Genomic Advances to Increase Feed Efficiency & Carcass

Quality of Grow Finish Pigs

Graham Plastow

Livestock Gentec, Dept. of Agricultural Food and Nutritional Science,

University of Alberta

Banff Pork Seminar 2019

Who Are We?

Livestock Gentec is a not-

for-profit centre developing

genomic technologies for

the Livestock Industry.

Canadian and International

• Swine, Beef, Dairy, Other species

• Research and Industry Populations

• Feed Efficiency, Quality, Health

– Climate Change

– Feeding the growing populationMeeting the demand for animal protein

– Antimicrobial Resistance

• Omics and other technologies that address our priorities

Past Improvement

• Feed efficiency

1972

2007

FCR: 3.8

FCR: 2.6

220 pounds

275 pounds

836 pounds

715 pounds

32%

David Casey, BPS 2010

Advances in Pork Production 21: 289-294

330 lbs of feed x 1,000 market hogs =

165 Tons of Feed Saved!!!!!

65% from

lean growth

selection

35% from measuring feed intake

David Casey, BPS 2010

Past Success

1980 N. Carolina State University

1980 2005

Images courtesy of T. See

Source: Fix et al. 2010 Livestock Sci. 128:108-114 ;

Ludu and Plastow 2013 Genome 56: 556-566.

Past Success

1980

How were these results achieved?

Candidate Genes and WGA

Candidate genes

Is there still a need for candidate gene

approaches in the era of genome-wide

association studies?

Stefan Wilkening, Bowang Chen, Justo Lorenzo

Bermejo, Federico Canzian

Genomics 93 (2009) 415–419

“the black box of quantitative genetics”

Robert

Bakewell

“the whole of

the art

is in chusing the best males to the best females”

1725-1795

Candidate Genes and WGA

Candidate genes

Is there still a need for candidate gene

approaches in the era of genome-wide

association studies?

Stefan Wilkening, Bowang Chen, Justo Lorenzo

Bermejo, Federico Canzian

Genomics 93 (2009) 415–419

“the black box of quantitative genetics”

∆G = h2 * sd * a * (i/t)

“the breeders equation”

Why Genomics Now?

• In 50 years the world will need 100% more

food than we produce today.– With less land, water, and energy.

– With the impact of climate change.

Meat on the Menu

• Global food demand

forecast, 1961=100.

• Based on The Economist

with data from the Food and

Agriculture Organization.

• 3 billion people trying to

move into the middle class in

emerging economies will

drive meat demand.

MeatDairyCerealsStarchy Roots

Why

Genomics?

“the black box of quantitative

genetics”

“Any sufficiently advanced technology is indistinguishable from magic” Arthur C Clarke

Where does genomics help?

•“Challenging Traits” or phenotypes•Difficult and expensive to measure• Lowly heritable• Sex-limited • Expressed late in life

∆G = h2 * sd * a * (i/t)

“Challenging Traits”

Why Genomics?

• Parentage or the origin of a piece of meat

• To identify carriers of a genetic disease

• To manage “genetic health”/diversity

• To determine the genetic potential of an animal at birth

• E.g.

– The carcass grade of a sl. pig

–A sire’s ability to breed prolific daughters

–An animals ability to tolerate environmental or disease challenge (robustness)

– Feed Efficiency

Genomics ACGT• How many letters (nucleotides) in the genetic

code?– Answer 4

• How many letters in an animal’s genome?– Answer:

The Halothane gene

A revolution in animal breeding

Although the “halothane gas test” was useful it could not detect carriers of this recessive “stress gene”.

Identification of a Mutation in Porcine

Ryanodine Receptor Associated with

Malignant Hyperthermia

Junichi Fujii, Kinya Otsu, Francesco

Zorzato, Stella De Leon, Vijay K.

Khanna, Janice E. Weiler, Peter J.

O'Brien and David H. MacLennan

Science 253: 448-451 1991

University of Toronto & the University of

Guelph

Canadian Livestock Genomics

A world first 1991

• The world’s first QTN applied over millions of pigs worldwide

• HAL1843™

• One in 3 billion base pairs

• Improved pig welfare

• Improved pig meat quality (muscle pH)

• 1992 Hal1843™ Approx $50 per animal (datapoint)

