 144) 6not passing Bonferroni correction, QTL significantly associated at FDR < 0.05

CHAPTER 4: TIME- AND POPULATION-DEPENDENT GENETIC PATTERNS

UNDERLIE BOVINE MILK SOMATIC CELL COUNT

Asha M. Milesa and Heather J. Husona1 aDepartment of Animal Science, Cornell University, Ithaca, NY 14853

1Corresponding author: Heather J. Huson, 201 Morrison Hall, 507 Tower Road, Ithaca, NY

14853. Phone: (607) 255-2289. E-mail: [email protected]

In Prep

ABSTRACT

The objective of this study was to determine whether genetic regulation of bovine milk somatic cell count (SCC) varied throughout the course of an individual lactation and to identify quantitative trait loci (QTL) which may differentiate populations of chronically mastitic and robustly healthy cows. Milk SCC has long been used as a proxy for clinical mastitis diagnosis in management and genetic improvement strategies to control the disease.

Cows (N = 471) were genotyped on the Illumina BovineHD 777K beadchip and composite milk samples collected for SCC at 0-1 DIM, 3-5 DIM, 10-14 DIM, 90-110 DIM, and 210-230

DIM, each time point representing key physiological transitions for the cow. Median lactation somatic cell score (SCS) and area under the SCS curve were calculated from farm test data. A total of 8 genome-wide associations were performed and 167 single nucleotide polymorphisms (SNPs) spanning the genome were significantly associated (False Discovery

Rate < 0.05). Of these associated regions, 27 out of 48 associated QTL were novel for clinical mastitis or SCC. The linkage disequilibrium (LD) block surrounding the associated QTL or a

1 Mb window in the absence of LD was interrogated for candidate genes, and many of those identified were related to multiple arms of the immune system including toll-like receptor signaling, macrophage activation, B cell maturation, T cell recruitment, and the complement pathway. These genes included EXOC4, BAMBI, ITSN2, IL34, FCN3, and CD8a/b. In addition, we identified populations of robustly healthy (SCS ≤ 4 from 10-14 DIM until study end), chronically mastitic (SCS > 4 from 10-14 DIM until study end), and average cows with fluctuating SCS, and calculated fixation indices to identify regions of the genome differentiating these three populations. A total of 12 SNPs were identified showing moderate allelic differentiation (FST ≥ 0.4) between Chronic, Healthy, and Average populations of

cows. Candidate genes in the region surrounding differentiated QTL were related to cell signaling and immune response, such as JAKMIP1 and MADCAM1. The wide range of significantly associated QTL spanning the genome and diversity of gene functions reinforces that mastitis is a complex trait and suggests that selection based on lactation stage-specific

SCS rather than a generalized score may lead to greater success in breeding mastitis-resistant cows.

Key words: genome wide association, mastitis, somatic cell score, single nucleotide polymorphisms

INTRODUCTION

Mastitis, a condition characterized by inflamed mammary tissue and the costliest disease facing U.S. dairy producers, can be attributed to problems with herd and farm management, milking management, the milking machine, and the cow itself (Mein et al., 2004;

Rollin et al., 2015). In 1994, milk somatic cell count (SCC), a common proxy for mastitis diagnosis, was added to national bovine genetic evaluations specifically to address the impact of elevated SCC on cow production and health. Though the genetic evaluation of dairy cattle has long been used to improve production and health traits, improvements by linkage based marker assisted selection have been hindered by the small proportion of genetic variation explained by each marker in regards to complex traits, as well as the high costs associated with marker validation (Soller, 1994; Andersson, 2001; Misztal, 2006). In addition, the use of bull genotypes and indirect phenotyping of their daughters, the lack of standardized reporting in the

United States, and the use of population and lactation average data in trait definition, potentially introduce bias to mastitis phenotypes and their genetic associations (Interbull Code of Practice,

2018). As genotyping technologies become increasingly affordable, more studies have emerged utilizing single nucleotide polymorphism (SNP) chip, genotype imputation, RNAseq, and whole genome sequence data, though these studies do not directly phenotype study animals and primarily rely on the less time-intensive method of de-regressing breeding values for trait definition (Biffani et al., 2017; Fang et al., 2017; Cai et al., 2018). To date, no genome wide association (GWA) study has been conducted utilizing high density cow genotypes and direct longitudinal SCC phenotyping on the same animals.

While SCC is a well-established proxy for mastitis diagnosis, lactation somatic cell score (SCS) averages and de-regressed breeding values are generalized phenotypes which may

not detect the specific mechanisms underlying the recruitment of somatic cells to the udder.

Inflammatory states post-partum have been well documented, and numerous studies have established that inflammatory responses are elevated following parturition given the stresses of lactogenesis, uterine involution, and changes in metabolic state, even in the absence of disease

(Bionaz et al., 2007; Graugnard et al., 2012; Akbar et al., 2015). Elevated milk SCC due to postpartum inflammation may create an upward skew in lactation average data, resulting in an inaccurate characterization of mastitis susceptibility. The objective of this study was to longitudinally characterize SCC in a cohort of cows and use high density genotype data to examine the genetics driving elevated SCC across different stages of lactation, and to assess the genetic differentiation among populations of chronically elevated and robustly healthy cows.

This study was designed specifically to address the limitations of using average lactation somatic cell score for genomic selection of mastitis resistant dairy cattle, given that there are non-pathological reasons for udder inflammation and that SCC is not exclusively an indicator of disease.

MATERIALS AND METHODS

Phenotyping. From June 2015 to July 2016, a convenience sample of 523 dairy cows from two farms in upstate New York were enrolled in a prospective cohort study involving six composite milk sample collections representing key physiological time points in lactation as previously described (Miles et al., 2019). The physiological conditions sampled for were parturition (0-1 days in milk (DIM)), baseline milk after clearance of colostrum (3-5 DIM), peak mastitis incidence and negative energy balance (10-14 DIM), return to neutral/positive energy balance

(50-60 DIM), peak production and early pregnancy (90-110 DIM), and mid to late lactation

(210-230 DIM). A California Mastitis Test (CMT) was used to assign a binary score to colostrum samples (0-1 DIM); positive if the CMT solution reacted to create distinct thickening of the sample, or a negative score if the sample remained liquid (Schalm and Noorlander, 1957)

. All remaining milk samples were outsourced to a milk laboratory for SCC by flow cytometry as previously described (FossomaticTM, Dairy One, Ithaca, NY (Miles et al., 2019). Both farms participated in monthly Dairy Herd Improvement testing and milk test data from the study period was pulled from herd management software (DairyComp 305, Valley Agricultural

Software, Tulare, CA). Farm milk test data was generated by the same laboratory that processed the researcher-collected samples. All SCC data were logarithmically transformed to a linear

푆퐶퐶 푙푛( ) somatic cell score by the following equation: 푆퐶푆 = 105 + 3 in R Studio version 3.3.2 (R 푙푛(2)

Core Team, 2016). Median SCS and area under the SCS curve (AUC) were calculated from the farm milk test data. Farm owner signed consent was obtained prior to sampling and this study was approved by Cornell University’s Institutional Animal Care and Use Committee under authorization reference number 2014-0121.

