Evidence for a genetic role in bovine respiratory disease

Not all cattle that are exposed to BRD pathogens respond the same way (Gershwin et al., Reference Gershwin, Van Eenennaam, Anderson, McEligot, Shao, Yarnes, Toeff-Rosenstein, Taylor and Neibergs2015). Some cattle are more likely to die whereas others remain healthy when exposed to the same level and type of pathogen. When differences in susceptibility to BRD are identified in animals that are managed the same, it indicates that some of the susceptibility to disease is due to differences in cattle's ability to resist the disease or their genetic predisposition to disease (Neibergs et al., Reference Neibergs, Seabury, Wojtowicz, Wang, Scraggs, Kiser, Neupane, Womack, Van Eenennaam, Hagevoort, Lehenbauer, Aly, Davis and Taylor2014). Differences in morbidity and mortality found between cattle breeds and between sire or family lines support that there is a genetic component to BRD (Muggli-Cockett et al., Reference Muggli-Cockett, Cundiff and Gregory1992; Snowder et al., Reference Snowder, Van Vleck, Cundiff and Bennett2005; Cusack et al., Reference Cusack, McMeniman and Lean2007; Heringstad et al., Reference Heringstad, Chang, Gianola and Osteras2008; Neibergs et al., Reference Neibergs, Seabury, Wojtowicz, Wang, Scraggs, Kiser, Neupane, Womack, Van Eenennaam, Hagevoort, Lehenbauer, Aly, Davis and Taylor2014, Reference Neibergs, Zanella, Casas, Snowder, Wenz, Neibergs and Moore2011). By identifying the DNA regions that are associated with an enhanced ability to resist BRD, genomic selection may be used to select cattle that are more likely to stay healthy when faced with a pathogen challenge.

Heritability estimates for BRD susceptibility vary based on how the trait was measured, if the trait was binary or ordinal and if the estimates were based on pedigree information or genomic (single nucleotide polymorphism or SNP) data. Heritability estimates range from 0.02 to 0.29 for BRD susceptibility in beef and dairy cattle (Lyons et al., Reference Lyons, Freeman and Kuck1991; Snowder et al., Reference Snowder, Van Vleck, Cundiff and Bennett2005; Heringstad et al., Reference Heringstad, Chang, Gianola and Osteras2008; Taylor et al., Reference Taylor, Fulton, Lehenbauer, Step and Confer2010; Neibergs et al., Reference Neibergs, Seabury, Wojtowicz, Wang, Scraggs, Kiser, Neupane, Womack, Van Eenennaam, Hagevoort, Lehenbauer, Aly, Davis and Taylor2014; Buchanan et al., Reference Buchanan, MacNeil, Raymone, McClain and Van Eenennaam2016; Gonzalez-Pena et al., Reference Gonzalez-Pena, Vukasinovic, Brooker, Przybyla and SDeNise2019). Higher heritability estimates tend to be identified in data sets with precise phenotypes and phenotypes that are ordinal rather than binary (Buchanan et al., Reference Buchanan, MacNeil, Raymone, McClain and Van Eenennaam2016).

Genomic selection uses genotypes as a tool to predict the future performance of offspring to select animals that will be part of the breeding herd. Genetic selection uses performance records on an individual or its ancestors to predict future performance and select breeding animals. Genomic-enhanced selection is the combination of genomic and genetic selection and is particularly powerful in traits that occur late in life, have low heritability estimates, or for traits that are expensive to collect (Garcia-Ruiz et al., Reference Garcia-Ruiz, Cole, Van Raden, Wiggans, Ruiz-Lopez and Van Tassell2016). BRD occurs throughout the life of cattle, has low to moderate heritability and recording of sick cattle, and determination of the BRD pathogens present is expensive, suggesting that genomic-enhanced selection is well-suited for BRD. In addition, genomic-enhanced selection increases the accuracy of prediction, facilitating more rapid improvement of breeding objectives (Hayes et al., Reference Hayes, Bowman, Chamberlain and Goddard2009; Meuwissen et al., Reference Meuwissen, Hayes and Goddard2016).

