Genetic parameter estimation for beef bull semen attributes

Abstract

Improvements in bull reproductive performance are necessary to optimize the efficiency of cattle production. Female fertility has been enhanced through assisted reproductive technologies as well as genetic selection; however, improving beef bull fertility has been largely ignored. Phenotypes routinely collected at bull semen collection facilities are believed to affect fertility and provide the phenotypes necessary for a genetic evaluation. The first objective of this study was to determine the significant fixed effects for modeling beef bull fertility using data from bull semen collection facilities. The second objective was to estimate variance components, heritabilities, repeatabilities, and correlations between beef bull semen attributes. Beef bull fertility phenotypes including volume (VOL), concentration (CONC), number of spermatozoa (NSP), initial motility (IMot), post-thaw motility (PTMot), 3-h post-thaw motility (3HRPTMot), percentage of normal spermatozoa (%NORM), primary abnormalities (PRIM), and secondary abnormalities (SEC) were obtained from two bull semen collection facilities. A total of 1,819 Angus bulls with 50,624 collection records were analyzed. Of the fixed class and covariate effects tested, the significant class effects were collection location and collection day within year and the significant covariate effects included age at collection, days since previous collection, and cumulative comprehensive climate index (CCI). For this study, the CCI was calculated for a 75-d period including the 61-d spermatogenesis cycle and 14-d epididymal transit time. The 75 d prior to collection accounted for the environmental stress a bull may have experienced over the course of development of the spermatozoa, which was more significant than the CCI calculated for collection day or spermatogenesis start date. Pre-thaw beef bull semen traits had low heritability estimates of 0.11 ± 0.02 (VOL), 0.09 ± 0.02 (CONC), 0.08 ± 0.02 (NSP), and 0.12 ± 0.03 (IMot). Heritabilities of post-thaw beef bull semen attributes were more variable at 0.10 ± 0.02 (PTMot), 0.05 ± 0.04 (3HRPTMot), 0.10 ± 0.04 (%NORM), 0.03 ± 0.03 (PRIM), and 0.18 ± 0.04 (SEC). Correlations of breeding values for these traits with scrotal circumference (SC) expected progeny difference (EPD) are low. The low to moderate heritability estimates indicate that genetic improvement can be made in beef bull semen quality traits if new tools are developed to augment the scrotal circumference EPD that are currently available within the industry.

Introduction

Reproductive efficiency impacts beef producers’ profitability and often determines whether beef producers will reach their production goals (Harris, 1970). Though most of the selection tools for fertility in the beef industry focus on female reproduction, only focusing on female reproductive capabilities limits the potential for increased performance and genetic improvement (Petrunkina and Harrison, 2011). Braundmeier and Miller (2001) acknowledge that using subfertile or infertile bulls affects the bull’s ability to cause conception. DeJarnette (2004) determined reproductive failure could be attributable to bull subfertility, and the fertility of a bull is dependent on the semen quantity, semen quality, and health status of the bull.

While fertility traits are generally considered to be lowly heritable, heritability estimates in the literature are, on average, largely moderate (0.20 to 0.31; Taylor et al., 1985; Mathevon et al., 1998; Kaps et al., 2000; Kealey et al., 2006; Gredler et al., 2007; Druet et al., 2009; Suchocki and Szyda, 2015; Berry et al., 2019). Dairy bull fertility phenotypes obtained from bull semen collection facilities have been used to estimate genetic parameters (Taylor et al., 1985; Mathevon et al., 1998; Druet et al., 2009; Suchocki and Szyda, 2015; Berry et al., 2019); however, the use of phenotypes obtained from beef bull semen collection facilities has been limited (Kaps et al., 2000; Kealey et al., 2006). Most of the studies of beef bull semen attributes have utilized breeding soundness examinations from yearling bulls (Smith et al., 1989; Christmas et al., 2001; Garmyn et al., 2011). Of the studies that have estimated the genetic parameters for beef bull semen attributes, few have incorporated genomic information. Including genomic information when evaluating lowly heritable polygenic traits improves the reliability of the estimates (Guarini et al., 2018).

The objective of this study was to determine the effects which contributed significantly to beef bull fertility, with the end goal of estimating variance components for beef bull semen attributes.

Materials and Methods

Data on beef bull semen attributes were collected from two beef bull semen collection facilities. Bull stud A provided a total of 48,611 collection records compiled between 2004 and 2018 on 1,626 bulls. Data obtained from stud B included 4,970 total collection records for 296 bulls collected between January 2008 and December 2018. Each collection record included the phenotypes volume (VOL), concentration (CONC), initial motility (IMot), post-thaw motility (PTMot), 3-h post-thaw motility (3HRPTMot), percentage of normal spermatozoa (%NORM), primary abnormalities (PRIM), and secondary abnormalities (SEC). The definitions of the beef bull semen attributes are outlined in Table 1. In addition to beef bull fertility phenotypes, the studs also provided registration number, name, breed, birth date, and owner.

Table 1.

Beef bull semen attributes, units of measure, and definitions utilized in the genetic evaluation¹

Traits	Units	Definition
VOL	mL	Total amount of the ejaculate, measured by milliliters
CONC	Million/mL	Relative amount of sperm cells per ejaculate, measured by a colorimeter
NSP	Million	Calculated by multiplying sperm concentration and semen volume; expressed in millions
IMot2	%	Percentage of progressively swimming spermatozoa in the ejaculate immediately after collection
PTMot2	%	Percentage of progressively swimming spermatozoa in the ejaculate, measured within 1 h of thawing
3HRPTMot2	%	Percentage of progressively swimming spermatozoa in the ejaculate, measured within 3 h after thawing
%NORM2	%	Percent morphologically normal spermatozoa
PRIM2	%	Percentage of spermatozoa with a defect to the head
SEC2	%	Percentage of spermatozoa with a defect to the tail

¹All PTMot measures are observed the day after the sample is collected. All traits are measured within 1 h of thawing with the exception of 3HRPTMot.

²Trait is measured subjectively by a trained laboratory technician.

Stud A and stud B had ejaculate quality standards a collection had to meet in order for the sample to be frozen and then distributed to the owner. The ejaculate had to meet freezing quality requirements in order to be frozen and tested for post-thaw semen quality measures. The freezing requirements were that a sample of the ejaculate had to have a progressive motility of at least 50% and no more than 30% abnormalities. If the pre-thaw standards were met, the post-thaw semen quality standards that were required for the semen to be considered viable were at least 50% progressive motility, no more than 30% abnormalities, and of the abnormalities, there could be no more than 19% of spermatozoa with a primary abnormality. As noted in Table 1, a PTMot and a 3HRPTMot were obtained for some of the samples. Post-thaw motility was measured within 1 h of thawing. After thawing, the sample was kept 37 °C. Three-hour post-thaw motility was measured within 3 h of thawing a sample of the ejaculate.

Summary statistics for all phenotypes were obtained from SAS v. 9.4 software (SAS Institute Inc., Cary, NC, USA). Data were edited with R software (R Core Team, 2013, Vienna, Austria) using the following criteria: VOL and CONC had to be greater than zero, and measurements recorded as percentages must be between 0 and 100. Records with a missing registration number were excluded from further analysis.

