Identifying local adaptation and creating region-specific genomic predictions in beef cattle

IDENTIFYING LOCAL ADAPTATION AND CREATING REGION-SPECIFIC GENOMIC PREDICTIONS IN BEEF CATTLE

Sponsoring Institution

National Institute of Food and Agriculture

Project Status

EXTENDED

Funding Source

AFRI COMPETITIVE GRANT

Reporting Frequency

Annual

Accession No.

1008909

Grant No.

2016-68004-24827

Project No.

MOC00051167

Proposal No.

2015-08789

Multistate No.

(N/A)

Program Code

A5161

Project Start Date

Mar 1, 2016

Project End Date

Feb 28, 2019

Grant Year

2017

Project Director
Decker, J. E.

Recipient Organization
UNIVERSITY OF MISSOURI
(N/A)
COLUMBIA,MO 65211

Performing Department
Office of Sponsored Programs A

Non Technical Summary
Cattle poorly adapted to their environment result in lost revenue and jeopardize the stability of the food supply. Large scale genetic data (i.e. genomic data) now allows us to rigorously analyze genetic adaptations and avoid the breeding of animals that will not thrive.We will use genomic methods to precisely identify DNA variants responsible for local adaption to tolerate biotic and abiotic threats to production, enhance resilience of beef production, and enhance beef quality and quantity. The project will achieve three objectives:Identify DNA variants responsible for regional genetic adaptationCreate geographic region-specific genomic predictions, focusing on adaptation variants from Objective 1.Educate the next generation of beef producers to fully embrace and properly use animal breeding tools.Analyzing more than 170,000 cattle with ~15 million high-accuracy DNA variants, we will use selection mapping to identify detailed chromosome regions (e.g. genes) responsible for local adaptation. We will also identify gene-by-environment interactions using multiple statistical methods. When local genetic adaptations exist, ranking animals using a regional genetic evaluation will be different from national cattle evaluations. Focusing on loci under regional genetic adaptation selection or with gene-by-environment interactions, we will develop region-specific genomic predictions. These genomic predictions will allow rapid identification of cattle best suited to an environment. Beef producers have not embraced appropriate animal breeding practices and new technologies, limiting genetic improvement of efficient, high quality beef cattle. We will create engaging curriculum for youth and undergraduate education, including internships, to train the next generation of beef producers.

Animal Health Component

15%

Research Effort Categories

Basic

20%

Applied

80%

Developmental

(N/A)

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
303	3310	1080	100%

Knowledge Area
303 - Genetic Improvement of Animals;

Subject Of Investigation
3310 - Beef cattle, live animal;

Field Of Science
1080 - Genetics;