Genomic evaluation in dairy cattle was first released in 2008 and has grown swiftly and has resulted in the rapid annual genetic improvement (Garcia-Ruiz et al., Reference Garcia-Ruiz, Cole, Van Raden, Wiggans, Ruiz-Lopez and Van Tassell2016; Council on Dairy Cattle Breeding, 2019). Most of this increase in genetic improvement is through genomic-enhanced selection coupled with the use of artificial insemination (Council on Dairy Cattle Breeding, Reference Buczinski, Ollivett and Dendukuri2019). In contrast, <10% of beef cattle are bred by artificial insemination so the use of genomic-enhanced selection has not resulted in a doubling of genetic progress in the beef industry as has been experienced by the dairy industry. Health traits such as BRD are not yet included in selection indexes in beef cattle in the United States. However, the use of health traits for genomic selection has been incorporated into beef operations that are vertically integrated so that they may realize the financial and animal welfare benefit of producing cattle less susceptible to BRD (Buchanan et al., Reference Buchanan, MacNeil, Raymone, McClain and Van Eenennaam2016).

The use of genomic selection to improve health traits is not new. Health traits have been improved through selection as genomic susceptibility to viral, bacterial, and parasitic diseases is well documented (Bishop et al., Reference Bishop, Axford, Nicholas and Owen2011). The identification of loci associated with enhanced resistance to trypanotolerance in N'Dama cattle and mastitis in Holstein cattle are two examples (Klungland et al., Reference Klungland, Sabry, Heringstad, Gro Olsen, Gomez-Raya, Våge, Olsaker, Ødegård, Klemetsdal, Schulman, Vilkki, Ruane, Aasland, Ronningen and Lien2001; Hanotte et al., Reference Hanotte, Ronin, Agaba, Nilsson, Gelhaus, Horstmann, Sugimoto, Kemp, Gibson, Korol, Soller and Teale2003; Tal-Stein et al., Reference Tal-Stein, Fontanesi, Dolezal, Scotti, Bagnato, Russo, Canavesi, Friedmann, Soller and Lipkin2010; Sahana et al., Reference Sahana, Guldbrandtsen, Thomsen and Lund2013). The availability of high-density genotyping assays that cover the genome at relatively inexpensive prices has paved the way for genome-enhanced selection. In addition to providing information for selection, the genotyping assays provide a platform to identify causal mutations that are involved in the regulation of gene expression or gene translation (Tam et al., Reference Tam, Patel, Turcotte, Bosse’, Pare’ and Meyre2019). The identification of causal mutations provides an opportunity to better understand the mechanisms of host susceptibility.

Phenotypes

For genomic-enhanced selection to be effective, it requires that a definition of the trait be standardized in the cattle industry. Selecting a standardized BRD phenotype that allows producers to accurately assess disease status with minimal effort would be ideal. For BRD phenotypes, the trait definition has varied depending on how the cattle were diagnosed or how many times an animal was treated for BRD. These measures have been used because they are easily applied within the industry. Currently, a standardized phenotype has not been adopted across the cattle industry to identify BRD, but BRD assessments have been proposed that include scoring rubrics, ultrasonography, and auscultation. To establish a standardized phenotype, cattle industry groups are beginning to meet to discuss this important first step. For example, in beef cattle, the Beef Improvement Federation formed a committee to evaluate and recommend BRD diagnostic criteria and ways to report BRD through feedlots. In dairy, discussions have been ongoing to include BRD data as part of the health traits reported for predicted transmitting abilities used for selection of replacement animals in commercial genotyping platforms as well as national databases maintained by the Council on Dairy Cattle Breeding. Predicted transmitting ability (PTA) for respiratory disease is the prediction of the susceptibility of the offspring to respiratory disease expressed as a deviation from the mean.