A total of 480 records were removed from stud A for having VOL or CONC measures at or below zero. For stud B, a total of 335 records were removed for either having a VOL or CONC phenotype recorded as a zero. No data were removed for any of the percentage traits, which included IMOT, PTMot, 3HRPTMot, %NORM, PRIM, and SEC, from either stud. For stud A, 48,131 collection records from 1,570 bulls were utilized for the analysis. From stud B, 2,493 collection records from 249 bulls were included in the analysis. Cumulatively, 50,624 collection records from 1,819 bulls were used to evaluate beef bull semen attributes. In analyses utilizing data combined across studs, only data recorded at both locations were used, which included VOL, CONC, IMot, NSP, PTMot, 3HRPTMot, %NORM, PRIM, and SEC.

A five-generation pedigree was obtained from the American Angus Association for 1,819 bulls. The five-generation pedigree consisted of 6,521 males and 17,136 females. The American Angus Association also provided scrotal circumference and expected progeny differences (EPDs) for 1,819 bulls. In addition, single-nucleotide polymorphism (SNP) data for 1,163 bulls were obtained from the American Angus Association. Imputation was performed by genotyping providers to the American Angus Association, so that all genotyped bulls had 54,609 SNP. Data were edited to remove SNPs with a call rate of <0.90 (n = 3,921) and minor allele frequency of <0.05 (n = 13,478). Animals with a call rate of < 0.90 were removed (n = 3). In addition, 620 SNP were removed due to the Mendelian conflict. A total of 29.75% of the SNP were removed during quality control; thus, 38,515 SNPs from 1,160 animals were available for analysis after quality control filtering.

There are a variety of environmental factors that affect bull fertility traits, including age (Taylor et al., 1985; Barth and Waldner, 2002; Senger, 2012), social interactions (Beerwinkle, 1974; Fritz et al., 1999; Van Eenennaam et al., 2007), temperature (Arnold and Dudzinski, 1978; Barth and Waldner, 2002; Garcia-Oliveros et al., 2020), and collection interval (Fuerst-Waltl et al., 2006). Putative predictor variables were extracted from the dataset to model environmental factors and utilized in a model selection procedure. The putative predictor variables included the collection location, collection barn, owner, season of collection (spring, summer, fall, and winter), day of collection, year of collection, day within year of collection, age of bull at collection, days since previous collection, weight, hip height, scrotal circumference, and collection number as a sequential number from the first collection to the last collection per bull. The season of collection was defined as spring comprising of March through May, summer comprising of June through August, autumn comprising of September through November, and winter comprising of December through February. Climatology data were obtained from the National Oceanic and Atmospheric Administration (NOAA; Rossow et al., 2016). The nearest NOAA station to stud A was approximately 20 miles away. The nearest NOAA station from stud B was approximately 10 miles away. Data obtained from NOAA included daily averages for temperature (Fahrenheit), humidity (%), pressure (inches of Mercury), wind speed (miles per hour), and snowfall (inches). Solar radiation was not available in the NOAA weather data, so yearly average solar radiation measurements (kW/m) for the same location where the NOAA data were obtained were accessed from the National Renewable Energy Laboratory (NREL; DOE, 2019). The temperature humidity index (THI; Mader et al., 2006) and comprehensive climate index (CCI; Mader et al., 2010) were calculated using the obtained weather variables. The THI and CCI were determined for the day of collection and 75-d prior to collection. In addition, a cumulative THI and CCI were calculated for a 75-d period including the 61-d spermatogenesis cycle and 14-d epididymal transit time. The 75 d prior to collection accounted for the environmental stress a bull may have experienced over the course of development of the spermatozoa. CCI and THI were determined to be collinear, and thus Akaike information criterion was utilized to determine whether to use CCI or THI in the model. For the remainder of the fixed effects, backward elimination was used to determine fixed effects contributing significantly to the model for each beef bull semen attribute.

Table 2 outlines the fixed effects that were significant for each beef bull fertility trait. For VOL, collection day within year, age, and days since previous collection were determined to be significant. Statistically significant effects for CONC were collection location and collection day within year. Initial motility and NSP both yielded significant effects of collection location, collection day within year, age, and days since previous collection. Collection location was not significant when modeling the %NORM or PRIM. Days since previous collection was not significant for any of the post-thaw beef bull semen attributes. In addition, cumulative CCI was not significant at a P-value of <0.05 for any of the traits, but when the P-values for CCI on collection day, spermatogenesis start day, and cumulative CCI were compared, cumulative CCI had the most significant P-value.

Table 2.

Levels of significance for the fixed effects used in the analysis of beef bull semen attributes

	Trait
Fixed effect	VOL	CONC	IMot	NSP	PTMot	3HRPTMot	%NORM	PRIM	SEC
n (50,624)	17,026	16,951	17,029	16,951	11,690	3,378	7,672	7,686	7,728
AIC1	85,270.8	23,4998.3	126,589.1	311,531.6	81,398.4	19,649.9	47,311.9	43,643.6	44,657
Collection location	0.7701	<0.0001	0.0016	<0.0001	<0.0001	—	0.9817	0.149	<0.0001
Collection day within year	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001
Age	<0.0001	0.5752	0.0004	<0.0001	0.0184	0.0007	0.0023	0.046	0.0771
Days since previous collection	<0.0001	0.4152	0.0022	<0.0001	0.8981	0.7902	0.3268	0.5217	0.7807
Cum CCI	0.2798	0.217	0.6691	0.4196	0.7352	—	0.8383	0.2431	0.9834

¹AIC, Akaike Information Criterion.

After examining the significant effects found in the current study and comparing them with previously published studies (Taylor et al., 1985; Berry et al., 2019), the following model was utilized for the genetic evaluation for all beef bull semen attributes:

$${\displaystyle \begin{array}{l}{\displaystyle y_i=b_0+b_{1}Location+b_{2}DayYear+b_{3}Age+b_{4}DaysSince+b_{5}CumCCI}\end{array}}$$

The model includes collection location (n = 2), collection day within year, age at collection, days since previous collection, and cumulative CCI. Collection location and collection day within year were fit as class variables. There were two different bull semen collection facilities that provided phenotypic records. There were 2,734 classes for collection day within year. The fixed effects fit as covariates were age at collection, days since previous collection, and cumulative CCI. Age at collection ranged from 296 to 4,884 d with an average of 1,134 d. The average days since previous collection was 8 with a range of 1 to 2,756 d. The cumulative CCI ranged from −207 to 2,390 with an average of 1,076. Table 2 outlines the levels of significance for all traits included in the model.