Keywords

Goals / Objectives
Of the major livestock sectors, beef cattle are the most exposed to abiotic and biotic stressors. While poultry, swine, and dairy operations typically keep the animals in consistent and protected confinement, provided necessary medical care, and provide a balanced total mixed ration diet allowing the environment of the animals to be maximally controlled, beef cows are grazed on range or pastures where they are exposed to a myriad of environmental stressors (National Research Council. Committees On Animal Nutrition, 1981).Beef cattle that are poorly adapted to the environment in which they live result in lost revenues and jeopardize the stability of the United States food supply. One prominent example in the U.S. beef industry is susceptibility to fescue toxicosis. Fescue has an endophyte fungus which produces ergot-like alkaloids which are toxic to animals. These toxins cause vasoconstriction, reduced appetite, poor reproduction, slowed growth, and exacerbate heat stress. In 1993, it was estimated that fescue toxicosis cost the U.S. beef industry $609 million annually (Hoveland, 1993). After adjusting for inflation, this translates to $998 million in 2014 dollars, but this adjustment ignores increased feeder calf and crop prices, so the true current cost is even greater. Losses also occur due to poor adaptation to high altitude, humid, arid, hot, and cold environments (Hohenboken et al., 2005). However, these issues can be alleviated, because there is ample variation in taurine cattle populations to make improvement. One of the ways for animals to cope with fescue and heat stress is to shed their winter hair coat in the spring, and hair shedding is a heritable trait (Gray et al., 2011). Genetic variance is appreciable for other adaptive traits, such as pulmonary arterial pressure with a heritability of 0.34 (Shirley et al., 2008). While indicine cattle have been used to breed cattle adapted to hot climates, this is not optimal and does not address when the stressors are not heat. Further, beef from indicine cattle has reduced palatability (O'Connor et al., 1997), thus decreasing demand and limiting the profits and sustainability of producers who use indicine genetics. As genetic variation is present within taurine cattle for adaptation to the various U.S. environments, the sustainable strategy is to utilize existing variation through artificial selection. What is needed are breeding decision tools to identify taurine animals adapted to regional environments, even when we do not necessarily know the traits leading to that adaptation.The beef industry has the most distributed breeding decision structure compared to all other major agricultural species, plants and animals included. Rather than germplasm and breeding populations controlled by large companies with highly trained breeders making the selection decisions, in the beef industry there are approximately 742,500 cow-calf operations and every operation makes their own breeding decisions. (In the United States there are 29.7 million beef cows divided by 40 head per operation into 742,500 operations. See http://www.ers.usda.gov/topics/animal-products/cattle-beef/background.aspx and http://www.beefusa.org/beefindustrystatistics.aspx.) Furthermore, these beef producers do not typically use optimal breeding practices. For example, a BEEF magazine article from January 29, 2014 reported a survey that found beef producers are more likely to select animals using actual trait records than use the more accurate genetic predictions (expected progeny differences, EPDs), and they are also more likely to select on individual traits and EPDs than use optimal economic selection indexes. The accuracy of selection decisions is higher when genetic predictions are used, thus response to selection will be more rapid. Vice versa, the accuracy of selection on actual trait measurements is limited by the heritability of the trait and is further reduced by the failure to correct for confounding environmental effects and is therefore far less accurate. Use of these faulty practices would likely lead to the dismissal of a plant or animal breeder in other agricultural sectors. A major educational effort is needed to enable and persuade the next generation of beef producers to understand and adopt these optimal breeding practices to ensure that beef production in the United States increases economic prosperity, national resilience and food security.Overview of ObjectivesThe overall goal of this project is to create tools, genomic predictions, which will allow beef producers to identify and breed cattle which are adapted to their production environment. Briefly summarized, our objectives are to:1. Identify genomic variants or loci responsible for regional adaptation.2. Create geographic region-specific genomic predictions, focusing on regional adaptation variants from Objective 1.3. Educate the next generation of beef producers to fully embrace and properly use animal breeding tools.

