Commercial application of marker- and gene-assisted selection in livestock: Strategies and lessons 1,2 J. C. M. Dekkers 3 Depar tment of Animal Science , Iowa State Univer sity, Ames 50011 -3150 ABSTRACT: During the past few decades, advances in molecular genetics have led to the identification ofmul tip le gen es or geneti c mar ker s associate d wit h genes tha t aff ect tra its of int erest in live sto ck, inc ludinggenes for single -gene tra its and QTL or genomi c region s tha t affe ct qua nti tat ive traits. This has pro vided oppor- tunities to enhance response to selection, in particular for traits that are difficult to improve by conventional selection (low heritability or traits for which measure- ment of phenotype is difficult, expensive, only possible lat e in life , or not pos sible on sel ect ion can didate s). Examples of genetic tests that are available to or used in indus try programs are document ed and classified into cau sat ive mut ati ons (direct mar ker s), linked mar k- ers in population-wide linkage disequilibrium with the QTL (LD markers), and linked markers in population- wide equilibrium with the QTL (LE markers). In gen- eral, although molecular genetic information has been Key Words: Breeding Programs, Markers-Assisted Selection, Quantitative Trait Loci 2004 American Society of Animal Science. All rights reserved. J. Anim. Sci. 2004. 82(E. Suppl.):E313–E328 Introduction Substantial advances have been made over the past decades through the application of molecular genetics 1 Information was provid ed by the following people and organiza- tions: D. Funk (Americ an Breeder s Ser vic e), J. McEwan (Ag Re- search), J. Hetzel (Genetic Solutions), M. Cowan (Genetic Visions), N. Buys (Gentec), E. Mullaart (Holland Genetics), J. Fulton and J. Arthur (Hy-Line Int.), R. Spelman (Livestock Improvement Com- pany ), G. T. Nieuwhof (Meat and Livestock Commission) , M. Lohuis andJ. Veen hui zen(Monsa nto), G. Plas tow(Sygen Int .),E. Knol (TOP- IGS), S. Dominik (CSIRO), R. Fernando (Iowa State Univ.), J. Gibson (ILRI), B. Hayes (Victorian Inst. Anim. Sci.), M. Rothschild (Iowa Sta te Uni v.) , S. Sch mutz (Un iv. Sas kat chewan ), andK. Weig el (Un iv. of Wisconsin). Financial support from the State of Iowa, Hatch and Multi-State Research funds. 2 This article was presented at the 2003 Joint ADSA-ASAS-AMPAmeeting as part of the Breeding and Genetics symposium “Molecu- lar Genetics.” 3 Corresp ondenc e: 225CKildee Hall (phon e: 515- 294-7 509;fax: 515- 294-9150; e-mail: [email protected]). Receive d Octob er 16, 2003. Accepted February 11, 2004. E313 used in industry programs for several decades and is growing, the extent of use has not lived up to initial expectations. Most applications to date have been inte- grated in existing programs on an ad hoc basis. Direct markers are preferred for effective implementation ofmar ker -assisted sel ect ion , followed by LD and LE marker s, the latte r requi ring within -family analysis and selection. Ease of application and potential for ex- tra-genetic gain is greatest for direct markers, followed by LD markers, but is antagonistic to ease of detection, which is greatest for LE markers. Although the success of these applications is difficult to assess, several have been hampered by logistical requirements, which are substantial, in particular for LE markers. Opportuni- tie s for t he use of mo lec ula r inf ormati on exis t, but the ir succe ssful implementat ion requi res a compr ehens ive int egrate d str ate gy tha t is clo sel y ali gned wit h bus ine ss goals. The current attitude toward marker-assisted se- lection is therefore one of cautious optimism. in the identification of loci and chromosomal regions that contain loci that affect traits of impor tanc e in livestock production (Andersson, 2001). This has en- abled opportunities to enhance genetic improvement programs in livestock by direct selection on genes or genomic regions that affect economic traits through mar ker-assis ted selection and gene int rogression (Dekkers and Hospital, 2002). To this end, many theo- retical studies have been conducted over the past sev- era l dec ade s to evaluate strate gie s for the use of mol ec- ular genetic information in selection programs. The extra responses to selection that have been predicted by several studie s (e. g., Meuwissen and Goddar d, 1996) have resulted in great optimism for the use ofmolec ular genetic infor mati on in industry bree dingprograms. Objectives of this paper are to assess the extent to which and in which ways marker and gene information has been use d in co mmercial livestock im- provement programs, to assess the successes and limi- tations that have been experienced in such applica- tions, and to discuss strategies to overcome these limi- tations. I will start with a discussion of the principles
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
used in industry programs for several decades and isgrowing, the extent of use has not lived up to initialexpectations. Most applications to date have been inte-grated in existing programs on an ad hoc basis. Directmarkers are preferred for effective implementation of marker-assisted selection, followed by LD and LEmarkers, the latter requiring within-family analysisand selection. Ease of application and potential for ex-tra-genetic gain is greatest for direct markers, followedby LD markers, but is antagonistic to ease of detection,which is greatest for LE markers. Although the success
of these applications is difficult to assess, several havebeen hampered by logistical requirements, which aresubstantial, in particular for LE markers. Opportuni-ties for the use of molecular information exist, but theirsuccessful implementation requires a comprehensiveintegrated strategy that is closely aligned with businessgoals. The current attitude toward marker-assisted se-lection is therefore one of cautious optimism.