Variance components were estimated utilizing an average information restricted maximum likelihood (AIREML) algorithm (Misztal et al., 2014). Both pedigree and genomic data were included in the single-step genomic best linear unbiased prediction (ssGBLUP) analysis. The H⁻¹ matrix was utilized instead of the traditional numerator relationship matrix, A⁻¹ to incorporate genomic information from genotyped animals. The H⁻¹ matrix includes both pedigree and genomic relationship information and is constructed as follows:

$${\displaystyle \begin{array}{l}{\displaystyle {\mathbf{H}}^{-1}={\mathbf{A}}^{-1}+\left[\begin{array}{cc}0& {\displaystyle 0}\\ 0& {\displaystyle {\mathbf{G}}^{-1}-{\mathbf{A}}_{22}^{-1}}\end{array}\right]}\end{array}}$$

where A⁻¹ is the inverse of the pedigree-based numerator relationship matrix, A⁻¹₂₂ is a subset of the numerator relationship matrix for the genotyped individuals, and G is the genomic relationship matrix for the genotyped individuals. Following the univariate analyses, a linear multivariate animal model with repeated records was fitted utilizing the BLUPF90 family of programs (Misztal et al., 2014) to obtain both variance and covariance parameter estimates. The multivariate models for the various traits are outlined below.

Variance components, heritabilities, repeatabilities, and genetic correlations for VOL, CONC, and NSP were estimated using the following trivariate model:

$${\displaystyle \begin{array}{l}{\displaystyle \left[\begin{array}{l}{\mathit{y}}_1\\ {\mathit{y}}_2\\ {\mathit{y}}_3\end{array}\right]=\left[\begin{array}{l}{\mathit{X}}_1{\mathit{b}}_1\\ {\mathit{X}}_2{\mathit{b}}_2\\ {\mathit{X}}_3{\mathit{b}}_3\end{array}\right]+\left[\begin{array}{l}{\mathit{Z}}_{\mathit{a}1}{a}_1\\ {\mathit{Z}}_{\mathit{a}2}{\mathit{a}}_2\\ {\mathit{Z}}_{\mathit{a}3}{\mathit{a}}_3\end{array}\right]+\left[\begin{array}{l}{\mathit{W}}_{\mathit{p}\mathit{e}1}\mathit{p}{\mathit{e}}_1\\ {\mathit{W}}_{\mathit{p}e2}\mathit{p}{\mathit{e}}_2\\ {\mathit{W}}_{\mathit{p}\mathit{e}3}\mathit{p}{\mathit{e}}_3\end{array}\right]+\left[\begin{array}{l}{\mathit{e}}_1\\ {\mathit{e}}_2\\ {\mathit{e}}_3\end{array}\right]}\end{array}}$$

where y_i is a vector of phenotypic observations, X_i is an incidence matrix relating the phenotypic observations to the fixed effects in the model, b_i is a vector of the fixed effects, Z_ai is an incidence matrix relating the phenotypic observations to the additive direct genetic effects, a_i is a vector of additive direct genetic effects, W_pei is an incidence matrix relating the phenotypic observations to the additive permanent environment genetic effects, pe_i is a vector of additive permanent environment genetic effects, and e_i is a vector of random residuals. VOL, CONC, and NSP were analyzed in a trivariate model because these three traits are measured pre-freezing. In addition, NSP is a function of VOL and CONC.

Initial motility, PTMot, and 3HRPTMot would not converge in a trivariate model. Percentage of normal spermatozoa, PRIM, and SEC also would not converge using a trivariate model. Instead, pairwise combinations of traits in bivariate models were performed within the motility and normality groups of traits. In addition, VOL, CONC, and NSP were included in bivariate models with the remaining traits in order to estimate genetic correlations among all the traits. The following model was used to estimate variance components, heritabilities, repeatabilities, and genetic correlations for IMot, PTMot, 3HRPTMot, %NORM, PRIM, and SEC:

$${\displaystyle \begin{array}{l}{\displaystyle \left[\begin{array}{l}{\mathit{y}}_1\\ {\mathit{y}}_2\end{array}\right]=\left[\begin{array}{l}{\mathit{X}}_1{\mathit{b}}_1\\ {\mathit{X}}_2{\mathit{b}}_2\end{array}\right]+\left[\begin{array}{l}{\mathit{Z}}_{\mathit{a}1}{\mathit{a}}_1\\ {\mathit{Z}}_{\mathit{a}2}{\mathit{a}}_2\end{array}\right]+\left[\begin{array}{l}{\mathit{W}}_{\mathit{p}\mathit{e}1}\mathit{p}{\mathit{e}}_1\\ {\mathit{W}}_{\mathit{p}\mathit{e}2}\mathit{p}{\mathit{e}}_2\end{array}\right]+\left[\begin{array}{l}{\mathit{e}}_1\\ {\mathit{e}}_2\end{array}\right]}\end{array}}$$

where y_i is a vector of phenotypic observations, X_i is an incidence matrix relating the phenotypic observations to the fixed effects in the model, b_i is a vector of the fixed effects, Z_ai is an incidence matrix relating the phenotypic observations to the additive direct genetic effects, a_i is a vector of additive direct genetic effects, W_pei is an incidence matrix relating the phenotypic observations to the additive permanent environment genetic effects, pe_i is a vector of additive permanent environment genetic effects, and e_i is a vector of random residuals. The analyses were completed using the BLUPF90 suite of programs (Misztal et al., 2014).

Estimated breeding values (EBVs) for each individual were utilized for further inquiry. For the calculation of EBVs, the following univariate animal model was used:

$${\displaystyle \begin{array}{l}{\displaystyle \left[{\mathit{y}}_1\right]=\left[{\mathit{X}}_1{\mathit{b}}_1\right]+\left[{\mathit{Z}}_{\mathit{a}1}{\mathit{a}}_1\right]+\left[{\mathit{W}}_{\mathit{p}\mathit{e}1}\mathit{p}{\mathit{e}}_1\right]+\left[{\mathit{e}}_1\right]}\end{array}}$$

EBVs provide the opportunity for selection decisions to be made using all available information on an animal. A univariate animal model was used for the calculation of EBVs because the univariate results were very similar to the multivariate results. In addition, the correlation between the bivariate EBV results from each of the analyses for a trait was estimated, and only the IMot EBV (0.98) and SEC EBV (0.99) results were highly correlated, and thus the univariate model was used to calculate the EBV. The correlations between the bivariate EBV results for the remaining traits ranged from 0.65 to 0.86. The overall average correlation between the individual trait bivariate results was 0.85.

The univariate EBV results for the beef bull semen attributes were correlated with scrotal circumference (SC) EPD obtained from the American Angus Association. The correlations were obtained using the BLUPF90 family of programs (Misztal et al., 2014).

Results and Discussion

Summary statistics for the response variables are reported in Table 3. Interestingly, the mean IMot was 49.18% and the median was 50%, which is just below the 50% minimum, a collection must have to be considered for freezing. However, of the 44,431 collection records, 67% of the collections met the initial freezing requirements and had post-thaw quality control records. Comparing the mean to the median, it was determined that ejaculates with poor IMot had a very low percentage of motile spermatozoa, which skewed the average IMot down. The average PTMot was 43.54% with a median of 50%. Similar to IMot, some very poor PTMot measures were recorded causing the average PTMot to be lower than expected. This was verified by comparing the mean with the median. The average 3HRPTMot was substantially lower, as expected, at 15.73% with a lower median of 15%. It is difficult to speculate whether the 3HRPTMot measures are good or bad indicators of fertility because of the high standard errors for the genetic correlations with other beef bull semen attributes. The high standard errors are likely a result of the trait not being recorded for every collection. In addition, collections that did not meet freezing quality standards would not have any post-thaw measures recorded, resulting in biased data.