For genomic studies, scoring rubrics and the use of treatment records have been used to define BRD cases. These scoring rubrics were initially designed for dairy calves and attribute the severity of symptoms to a score, so that a high score characterizes cattle with BRD clinical signs. One common scoring rubric is the Wisconsin calf health rubric (McGuirk, Reference McGuirk2008; McGuirk and Peek, Reference McGuirk and Peek2014). The Wisconsin calf health rubric scores each of the BRD clinical signs with a value of zero to three and sums them together to reach a cumulative score. The clinical scoring rubric is based on rectal temperature, presence of spontaneous or induced cough, nasal discharge, and the greatest of the score based on ocular discharge or head and ear position. The maximum cumulative score is 12 and the minimum score is zero and a case is defined as an animal with scores ≥5. A validation of this method was recently completed where the sensitivity and specificity of identifying BRD calves were 62.4 and 74.1%, respectively (Buczinski et al., Reference Buczinski, Ollivett and Dendukuri2015). The Wisconsin calf health rubric has also been applied to cattle in feedyards to identify cattle with BRD at the producer level (McGuirk, Reference McGuirk2008; McGuirk and Peek, Reference McGuirk and Peek2014; Neupane et al., Reference Neupane, Kiser and Neibergs2018). When compared with detecting lung consolidation, the Wisconsin calf health rubric performed better than lung auscultation, but not as well as thoracic ultrasound. Thoracic ultrasonography had an increased sensitivity (79.4%) and specificity (93.9%) but its direct use in a producer setting to detect BRD cattle is limited (Buczinski et al., Reference Buczinski, Ollivett and Dendukuri2015).

Three scoring rubrics have been tested that use similar clinical signs as the Wisconsin calf health rubric but score the clinical signs as present or absent (with the exception of nasal discharge which has three levels in one scoring system) and have removed induced cough as a measure of clinical symptoms and added respiratory quality (Love et al., Reference Love, Lehenbauer, Kass, Van Eenennaam and Aly2014). The three scoring rubrics proposed by Love et al. (Reference Love, Lehenbauer, Kass, Van Eenennaam and Aly2014) performed similarly in their ability to detect cases and controls and agreed (κ > 0.8) with the Wisconsin calf health rubric.

Genome-wide association analyses

Genome-wide association analyses (GWAA) are used to identify genomic regions associated with cattle performance without limiting investigations to genes whose functions have been characterized. The analyses compare the allele and genotypic frequencies of cattle with BRD and those without BRD to identify regions that have large genetic effects on the susceptibility of the disease. Genome-wide association studies have been performed on both beef and dairy cattle to identify genomic regions associated with BRD susceptibility.

Lung lesions

BRD results in damage to the lungs that may present as lesions at the time of harvest. The use of lung lesions as an indication of BRD in a case–control design for a GWAA has been performed in two studies in the United States and one study in Israel. The case–control design for lung lesions as a phenotype has the misfortune of not identifying all cattle that have been affected with BRD. Not all cattle that have been affected with BRD will present with lung lesions depending on the course of disease, and so like the use of clinical diagnostics, the separation of animals into cases and controls reflects the evidence of disease at a specific point in time. The first United States study of lung lesions consisted of pooled samples where each pool comprised 96 animals. Sixty pooled samples of cattle with lung lesions (cases) and 60 pooled samples of cattle without lung lesions (controls) were evaluated with an Illumina BovineHD BeadChip (Keele et al., Reference Keele, Kuehn, McDaneld, Tait, Jones, Smith, Shackelfore, King, Wheeler, Lindholm-Perry and McNeel2015). Associations were identified by using a 5% genome-wide error rate threshold (P ≤ 1.49 × 10⁻⁷) or a false discovery rate (FDR) of 5% (P < 5.38 × 10⁻⁶). The cattle in this study consisted of a ‘natural’ cattle population, where cattle did not receive antibiotics, hormones, or animal byproducts and were traceable to place of birth, and a ‘conventional’ population where cattle may have received antibiotics, hormones, animal byproducts, and were not traceable to place of birth. Lung lesions were present in >30% of cattle in the natural population, while 9.8% of those in the conventional cattle population had lung lesions (Keele et al., Reference Keele, Kuehn, McDaneld, Tait, Jones, Smith, Shackelfore, King, Wheeler, Lindholm-Perry and McNeel2015). Fourteen SNPs were associated with lung lesions at the 5% genome-wide error rate (P ≤ 1.49 × 10⁻⁷) and 85 loci were associated with an FDR of 5% (P ≤ 5.38 × 10⁻⁶; Table 1; Keele et al., Reference Keele, Kuehn, McDaneld, Tait, Jones, Smith, Shackelfore, King, Wheeler, Lindholm-Perry and McNeel2015).