in the identification of loci and chromosomal regions
that contain loci that affect traits of importance inlivestock production (Andersson, 2001). This has en-
abled opportunities to enhance genetic improvementprograms in livestock by direct selection on genes orgenomic regions that affect economic traits through
marker-assisted selection and gene introgression(Dekkers and Hospital, 2002). To this end, many theo-retical studies have been conducted over the past sev-
eral decades to evaluate strategies for the use of molec-ular genetic information in selection programs. The
extra responses to selection that have been predicted
by several studies (e.g., Meuwissen and Goddard,1996) have resulted in great optimism for the use of
molecular genetic information in industry breeding programs. Objectives of this paper are to assess the
extent to which and in which ways marker and geneinformation has been used in commercial livestock im-provement programs, to assess the successes and limi-
tations that have been experienced in such applica-tions, and to discuss strategies to overcome these limi-
tations. I will start with a discussion of the principles
Figure 1. Response lost over one, two, and three generations and in cumulative discounted response over 10generations at 10% interest (CDR10) from tandem vs. index selection on a QTL with an initial frequency of thefavorable allele of 0.1, and additive effect a (in genetic standard deviations, σg). Selection is for a quantitative traitwith selected proportions 10 and 25%, and accuracies of polygenic EBV of 0.8 and 0.5 for sires and dams, respectively.
tion of some parents that do not carry the target allele,
overall response is expected to be greater. In particu-
lar, if multiple genes or QTL regions must be intro-
gressed simultaneously, the requirement that selected
parents carry the target allele for all QTL is infeasible
in livestock and not necessary for successful introgres-
sion (Chaiwong et al., 2002).
Industry Application of Marker-Assisted Selection
Examples of Commercially Availableor Utilized Genetic Tests
A nonexhaustive summary of gene or marker tests
that are currently available or used in commercial
breeding programs is given in Table 1, with tests cate-
gorized by the type of trait and the type of marker. A
substantial number of genetic tests are available.
Some applications of selection for individual genes oc-
curred prior to the era of molecular genetics, including
selection on observable genetic defects and appear-
ance, the halothane test as a physical test for the RYR
gene, and use of the B-blood group as a physiological
LD marker for selection for disease resistance in poul-
try, which started in the 1960s (Hansen et al., 1967;
Hansen and Law, 1970; J. Arthur, Hy-Line Int., Dallas
Center, IA, personal communication). Several tests are
used for within-house selection only (e.g., PICmarq
markers used by the Pig Improvement Co., G. Plastow,
Sygen Int., Berkeley, CA, personal communication),
whereas others are available through commercial gen-
otyping services. To date, the majority of publicly
available tests are for direct or LD markers.