Table 3.

Number of records (N), mean, SD, minimum, and maximum for each Angus bull semen attribute

	N	Mean	Median	SD	Minimum	Maximum
VOL	44,431	7.96	7.10	4.22	0.10	74.00
CONC	44,038	1,021.16	972.00	495.02	10.00	3,906.00
IMot	44,418	49.18	50.00	16.16	0.00	100.00
NSP	44,038	8,004.18	6,877.50	5,519.61	64.50	69,795.00
PTMot	29,877	43.53	50.00	13.78	0.00	75.00
3HRPTMot	8,299	15.72	15.00	12.48	0.00	60.00
%NORM	19,455	75.18	76.00	8.37	2.00	100.00
PRIM	19,452	13.00	12.00	7.53	0.00	100.00
SEC	19,521	12.09	11.00	7.70	0.00	100.00

Model selection

Table 2 outlines the fixed effects that were significant for each beef bull fertility trait. For VOL, collection day within year, age, and days since previous collection were determined to be significant. Statistically significant effects for CONC were collection location and collection day within year. Initial motility and NSP both yielded significant effects of collection location, collection day within year, age, and days since previous collection. Collection location was not significant when modeling the %NORM or PRIM. Days since previous collection was not significant for any of the post-thaw beef bull semen attributes. In addition, cumulative CCI was not significant at a P-value < 0.05 for any of the traits, but when the P-values for CCI on collection day, spermatogenesis start day, and cumulative CCI were compared, cumulative CCI had the most significant P-value.

After examining the significant effects found in the current study and comparing them with previously published studies (Taylor et al., 1985; Berry et al., 2019), the following model was utilized for the genetic evaluation for all beef bull semen attributes:

$${\displaystyle \begin{array}{ll}{\displaystyle y_i=}& b_0+b_{1}Location+b_{2}DayYear+b_{3}Age\\ & +b_{4}DaysSince+b_{5}CumCCI\end{array}}$$

The model includes collection location (n = 2), collection day within year, age at collection, days since previous collection, and cumulative CCI. Collection location and collection day within year were fit as class variables. There were two different bull semen collection facilities that provided phenotypic records. There were 2,734 classes for collection day within year. The fixed effects fit as covariates were age at collection, days since previous collection, and cumulative CCI. Age at collection ranged from 296 to 4,884 d with an average of 1,134 d. The average days since previous collection was 8 with a range of 1 to 2,756 d. The cumulative CCI ranged from −207 to 2,390 with an average of 1,076. Table 2 outlines the levels of significance for all traits included in the model.

Variance components

The additive direct genetic, permanent environment, and residual variances for the ssGBLUP analyses are presented in Table 4. The reported variance components, heritabilities, and repeatabilities results for VOL, CONC, and NSP were obtained from the trivariate analysis. For the remainder of the beef bull semen attributes, estimates from all bivariate analyses were averaged and reported. Limited published research makes the comparison to previous studies difficult. Taylor et al. (1985) published a direct additive genetic variance for VOL that was slightly lower than the estimate in the current study. However, Taylor et al. (1985) did not account for a permanent environment effect, and the environmental variance was much lower than the estimate in the current study. Additionally, Berry et al. (2019) published a similar genetic variance estimate for VOL of 0.72 mL².

Table 4.

Variance component estimates for direct additive genetic $({\mathrm{\sigma }}_{\mathrm{a}}^2$), permanent environment$({\mathrm{\sigma }}_{\mathrm{p}\mathrm{e}}^2)$, and environmental $({\mathrm{\sigma }}_{\mathrm{e}}^2)$variance using a multivariate ssGBLUP for semen traits¹

	${\mathrm{\sigma }}_{\mathrm{a}}^2$	${\mathrm{\sigma }}_{\mathrm{p}\mathrm{e}}^2$	${\mathrm{\sigma }}_{\mathrm{e}}^2$	$h^2$	Minimum $h^2$	Maximum $h^2$	$r^2$	Minimum $r^2$	Maximum $r^2$
VOL, mL2	1.60	2.76	10.22	0.11 ± 0.02	NA	NA	0.30 ± 0.01	NA	NA
CONC, m/mL2	18,560.20	52,040.82	142,806.10	0.09 ± 0.02	NA	NA	0.33 ± 0.01	NA	NA
NSP	2.05 × 106	5.68 ×106	1.82 ×107	0.08 ± 0.02	NA	NA	0.30 ± 0.01	NA	NA
IMot, %2	29.90	74.68	155.55	0.12 ± 0.03a	0.11 ± 0.03	0.12 ± 0.02	0.40 ± 0.01g	0.40 ± 0.01	0.40 ± 0.01
PTMot, %2	28.81	56.44	106.75	0.15 ± 0.03b	—	—	0.44 ± 0.01h	—	—
3HRPTMot, %2	7.38	19.16	112.56	0.09 ± 0.04c	—	—	0.34 ± 0.02i	—	—
%NORM, %2	8.61	40.98	44.03	0.09 ± 0.04d	0.05 ± 0.03	0.14 ± 0.04	0.53 ± 0.01j	0.52 ± 0.01	0.54 ± 0.01
PRIM, %2	1.57	34.59	24.22	0.03 ± 0.03e	0.03 ± 0.03	0.02 ± 0.03	0.60 ± 0.01k	0.60 ± 0.01	0.60 ± 0.01
SEC, %2	10.29	17.70	29.75	0.18 ± 0.04f	0.18 ± 0.04	0.18 ± 0.04	0.48 ± 0.01l	0.48 ± 0.01	0.48 ± 0.01

¹VOL, CONC, and NSP were analyzed with a trivariate approach, and the remainder of the traits were modeled with a bivariate model. If a bivariate model was used, the range of the estimates is reported. Heritability $(h^2)$and repeatability $(r^2)$estimates and standard errors are also reported.

²“—" only one estimate reported due to convergence issues.

^aReported results are the average heritabilities from the bivariate analyses with PTMot and 3HRPTMot.

^bReported results are the heritabilities from the bivariate analyses with IMot. The bivariate model with 3HRPTMot did not converge.

^cReported results are the heritabilities from the bivariate analyses with IMot. The bivariate model with PTMot did not converge.

^dReported results are the average heritabilities from the bivariate analyses with PRIM and SEC.

^eReported results are the average heritabilities from the bivariate analyses with %NORM and SEC.

^fReported results are the average heritabilities from the bivariate analyses with %NORM and PRIM.

^gReported results are the average repeatabilities from the bivariate analyses with PTMot and 3HRPTMot.

^hReported result is the repeatabilities from the bivariate analyses with IMot. The bivariate model with 3HRPTMot did not converge.

ⁱReported results are the average repeatabilities from the bivariate analyses with IMot. The bivariate model with PTMot did not converge.

^jReported results are the average repeatabilities from the bivariate analyses with PRIM and SEC.

^kReported results are the average repeatabilities from the bivariate analyses with %NORM and SEC.

^lReported results are the average repeatabilities from the bivariate analyses with %NORM and PRIM.