Genotyping and Quality Control. Genomic DNA was extracted from whole blood taken from the coccygeal vessel, collected in 10 mL K2EDTA anticoagulant vacutainers, and stored at 4ºC or -20ºC until DNA extraction. Extractions were performed according to the Gentra Puregene

Blood Kit protocol (Gentra Systems, Inc. Minneapolis, MN, USA) using laboratory-made buffers. A total of 471 cows were submitted for SNP genotyping on the Illumina BovineHD

777K beadchip (Illumina, Inc., San Diego, CA) by GeneSeek (Neogen Genomics, Lincoln,

NE). Quality assurance measures filtered out samples with SNP and individual call rate < 0.9, minor allele frequency < 0.05, and allele number > 2. The relatedness of all pairs of individuals

was assessed via Identity-by-descent (IBD) estimates from genotype data and individuals with

IBD estimate ≥ 0.9, demonstrating significant relatedness, were removed. After all quality control filtering was applied, 458 cows with 581,663 SNPs remained for analysis.

Population Analysis. To identify distinct populations of cows, SCS at 10-14 DIM was plotted against SCS at 50-60 DIM (Figure 4.1a) and quadrants drawn at a threshold of SCS = 4 (SCC

= 200,000 cells/ml), a well-established threshold at which intramammary infection is likely present (McDermott et al., 1982; Dohoo and Leslie, 1991; Schepers et al., 1997). Cows with

SCS ≤ 4 at both 10-14 DIM and 50-60 DIM were designated “healthy” (green squares), those with SCS ≤ 4 10-14 DIM but SCS > 4 at 50-60 DIM were designated “new cases” (black circles), those with SCS > 4 at 10-14 DIM but SCS ≤ 4 at 50-60 DIM were labeled “cures”

(blue stars), and cows with SCS > 4 at both time points were labeled “chronic” (red triangles).

The progression of these 4 groups of cows over their lactation was tracked by plotting SCS at

50-60 DIM against SCS at 90-110 DIM (Figure 4.1b), and SCS at 90-110 DIM against SCS at

210-230 DIM (Figure 4.1c). Populations were defined as cows that remained in the “Healthy” quadrant the entire study period (solid green squares, n = 239), those with “Chronic” elevated

SCC throughout the study period (solid red triangles, n = 12), and “Average” cows whose SCC fluctuated across quadrants during the study (n = 168). Wright’s F Statistic (FST) by marker was calculated to measure the genetic divergence between the subpopulations of “Healthy” and

“Average”, “Healthy and “Chronic”, and “Average” and “Chronic” with 0 representing no allelic divergence between populations and values up to 1 suggesting increasing population differentiation (Weir and Cockerham, 1984). Any SNP with FST ≥ 0.4 was considered moderately differentiated given the high prevalence of inbreeding within Holstein cattle, and surrounding regions investigated for candidate genes as described below (Forutan et al., 2018).

), ), cows with SCS < 4

4 at 10 to 14 DIM and

□

< < 4 at all 4 time points; solid red

SCS Linear SCSLinear at A) 10 to 14 plotted DIM

4 at 10 to 14 DIM but < 4 at 50 to 60 DIM are

cows that had

SCS

with with

) ) represent

t t 90 to 110 DIM, and C) 90 to 110 DIM plotted against 210 to 230

), ), cows

∆

4 at all 4 time 4 all points. at 4

squares squares (■

SCS

60 60 DIM are designated with a black circle (o), cows with SCS

4 4 at 50 to

Overview of Overview somatic cell (SCS)score based population stratification.

50 50 to 60 DIM, B) 50 to 60 DIM plotted agains

50 50 to 60 DIM are designated with a red triangle ( Figure Figure 4.1. against DIM. Cows with SCS < 4 at 10 to 14 DIM and 50 to 60 DIM are designated with a green square ( at 10 to 14 DIM and designated with a blue star (*). Solid green cows represent (▲) had that triangles

Genome-Wide Association. Efficient Mixed Model Linear analysis (EMMAX) models were used to allow the inclusion of the IBD matrix to correct for any population structure in Golden

Helix SNP & Variation Suite software (Kang et al., 2010). Additive, dominant, and recessive inheritance models were considered along with the variables of farm, parity, and milk yield given their potential confounding effects on traits of interest. Lactation length was included as a covariate in all genome wide associations (GWA) for total lactation traits (median SCS, AUC) to address the influence of varying time at risk on longitudinal SCS measures. P-values were adjusted for multiple testing using a Bonferroni correction and False Discovery Rate (FDR) threshold of

0.05.

Model Selection and Candidate Gene Investigation. To identify QTL associated with elevated

SCC and potential subclinical mastitis at time of calving, a case-control GWA using the CMT score at 0-1 DIM was performed. To determine the influence of lactation stage on elevated SCC events, 5 GWA were performed for the quantitative trait of SCS at 3-5 DIM, 10-14 DIM, 50-

60 DIM, 90-110 DIM, and 210-230 DIM. To establish a point of comparison for these lactation stage analyses, a GWA for the traditional phenotype of average lactation SCS was performed.

To account for variation in SCS not represented by a median score, a GWA was performed for

SCS AUC. A genomic inflation factor lambda (median observed P-value divided by the median expected P-value) and Quantile–Quantile (QQ) plots of the log10 (Expected P-values) against the log10 (Observed P-values) were used to select the most appropriate models for each trait.

Regions with SNPs passing FDR were interrogated for candidate genes, and any gene in linkage disequilibrium (LD) with an associated marker (r2 ≥ 0.8), or in a 500 kb upstream or

downstream range of said marker if LD was not present, was identified using the NCBI RefSeq

Database (O'Leary et al., 2016). All genome coordinates given use the most recent bovine genome assembly ARS UCD 1.2.

RESULTS

After all quality assurance measures were applied, 454 cows with 581,663 SNPs remained for analysis. Descriptive statistics for SCS phenotypes are reported in Table 4.1. A comprehensive list of differentiated or significantly associated markers, their positions, and surrounding candidate genes can be found in Supplementary Tables 4.1 and 4.2, respectively.

Population Analyses. FST by marker comparing the different populations is shown in Figure

4.2. There was no apparent allelic differentiation between Average and Healthy populations

(Figure 4.2a). There were 12 SNPs indicating divergence (FST ≥ .4) between Average and

Chronic populations on Bos taurus autosomes (BTA) 4, 6, and 7 (Figure 4.2b). In a comparison of Healthy and Chronic populations, 3 SNPs were differentiated (FST ≥ .4) on BTA 4 and 5

(Figure 4.2c). All 9 moderately differentiated SNPs on BTA 4 resided in an LD block including two SNPs located at BTA 4:18655887 and 18669066 which were identified in both the Average and Healthy cow comparisons to the Chronic cows (Figure 4.3). For each SNP, chromosome, position, major and minor alleles, and minor allele frequencies (MAF) stratified by Healthy,

Average, and Chronic populations are summarized in Table 4.2.