Although there are a large number of scientific re-
ports on detection of QTL for livestock (e.g., Bidanel
and Rothschild, 2002; Bovenhuis and Schrooten, 2002;
Hocking, 2003), most of these were identified in experi-
mental populations using crosses between breeds or
lines (Andersson, 2001). Such studies identify QTL
that differ in frequency between breeds but results
cannot be used directly for selection within breeds.They can, however, provide an important stepping
stone for identification of LD markers for QTL that
segregate within breeds using positional candidate
gene approaches (Rothschild and Soller, 1997). An ex-
ample is the detection of additional mutations in the
RN gene, known as PRKAG3, which have been found to
segregate in commercial lines of pigs using a positional
candidate gene approach in a QTL region that was
detected in a cross between two commercial breeds
(Ciobanu et al., 2001). The use of experimental crosses
explains the abundance of the use of direct and LD
Table 1. Examples of gene tests used in commercial breeding for different species (D =dairy cattle, B = beef cattle, C = poultry, P = pigs, S = sheep) by trait category and typeof marker
Linkage Linakge
disequilibrium e quilibrium
Trait category Direct marker marker marker
Congenital defects BLAD (Da)
Citrulinaemia (D,Bb)
DUMPS (Dc)CVM (Dd)
Maple syrup urine (D,Be)
Mannosidosis (D,Bf )
RYR (Pg ) RYR (Ph)
Appearance CKIT (Pi) Polled (Bn)
MC1R/MSHR (P j,Bk,Dl)
MGF (Bm)
Milk quality κ-Casein (Do)
β -lactoglobulin (Do)
FMO3 (Dp)
Meat quality RYR (Pg ) RYR (Ph)
RN/PRKAG3 (Pq) RN/PRKAG3 (Pr)
A-FABP/FABP4 (Ps)
H-FABP/FABP3 (Pt)
CAST (Pu
, B v
)>15 PICmarq (Pw)‡
THYR (Bx)
Leptin (By)
Feed intake MC4R (Pz)
Disease Prp (Saa) B blood group (Cbb)
F18 (Pcc) K88 (Pdd)
Reproduction Booroola (See) Booroola (Sff )
Inverdale(Sgg ) ESR (Phh)
Hanna (Sii) PRLR (P jj)
RBP4 (Pkk)
Gro wth and composition M C4R (Pz) CAST (Pu) QTL (Pll)
IGF-2 (Pmm) IGF-2 (Pnn)
Myostatin (Boo) QTL (Bpp)
Callipyge (Sqq) Carwell (Srr)
Milk yield and composition DGAT (Dss) PRL (Dtt) QTL (Duu)
GRH (D vv)
κ-Casein (Do)
aShuster et al. (1992); bDennis et al. (1989); cSchwenger et al. (1993); dBorchersen (2001); eDennis andHealy (1999); f Berg et al. (1997), Leipprandt et al. (1999); g Fuji et al. (1991); hHanset et al. (1995); iMarklundet al. (1998); jKijas et al. (1998); kKlungland et al. (1995); lJoerg et al. (1996); mSeitz et al. (1999); nSchmutzet al. (1995); oMedrano and Aquilar-Cordova (1990), Rincon and Medrano (2003); pLunden et al. (2002);qMilan et al. (2000); rCiobanu et al. (2001); sGerbens et al. (1998); tGerbens et al. (1999); uCiobanu et al.(2004); vBarendse (2001); wG. Plastow (Sygen Int., Berkeley, CA, personal communication); xBarendse etal. (2001); yBuchanan et al. (2002); zKim et al. (2000); aaBelt et al. (1995); bbHansen et al. (1967), Hansenand Law (1970); cc Vogeli e t al. (1997), Meijerink et al. (2000); ddJørgensen et al. (2004); eeWilson et al. (2001);ff Lord et al. (1998); gg Galloway et al. (2000); hhRothschild et al. (1996); iiMcNatty et al. (2001); jj Vincent etal. (1998); kkRothschild et al. (2000); llM. Lohuis (Monsanto Co., St. Louis, MO, personal communication);mmGeorges et al. (2003); nnJeon et al. (1999), Nezer et al. (1999); ooGrobet et al. (1998); ppJ. Hetzel (GeneticSolutions, Brisbane, Australia, personal communication); qqFreking et al. (2002); rrNicoll et al. (1998);ssGrisart et al. (2002); ttCowan et al. (1990); uuSpelman et al. (1996), Arranz et al. (1998), Coppieters et al.(1998), Georges et al. (1995), Zhang et al. (1998); vvBlott et al. (2003).
‡Applies to both direct and linkage disequilibrium columns.
markers compared with LE markers for species such
as pigs, beef cattle, and poultry (Table 1). An alterna-
tive is to follow a breed-cross QTL analysis with an
LE QTL analysis within commercial lines in identified
regions, which has shown to be successful in pigs (Ev-
ans et al., 2003) and has been used in beef cattle (J.