All variance estimates in the literature for CONC are much higher than the estimates for this study, which are presented in Table 4. Taylor et al. (1985) estimated the genetic and environmental variances for CONC to be 7.45 × 10⁷ and 1.24 × 10⁸ million/mL², respectively. Kaps et al. (2000) used a population of Simmental bulls with a single breeding soundness examination record to estimate variance components for CONC. The direct additive genetic effect was 4.36 × 10⁶ million/mL², and environmental variance was 1.22 × 10⁷ million/mL². Mathevon et al. (1998) estimated variances for the genetic effect, permanent environment effect, and environment effect to be 1.86 × 10⁸, 2.6 × 10⁴, and 3.8 × 10⁴ million/mL², respectively.

Varying motility measures with variance component estimates are presented in the literature (Mathevon et al., 1998; Kealey et al., 2006; Corbet et al., 2013), and most are much higher than the variance components displayed in Table 4. The exception is the variances published by Berry et al. (2019), which reported estimates of 30.25%² and 21.72%² for the genetic variance of the percentage of living spermatozoa pre-freezing and the genetic variance post-freezing, respectively. Berry et al. (2019) did not publish variance component estimates for permanent environment or environmental variance.

Variance component estimates in the literature (Kealey et al., 2006; Corbet et al., 2013) for the %NORM are much higher than the estimates displayed in Table 4. Kealey et al. (2006) published genetic and environmental variances for PRIM and SEC. While the PRIM variance components are different than those displayed in Table 4, the SEC variance components were comparable. Kealey et al. (2006) published a genetic variance estimate of 17.50%² and environmental variance of 36.04%² for SEC; however, there were not repeated records in the study.

Heritability estimates

Heritability estimates from multivariate ssGBLUP analysis are presented in Table 4. The heritability estimate for VOL was low (0.11 ± 0.02) but within the range of published literature estimates. Kaps et al. (2000) reported a heritability for VOL of 0.04 from a study of young Simmental bulls; however, the standard error for the estimate was extremely large (0.54), so the estimate was not different from zero. Kealey et al. (2006) published an estimate of 0.09 ± 0.08 with a population of young Line 1 Hereford bulls. Kaps et al. (2000) obtained VOL measurements from breeding soundness examinations, whereas Kealey et al. (2006) obtained measurements from semen collected from Line 1 Hereford bulls specifically for the study. VOL estimates in the literature that were obtained from dairy bull semen collection facilities have yielded more moderate heritability estimates of 0.20 (Berry et al., 2019), 0.22 (Druet et al., 2009), and 0.26 (Suchocki and Szyda, 2015). It is difficult to speculate the justification for the differences due to the lack of information available in the literature. While the sample size and method of phenotype collection in the current study are most similar to the dairy bull studies, the heritabilities are more similar the data obtained from beef bull studies.

The heritability estimate for CONC in this study was lower than published heritability estimates ranging from 0.13 to 0.52. Interestingly, beef bull heritability estimates in the literature for CONC were lower (0.13, Knights et al., 1984; 0.26, Kaps et al., 2000; and 0.16, Kealey et al., 2006) than those estimated using data from dairy bull semen collection facilities (0.39, Ducrocq and Humblot, 1995; 0.52, Mathevon et al., 1998; and 0.34, Suchocki and Szyda, 2015). Thus, the comparably lower heritability estimates from the current study could be attributed to the fact that the data were collected from beef bulls. All studies referenced here had a smaller population size than the current study, with the exception of Ducrocq and Humblot (1995), which used a population of 2,387 Normande dairy bulls.

The number of spermatozoa (NSP) had a low heritability estimate of 0.08 ± 0.02. Taylor et al. (1985) noted that estimating the heritability of a trait that is a function of two other traits, such as NSP, could impact that ability to accurately provide genetic estimates because the trait that is a function of others is influenced by the phenotypes collected on the original two traits. The heritability estimate for NSP in the current study is slightly higher than the estimate of 0.03 reported by Taylor et al. (1985), which utilized dairy bull phenotypes from a bull semen collection facility. More recently, Suchocki and Szyda (2015) and Berry et al. (2019) published heritability estimates of 0.27 and 0.38 for NSP with data obtained from dairy bull semen collection facilities. Differing estimates could be attributed to using dairy bulls, smaller population size in the literature, or the model used for genetic evaluation. While Suchocki and Szyda (2015) and Berry et al. (2019) both utilized animal models, the fixed effects included in their models were different than effects included in the current study. Suchocki and Szyda (2015) only included fixed effects of bull age and season, whereas Berry et al. (2019) additionally included fixed effects of inbreeding, days since last ejaculation, and breed.

Initial motility had a heritability estimate of 0.12 ± 0.03. Initial motility heritabilities published in the literature range from 0.05 to 0.43. The studies that published low heritability estimates (0.13, Knights et al., 1984; 0.08, Smith et al., 1989; 0.07, Christmas et al., 2001; 0.05, Garmyn et al., 2011) come from beef bull motility phenotypes obtained from yearling breeding soundness examinations. Motility phenotypes obtained from beef and dairy bull semen collection facilities had higher heritabilities (0.22 to 0.43; Ducrocq and Humblot, 1995; Mathevon et al., 1998; Kealey et al., 2006; Suchocki and Szyda, 2015; Berry et al., 2019).

Post-thaw motility had a heritability estimate of 0.15 ± 0.03, which was slightly higher than the IMot heritability estimate. The PTMot estimate is lower than those found in the literature, which varies from 0.21 to 0.25 (Kealey et al., 2006; Druet et al., 2009; Berry et al., 2019). While there are only three post-thaw estimates found in the literature, the estimates come from both beef and dairy bulls from semen collection facilities. The population size for the current study was larger than the populations used in these studies; therefore, the standard error of the estimate in the current study was smaller than the other studies.

Three-hour post-thaw motility was lowly heritable (0.09 ± 0.04). Three-hour post-thaw motility had the smallest number of phenotypes (n = 8,299) with only 561 bulls contributing phenotypes to the trait. There are no comparable estimates published in the literature; however, if the estimate is compared to the previously discussed published motility estimates, the estimate is much lower. This may be a function of the trait if the variance in the population collapses quickly after thawing or an artifact of the number of phenotypes available for analysis.

The %NORM yielded one of the more variable heritability estimates from the bivariate analyses (0.09 ± 0.04). The genetic variance in the bivariate model with PRIM was much lower than the bivariate model with SEC. In studies using beef bulls, Smith et al. (1989) and Gredler et al. (2007) published similar heritability estimates of 0.07 and 0.10, respectively. Gredler et al. (2007) utilized data obtained from Austrian Simmental bulls collected for artificial insemination (AI). In the Kealey et al. (2006) study using semen collected from Line 1 Hereford bulls, an estimate of 0.35 was published. Differences in literature estimates could be attributed to differing population sizes (n = 841, Kealey et al., 2006; n = 301, Gredler et al., 2007), as both reports used a smaller population size than the current study.