Project Methods
Using high-density SNP genotype data for over 130,000 Angus, 15,000 Hereford, 25,000 Simmental, 4,500 Gelbvieh, 12,000 Red Angus and 2,240 Limousin samples, we will impute the existing SNP data up to more than ~15M SNPs. Using 15M SNPs in our genome-wide analyses will enable a detailed analysis of the genome.Based on the geographic regions described in Leighton, Willham, and Berger (1982), we will initially define 9 U.S. regions. We can also use the PRISM data to refine geographic regions using k-means clustering of reported variables to group 4 km resolution locations into larger regions with similar environments. We will assign membership of a genotyped animal to a region based on multiple criteria.All selection scan analyses will be conducted within a single breed; analyses will be repeated for each of the 6 breeds. If an animal is not adapted to the environment, the animal will underperform and will be excluded from the breeding population. This selection causes the frequency of these DNA variants to be significantly different between regional populations.We will use hapFLK and TreeSelect to identify genomic loci that have diverged between geographic regions due to local genetic adaptation selection.We will fit 8 environmental variables from the PRISM data in which an animal was born as the dependent variable in a linear mixed-model GWAS. This analysis will identify DNA variants significantly associated with an environmental variable, and thus responding to selection for that environmental stressor.Focusing on selection scans at whole genome sequence resolution is powerful for several reasons. First, it allows us to use existing genotype data without the need to phenotype and genotype new samples as part of a controlled experiment. Second, selection mapping allows us to utilize the past 150 years of history and local adaptation of cattle in the United States. Third, selection mapping allows us to identify local adaptation loci that have responded to pressures that we could not measure or would not hypothesize as driving local adaptation. We can further expand our knowledge of the biology of local adaptation in beef cattle based only on genomic data.We will further look for gene-by-environment interactions and will use local adaptation selection scan and the gene-by-environment results to develop genomic predictions for adaptation to the variety of production environments that exist in the United States.We will conduct a genetic prediction (EPD) environmental interaction analysis by looking for regional differences in EPD values for cows that repeatedly produce in a region versus those that fail to rebreed 425 days after delivering their first calf.We will use GCTA to estimate the gene-by-environment variance between the recorded phenotypes and the environmental data from the PRISM dataset.Using the geographic region assignments, we will perform within-region breed-specific genome-wide association analyses of traits most exposed to abiotic and biotic stress. We will use the BOLT-LMM software for single trait GWAA, as it has decreased requirements and improved power.To further increase our power, we will perform multiple-trait meta-analyses of univariate GWAA or we will perform multivariate GWAA.Because we have stratified these GWAA by environment, we can use the random effects meta-analysis model in the METASOFT software to test if effect sizes between regions are significantly heterogeneous with Cochran's Q statistic.We will analyze gene-by-environment interactions using a linear mixed-model genotype-by-environment interaction GWAA using the GWAF package in R, focusing on environmental factors with the largest gene-by-environment variance components from the GCTA analysis.One of the phenotypes most involved in regional adaptation is an animal's ability to shed its winter coat at the appropriate time. We will collect DNA samples on 6,921 animals with hair coat scores. Cattle with phenotypes and blood samples will be genotyped using a SNP assay containing ~35,000 SNPs used for imputation and ~195,000 SNPs that are loss-of-function mutations and other functional variants. We will impute the samples to full genome sequence level data. Linear mixed-model GWAA will identify variants that are strongly associated with hair shedding.The ultimate goal and deliverable of the research portion of this project are region-specific genomic predictions that will be used to identify and select regionally adapted beef cattle. When gene-by-environment interactions are present, the ranking of animals in a regional genetic evaluation will be different from national cattle evaluations.We will create three curricula to train the next generation of livestock breeders. Current curricula focus on genetics basics, but do not fully broach bad habits of beef cattle breeding. The deficit model of science communication has not proven effective. Rather, we will build our education model on the Unified Theory of Acceptance and Use of Technology. Thus, our curriculum will focus on:Trust and effectiveness of beef breeding best practices and technologies.Simplicity of using selection indexes and genomic technologies.The profit and sustainability outcomes of using best practices and technology.We will build a youth curriculum of 4 modules which will include instruction in applied and practical beef breeding.As part of this project we will sponsor a national youth essay contest. Participants will be asked to answer the prompt "What does it mean to be a beef breeder in the 21st century".We will construct an undergraduate curriculum covering the same types of topics as the youth curriculum (see above), but we will design the curriculum so that it is appropriate for a university setting. Curriculum will be to meet the outlined learning outcomes of: 1) reinforcement of animal breeding concepts; 2) determination of animal breeding best practices under different scenarios; 3) improved written and oral communication skills; and 4) experience with different social media outlets for information dissemination.In order to impact the social influence aspect of beef breeding best practice and technology adoption and increase student involvement, each student enrolled in the course will be required to post one item regarding beef breeding on the social media platform of their choice.Each student will also be required to write an essay for the course on a relevant beef breeding topic, helping them to think deeply about beef genetics and develop communication skills. Essays will be published on the A Steak in Genomics blog. Students will give an oral presentation summarizing their essay to improve oral communication skills.Collaborator Smith and co-workers developed a successful reproductive management internship program. As part of the internship, the students will attend 4 beef genetics training sessions. Furthermore, interns will act as genetic consultants to the participating beef farms and ranches.How results will be analyzed, assessed, and interpreted Importantly, the research statistics use null and alternative hypotheses allowing us to produce robust, significant, biologically meaningful results. We will create genomic predictions using only the SNPs identified as involved with local adaptation to increase signal and reduce noise. Genomic predictions will be assessed using common measures of accuracy and predictive ability.The overall evaluation will seek to determine: 1) Knowledge gained from curriculum and internship training, 2) How teachers and students perceive the efficacy and effectiveness of beef breeding curriculum, 3) Changes in student learning before and after engaging in the curriculum, 4) Perceptions of the efficacy/effectiveness of the breeding curriculum, 5) Student outreach for beef breeding curriculum, 6) Efficacy/effectiveness of breeding internship program.