Genome-Wide Associations. Final models selected are summarized in Table 4.3 with respective

Manhattan plots shown in Figure 4.4. A dominant inheritance model with farm and parity covariates significantly associated 9 SNPs (FDR < 0.05) with CMT score at time of calving on

BTAs 4 and 13 (Figure 4.4a). Recessive inheritance models with no covariates identified QTL associated with SCS at different stages in lactation spanning the genome. A total of 23 SNPs

100

2.1 2.4 2.6 2.5 2.8

IQR

574.6

rtile range range rtile (IQR),

2.8 3.5 4.9 6.1 5.6 4.5

179557

Variance

1.8 3.7 2.5 1.4 1.3 1.9

512.4

Median

2.2 4.1 2.9 1.9 1.9 2.3

577.2

Mean

451 454 454 451 448 438 418

230 DIM 230

Descriptive statistics of each SCS phenotype. Mean, median, variance, SCS interqua median, phenotype. statistics Mean, each of Descriptive

14 DIM 14 DIM 60 DIM 110

- - -

5 DIM 5

at 50 at

sample size (n) across all cows are (n) reported. size across are sample cows all

Table and SCS Lactation Under Curve Area SCS Median Lactation SCS3 at SCS10 at SCS SCS90 at SCS210 at

101

≥ 4) are indicated purple. in are 4) ≥

th moderate differentiation (F allelic moderate th

by by marker comparing a) Average and Healthy, b) Average and Chronic, and c) Healthy and Chronic

Figure Figure 4.2. SNPs populations. wi

C A

102

≥ 4) are indicated purple. in are 4) ≥

elic differentiation (F differentiation elic

) Average and Chronic, b) Healthy and Chronic populations, and c) linkage linkage c) and populations, Chronic and b) Chronic, Healthy and Average )

by marker on BTA 4 comparing a comparing 4 BTA on marker by

3. 3.

Figure Figure all SNPs with moderate disequilibriumthis in region.

B A 103

0.45 0.45 0.45 0.45 0.45 0.45 0.50 0.45 0.45 0.32 0.32 0.55 0.27

Chronic

0.07 0.07 0.07 0.07 0.07 0.07 0.06 0.08 0.08 0.04 0.04 0.13 0.06

Average

Minor Allele Frequency MinorAllele

0.10 0.10 0.10 0.10 0.09 0.10 0.11 0.10 0.10 0.06 0.06 0.16 0.03

Healthy

A A A A A A G G A A A A

Allele

Minor

G G A G C G G A A G G G G

Allele Major Major

1 1 1 1 1 1 1 1 1 1 2

0.40

ele frequency by population are given. Bolded rows indicate SNPs rows Bolded given. are population by frequency ele

/0.40

/ /

0.47 0.47 0.47 0.47 0.60 0.44 0.44 0.42 0.42 0.42 0.42

0.47

Position

18718084 18655237 18655887 18662663 18669066 18685097 18698840 18705074 18707492 43160366 43877784 103041074 13984552

> 0.4) between Average and Chronic and between 0.4) Average > populations Chronic and between 0.4) Healthy > populations

ST ST

4 4 4 4 4 4 4 4 4 7 7 6 5

study e study

BTA

analysis identifies markers differentiating Holstein cow populations based on their SCS. Chromosome, based populations SCS. markers differentiating cow their on identifies Holstein analysis Chromosome,

c populations. and c Average cowsHealthy both from

87868

14 DIM until until DIM 14

14 DIM until study DIM 14 end

00005562

values, major and minor alleles, as minor as all as well minor alleles, and major values,

NGS

Marker based based F Marker

eHD0700012914

BFGL

SCS > 4 from 10 from 4 > SCS moderately differentiated SNPs differentiated (F moderately SNPs differentiated (F moderately 10 from 4 ≤ SCS SCS period study during Cowsfluctuating with

BovineHD0400005579 Table position,F Chroni differentiating Marker BovineHD04 BovineHD0400005563 BovineHD0400005564 BovineHD0400005565 BovineHD0400005570 BovineHD0400005575 BovineHD0400005577 BovineHD0400005578 Bovin ARS BovineHD0600029283 BovineHD0500004192 1 2 3 4 5

104

< 0.05 <

0 0 0 0 0 0

10 24

P P

Bonferroni

9 0 8 0 0

23 25

106

FDR

0.54 0.40 0.12 0.31 0.43 0.35 0.49 0.55

Pseudo

heritability

1.00 1.00 0.98 1.00 0.99 0.99 1.01 0.99

Pseudo

lambda

ntile plots in Supplementary Figure 1 ntileFigure plots Supplementary in

Qua

Quantile

Inheritance

Recessive Dominant Recessive Recessive Recessive Recessive Recessive Recessive

control

Model Type Model

Linear Case Linear Linear Linear Linear Linear Linear

451 454 454 448 438 418 454 454

wide association models selected. Summary of final models including sample size (N), model type, type, (N), model size including models sample final of selected. models wideSummary association

Genome

230 DIM 230

SCSTrait

14 DIM 14 60 DIM 60 DIM 110

- - -

5 DIM 5

genomic inflation factor for model quality control, quality model for factor inflation genomic at multiple corrections testing Discovery False or (FDR) Bonferroni Rate either passing SNP of associations number

10 Table SNPs. associated significantly trait of number total patterns, and quality heritabilities, measures, control inheritance CMT 3 50 90 210 SCS Median AUC 1 2

105

were significantly associated with SCS at 10-14 DIM on BTAs 3, 5, 6, 11, and 29 (Figure 4.4c);

8 SNPs were significantly associated with SCS at 50-60 DIM on BTAs 1, 3, 4, and 26 (Figure

4.4d); and 106 SNPs were significantly associated with SCS at 210-230 DIM on ten chromosomes spanning the genome (Figure 4.4f). There were no significant associations with

3-5 DIM or 90-110 DIM (Figure 4.4b and e, respectively). Recessive inheritance models with covariates of parity, farm, and lactation length significantly associated 25 SNPs with median

SCS on 12 chromosomes across the genome (Figure 4.4g); no SNPs were significantly associated with AUC.