The abnormality traits had varying heritability estimates of 0.03 ± 0.03 for PRIM and 0.18 ± 0.04 SEC in both bivariate models. The PRIM heritability estimates in the literature that were obtained from beef bulls range from 0.27 to 0.35 (Smith et al., 1989; Christmas et al., 2001; Kealey et al., 2006; Garmyn et al., 2011), which are higher than in this study. Comparable studies reporting post-thaw semen attributes from phenotypes obtained from bull semen collection facilities are difficult to find; however, Druet et al. (2009) published moderate heritability estimates of 0.35 for percentage of spermatozoa with an abnormal head, a trait comparable to PRIM. Similar to SEC, Druet et al. (2009) published an estimate for percentage of spermatozoa with an abnormal tail and percentage of spermatozoa with an abnormal cytoplasmic droplet of 0.19 ± 0.12 and 0.19 ± 0.08, respectively. Smith et al. (1989) published an SEC estimate not different from zero (0.02 ± 0.05); however, other beef bull studies reported moderate heritability estimates of 0.26 (Christmas et al., 2001), 0.33 (Kealey et al., 2006), and 0.23 (Garmyn et al., 2011).

While the reported heritability estimates are low to moderate, the majority of the standard errors are small and thus estimates are different from zero. While fertility traits are generally lowly heritable, it does not mean that these traits should be ignored and genetic improvement cannot be made. An example of another lowly heritable trait that has been well studied and improved is productive life in Holstein dairy cattle. Productive life is a trait that is indicative of how long a cow stays in the herd after her first calving (Strandberg, 1992) and has an average heritability estimate of 0.085 (Bennet, 2009). The Council on Dairy Cattle Breeding (CDCB) estimated that the average breeding value of productive life was −12.73 in 1960. In 2016, the average EBV for productive life was 4.53. In 56 yr, the dairy cattle industry was able to improve productive life by 17.26 d through recording phenotypes and utilizing selection tools. Similar improvements in male fertility can be made if the necessary genetic selection tools are developed and deployed within the industry.

Repeatability estimates

The average number of collections per bull was 28 with a range of 1 to 1,181. Repeatability estimates are provided in Table 4. All estimates were moderately to highly repeatable with small standard errors, meaning individual AI beef bull semen attributes are consistent within bull.

Published repeatability estimates for bull semen attributes are similar to the estimates found in the current study. The estimated repeatability for VOL was similar to that published in other studies. Using dairy bull populations, Haque et al. (2001), Gredler et al. (2007), and Druet et al. (2009) published similar repeatability estimates of 0.34, 0.29, and 0.33, respectively. Taylor et al. (1985) published a slightly lower estimate of 0.23 using dairy bulls. Mathevon et al. (1998) estimated the repeatability of VOL in young and mature dairy bulls and published estimates were 0.45 and 0.51, respectively. Additionally, Stälhammar et al. (1989) published a repeatability of 0.57 in a population of dairy bulls.

The repeatability of CONC was estimated to be 0.33 with a small standard error in the current study. Taylor et al. (1985), Druet et al. (2009), and Gredler et al. (2007) estimated similar values ranging from 0.32 to 0.37. Haque et al. (2001) found a lower repeatability of 0.20 in a population of 30 dairy bulls. On the other hand, Mathevon et al. (1998) found a higher estimate of 0.61 in a population of mature dairy bulls.

NSP had a similar repeatability to VOL and CONC, unsurprisingly, as NSP is a function of the two. The repeatability estimate for NSP is provided in Table 4. Gredler et al. (2007) and Druet et al. (2009) found a slightly lower estimate of 0.24. Both studies previously mentioned used dairy bull phenotypes collected from a bull semen collection facility. Mathevon et al. (1998) published a repeatability estimate for total spermatozoa, a phenotype similar to NSP. For mature dairy bulls, Mathevon et al. (1998) reported an estimate of 0.54.

Initial motility, PTMot, and 3HRPTMot had repeatability estimates of 0.40, 0.33, and 0.29, respectively. Literature estimates for motility repeatability range from 0.08 to 0.64. The lowest repeatability was estimated using motility score phenotypes (Gredler et al., 2007) from dairy bulls. The highest estimate was published by Mathevon et al. (1998) in a population of mature dairy bulls. Haque et al. (2001) used motility phenotypes for mass activity, which is a scoring system evaluating the overall motility of the sample. Stälhammar et al. (1989) published repeatability estimates for motility pre-freezing and post-freezing with estimates of 0.54 and 0.53, respectively.

Repeatability estimates for %NORM, PRIM, and SEC are difficult to find in the literature. The percentage of normal spermatozoa had a repeatability of 0.53 with a standard error of 0.01. The moderate repeatability indicates phenotypes are consistent across collections. PRIM had the highest repeatability estimate of the study (Table 4). SEC had a repeatability estimate of 0.48, suggesting a bull will routinely have a similar PRIM and SEC measures from collection to collection.

Estimated breeding values

EBVs were obtained for each beef bull semen attribute from a univariate analysis. Summary statistics for EBV for the bulls with phenotypes are provided in Table 5. The averages for each beef bull fertility EBV are near zero for all traits, but the minimum and maximum values for the EBV indicate the values have a wide range. Notably, CONC, NSP, and IMot have negative average EBV, which would indicate that bulls with phenotypes are slightly lower than the population as a whole (including animals in the pedigree without phenotypes).

Table 5.

Number of records (N), mean, SD, minimum and maximum for EBVs, and average accuracies for beef bull semen attributes for bulls with phenotypes in the genetic analysis

	N	Mean	SD	Minimum	Maximum	Beef improvement federation accuracy	Classical animal breeding accuracy
VOL	1,819	0.04	2.45	−7.10	11.00	0.21	0.61
CONC	1,819	−9.97	104.55	−335.46	428.38	0.16	0.54
IMot	1,819	−0.03	4.38	−16.97	11.62	0.18	0.57
NSP	1,819	−1.15	67.47	−208.11	327.73	0.16	0.54
PTMot	1,819	0.23	3.93	−18.30	11.57	0.18	0.57
3HRPTMot	1,819	0.07	1.00	−4.44	4.54	0.11	0.45
%NORM	1,819	0.13	2.04	−16.25	7.33	0.16	0.53
PRIM	1,819	0.19	2.99	−10.39	30.01	0.05	0.30
SEC	1,819	0.38	9.45	−29.20	114.16	0.24	0.64

Genetic trends

Genetic trends are shown in Supplementary Figures S1–S9. As expected, the genetic trends for all traits are variable. There are currently no selection tools for producers to utilize for the improvement of beef bull semen attributes. VOL appears to increase from 1997 to 2004 and then shows a slight decrease. The trend for VOL could be a result of indirect selection, but it is more probable that the trend is due to sample size. Until 2005, there were less than 50 bulls between each collection facility born each year contributing to the average; however, in 2005, the number of bulls doubled. In subsequent years after 2005, the population remained greater than 100 bulls in each birth year. For IMot, the smooth line indicates a decrease in the trait from 1997 to 2004; however, the standard error bars shown in the figure indicate that the estimates are very variable, and this could be attributed to the extremely small sample size in 1997 and 1998 (n = 3 and 1 bulls, respectively). Similar to PTMot, 3HRPTMot shows an initial decrease but then increases until 2017. PRIM and SEC show inverse genetic trends. When the average EBV for PRIM was low, the average EBV for SEC was high and vice versa. Providing producers with genetic selection tools for beef bull semen attributes could make the genetic trends more favorable in the future.