Progress 03/01/16 to 02/28/17

Outputs
Target Audience:Using curriculum development, internships, essay contests, popular press, and social media we will reach a broad audience. That audience will include breed associations, allied industry partners, extension professionals, farmers, ranchers, graduate students, undergraduate students, and youth interested in beef cattle. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?This grant has helped train four PhD students, one Masters student, and one undergraduate student. How have the results been disseminated to communities of interest?In the first year of the grant, PD Decker gave 21 presentations to approximately 1,000 people on beef genetics, genomics or local adaptation. Highlights include presentations at the GeneSeek Innovation Seminar, American Simmental Association Fall Focus, Beefmaster Breeders United Convention, and Montana Stockgrowers Convention. We have published a fact sheet titled "Hair shedding scores: A tool to select heat tolerant cattle" on eBEEF.org, an eXtention Community of Practice (http://articles.extension.org/pages/74069/hair-shedding-scores:-a-tool-to-select-heat-tolerant-cattle). We also keep producers and industry professionals aware of project developments through PD Decker's blog, A Steak in Genomics. A Steak in Genomics had 11,349 page views in the first year of the grant. What do you plan to do during the next reporting period to accomplish the goals?YEAR TWO: Objective 1. In year two, we will finish creating our pipeline to impute SNP assay genotypes to tens of millions of variants from whole-genome resequencing data. Our graduate research assistants have already demonstrated the ability to develop computer pipelines. We will also complete selection scans using FLK, TreeSelect, and associations with environmental variables. Most of the effort in year two will be towards interpreting the results of these analyses. YEAR TWO: Objective 2. We will finish up analyses looking at which traits make cows successful or unsuccessful and look for gene-by-environment interactions with phenotypic records of these traits. We will also use variance analysis to look for traits with a strong genetic-by-environment variance term. Researchers at the University of Missouri and Texas A&M will also focus on gene-by-environment genome-wide association analyses in year two. One of the results of the cow failure/success outcome in year one has helped us identify fat thickness as an important trait to analyze for gene-by-environment associations. In the second half of year two, we will perform genome-wide association analyses of hair shedding. Finally, we will begin creating region-specific genomic predictions of production traits strongly affected by the environment. YEAR TWO: Objective 3. We will analyze and interpret the results of our producer survey. This will not only impact the design of our youth curriculum, but will also help refine our undergraduate curriculum. Design of youth curriculum will begin in earnest in year two. We will also test and refine this curriculum in year two. In the summer of year two, we will also have the opportunity to teach our undergraduate curriculum at South Dakota State University. We will also train our second set of interns. Finally, we will published the results of our first essay contest and in December through February of year two we will deploy our second essay contest.