DISCUSSION

Population Analyses. In a comparison of Healthy and Average populations, the low FST by marker values suggest there is not great genetic difference between cows with consistently low

SCC and those who experience fluctuations in SCC as their lactation progresses. However, FST by marker analysis reveals genomic regions differentiated Chronic cows from both Average and Healthy cows, suggesting that cows which chronically experience elevated SCC are genetically different from cows with more consistently low milk SCC. Of particular interest is the QTL identified on BTA 4, which indicated differentiation of Chronic from all other cows

(Figure 4.3). The breaks in this LD block suggest this region may be a recombination site; only

2 of the 9 markers differentiating Chronic and Average also differentiated Healthy cows, but

LD suggests they are inherited together in the study population. An overlapping QTL has been previously mapped for milk SCS in a large Holstein study using microsatellite data and a granddaughter design (Zhang et al., 1998), reinforcing the relevance of these SNP markers for use in selecting for mastitis resistant cows. The SNP at BTA 6:103041074 has not been previously

106

Figure 4.4. Manhattan plots showing associated SNPs at a) 0-1 DIM, b) 3-5 DIM. c) 10-14 DIM, d) 50- 60 DIM, e) 90-110 DIM, f) 210-230 DIM, and g) Median SCS for entire lactation. Significantly associated SNPs (FDR < 0.05) are indicated in blue. Bonferroni multiple correction threshold indicated by the black line. 107

associated with mastitis-related traits, but falls directly within the protein-coding gene Janus

Kinase and Microtubule Interacting Protein 1 (JAKMIP1), which is suspected to play a role in the microtubule-dependent transport of the GABA-B receptor and is commonly expressed in lymphoid tissues (Costa et al., 2007). The QTL identified on BTA 7 differentiating Average and Chronic cows overlaps with QTL previously mapped for milk somatic cell score in German

Holsteins (Kuhn et al., 2003). Other QTL in this region have been mapped for milking speed and milk yield, both of which have been associated with mastitis risk (Boichard et al., 2003;

Schrooten et al., 2004; Japertiene et al., 2007; Marete et al., 2018b). Within this QTL, a SNP at

BTA 7:43160366 lies directly within Mucosal Vascular Addressin Cell Adhesion Moledule 1

(MADCAM1), which encodes an endothelial cell adhesion molecule involved in directing leukocytes towards inflammation sites (Vavricka et al., 2018). A second SNP at

BTA7:43877784 is within Adenomatous Polyposis Coli Protein 2 (APC2), a regulator of the

WNT2 signaling pathway which has been associated with the pathogenesis of several human cancers (Katoh, 2003). We propose that these differentiated markers are candidates for inclusion in selection indices for mastitis-resistant cows, given prior association to mastitis- related traits through QTL mapping and the immune-related functions of the genes within which these SNPs reside. Selection strategies based on this genomic region should be made with caution, given that it has been mapped for the traits of milk yield and milking speed, and further investigation is required to ensure selection for mastitis resistance is not antagonistic to milk yield, and vice versa.

Genome-wide Associations. To assess the genetics underlying postpartum elevated SCC, a case-control GWA for CMT score at 0-1 DIM (negative n = 398, positive n = 125) was

108

performed, and SNPs significantly associated on BTA 4 and 13, the former of which has not been previously mapped for any health traits (Figure 4.4a). The associated marker at BTA

4:97507560 lies directly within Exocyst Complex Component 4 (EXOC4), a protein complex which has been implicated in mechanisms of host defense due to its role in long-range cell signaling and phagosome maturation (Martin-Urdiroz et al., 2016). A significantly associated

SNP located at BTA 13:35930715 overlaps with a previously mapped QTL for SCS and milk yield in Danish Holsteins, which suggests this QTL may have pleiotropic effects related to milk production traits (Lund et al., 2008). Nearby candidate genes include Lysozyme Like 1 (LYZL1), an antimicrobial protein, BMP And Activin Membrane Bound Inhibitor (BAMBI), which has been implicated in macrophage and T cell regulation (Sun et al., 2019). We previously associated positive CMT scores at time of calving with increased odds of experiencing an elevated SCC event, suggesting that not only may CMT score at time of calving be a useful measure for selective management of high risk cows, but that these associated markers are strong candidates for inclusion in health trait indices for genomic selection (Miles et al., 2019).

However, the potential pleiotropy of the associated QTL on BTA 13 makes it difficult to use in genomic selection due to the potentially inverse relationships of SCS and milk yield.

The GWA for milk SCS at 10-14 DIM, representing peak mastitis incidence, significantly associated SNPs at both novel and previously mapped QTL for mastitis traits

(Figure 4.4c). A significantly associated SNP at BTA 3:89806231 overlaps with a QTL previously associated with SCS in a GWA study using imputed SNP chip data in French

Holsteins (Marete et al., 2018b). The SNPs significantly associated with SCS at 10-14 DIM at

BTA 5:955228 and 6:111916870 have had no prior association with milk SCS, whereas the associated SNP at BTA 6:30947112 overlaps with QTL previously associated with both clinical

109

mastitis and SCS (Klungland et al., 2001; Bennewitz et al., 2004; Michenet et al., 2016). Two large QTL were significantly associated on BTA 11, and while neither have been previously mapped for SCS or clinical mastitis, both overlap with previously mapped QTL for milk yield, which has a well-established relationship to mastitis, in a study of Danish Holsteins (Kucerova et al., 2006). The QTL at BTA29: 21528226-21543306 significantly associated with SCS at 10-

14 DIM has not been associated with mastitis before, though does overlap with a QTL previously mapped for milk yield (Kucerova et al., 2006). While not within any genes, the SNPs in this associated block neighbor the Small VCP Interacting Protein (SVIP), which is an inhibitor of the Endoplasmic Reticulum-Associated Degradation (ERAD) pathway, and thus suspected to protect the cell from excessive degradation (Wang et al., 2011). The identification of novel QTL for SCS during peak mastitis incidence (10-14 DIM) suggests that some genetic drivers of mastitis are not detected with de-regressed breeding values or lactation averages.

Perhaps only QLT not previously mapped for milk yield should be used in selection for mastitis resistance, given the multiple reports of positive genetic correlation between clinical mastitis and milk yield, and more precise SCS phenotyping is required to make this distinction (Pösö and Mäntysaari, 1996; Heringstad et al., 1999; Hansen et al., 2002).

To assess the genetics underlying milk SCC at the return to neutral/positive energy balance and the end of the voluntary breeding period, we performed GWA for SCS at 50-60

DIM and significantly associated a marker block at BTA 1: 65045803-65045168 which has not been previously mapped for mastitis-related traits (Figure 4.4d). The SNPs in this block were within the G Protein-Coupled Receptor 156 (GPR156) gene, a cell-surface receptor involved in GABA signaling, known to influence many pathways including immune response

(Liu et al., 2019). Significantly associated QTL on BTA 3: 118976220-119034557 have not

110

been previously mapped for mastitis related traits, but were associated with udder height and fore udder attachment, traits we previously associated with increased odds of elevated SCC

(Cole et al., 2011; Miles et al., 2019). A significantly associated SNP at BTA 4: 19709677 overlaps with QTL previously mapped for SCS in multiple studies, and one at BTA 26:

30956828 was previously mapped for both SCS and clinical mastitis (Zhang et al., 1998;

Rupp and Boichard, 1999; Kuhn et al., 2003; Longeri et al., 2006; Lund et al., 2008). Nearby candidate genes include MAX Interactor 1 (MXI1), a dimerization protein implicated in hematopoietic stem cell differentiation and implicated in the inflammatory response (Yoo et al., 2007). We posit that examining SCS at this time point removes bias introduced in the phenotype by commonly observed non-pathogenic causes of inflammation in early lactation, such as lactogenesis, uterine involution, and negative energy balance. Furthermore, the confidence of the associations on BTAs 4 and 26 are strongly validated given their mapping in other independent studies.