While some trends may be speculated, there is not a common trend among all fertility traits. It does not appear that the traits evaluated have improved or declined as a whole over the past 20 yr. The trends shown in Supplementary Figures S1–S9 generally indicate that fertility has remained the same.

Correlations

Genetic and phenotypic correlations between Angus bull semen attributes are presented in Table 6. The majority of the genetic and phenotypic correlations were low, with several notable exceptions. NSP was strongly and positively phenotypically correlated with VOL and CONC (0.66 ± 0.01; 0.61 ± 0.01). This would be expected as NSP is a function of VOL and CONC. The genetic correlation between VOL and NSP mirrored the phenotypic correlation, which was strong and favorable (0.75 ± 0.08). Literature estimates for phenotypic and genetic correlations between VOL and NSP are varied. Gredler et al. (2007) estimated the genetic correlation to be 0.83 ± 0.13 in a population of dual-purpose Simmental bulls. Similarly, Berry et al. (2019) estimated phenotypic and genetic correlations to be 0.63 (no standard error reported) and 0.66 ± 0.16, respectively. Taylor et al. (1985) and Druet et al. (2009) published lower genetic correlation estimates (0.45; no standard error and 0.47 ± 0.18, respectively); however, the phenotypic correlations were more similar to the phenotypic correlation estimates presented in the current study. Genetic correlations between CONC and NSP in the literature are higher than the genetic correlation reported in this study 0.55 ± 0.13. Gredler et al. (2007) estimated the genetic correlation to be 0.60 ± 0.17, and Taylor et al. (1985) published a slightly higher correlation of 0.72; no standard error. However, estimates from the literature are similar to our estimated phenotypic correlation. All phenotypic correlations in the literature range from 0.51 to 0.71 (Taylor et al., 1985; David et al., 2006; Gredler et al., 2007; Druet et al., 2009; Berry et al., 2019), and our estimate was 0.61 ± 0.01.

Table 6.

Genetic correlations (above the diagonal) and phenotypic correlation (below the diagonal) between beef bull semen attribute traits using a multivariate ssGBLUP model¹

	VOL	CONC	NSP	IMot	PTMot	3HRPTMot	%NORM	PRIM	SEC
VOL		−0.10   0.17	0.75   0.08	0.23   0.16	0.17   0.16	0.54   0.69	0.18   0.22	0.52   0.61	−0.13   0.17
CONC	−0.06   0.01		0.55   0.13	0.09   0.19	−0.15   0.19	0.04   0.59	−0.11   0.24	dnc	0.04   0.19
NSP	0.66   0.01	0.61   0.01		0.23   0.18	0.02   0.19	0.31   0.61	0.13   0.24	0.08   0.73	−0.10   0.20
IMot	0.13   0.01	0.12   0.01	0.16   0.01		0.92   0.04	dnc 2	0.77   0.09	0.33   0.20	0.63   0.82
PTMot	0.16   0.01	−0.03   0.01	0.12   0.01	0.32   0.01		dnc	−0.74   0.30	−0.02   0.61	0.11   1.97
3HRPTMot	0.22   0.01	−0.24   0.01	−0.01   0.01	0.04   0.01	0.74   0.01		−0.44   0.15	−0.25   0.17	dnc
%NORM	0.09   0.01	0.12   0.01	0.16   0.01	0.20   0.01	0.03   0.01	0.15   0.02		0.03   2.07	−0.81   0.10
PRIM	−0.09   0.01	−0.03   0.01	−0.10   0.01	−0.21   0.01	−0.33   0.01	−0.55   0.01	−0.60   0.01		−0.60   1.13
SEC	−0.01   0.01	−0.13   0.01	−0.09   0.01	−0.15   0.01	0.30   0.01	0.54   0.01	−0.58   0.01	−0.31   0.01

¹Standard error estimates are below correlation estimates. Bold numbers are the actual correlations and the unbolded numbers are the standard errors for the correlations.

²dnc, model did not converge.

The phenotypic and genetic correlations between pre-freeze and post-thaw semen characteristics were generally low. Interestingly, CONC was negatively phenotypically and genetically correlated with PTMot, indicating as CONC increases, PTMot decreases; however, the large standard errors make the estimate essentially not different from zero. The phenotypic correlation between CONC and 3HRPTMot mirrored the correlation between CONC and PTMot. Several of the genetic correlation estimates among pre-freeze and post-thaw semen traits had large standard errors, and the estimates were not different from zero. The large standard errors could be attributed to the small sample size of the post-thaw semen traits.

Table 6 presents that IMot was strongly genetically correlated with all post-thaw semen characteristics. An increase in IMot would be associated with an increase in PTMot, 3HRPTMot, and %NORM. One would expect that a high IMot would lead to more motile spermatozoa after thawing with fewer abnormalities because abnormalities should affect the ability of the spermatozoa to move progressively. Previous literature has reported strong, positive phenotypic and genetic correlations between IMot and PTMot (Druet et al., 2009; Berry et al., 2019). In addition, previous studies have reported strong, positive phenotypic and genetic correlations between motility and %NORM (Druet et al., 2009; Berry et al., 2019). Varying and limited correlation estimates between IMot and spermatozoa abnormalities have been reported in the literature. While Smith et al. (1989) estimated the phenotypic and genetic correlations between motility and PRIM as −0.31 (no standard error) and −0.36 ± 0.55, respectively, Kealey et al. (2006) reported a conflicting genetic correlation of 0.57 (no standard error). Literature estimates for correlations between motility and SEC are also contradictory. Smith et al. (1989) estimated a phenotypic correlation of −0.22; no standard error and a genetic correlation of 0.71 ± 0.89. Conversely, Kealey et al. (2006) estimated a genetic correlation between motility and SEC of −0.54; no standard error. A negative correlation between motility and PRIM and SEC traits would be expected as the movement of the sperm cell is dependent upon a normally functioning head and tail, which is consistent with the genetic correlation results presented in Table 6.

Three-hour post-thaw motility was strongly and positively phenotypically correlated with PTMot (Table 6), implying that as PTMot increased, 3HRPTMot increased. The positive correlation indicated that bulls with high PTMot continued to have high motility for an extended period of time after thawing. However, only 28% of the collections had a 3HRPTMot phenotype. There are currently no literature estimates for an extended PTMot measure making it impossible to compare current estimates with previous findings.

PRIM had a strong, negative phenotypic and a slightly less negative genetic correlation with 3HRPTMot indicating that as 3HRPTMot increased, PRIM decreased. A negative correlation would indicate that as motility increased, fewer abnormalities of the head would be observed. While there are no estimates in the literature pertaining to an extended PTMot measure, estimates from the current study can be compared with other motility estimates in the literature. Druet et al. (2009) reported a strong, negative genetic correlation between motility and percentage of spermatozoa with an abnormal head, but a positive phenotypic correlation, which contradicts the estimates from the current study. As discussed previously, Smith et al. (1989) reported a negative genetic correlation between motility and PRIM, but Kealey et al. (2006) reported a strong, positive genetic correlation. Again, the current study is most comparable to Kealey et al. (2006) because the phenotypes were obtained from frozen semen samples from Hereford bulls.