Impacts
What was accomplished under these goals? Objective 1 We have downloaded and analyzed data from the PRISM database. In our exploratory analyses, we realized that when creating geographic clusters, many of the climate measures provided redundant information. In our analysis, we used 30-year Normal Annual Precipitation, 30-year Normal Annual Temperature, and Terrain Elevation. Using three clusters was the most strongly support from k-means analysis. In order to compare with historical analyses and to have more homogenous regions, we also clustered the United States into nine regions based on these three variables. We can match data to its geographic location either via its address or via zip code. We have also created a computer pipeline in which we can process, quality control, phase and merge large genotype data sets automatically in a manner of minutes. We have already run the genotype data from the American Simmental Association through this pipeline. In three minutes and eighteen seconds, we processed seven assays, which included 11,418 samples. It takes between 15 to 50 seconds to process an individual assay. We have started to run selection scans with the 10,935 Simmental samples genotyped at 919,968 SNPs that passed our quality filters and imputation steps. We first run principal component analysis. This exploratory analysis allows us to make sure no errors occurred during preprocessing of the data and there are no anomalies in the data. For instance, if the allele coding is wrong for one of the assays, we will spot this issue when visualizing the principal component analysis. Further, the principal component analysis identifies population and family structure within the data. The EIGENSOFT software package contains a program that can take the SNP loadings from the principal component analysis and test whether the SNPs have diverged in allele frequency more than expected due to drift, and are thus under selection. These loci are under selection between different breeds (Angus versus Simmental) or between different families within the Simmental breed. If we find overlap between these between-family selected variants and the between-region selected variants, we know that we need to be cautious in assigning a local adaptation function to these variants, as they may simply be family differences. Objective 2 We have completed the initial analysis comparing cows that successfully reproduce for three consecutive years versus those that fail to rebreed after having one calf with data from the American Hereford Association. Birth weight (BW), calving ease direct (CED), and calving ease maternal (CEM) which are measures of dystocia (calving difficulties) are associated with sustained reproductive success. If a cow's reproductive tract is damaged during parturition, it is more difficult for her tract to heal and for her to rebreed in a timely manner. Thus, these results agree with our current understanding. However, we see cattle producers put considerable emphasis on calving ease direct, but these results suggest that calving ease maternal is just as important. The association between maternal milk and sustained reproductive success may be an effect of breeders selection decisions for increased milking potential of their cows. We often hear beef breeders discussing the selection of "easier fleshing" cattle. We interpret the fat thickness EPD's association with sustained reproductive success to be a quantitative measure of "easier fleshing". Although fat thickness is typically considered a carcass trait, these results suggest that it may have an important impact on a cow's ability to store energy as fat. However, we did not see any regional trends when we plotted the difference in average EPD values between the failure and success cows. This implies that there is no difference is the additive genetic merit (EPD) between cows in different regional environments. However, national cattle evaluations used to estimate EPDs only model additive genetic effects and any nonadditive or genetic interaction effects add to the residual variance. Thus, gene-by-environment interactions would not be included in the EPDs and would not be observed. However, if we plotted phenotypic differences (rather than EPD differences), we may observe regional differences. Through our team's online presence (A Steak in Genomics) and with help from our breed association partners, we have recruited approximately 8,000 cattle to participate in our hair shedding research. We have created what we believe is a win-win situation. Beef breeders collect hair shedding scores on their cattle for three years in a row and send a DNA sample (either hair card or blood card) to the University of Missouri. We then send these samples off for genotyping, and use the hair shedding data and genotypes to perform a genome-wide association study of hair shedding. We also share these genotypes with the participating breed associations, and these breed associations use the data to produce genomic-enhanced EPDs. Thus, by participating in the project, the beef producers get free genomic-enhanced EPDs. We have already received the first year of hair shedding scores, DNA cards, and genotyping at GeneSeek has been completed for 2,186 of these samples. We continue to work with producers to complete submission of hair shedding scores and DNA samples. We are steadily inventorying these samples at the University of Missouri and submitting them for genotyping at GeneSeek. Objective 3 In order to create a youth curriculum that accurately corrects misconceptions in the beef industry, we first needed a clear picture of the attitudes and practices of beef breeders. We have created a survey that is a stable, reliable, and valid instrument to capture attitudes and barriers toward EPD use. The survey was developed in the fall of 2016 and was tested with producers at a regional field day in November to estimate reliability on the summated constructs and to gain feedback. In December and January, we contacted a group of agriculture teachers who were beef producers. We conducted a test-retest (same survey, two different times) with those producers to estimate the stability of the non-summated constructs and again asked for feedback. The survey is now completely developed and we are in the process of launching response collection. This survey will provide insight into the barriers of EPD use and will inform many decision makers and industry professions on how to help beef producers. We launched our first essay contest. We secured sponsorships from GeneSeek and Zoetis so that we could offer a first prize of $500, a second prize of $300, and a third prize of $200. The blog post (http://blog.steakgenomics.org/2016/12/essay-contest.html) announcing the essay contest has received 1,060 page views. Twenty youth have submitted essays to the contest, answering the prompt "What does it mean to be a beef breeder in the 21st century?" Members of the grant team, advisory board members, and extension professionals will judge essays. The winning essay will be published in one of BEEF Magazine's online newsletters and the second through fifth place essays will be published on A Steak in Genomics. Development has started on our undergraduate curriculum. The syllabus and lesson plans for three of the nine modules have been created. The curriculum will be taught for the first time the summer of 2017 at South Dakota State University. In 2016, we supported our first cohort of F.B. Miller interns. These interns were taught the basics of how EPDs work, why selection based on EPDs is superior to phenotypic selection, real-world examples of increased production from using EPDs, the advantages of genomic enhanced EPDs, maximizing profit by selecting animals using economic selection indexes, the simplicity of using selection indexes, and practice in selecting artificial insemination sires using selection indexes.