Cost of Genotyping

C N S I I D P L I Y

C N S I I N P L I Y

NH2 COO

H

Transmembrane

domains

I II III IV V VI VII

Allele 1

homozygote

sequence

Allele 2

homozygote

sequence

293 295 297 299 300a

b 1/1 2/2 1/2

542

466

MC4R mutation and TestKim et al., 2000. Mammalian Genome 11:131-135

Comparison Between

Genotypes

Backfat 1.1 mm less

Feed intake -.17 kg/day

Feed efficiency -.09

Days to market 2.8 more

Benefits Producer Consumer

2/2 1/1 1/2

Genotype 11 12 22 p

ADFI kg/day

1.94 2.03 2.11 <0.01

Days to 110kg

167.9 166.9 164.6 <0.001

MC4R mutation and TestKim et al., 2000. Mammalian Genome 11:131-135

• Commercial trials in UK

N ccw P2 L%

Random group 1833

72.5kg

11.8mm

57.4%

Lean genotype

2137

73.5kg

10.7mm

58.8% Improvement -1.1mm +1.4%

Demonstrating value

Reasons for limited use of

genetic markers in industry

breeding programs

•Few markers were available

• and they explained a fairly limited % of variance

•Marker effects often not consistent or not transferable

to commercial populations

•High genotyping costs

Slide courtesy of Jack Dekkers, ISU

• 1992 Hal1843™ Approx $50 per animal (datapoint)

• 2007 Illumina Bov50kSNP (co-developed at UofA)

Approx $250 per animalor <0.5c per datapoint

A 10,000 fold reduction per datapoint

Cost of Genotyping

• 1992 Hal1843™ Approx $50 per animal (datapoint)

• 2007 Illumina Bov50kSNP (co-developed at UofA)

Approx $250 per animalor <0.5c per datapoint

A 10,000 fold reduction per datapoint

Cost of Genotyping

Today1/100th cent

Tomorrowthe $100 genome

How to use high-density SNP data?

Conduct Association Analysisfor each SNP

Genotype large # of

Individuals

for large numbers of SNPs

+ collect their phenotypes

Slide courtesy of Jack Dekkers, ISU

How to use high-density SNP data?

Conduct Association Analysisfor each SNP

Use only significant SNPs

for MAS

Genotype large # of

Individuals

for large numbers of SNPs

+ collect their phenotypes

Problem:• Small effects are missed

Slide courtesy of Jack Dekkers, ISU

Carcass and Pork Quality

～4000 piglets born alive

100 DurocBoars

400 F1 SowsLW*Landrace

~ 2000 finish pigs slaughter

Illumina pig 60K

SNPs Panel

Non-genetic analysis

Genetic parameter estimation

GWAS

Genomic prediction

• Pedigree• Growth traits• Ultrasound tests

> 50 measurements• Carcass components• Meat quality• Crude fat

components• Panelist sensory tests

42,721 SNPs for 1976 F2 pigs. Bayes B (GenSel), 1 Mb window. Factors: population, sex & contemporary group (year, group, slaughter batch).

Traits Regions (SSC_Mb) No. of SNPs % GenVar. % PheVar.

Hot carcass weight 7_102 20 4.84 0.90

pH24 15_133 26 9.26 1.30

Peak shear force

2_4:5 37 7.14 1.43

2_109 17 6.53 1.31

17_20 18 1.63 0.33

Backfat depthX_139: 140 20 16.41 0.92

1_176: 179 56 62.20 3.48

Crude fat 1_176: 178 38 11.83 0.11

Results

Zhang et al. BMC Genetics 2015, 16:33.Zhang et al., PLoS ONE 2016, 11(2): e0145082.

Expectation

Reality

Adaptive Lasso

Yang et al 2017 Can. J. Animal Science 97:721–733.

Residual Feed Intake

Selection Lines

Jack Dekkers

Department of Animal Science

• ~35% of differences in feed efficiency

are independent of growth and backfat = RFI

• RFI is a heritable trait and responds to selection

• Pigs that are selected for increased efficiency based on RFI do NOT

have greater behavioral, physiological, and health problems, or are

more susceptibility to stress and disease?

• In contrast pigs selected for efficiency based on RFI:

– Are calmer and less fearful

– Are less responsive to physiological stress

– Are less affected by PRRS infection

– Appear to have a more effective efficient immune response

– Are not more affected by heat stress

– Are better able to withstand the stresses of gestation and lactation

– Are better able to direct resources where needed – greater “metabolic flexibility”

