Implementation of Genetic Selection for Grazing Distribution to Make Cattle Grazing in the Western US More Sustainable

Results from Genome Wide Association Analyses. Five SNP were associated with each of the six terrain-use traits: slope, elevation, vertical climb, distance from water, rough index, and rolling index. Several SNP were associated with more than one trait and therefore, 26 unique SNP effects were detected. Single nucleotide polymorphisms associated with terrain-use traits were observed on chromosomes 1, 4, 5, 8, 10, 11, 12, 13, 16, 17, 18, 21, 22, 23, 24, 27, and 29, though the largest number of variants was found on chromosome 10 (n = 5). The posterior inclusion probabilities for these 26 SNP ranged from 0.02 to 0.18 (Figure 1). Marker rs41600226 on chromosome 23, associated with elevation, had the highest PIP across all association analyses. The mean PIP for all SNP considered was 0.006 for elevation and vertical climb whereas the mean PIP for slope, distance travelled from water, rolling index and rough index was 0.005.

Of the 26 SNP that were associated with terrain-use traits, 10 were intronic variants and 16 were intergenic. Further examination of the candidate SNP using Cattle QTL database revealed that 24 QTL had been previously identified and associated with production and health traits in dairy and beef cattle including: birth weight, carcass weight, residual feed intake, milk yield, and bovine tuberculosis susceptibility. Putative candidate genes were related to diverse biological functions including: gene expression, intracellular transport, and glucose metabolism.

Genome Wide Associations Analyses Discussion. The genome-wide association study revealed QTL that may be important for terrain-use in extensive rangeland pastures. The posterior inclusion probabilities for the SNP in the six GWAS were considered low; however, Manhattan plots for five of the six traits had notable peaks and a review of the literature revealed no standard threshold for determining significance based on PIP. Furthermore, previous mammalian studies suggested that PJP varies greatly depending upon the trait of interest. In a previous study, cooperators in this study (Speidel et al. 2018 J. Anim. Sci. 96, 846-853) examined the genetic architecture of Heifer Pregnancy and Stayability in Red Angus cattle and set a PIP threshold for each trait. Speidel et al. (2018) reported lower PIP for Heifer Pregnancy (6%) than Stayability (100%). More specifically, the maximum PIP for heifer pregnancy was 6% whereas the max PIP for stayability was 100%. These authors hypothesized that the low inclusion of markers for heifer pregnancy was due to limited phenotypic data on genotyped individuals (n = 567) and the low heritability of the trait (h2 = 0.12). Perhaps, the PIP for the terrain-use traits presented in this study were influenced by similar factors. The moderate sample size of this study limited the statistical power for QTL detection and results suggested that terrain-use traits may be lowly heritable. Yet, this study was challenged by landscape diversity among ranches, breed of cattle, and our ability to adequately quantify the difficulty cattle incur when utilizing steep slopes, climbing and descending rugged topography and traveling long distances from water. Terrain use is a multi-dimensional trait that should integrate slope, vertical climb (or elevation) and distance traveled from water. Although the rough and rolling terrain indices combine these traits into one value, alternative approaches to integrate slope, vertical climb and distance traveled from water into a single trait deserves further study.

Twenty-four genomic windows, spanning 13 chromosomes, were associated with the six terrain-use traits and six of the windows were associated with multiple traits. Within the genomic windows, 25 lead SNP were identified. Parallel with the PIP, the genomic windows explained a low proportion of genetic variance for the terrain-use traits. Across the six traits, the maximum proportion of genetic variance explained by a single window was 0.0057%. Again, these estimates may have been influenced by the heritability and marker density (i.e., the number of markers captured in the genomic window). A greater proportion of genetic variance can be explained using dense genotyping arrays as there is greater linkage disequilibrium between causative mutations and the surrounding markers. Another important consideration regarding the proportion of genetic variance explained was the genetic architecture underlying quantitative traits. Complex quantitative traits that are influenced by many QTL with small effects (i.e., QTL explain a small percentage of total genetic variance on an individual basis. Genomic regions containing causative mutations may encompass few SNP with large effects whereas the remaining regions contained many markers with small effect.

A comparison of the candidate SNP to the genomic windows revealed 32 QTL and 29 putative candidate genes that may play a role in beef cow terrain-use. Thirty of the QTL were documented in Cattle QTLdb and associated with numerous beef cattle production traits. The 29 putative candidate genes identified in this analysis were not concordant with the five genes identified by our previous Western SARE study SW09-054 (Bailey et al. 2015, Rangeland Ecol. Manage. 68, 142-149). Confirmation of the previously identified genes may have been hindered by differences in SNP density (high-density genotypes vs. BovineSNP50 genotypes), statistical methodology (single-SNP regression vs. Bayesian multiple-SNP regression), and extreme topography variation among rangeland beef cattle operations as new data were added to these resources. Loci on chromosome 29 were detected in both studies and warrant fine mapping efforts to understand and identify the potential causal mutations.