SEC had a moderate, positive phenotypic correlation with 3HRPTMot, which suggested that as 3HRPTMot increased, SEC increased. However, the genetic correlation was strong and negative. Similarly, Kealey et al. (2006) presented a strong, negative genetic correlation between motility and SEC. A negative correlation would be expected, as 3HRPTMot would be expected to increase with fewer abnormalities of the tail present in the sample.

As expected, Table 6 presents that the %NORM was strongly, negatively phenotypically correlated with both PRIM and SEC. While the genetic correlation between %NORM and SEC was negative, there was a slight, positive genetic correlation between %NORM and PRIM. However, the standard error is extremely high and could be attributed to sample size. Smith et al. (1989) estimated that %NORM was strongly, negatively genetically correlated with PRIM. However, Smith et al. (1989) reported that %NORM was positively genetically correlated with SEC (0.16 ± 1.54).

Primary and SEC exhibited a negative phenotypic correlation with each other, which indicates that a sample with more PRIM tends to have fewer SEC. The genetic correlation was strong and negative; however, the standard error was greater than one. Kealey et al. (2006) reported a similar genetic correlation of −0.87; no standard error from a population of Line 1 Hereford bulls; however, Smith et al. (1989) reported a genetic correlation between PRIM and SEC of 0.14 ± 0.64. The corresponding phenotypic correlation estimated by Smith et al. (1989) was 0.17; no standard error. Smith et al. (1989) utilized breeding soundness examination data from yearling beef bulls for the genetic evaluation, whereas Kealey et al. (2006) used phenotypes from frozen semen samples collected from Line 1 Hereford bulls. Thus, it would be expected that the estimate in the current study would mirror that of Kealey et al. (2006).

Notably, some genetic correlations had exceedingly large standard error estimates. The large standard error estimates indicate the correlation estimate is not reliable. High standard errors were obtained for some of the correlations between post-thaw traits. The post-thaw traits had fewer observations and were biased because the sample had to meet certain quality standards for post-thaw phenotypes to be recorded. These two reasons could account for the large standard errors.

Genetic correlation among scrotal circumference EPDs and beef bull fertility EBV

Correlations among scrotal circumference EPDs and beef bull semen attribute EBV were low (Table 7). The average Beef Improvement Federation accuracy for the SC EPD was 0.52 indicating that the SC EPDs are fairly reliable estimates. Thus, the low correlations are not necessarily the fault of unreliable estimates. Literature estimates for the correlations among SC EPD and beef bull fertility EBV are limited, thus genetic correlations among fertility phenotypes are used for comparison. Generally, literature estimates indicate the correlations among scrotal circumference and fertility phenotypes are stronger than what was determined in the current study. The weakest correlation in the current study was between VOL EBV and scrotal circumference EPD (−0.01 ± 0.02), and the estimate was not different from zero. The strongest correlation found was between scrotal circumference EPD and the EBV for CONC (0.15 ± 0.02). According to the correlation, as scrotal circumference EPD increases, so does CONC EBV. Similarly, the correlation between scrotal circumference and NSP was favorable (0.12 ± 0.02) indicating as scrotal circumference EPD increases, so does NSP EBV, but the relationship was weak. Literature estimates for the correlation of VOL, CONC, and NSP to SC were not found. All correlations between scrotal circumference and motility measures were close to zero but negative. Smith et al. (1989) estimated the genetic correlation between motility and scrotal circumference to be 0.04, which is substantially lower than other literature estimates of 0.56 (Christmas et al., 2001), 0.36 ± 0.29 (Garmyn et al., 2011), and 0.56 ± 0.10 (Corbet et al., 2013). The correlation between scrotal circumference EPD and %NORM EBV was −0.06 ± 0.02. This estimate differs from genetic correlations in the literature, which are positive, moderate, and favorable (Smith et al., 1989; Corbet et al., 2013). A majority of the literature estimates for the genetic correlation between scrotal circumference and PRIM are strong and negative (Christmas et al., 2001; Garmyn et al., 2011). The estimate in the current study is positive, albeit very low. In contrast, the correlation between scrotal circumference EPD and SEC EBV is not only negative but also small. While our estimates are much lower than those in the literature, the direction of the estimates concurs with Christmas et al. (2001) and Garmyn et al. (2011). Though Christmas et al. (2001), Garmyn et al. (2011), and Corbet et al. (2013) studies have shown that SC is associated with motility, %NORM, and abnormalities, the results from the current study indicate that SC is not a good predictor of fertility, which necessitates the need for genetic selection tools for beef bull fertility.

Table 7.

Correlations between scrotal circumference EPDs obtained from the American Angus Association and beef bull semen attribute EBVs

Trait	Correlation +/− SE
VOL	−0.01 ± 0.02
CONC	0.15 ± 0.02
IMot	0.12 ± 0.02
NSP	−0.06 ± 0.02
PTMot	−0.07 ± 0.02
3HRPTMot	−0.07 ± 0.02
%NORM	−0.06 ± 0.02
PRIM	0.07 ± 0.02
SEC	−0.08 ± 0.02

Conclusions

Model selection revealed that collection location, collection day within year, age at collection, days since previous collection, and cumulative CCI all impact beef bull fertility. Utilizing the determined fixed effects in a genetic model, genetic parameters were determined. Although the reported heritabilities in the current study were low, it should not diminish the importance of studying the genetics of beef bull semen attributes. Similarly to other lowly heritable traits, recording phenotypes and providing producers with genetic selection tools could allow genetic improvement. The nine different fertility phenotypes in the current study had varying correlations. While the majority of traits were highly genetically correlated, traits such as CONC and motility or PTMot and PRIM were very lowly genetically correlated. Beef bull fertility EBVs were correlated with SC EPD to determine if SC could potentially be an indicator of beef bull fertility utilizing the BLUPF90 family of programs (Misztal et al., 2014). However, the low correlations indicate that SC EPD did not affect beef bull fertility EBV. Findings in the current research provide evidence that beef bull fertility could be impacted by genetic selection tools and current predictor traits available are not enough. The genetic potential of a bull’s fertility is dependent upon fertility phenotypes being collected and evaluated rather than utilizing indicator traits to select for beef bull fertility. Developing tools to improve bull fertility in the beef industry would increase beef production, improve producer profitability, improve the efficiency and sustainability of the livestock industry, and could provide benefits to male fertility research in other species.

Glossary

Abbreviations

%NORM

percentage of normal spermatozoa

3HRPTMot

3-h post-thaw motility

CCI

comprehensive climate index

CONC

concentration

EBV

estimated breeding value

EPD

expected progeny difference

GBLUP

genomic best linear unbiased prediction

IMot

initial motility

NOAA

National Oceanic and Atmospheric Administration

NSP

number of spermatozoa

PRIM

primary abnormalities

PTMot

post-thaw motility

SEC

secondary abnormalities

SNP

single-nucleotide polymorphism

ssGBLUP

single-step genomic best linear unbiased prediction

THI

temperature humidity index

VOL

volume

Conflict of interest statement

None of the authors have any potential conflicts of interest.