– Are less affected by environmental differences

Conclusions

Residual Feed Intake

Selection Lines

Develop tools to improve feed efficiency

Genetic Markers

Selection

Yorkshire

Large White

# 2

011

-68004-3

0336

Low RFI line Hi RFI line

IlluminaPorcineSNP60

BeadChip

CONCLUSIONS – GWAS

• In general, few strong associations

• Traits appear highly polygenic, in particular RFI

• Largest effects for MC4R on all traits, except RFI

GENOMIC PREDICTION

Solution: Genomic selection

Genetic Evaluation using high-density SNPs

• All SNPs are fitted simultaneously, i.e. 60,000 vs. 1 at a time

• SNP effects are fitted as random vs. fixed effects

• enables all SNPs to be fitted simultaneously

• shrinks SNP effect estimates to 0 depending on evidence from data

Meuwissen et al. 2001 Genetics

Estimates of SNP effects bk

yi = m + S bk gik + eiSNP k

Implemented using a variety of

Bayesian methods (Bayes-A, -B, -C, C-p)

Or by using genomic vs. pedigree

relationships in animal model BLUP (GBLUP)

^ Use to estimate

breeding value of new

animals based on

genotypes alone

Genomic EBV = S bk gik^

^

Genomic EBV = S bk gik^

Example

Genomic EBV with 3 SNPswith estimated effects (b for # A alleles) of:

+10 for SNP 1

+ 5 for SNP 2

–10 for SNP 3

Estimate marker effectsGenotype

for >50,000 SNPs

Phenotype

Genotypefor >50,000

SNPs

Predict BV from marker genotypes at

early age

Genotypefor >50,000

SNPs

Tra

inin

g d

ata

Predict BV from marker genotypes at

early age

Genomic selectionGenetic Evaluation using high-density SNPs

Meuwissen et al. 2001

Genus Annual Report 2015 p16

This shows the impact of genomic selection in pigs with the largest effect coming from improving “a” in the breeders’ equation.

http://www2.topigsnorsvin.com/l/97812/2016-06-16/3rly3y

Both report annual genetic progress at the customer level has increased by over 50%

Omics ProfilingGenome, Transcriptome, Proteome, Metabolome, Metagenome

Selection Management Sorting

Biomarkers

Phenomics - Automated Data Collection e.g. Wearables

Laboratory or Pen-side

Feed efficiency, Disease Resilience, Animal Welfare, Product Quality, Food Safety

A Vision of Precision Livestock Agriculture

N Cook, Alberta Agriculture & Forestry, Lacombe

J-P Laforest, Université Laval F Fortin, CDPQ, QC

Goals

•How do animals change over time?

•How do they respond to different environments/nutrition?

•How do they respond to stress or disease?

•How do the different omes interact?

•How can these signals allow us to “talk with the animals” and improve production and sustainability?

Resilience

StressSocial

Environmental

Metabolic

Immunological

Human Interaction

Productivity loss

Measure using FI

phenotypes

Regrouping after

weaning, to finishing, to

slaughter

Temperature, humidity, light,

dust, gases, ammonia,

Myotoxins, sound, etc

Food/water restriction or deprivation

Disease, vaccination

Snaring,moving

bleeding, weighing, tattooing

NOT Exclusive (Interactions)

Austin Putz after

Genomic evaluation using Crossbred Data

Crossbred

Performance

in Field

Environment

Purebred

Performance

in High Health

Environment

rpc < 1

Dr. Jack Dekkers

Dr. Bob Kemp

Dr. Graham Plastow

Genomic Selection Applications

• within breed (wbGS), where accuracy is obtained by maintaining huge within-breed reference populations;

• or across breed (abGS) where accuracy is obtained from across-breed reference populations and high-density GS methods that focus on causative genomic regions.

Meuwissen, Hayes and Goddard (2016) Animal Frontiers, 6: 6–14.

Solutions

• Identify causative mutations – work consistently across populations

Solutions

• Identify causative mutations – work consistently across populations

• Combining coding sequence variants with dense SNP markers in Bayes RC

• Increased power to detect causal variants and increased accuracy of genomic prediction

• Most apparent in populations not closely related to the reference population

Finding Causative Mutationsfrom genotype to phenotype

Conclusions

• Genomic Selection provides a tool for increasing rate of genetic improvement at the commercial level

• It provides a mechanism to incorporate new traits (those that have been expensive and difficult to measure)

• New approaches to phenotype collection to enable full application especially at the commercial level

• The “internet of things” and machine learning point to new opportunities for successfully mining the resulting “big data”

54

Acknowledgements

• Jack Dekkers for sharing slides on genomic selection and the ISU RFI selection lines.

• All of our collaborators around the world.

• This work at the University of Alberta was supported by the following organizations: Alberta Innovates, Alberta Agriculture and Forestry, PigGen Canada, Genesus, Hypor, Topigs Norsvin, Genome Alberta, Genome Canada, Swine Innovation Porc, MITACS, and NSERC.

Thanks!