Publications

Type: Journal Articles Status: Published Year Published: 2016 Citation: Taylor, Jeremy F., Kristen H. Taylor, and Jared E. Decker. "Holsteins are the genomic selection poster cows." Proceedings of the National Academy of Sciences (2016): 201608144.
Type: Journal Articles Status: Published Year Published: 2016 Citation: Taylor, Jeremy F., Lynsey K. Whitacre, Jesse L. Hoff, Polyana C. Tizioto, JaeWoo Kim, Jared E. Decker, and Robert D. Schnabel. "Lessons for livestock genomics from genome and transcriptome sequencing in cattle and other mammals." Genetics Selection Evolution 48, no. 1 (2016): 59
Type: Theses/Dissertations Status: Published Year Published: 2016 Citation: Wilson, Miranda. "Multivariate genome-wide association studies and genomic predictions in multiple breeds and crossbred animals." MS thesis., University of Missouri--Columbia, 2016.
Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Wilson, Miranda L, Robert D. Schnabel, Robert Weaber, Jeremy F. Taylor, and Jared E. Decker. "Using Haplotype Based Models for Genomic Predictions in Crossbred Animals." In The Allied Genetics Conference. Orlando, FL. July 13-17, 2016.
Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Whitacre Lynsey K., Mark L. Wildhaber, Gary S. Johnson, J.M. Downs, T. Mhlanga-Mutangadura, Vernon M. Tabor, D. Fenner, and Jared E. Decker. "Genomic Variation and Population Structure of the Threatened Neosho Madtom (Noturus placidus)." In The Allied Genetics Conference. Orlando, FL. July 13-17, 2016.
Type: Websites Status: Published Year Published: 2016 Citation: A Steak in Genomics. http://blog.steakgenomics.org/
Type: Websites Status: Published Year Published: 2016 Citation: eBEEF.org, Beef Genetics eXtension Community of Practice.
Type: Other Status: Published Year Published: 2017 Citation: Decker, Jared E. and Jane Parish. Hair shedding scores: A tool to select heat tolerant cattle. eBEEF.org, an eXtention Community of Practice http://articles.extension.org/pages/74069/hair-shedding-scores:-a-tool-to-select-heat-tolerant-cattle.