Our previous study Western SARE 09-054 (Bailey et al. 2015) used BovineHD genotypes and single-SNP regression analysis to identify QTL associated with beef cattle terrain-use. Significant markers identified in this analysis were used to develop a custom-genotyping panel of 50 SNP genotypes for subsequent association analyses with phenotypes. The genotype data in this study were derived from BovineSNP50; therefore, the eight candidate SNP discovered by our previous study (SW09-054, Bailey et al. 2015) were not replicated in this study. In our study, Bayesian methodology was used instead of single-SNP regression because simultaneously fitting markers accounts for all SNP in linkage disequilibrium (LD) with the QTL which decreases the proportion of unexplained genetic variance. In addition, Bayesian multiple-regression results in fewer false positives because population structure was explained in the model. Differences in the genomic architecture (i.e., allele or genotypic frequencies) between study populations (ranches) most likely influenced QTL detection.

All of the ranches except Carter Ranch were in moderate (3,940 to 5,250 feet) to high-elevation regions (≥ 5250 feet) where a reduction in atmospheric pressure results in lower partial pressure of oxygen. This may help explain the association between the previously discussed candidate genes and the terrain-use traits. Cattle that are utilizing rugged terrain in high altitude regions may experience hypoxia as oxygen consumption increases in active muscle cells. In hypoxic conditions, mammals may undergo physiological changes to compensate for the lack of oxygen including angiogenesis, erythrocyte modification, heart remodeling, and accelerated gluconeogenesis. Furthermore, cattle maintained in high-elevation regions may develop high altitude disease as a result of hypoxia induced pulmonary arterial hypertension. The role of the hypoxia candidate gene and the relationship between pulmonary arterial pressure and terrain-use of cattle deserves further study.
As previously discussed, three of the putative candidate genes (INHBA, LRRC4B, and SEPT7) were associated with growth traits and feed efficiency in cattle. The functions of INHBA, LRRC4B, and SEPT7 are particularly interesting given the potential relationship between cow-size and terrain-use and locomotion and residual feed intake (RFI). Use of rugged terrain-use should require greater energy expenditure and be unfavorable for low RFI. Additional studies with an independent population are needed to further examination the relationship between mature cow weight/size, residual feed intake, and terrain-use traits.

The QTL detected in this study support the polygenic nature of complex traits regulating beef cow terrain-use traits. Detection of genotype to phenotype associations in the current study was likely limited by the moderate sample size and lack of uniformity in the data. Therefore, a large independent population of beef cows, composed of one breed, grazing on the same pastures is needed to refine terrain-use measurements and further elucidate the role of genetics in beef cattle terrain-use phenotypes.

Analysis of Previously Identified Candidate Genes. Our previous study (SARE SW09-54) suggested that the traits used in indices to quantify grazing distribution are moderately heritable; therefore, genomic selection could be used to improve grazing distribution. Five quantitative trait loci (QTL) and underlying candidate genes (ACN9, FAM48A, GRM5, MAML3, and RUSC2) were found to be associated with grazing distribution traits in cattle. We examined these candidate genes and identified additional SNP that can be incorporated into a previously developed 50-SNP panel used for genotype associations with grazing distribution phenotypes (Western SARE SW09-054). Sequencing of RNA (RNA-Seq) yielded 30 million reads (single-read) per sample from 6 tissues (aorta, LM muscle, lung, pulmonary artery, and right and left ventricle) collected as part of an altitude tolerance study of Angus cattle. These tissues were from 10 steers with outlying pulmonary arterial pressure observations and unique sires. Sequences were assembled to the annotated bovine reference genome (UMD3.1; release annotation 87) and analyzed using CLC Genomics Workbench (version 8.0). Variant detection was performed using two methods: (1) individual samples and (2) a pool of all samples. No variants were detected in GRM5, MAML3, and RUSC2; however, individual sample analysis identified 30 SNP within ACN9 and FAM48A, and pooled sample analysis identified 184 SNP within ACN9 and FAM48A. Twenty-one SNP were identified in both approaches. The Ensemble Variant Effect Predictor was used to determine the functional consequence of each SNP. Of the 21 SNP, 16 were intronic, four were exonic, and one was reported to be a downstream variant and a splice acceptor variant. The SNP discovered using RNASeq technology were compared to the exonic SNP in dbSNP within the 5 candidate genes. There were 1,663 exonic SNP in dbSNP in these genes. One synonymous SNP, located within ACN9 (rs382949979), was observed in both data from RNASeq and dbSNP. In summary, 21 SNP were discovered in two of the five candidate genes underlying QTL associated with grazing distribution that were identified in the previous study (SW09-054).