Table 1. Genome-wide association analysis of cattle that had lung lesions as an indication of susceptibility to bovine respiratory disease

The second study performed in the United States consisted of 920 steers that were also enrolled in a ‘natural’ cattle feeding program (Kiser et al., Reference Kiser, Lawrence, Neupane, Seabury and Neibergs2017). Half of these cattle were clinically diagnosed with BRD and removed from the natural program to receive treatment. Lung lesions were evaluated as described by Tennant et al. (Reference Tennant, Ives, Harper, Renter and Lawrence2014). Pseudo-heritability estimates for BRD as measured by lung consolidation was 0.25, 0.0 for the presence of fibrin tissue in the lung, 0.28 for the presence of lung consolidation and fibrin in the lung, and 0.13 for hyperinflated lungs (Kiser et al., Reference Kiser, Lawrence, Neupane, Seabury and Neibergs2017). Four loci were associated (P < 1 × 10⁻⁵) with lung consolidation, three loci were associated with lungs that contained both consolidation and fibrin tissue, and 10 loci were associated with hyperinflation of the lungs (Table 1). Lung lesions were identified in >60% of lungs of cattle irrespective of whether the cattle showed clinical signs of BRD (Kiser et al., Reference Kiser, Lawrence, Neupane, Seabury and Neibergs2017). None of the associations identified by Keele et al. (Reference Keele, Kuehn, McDaneld, Tait, Jones, Smith, Shackelfore, King, Wheeler, Lindholm-Perry and McNeel2015) with a 5% genome-wide error rate were identified as associated in the Kiser et al. (Reference Kiser, Lawrence, Neupane, Seabury and Neibergs2017) study, but both studies identified a locus on BTA4 at 102 Mb associated with lung lesion (FDR < 0.05) and hyperinflated lungs (P < 1 × 10⁻⁵).

Scoring of lung lesions in Holstein male calves at harvest in Israel partitioned cattle into High (Glatt Kosher) and Low (Non-Kosher) groups (Lipkin et al., Reference Lipkin, Strillacci, Eitam, Yishay, Schiaini, Soller, Bagnato and Shabtay2016). A ‘Kosher’ animal is one in which the lung is intact after all lung lesions are removed. ‘Glatt Kosher’ refers to animals that have no lung adhesions and make up the High group. In contrast, ‘Non-Kosher’ (low group) cattle have lungs with torn adhesions or that are not intact after lesions are removed. Twenty-one to 31 male Holstein calves formed five High and two Low groups, which were pooled for genotyping with the Illumina BovineHD BeadChip. Nineteen regions were identified as associated (P < 2.5 × 10⁻¹) with Kosher status or the presence of lung lesions (Lipkin et al., Reference Lipkin, Strillacci, Eitam, Yishay, Schiaini, Soller, Bagnato and Shabtay2016; Table 1). None of the loci associated with Kosher status were shared with the two lung lesion studies in beef cattle in the United States. This is unsurprising as the phenotypes differed, cattle breeds used in these studies differed, and management differed which likely effected which pathogens were responsible for disease and thus the loci involved in susceptibility.