A GWA for SCS at 210-230 DIM, representing mid to late lactation and mid pregnancy, significantly associated SNPs on BTA 4 at 22557323 and 118912645 (Figure 4.4f). The former resides within Diacylglycerol Kinase Beta (DGKB), a regulator associated with stimulation of peripheral blood hematopoietic stem cells (Mishima et al., 2017). Both of these associated SNPs overlap with previously mapped QTL for both somatic cell count and score (Zhang et al., 1998;

Kuhn et al., 2003; Longeri et al., 2006; Ibeagha-Awemu et al., 2016). A region located on BTA

5:10674537-10679054 significantly associated with SCS at 210-230 DIM overlaps with previously mapped QTL for both SCS and clinical mastitis (Lund et al., 2008; Pimentel et al.,

2011). We identified a novel QTL on BTA 5:113340241-113358226, in which significantly associated SNPs lie inside the NFAT Activating Protein with ITAM Motif 1 (NFAM1) gene, a

111

promoter for cytokine-encoding genes and also believed to regulate B cell development and signaling (Ohtsuka et al., 2004). Also significantly associated with SCS at 210-230 DIM was a

QTL on BTA 6:4761931-4920678 which overlaps with QTL previously mapped for SCC

(Klungland et al., 2001). In addition, a significantly associated SNP in this region lies within

PR Domain Zinc Finger Protein 5 (PRDM5), a transcription factor associated with cell proliferation and the modulation of pro-inflammatory cytokines (Wu et al., 2017). A significantly associated SNP on BTA 7:95230312 was in near proximity to Endoplasmic

Reticulum Aminopeptidase 1 (ERAP1), polymorphisms of which have been shown to affect antigen presentation by MHC class 1 molecules and underlie the pathogenesis of various inflammatory diseases (Lopez de Castro et al., 2016). A large associated QTL on BTA

7:95730312-101060884 has been previously mapped for SCS, and contains the Solute Carrier

Organic Anion Transporter Family Member 4C1 and 6A1 (SLCO4C1 and SLCO6A1) genes, which have been implicated in the modulation of inflammatory responses (Toyohara et al.,

2009; Cole et al., 2011b). Two significantly associated SNPs on BTA 11: 66762518 and

68023914 fall within a previously mapped QTL for SCC (Schnabel et al., 2005). In addition, two novel QTL on BTA 11 were significantly associated and nearby candidate genes included

Lysocardiolipin Acetyltransferase 1 (LCLAT1) which has been suggested to control hematopoietic cell differentiation, Yippee Like 5 (YPEL5) which has been shown to negatively regulate innate immune responses, FOS Like 2 (FOSL2) whose signaling pathway is involved in many branches of the immune system, as well as the BRISC and BRCA1 A Complex Member

2 (BABAM2), an adapter of the BRISC complex which has been shown to down-regulate the immune response to bacterial lipopolysaccharides and is suspected to regulate interferon and tumor necrosis factor alpha signaling (Li et al., 2004; Xiong et al., 2008; Ciofani et al., 2012;

112

Zheng et al., 2013; Jeidane et al., 2016). Further candidate genes were identified in this region on BTA 11, including Intersectin 2 (ITSN2) which is implicated in T cell antigen receptor endocytosis (Burbage et al., 2018). A significantly associated QTL on BTA 18 was not previously mapped for mastitis-related traits, but we identified nearby candidate genes including Interleukin 34 (IL34), a cytokine involved in macrophage viability and differentiation and whose signaling pathways affect many mechanisms of the immune system, Fucose Kinase

(FCSK), a catalyzing enzyme for L-fucose utilization, which has been implicated in antigen recognition and immune cell-cell interactions, and Golgi Glycoprotein 1 (GLG1), which is part of a critical ligand-receptor interaction for early response to lipopolysaccharides in bovine

(Yuan et al., 2008; Li et al., 2009; Higgins et al., 2014; Lin et al., 2019). An associated QTL on

BTA 21 was previously mapped for clinical mastitis, and nearby candidate genes include

Endothelin 1 (ET1), a cell receptor involved in the regulation of inflammation (Schulman et al.,

2004; Elisa et al., 2015). Significantly associated SNPs on BTAs 22 and 23 have previously been mapped for SCC (Schulman et al., 2004; Maltecca et al., 2011). A significantly associated

QTL on BTA 23:14982097-15014257 has not been previously mapped for mastitis traits, but candidate genes in LD with associated SNPs include Triggering Receptor Expressed On

Myeloid Cells Like 1 through 3 (TREM1, TREM2, TREM3), a well-established positive regulator of inflammation, and Transcription Factor EB (TFEB), which has been implicated as a negative regulator of endothelial inflammation and T cell recruitment (Lu et al., 2017; Liu et al., 2018). By 210-230 DIM the confounding physiological factors of parturition, lactogenesis, negative energy balance, and inflammation during early pregnancy have been removed and milk SCS may be a more accurate reflection of mastitis. The use of generalized milk SCC and lack of pathogenesis-specific phenotyping may explain the extremely polygenic nature of this

113

trait and diverse arms of the immune system implicated in these analyses. Milk SCC with differential or microbial profiling would be required to refine the phenotype at this lactation stage and may result in more concentrated signals to specific genomic regions.

A GWA was performed for the traditional phenotype of lactation median SCS to provide a point of comparison for our lactation stage-specific analyses, and significantly associated

SNPs on 11 chromosomes spanning the genome (Figure 4.4g). Novel QTL were associated on

BTAs 2 and 4; nearby candidate genes included Sphingomyelin Phosphodiesterase Acid Like

3B (SMPDL3B), which has been shown to negatively regulate Toll Like Receptor signaling and to influence macrophage membrane fluidity, Thymocyte Selection Associated Family Member

2 (THEMIS2), which has been implicated in positive regulation of B cells and is believed to regulate macrophage TLR signaling, and Ficolin 3 (FCN3), a driver of the lectin complement pathway, as well as Synaptotagmin Like 1 (SYTL1), a mediator of neutrophil chemotaxis and granule exocytosis (Garred et al., 2009; Peirce et al., 2010; Heinz et al., 2015; Cheng et al.,

2017; Ramadass et al., 2019). Similarly, mutations in Non-Homologous End Joining Factor 1

(NHEJ1) have been implicated in immunodeficiency disorders (Al-Marhoobi et al., 2019).