Terrain Use Indices.  Livestock grazing distribution is a complex trait, which makes it difficult to evaluate in genome wide association studies. Slope and horizontal and vertical distance to water, alone and in combination, affect how cattle use mountainous rangelands, which makes terrain use a multidimensional trait. Assessment of terrain use by cattle is difficult because the metrics used to describe aspects of terrain use different scales of measurement. For example, slope is measured in percent or degrees, while distance from water is measured in meters. Ongoing studies are developing indices that can be used measure and evaluate terrain use of cattle n rugged rangeland. Tracking data from 15 to 18 cows at each of 4 ranches in New Mexico, Montana and Nevada were used in resource selection analyses. The ranches had different terrain types varying from rolling to very mountainous terrain. Slope, elevation, and distance to water at overlapping 100-m diameter randomly-selected plots were calculated at each ranch. Cattle use of each plot were used general linear model with a negative binomial distribution. Feature scaling was used to normalize these variables so that could be evaluated with relatively similar units. Except for the ranch with rolling terrain, cattle selected again steep slopes (linear relationship). Cattle selected against higher elevation in mountainous terrain, and the relationship was quadratic. Cattle selected again areas far from water at all ranches, but the relationship was quadratic rather than linear. These results suggest that feature scaling and resource selection analyses may be tools to help evaluate the multidimensional aspects of terrain use in cattle, but the relationships between terrain metrics will likely vary among ranches because of the unique terrain characteristics of each pasture.

Repeatability of Cattle Grazing Patterns. We used tracking collars to monitor cattle movements and evaluate grazing distribution phenotypes. However, few studies have evaluated the consistency of cattle movements over time in rugged rangeland pastures. We examined the repeatability of cattle movement patterns at five locations in New Mexico and Arizona: the Chihuahuan Desert Rangeland Research Center (CDRRC), Evans Ranch, Wilbanks Ranch, Hartley Ranch and Todd Ranch. These data were obtained from this study and our previous SARE study (SW09-054). Eight to 19 randomly-selected cows from herds of 40 to 200 cows were tracked with GPS collars at each ranch. Cows were tracked at either 10- or 15-minute intervals. Terrain use was summarized by week. A repeated measures analysis was conducted on each ranch using the weekly average of slope use, elevation use, and distance from water as the dependent variables. Intraclass correlations of weekly averages of the three terrain use metrics were used to assess temporal consistency of grazing distribution traits. Week was a fixed effect and cow was a random effect. Intraclass correlations of terrain use by individual cows varied among ranches. The Wilbanks Ranch had the strongest intraclass correlations for slope, elevation, and distance to water of 0.60, 0.50, and 0.77, respectively. Intraclass correlations for elevation at the Hartley and Todd Ranches were strong, 0.61 and 0.71, respectively, but correlations for slope and distance to water were weak to moderate (0.18 to 0.30). In contrast, intraclass correlation at the CDRRC and Evans Ranch were weak (0.00 to 0.08). Our results suggest that consistency of terrain use by cattle can vary by location; however, these relationships are positive and moderate to strong at most ranches. Factors such as cattle familiarity with pastures and the nature of the terrain features may explain part of this variability, however additional analyses examining how temporal changes in terrain use affect this phenotype are needed.

Terrain use and grazing distribution traits in cattle are difficult to evaluate because movement patterns are temporally variable due to ever changing climatic and forage conditions. Terrain-use can be monitored with GPS collars, but the length of tracking is constrained by battery life and on-site grazing and management plans. Cattle tracking studies have varied from days to months. We evaluated data from our previous SARE study (SW09-054) to determine if shorter, 30 to 60 day, sampling durations were as effective as a 3 month tracking periods for characterizing slope and elevation use, and vertical and horizontal distance traveled from water of individual cows. Fifteen Limousin cows from a herd of 250 cows were tracked at 15-minute intervals for 92 days during late winter and early spring in a 9,065 ha pasture that included both gentle and rugged terrain. Terrain-use during 30- and 60-day periods at the beginning and end of the tracking were compared to the full 92-day period. Slope, elevation, and vertical and horizontal distance from water from the full sample period were regressed on values from the shorter data subsets and correlations were calculated between sampling periods for each pair of terrain use metrics. The 60-day periods showed strong agreement with the full 92-day period with correlation coefficients varying from 0.90 to 0.97. Correlations between 30-day periods and the 92-day period varied from 0.30 for distance to water during the early period to 0.90 to 0.93 for all traits during the late 30-day period. These preliminary analyses suggest that 2-month tracking periods are equivalent to 3-month tracking periods for identifying differences in terrain-use among beef cows. These findings help facilitate collection of grazing distribution phenotypes from larger number of cattle for genetic improvement.
Only a few studies have examined the impact of cow size on grazing distribution. Tracking data from 74 mature cows collected in this project at the CDRRC (n = 14), Silver Spur Ranch (n = 28) and Fort Union Ranch (n = 32) were used to determine if there are relationships between cow body size and terrain use. After fitting the statistical model for ranch, residual correlations were used to examine the relationship between cow body condition score, linear measures of cow size, and terrain use metrics. No important residual correlations were detected between terrain use metrics and cow body measures (P > 0.05). These results suggest that terrain use of cattle is not related to cow size and body condition. Although genetic correlations need to be evaluated, the phenotypic correlations suggest that selection for terrain use should not affect cow size or body condition.