Clinical disease

Two GWAA have been performed using a case–control clinical BRD design to identify susceptibility loci with the Wisconsin calf health rubric. Pre-weaned Holstein calves were sampled from California or New Mexico and a GWAA performed using the Illumina BovineHD BeadChip. Four analytical approaches were used and the most significant loci from each analysis were evaluated for concordance (Neibergs et al., Reference Neibergs, Seabury, Wojtowicz, Wang, Scraggs, Kiser, Neupane, Womack, Van Eenennaam, Hagevoort, Lehenbauer, Aly, Davis and Taylor2014). Concordant results were identified for 373 SNPs in the California population consisting of 2014 calves, 370 SNPs in the 767 New Mexico calves, and 324 SNPs when the two populations were combined. The 10 loci concordant across all four analyses with the strongest evidence for an association are shown in Table 2 (Neibergs et al., Reference Neibergs, Seabury, Wojtowicz, Wang, Scraggs, Kiser, Neupane, Womack, Van Eenennaam, Hagevoort, Lehenbauer, Aly, Davis and Taylor2014). A second study using the same animals imputed the BovineHD BeadChip genotypes to whole genome sequence data to identify and refine the loci associated with BRD (Hoff et al., Reference Hoff, Decker, Schnabel, Seabury, Neibergs and Taylor2019). Imputation is the use of genotypes of a reference population to predict genotypes in silico in a similar population of animals. The in silico (imputed) genotypes are inferred based on regions of linkage disequilibrium between genotypes obtained through chemistry. After filtering of variants, over 9 million imputed variants were evaluated with an average imputation accuracy of 84.2% (Hoff et al., Reference Hoff, Decker, Schnabel, Seabury, Neibergs and Taylor2019). The data enhanced the HD analysis by refining the loci previously identified and increasing the proportion of variance explained by the SNP variants. Of the 100 most significant QTL in the California and New Mexico populations, 11 were shared (Table 2; Hoff et al., Reference Hoff, Decker, Schnabel, Seabury, Neibergs and Taylor2019). The differences in loci identified in the California and New Mexico populations were most likely due to genetic heterogeneity or the host response to the different pathogens responsible for BRD identified by bacterial and viral diagnostics of the calves in the study (Neibergs et al., Reference Neibergs, Seabury, Wojtowicz, Wang, Scraggs, Kiser, Neupane, Womack, Van Eenennaam, Hagevoort, Lehenbauer, Aly, Davis and Taylor2014; Hoff et al., Reference Hoff, Decker, Schnabel, Seabury, Neibergs and Taylor2019). This suggests that genomic prediction models may perform poorly in different environments where the pathogen profiles for BRD differ. Additional studies of BRD pathogen profiles in different environments need to be conducted to assist in improving the accuracy of genomic prediction for BRD susceptibility.

Table 2. Genome-wide association analysis of susceptibility of cattle to clinical bovine respiratory disease

^a SM McGuirk and SF Peek, 2014.

^b Top 10 most significant loci identified in each study.

^c Eleven loci shared across two pre-weaned Holstein calf populations.

A third GWAA of 45,425 SNP was used to analyze 67,289 animals that had genotypic and BRD treatment data using a single-step BLUP method taken from 326 dairy producers in the United States (Gonzalez-Pena et al., Reference Gonzalez-Pena, Vukasinovic, Brooker, Przybyla and SDeNise2019). In this cohort, the incidence of BRD was 21% in calves from birth to 1 year of age, with most cases occurring prior to 30 days of age (Gonzalez-Pena et al., Reference Gonzalez-Pena, Vukasinovic, Brooker, Przybyla and SDeNise2019). The mean reliability (accuracy) for the prediction of respiratory disease was 0.419 with a range of 0.189–0.99 and the heritability was 0.042. The PTA of bulls with a minimum of 100 offspring in the 326 dairies identified bulls with large differences in the PTA for BRD susceptibility. Bulls with the worst PTA (high values) for respiratory disease reported up to 50.4% of their progeny were affected by BRD (Gonzalez-Pena et al., Reference Gonzalez-Pena, Vukasinovic, Brooker, Przybyla and SDeNise2019). Selection of bulls with low BRD PTAs would be beneficial in reducing the incidence of disease with similar levels of management. This study formed the basis for the wellness trait testing commercially provided by Zoetis and so no descriptions of the loci associated with susceptibility to BRD were provided.

Differences in the beef and dairy cattle BRD studies include the prevalence of crossbred beef cattle compared with the high level of purebred dairy cattle and the age and immunocompetence of pre-weaned Holstein calves compared to older weaned beef calves in a feedlot in some studies. Crossbreds will have varying levels of linkage disequilibrium, whereas purebreds will be more similar between animals. As linkage disequilibrium will affect the accuracy of prediction for selection, using crossbred cattle in a study can decrease the statistical power to detect an association compared to using purebred cattle. To compensate, a larger sample size is needed for a crossbred population to detect a locus with the same effect on the phenotype as would be detected in a purebred population. The benefit of identifying an association in a crossbred cattle population is that it is likely that the SNP associated with the phenotype is closer to the causal mutation than in a purebred population. This is because the linkage disequilibrium in a crossbred population will generally not extend over as great of a distance as in a purebred population. To identify loci associated with BRD in crossbred cattle, large studies will be needed, ideally with standardized phenotypes that the industry has adopted, to facilitate the use of these loci for genomic selection.