Significantly associated SNPs on BTA 6 in a gene-scarce region were previously mapped for both SCS and clinical mastitis (Bennewitz et al., 2004; Sodeland et al., 2011). A significantly associated SNP at BTA 11:48421277 overlaps with a QTL previously mapped for SCS, and was positioned near T Cell Surface Glycoprotein CD8 Beta and Alpha Chain (CD8a and b), which is heavily involved in antigen presentation (Zhang et al., 1998; Parel and Chizzolini,

2004). A significantly associated QTL on BTA 15 was previously mapped for SCS, and nearby candidate genes include RAB39a Member RAS Oncogene Family (RAB39A) which has been implicated in macrophage activation in response to lipopolysaccharides as well as ATM

114

Serine/Threonine Kinase (ATM) which is involved in B cell antigen receptor expression

(Boichard et al., 2003; Rossi and Gaidano, 2012; Seto et al., 2013). Both associated SNPs on

BTA 19 have been previously mapped for SCS in multiple studies, and candidate genes include

Interferon Induced Protein 35 (IFI35) a proinflammatory damage-associated molecular pattern released by activated macrophages in response to lipopolysaccharides, Transmembrane Protein

106A (TMEM106A) which has been shown to upregulate and polarize macrophages thereby inducing proinflammatory cytokine release, and CD300 Molecule Like Family Member G

(CD300LG), widely expressed on hematopoietic cells and involved in lymphocyte transmigration in inflamed tissues (Bennewitz et al., 2004; Tal-Stein et al., 2010; Cole et al.,

2011b; Umemoto et al., 2013; Dai et al., 2015; Xiahou et al., 2017). A novel QTL was associated on BTA 20:26074612-26079633, and associated SNPs are located inside Integrin

Subunit Alpha 2 (ITGA2), which is suspected to mediate inflammation by regulating immune cell adhesion (Adorno-Cruz and Liu, 2019). Two SNPs on BTA 28 were significantly associated, one novel (BTA 28:19095996) and one near a previously mapped QTL for SCS in beef cattle (Imumorin et al., 2011). There were no commonalities among the median SCS and lactation stage GWA, suggesting that there are indeed different genetic mechanisms driving milk SCC over the course of a lactation.

CONCLUSIONS

A total of 48 QTL were identified via GWA for SCS, and of those 27 were novel, having never been mapped for SCC or clinical mastitis previously. Numerous candidate genes were identified relating to immunity, and their function varied across multiple branches of the immune system including stimulation of inflammatory cytokines, macrophage activation, B

115

cell maturation, T cell recruitment, and the lectin pathway of the complement system.

Redundancy is built into the immune system, and the identification of candidate genes belonging to different branches suggest that extremely mastitis susceptible cows may experience failures in multiple immune response mechanisms. We found that extremely susceptible cows (Chronic) are moderately differentiated from others (Healthy and Average) and that this allelic differentiation can be localized to BTAs 4, 6, and 7. We posit that variation in SCS across a lactation must be considered in genomic selection strategies given the non-pathological causes of udder inflammation and the differing etiologies of mastitis.

The wide range of significantly associated QTL and diversity of candidate gene functions suggest that selection based on phenotypes specific to lactation stage and pathogenesis rather than generalized SCS may lead to greater success in breeding mastitis-resistant cows.

ACKNOWLEDGEMENTS

We thank commercial dairy farms for their invaluable participation, and many Cornell

University graduate and undergraduate students associated with the Huson lab for their assistance with on farm sampling and laboratory work. Special thanks to Allison Herrick for her assistance in data compilation and Stephen Parry of the Cornell University Statistical

Consulting Unit.

116

REFERENCES

Adorno-Cruz, V., Liu, H., 2019. Regulation and functions of integrin alpha2 in cell adhesion

and disease. Genes Dis 6, 16-24.

Akbar, H., Grala, T.M., Vailati Riboni, M., Cardoso, F.C., Verkerk, G., McGowan, J.,

Macdonald, K., Webster, J., Schutz, K., Meier, S., Matthews, L., Roche, J.R., Loor, J.J.,

2015. Body condition score at calving affects systemic and hepatic transcriptome

indicators of inflammation and nutrient metabolism in grazing dairy cows. Journal of

Dairy Science 98, 1019-1032.

Al-Marhoobi, R., Al-Musalhi, M., Naseem, S.U., Wali, Y., Alsayegh, A., Al-Tamemi, S., 2019.

Combined Immunodeficiency, Hemolytic Anemia, and Growth Retardation Secondary

to a Homozygous Mutation in the NHEJ1 Gene. J Pediatr Hematol Oncol.

Andersson, L., 2001. Genetic dissection of phenotypic diversity in farm animals. Nature

Reviews Genetics 2, 130-138.

Bennewitz, J., Reinsch, N., Guiard, V., Fritz, S., Thomsen, H., Looft, C., Kuhn, C., Schwerin,

M., Weimann, C., Erhardt, G., Reinhardt, F., Reents, R., Boichard, D., Kalm, E., 2004.

Multiple quantitative trait loci mapping with cofactors and application of alternative

variants of the false discovery rate in an enlarged granddaughter design. Genetics 168,

1019-1027.

Biffani, S., Del Corvo, M., Capoferri, R., Pedretti, A., Luini, M., Williams, J.L., Pagnacco, G.,

Minvielle, F., Minozzi, G., 2017. An alternative experimental case–control design for

genetic association studies on bovine mastitis. animal 11, 574-579.

117

Bionaz, M., Trevisi, E., Calamari, L., Librandi, F., Ferrari, A., Bertoni, G., 2007. Plasma

Paraoxonase, Health, Inflammatory Conditions, and Liver Function in Transition Dairy

Cows. Journal of Dairy Science 90, 1740-1750.

Boichard, D., Grohs, C., Bourgeois, F., Cerqueira, F., Faugeras, R., Neau, A., Rupp, R.,

Amigues, Y., Boscher, M.Y., Leveziel, H., 2003. Detection of genes influencing

economic traits in three French dairy cattle breeds. Genet Sel Evol 35, 77-101.

Burbage, M., Gasparrini, F., Aggarwal, S., Gaya, M., Arnold, J., Nair, U., Way, M.,

Bruckbauer, A., Batista, F.D., 2018. Tuning of in vivo cognate B-T cell interactions by

Intersectin 2 is required for effective anti-viral B cell immunity. Elife 7.

Cai, Z., Guldbrandtsen, B., Lund, M.S., Sahana, G., 2018. Prioritizing candidate genes post-

GWAS using multiple sources of data for mastitis resistance in dairy cattle. BMC

Genomics 19, 656-656.

Cheng, D., Deobagkar-Lele, M., Zvezdova, E., Choi, S., Uehara, S., Baup, D., Bennett, S.C.,

Bull, K.R., Crockford, T.L., Ferry, H., Warzecha, C., Marcellin, M., de Peredo, A.G.,

Lesourne, R., Anzilotti, C., Love, P.E., Cornall, R.J., 2017. Themis2 lowers the

threshold for B cell activation during positive selection. Nat Immunol 18, 205-213.

Ciofani, M., Madar, A., Galan, C., Sellars, M., Mace, K., Pauli, F., Agarwal, A., Huang, W.,

Parkhurst, C.N., Muratet, M., Newberry, K.M., Meadows, S., Greenfield, A., Yang, Y.,

Jain, P., Kirigin, F.K., Birchmeier, C., Wagner, E.F., Murphy, K.M., Myers, R.M.,

Bonneau, R., Littman, D.R., 2012. A validated regulatory network for Th17 cell

specification. Cell 151, 289-303.

Cole, J.B., Wiggans, G.R., Ma, L., Sonstegard, T.S., Lawlor, T.J., Jr., Crooker, B.A., Van

Tassell, C.P., Yang, J., Wang, S., Matukumalli, L.K., Da, Y., 2011. Genome-wide

118

association analysis of thirty one production, health, reproduction and body

conformation traits in contemporary U.S. Holstein cows. BMC Genomics 12, 408.

Costa, V., Conte, I., Ziviello, C., Casamassimi, A., Alfano, G., Banfi, S., Ciccodicola, A., 2007.

Identification and expression analysis of novel Jakmip1 transcripts. Gene 402, 1-8.

Dai, H., Xu, D., Su, J., Jang, J., Chen, Y., 2015. Transmembrane protein 106a activates mouse

peritoneal macrophages via the MAPK and NF-kappaB signaling pathways. Sci Rep 5,

12461.

Dohoo, I.R., Leslie, K.E., 1991. Evaluation of changes in somatic cell counts as indicators of

new intramammary infections. Preventive Veterinary Medicine 10, 225-237.

Elisa, T., Antonio, P., Giuseppe, P., Alessandro, B., Giuseppe, A., Federico, C., Marzia, D.,

Ruggero, B., Giacomo, M., Andrea, O., Daniela, R., Mariaelisa, R., Claudio, L., 2015.

Endothelin Receptors Expressed by Immune Cells Are Involved in Modulation of

Inflammation and in Fibrosis: Relevance to the Pathogenesis of Systemic Sclerosis. J

Immunol Res 2015, 147616.

Fang, L., Sahana, G., Su, G., Yu, Y., Zhang, S., Lund, M.S., Sørensen, P., 2017. Integrating

Sequence-based GWAS and RNA-Seq Provides Novel Insights into the Genetic Basis

of Mastitis and Milk Production in Dairy Cattle. Sci Rep 7, 45560-45560.

Forutan, M., Ansari Mahyari, S., Baes, C., Melzer, N., Schenkel, F.S., Sargolzaei, M., 2018.

Inbreeding and runs of homozygosity before and after genomic selection in North

American Holstein cattle. BMC Genomics 19, 98-98.

Garred, P., Honore, C., Ma, Y.J., Munthe-Fog, L., Hummelshoj, T., 2009. MBL2, FCN1, FCN2

and FCN3-The genes behind the initiation of the lectin pathway of complement. Mol

Immunol 46, 2737-2744.

119

Graugnard, D.E., Bionaz, M., Trevisi, E., Moyes, K.M., Salak-Johnson, J.L., Wallace, R.L.,

Drackley, J.K., Bertoni, G., Loor, J.J., 2012. Blood immunometabolic indices and

polymorphonuclear neutrophil function in peripartum dairy cows are altered by level of

dietary energy prepartum1. Journal of Dairy Science 95, 1749-1758.

Hansen, M., Lund, M.S., Sorensen, M.K., Christensen, L.G., 2002. Genetic parameters of dairy

character, protein yield, clinical mastitis, and other diseases in the Danish Holstein

cattle. J Dairy Sci 85, 445-452.

Heinz, L.X., Baumann, C.L., Koberlin, M.S., Snijder, B., Gawish, R., Shui, G., Sharif, O.,

Aspalter, I.M., Muller, A.C., Kandasamy, R.K., Breitwieser, F.P., Pichlmair, A.,

Bruckner, M., Rebsamen, M., Bluml, S., Karonitsch, T., Fauster, A., Colinge, J.,

Bennett, K.L., Knapp, S., Wenk, M.R., Superti-Furga, G., 2015. The Lipid-Modifying

Enzyme SMPDL3B Negatively Regulates Innate Immunity. Cell Rep 11, 1919-1928.

Heringstad, B., Klemetsdal, G., Ruane, J., 1999. Clinical mastitis in Norwegian cattle:

Frequency, variance components, and genetic correlation with protein yield. Journal of

Dairy Science 82, 1325-1330.

Higgins, M.A., Suits, M.D., Marsters, C., Boraston, A.B., 2014. Structural and functional

analysis of fucose-processing enzymes from Streptococcus pneumoniae. J Mol Biol

426, 1469-1482.

Ibeagha-Awemu, E.M., Peters, S.O., Akwanji, K.A., Imumorin, I.G., Zhao, X., 2016. High

density genome wide genotyping-by-sequencing and association identifies common and

low frequency SNPs, and novel candidate genes influencing cow milk traits. Sci Rep 6,

31109.

120

Imumorin, I.G., Kim, E.H., Lee, Y.M., De Koning, D.J., van Arendonk, J.A., De Donato, M.,

Taylor, J.F., Kim, J.J., 2011. Genome Scan for Parent-of-Origin QTL Effects on Bovine

Growth and Carcass Traits. Front Genet 2, 44.

Japertiene, R., Juozaitiene, V., Kriauziene, J., Rudejeviene, J., Japertas, S., 2007. The

interrelationships between milkability traits and subclinical mastitis in cows. Pol J Vet

Sci 10, 255-261.

Jeidane, S., Scott-Boyer, M.P., Tremblay, N., Cardin, S., Picard, S., Baril, M., Lamarre, D.,

Deschepper, C.F., 2016. Association of a Network of Interferon-Stimulated Genes with

a Locus Encoding a Negative Regulator of Non-conventional IKK Kinases and IFNB1.

Cell Rep 17, 425-435.

Kang, H.M., Sul, J.H., Service, S.K., Zaitlen, N.A., Kong, S.-y., Freimer, N.B., Sabatti, C.,

Eskin, E., 2010. Variance component model to account for sample structure in genome-

wide association studies. Nature Genetics 42, 348.

Katoh, M., 2003. WNT2 and human gastrointestinal cancer (review). Int J Mol Med 12, 811-

816.

Klungland, H., Sabry, A., Heringstad, B., Olsen, H.G., Gomez-Raya, L., Vage, D.I., Olsaker,

I., Odegard, J., Klemetsdal, G., Schulman, N., Vilkki, J., Ruane, J., Aasland, M.,

Ronningen, K., Lien, S., 2001. Quantitative trait loci affecting clinical mastitis and

somatic cell count in dairy cattle. Mamm Genome 12, 837-842.

Kucerova, J., Lund, M.S., Sorensen, P., Sahana, G., Guldbrandtsen, B., Nielsen, V.H.,

Thomsen, B., Bendixen, C., 2006. Multitrait quantitative trait Loci mapping for milk

production traits in danish Holstein cattle. J Dairy Sci 89, 2245-2256.

121

Kuhn, C., Bennewitz, J., Reinsch, N., Xu, N., Thomsen, H., Looft, C., Brockmann, G.A.,

Schwerin, M., Weimann, C., Hiendleder, S., Erhardt, G., Medjugorac, I., Forster, M.,

Brenig, B., Reinhardt, F., Reents, R., Russ, I., Averdunk, G., Blumel, J., Kalm, E., 2003.

Quantitative trait loci mapping of functional traits in the German Holstein cattle

population. J Dairy Sci 86, 360-368.

Li, C.J., Li, R.W., Elsasser, T.H., Kahl, S., 2009. Lipopolysaccharide-induced early response

genes in bovine peripheral blood mononuclear cells implicate GLG1/E-selectin as a key

ligand-receptor interaction. Funct Integr Genomics 9, 335-349.

Li, Q., Ching, A.K., Chan, B.C., Chow, S.K., Lim, P.L., Ho, T.C., Ip, W.K., Wong, C.K., Lam,

C.W., Lee, K.K., Chan, J.Y., Chui, Y.L., 2004. A death receptor-associated anti-

apoptotic protein, BRE, inhibits mitochondrial apoptotic pathway. J Biol Chem 279,

52106-52116.

Lin, W., Xu, D., Austin, C.D., Caplazi, P., Senger, K., Sun, Y., Jeet, S., Young, J., Delarosa,

D., Suto, E., Huang, Z., Zhang, J., Yan, D., Corzo, C., Barck, K., Rajan, S., Looney, C.,

Gandham, V., Lesch, J., Liang, W.C., Mai, E., Ngu, H., Ratti, N., Chen, Y., Misner, D.,

Lin, T., Danilenko, D., Katavolos, P., Doudemont, E., Uppal, H., Eastham, J., Mak, J.,

de Almeida, P.E., Bao, K., Hadadianpour, A., Keir, M., Carano, R.A.D., Diehl, L., Xu,

M., Wu, Y., Weimer, R.M., DeVoss, J., Lee, W.P., Balazs, M., Walsh, K., Alatsis, K.R.,

Martin, F., Zarrin, A.A., 2019. Function of CSF1 and IL34 in Macrophage Homeostasis,

Inflammation, and Cancer. Front Immunol 10, 2019.

Liu, F., Zhang, X., Zhang, B., Mao, W., Liu, T., Sun, M., Wu, Y., 2018. TREM1: A positive

regulator for inflammatory response via NF-kappaB pathway in A549 cells infected

with Mycoplasma pneumoniae. Biomed Pharmacother 107, 1466-1472.

122

Liu, F., Zhang, Y.Y., Song, N., Lin, J., Liu, M.K., Huang, C.L., Zhou, C., Wang, H., Wang, M.,

Shen, J.F., 2019. GABAB receptor activation attenuates inflammatory orofacial pain by

modulating interleukin-1beta in satellite glial cells: Role of NF-kappaB and MAPK

signaling pathways. Brain Res Bull 149, 240-250.

Longeri, M., Polli, M., Strillacci, M.G., Samore, A.B., Zanotti, M., 2006. Short communication:

quantitative trait loci affecting the somatic cell score on chromosomes 4 and 26 in Italian

Holstein cattle. J Dairy Sci 89, 3175-3177.

Lopez de Castro, J.A., Alvarez-Navarro, C., Brito, A., Guasp, P., Martin-Esteban, A., Sanz-

Bravo, A., 2016. Molecular and pathogenic effects of endoplasmic reticulum

aminopeptidases ERAP1 and ERAP2 in MHC-I-associated inflammatory disorders:

Towards a unifying view. Mol Immunol 77, 193-204.

Lu, H., Fan, Y., Qiao, C., Liang, W., Hu, W., Zhu, T., Zhang, J., Chen, Y.E., 2017. TFEB

inhibits endothelial cell inflammation and reduces atherosclerosis. Sci Signal 10.

Lund, M.S., Guldbrandtsen, B., Buitenhuis, A.J., Thomsen, B., Bendixen, C., 2008. Detection

of quantitative trait loci in Danish Holstein cattle affecting clinical mastitis, somatic cell

score, udder conformation traits, and assessment of associated effects on milk yield. J

Dairy Sci 91, 4028-4036.

Maltecca, C., Gray, K.A., Weigel, K.A., Cassady, J.P., Ashwell, M., 2011. A genome-wide

association study of direct gestation length in US Holstein and Italian Brown

populations. Anim Genet 42, 585-591.

Marete, A., Sahana, G., Fritz, S., Lefebvre, R., Barbat, A., Lund, M.S., Guldbrandtsen, B.,

Boichard, D., 2018. Genome-wide association study for milking speed in French

Holstein cows. J Dairy Sci 101, 6205-6219.

123

Martin-Urdiroz, M., Deeks, M.J., Horton, C.G., Dawe, H.R., Jourdain, I., 2016. The Exocyst

Complex in Health and Disease. Front Cell Dev Biol 4, 24-24.

McDermott, M.P., Erb, H.N., Natzke, R.P., 1982. Predictability by somatic cell counts related

to prevalence of intrammary infection within herds. J Dairy Sci 65, 1535-1539.

Mein, G., Reinemann, D., Schuring, N., Ohnstad, I., 2004. Milking machines and mastitis risk:

a storm in a teatcup. In, National Mastitis Council Annual Meeting, Charlotte, NC.

Michenet, A., Barbat, M., Saintilan, R., Venot, E., Phocas, F., 2016. Detection of quantitative

trait loci for maternal traits using high-density genotypes of Blonde d'Aquitaine beef

cattle. BMC Genet 17, 88.

Miles, A.M., McArt, J.A.A., Leal Yepes, F.A., Stambuk, C.R., Virkler, P.D., Huson, H.J.,

2019a. Udder and teat conformational risk factors for elevated somatic cell count and

clinical mastitis in New York Holsteins. Preventive Veterinary Medicine 163, 7-13.

Miles, A.M., McArt, J.A.A., Leal Yepes, F.A., Stambuk, C.R., Virkler, P.D., Huson, H.J.,

2019b. Udder and teat conformational risk factors for elevated somatic cell count and

clinical mastitis in New York Holsteins. Prev Vet Med 163, 7-13.

Mishima, S., Matsuda, C., Ishihara, T., Nagase, M., Taketani, T., Nagai, A., 2017. Single

nucleotide polymorphisms of the DGKB and VCAM1 genes are associated with

granulocyte colony stimulating factor-mediated peripheral blood stem cell mobilization.

Transfus Apher Sci 56, 154-159.

Misztal, I., 2006. Challenges of application of marker assisted selection - a review. Animal

Science Papers and Reports 24, 5-10.

O'Leary, N.A., Wright, M.W., Brister, J.R., Ciufo, S., Haddad, D., McVeigh, R., Rajput, B.,

Robbertse, B., Smith-White, B., Ako-Adjei, D., Astashyn, A., Badretdin, A., Bao, Y.,

124