Table 3. Requirements and opportunities for the implementation of linkage equilibrium(LE) vs. linkage disequilibrium (LD) vs. direct (D) markers
Requirement or Relative
opportunity requirements
QTL detection requirements LE < LD << D
Marker development LE < LD << D
Phenotyping and data structure LE >> LD ∼ D
Genome-wide analysis opportunities LE >> LD >> D
Within-line confirmation requirements LE >> LD > DRoutine genetic evaluation requirements LE >> LD > D
Phenotyping and data structure LE >> LD ∼> D
Genotyping LE >> LD ∼> D
Genetic evaluation models LE >> LD ∼> D
Implementation logistics LE >> LD > D
Genetic gain opportunities (for given QTL) LE < LD ∼< D
Marketing opportunities (patents, product differentiation) LE << LD < D
of markers. These comparisons, which will be further
discussed below, also provide insight into the reasons
for the extent of success and limitations that have been
experienced in different commercial applications of MAS.
Marker Development and QTL Detection. Require-
ments for marker development are least for LE markers
and greatest for direct markers (Table 3; Andersson
2001). Whereas LE markers can be random anonymous
markers, direct markers require identification of the
causative mutation, and LD markers require close link-
age with the causative mutation, either identified as
targeted candidate gene polymorphisms or by high-den-
sity marker maps. In addition, LE markers allow for
genome-wide analysis of QTL based on a limited num-
ber of markers at 15- to 50-cM spacings. Genome-wide
analysis is also possible for LD markers, but this willrequire a very dense marker map, depending on the
extent of LD in the population (e.g., Meuwissen et al.,
2001). The latter seems to be wide in livestock popula-
tions (Farnir et al., 2000; McRae et al., 2002), such that
informative markers every 1 or 2 cM may be sufficient
to detect most QTL.
Associations between direct or LD markers and traits
can be identified based on a limited number of pheno-
typed and genotyped individuals, without a specific pop-
ulation or family structure. Detection of QTL using LE
markers, however, requires the presence of LD that
extend over 20 or more centimorgan. Such LD can be
created by crossing lines or found within families inoutbred populations. The latter requires large numbers
of phenotyped individuals with a specific family struc-
ture (e.g., Weller et al., 1990). The same is true for
estimation and confirmation of LE vs. LD marker ef-
fects in other (outbred) populations (e.g., Spelman and
Bovenhuis, 1998), resulting in greater phenotyping and
genotyping requirements at this stage for LE markers,
in particular if the initial QTL detection was based on
a cross between lines.
Marker-Assisted Genetic Evaluation. Requirements
for integration of marker data in routine genetic evalua-
tion procedures are also much greater for LE than forLD or direct markers, both with regard to requirementsfor the number and which individuals that must bephenotyped and genotyped, and with regard to methodsof analysis (Table 3). Use of LE markers in an outbredpopulation requires the phenotyping and genotyping of selection candidates and/or their relatives becauseeffects must be estimated on a within-family basis. Theextent of family data needed depends on recombinationrates between markers and QTL. Less data will beneeded and can be from more distant relatives if recom-bination rates are low. Direct and LD markers requirethe genotyping of only selection candidates because es-timates of genotype effects can be obtained from priorinformation or from a sample of individuals that haveboth genotype and phenotype information.
Data from LE markers can be incorporated intoBLUP animal model genetic evaluations using the ap-proach of Fernando and Grossman (1989), by fitting random effects for each QTL and allowing for differentQTL effects within families. This method, or extensionsthereof, has been applied to several commercial situa-tions in dairy cattle (Boichard et al., 2002; Bennewitzet al., 2003; E. Mullaart, personal communication). Ap-plication of these procedures requires substantial modi-fication of existing animal model genetic evaluation pro-cedures, estimation of variance components, and exten-sive computing resources. Data from LD or directmarkers on the other hand, can be incorporated in ex-isting genetic evaluation procedures as fixed genotypeor haplotype effects (Van Arendonk et al., 1999). If notall animals are genotyped, which will be the case inpractice, marker data must be supplemented with geno-type probabilities, which can be derived using pedigreeand marker data (e.g., Israel and Weller, 2002). Never-theless, computational requirements for incorporating LD or direct markers in genetic evaluation are muchless than for LE markers. Genetic evaluation require-ments are slightly greater for LD than for direct mark-ers because LD markers require identification and anal-ysis of marker haplotypes and confirmation of marker-QTL linkage phases.
Figure 2. Components of an integrated system for the use of molecular genetic information in breeding programsfor marker-assisted selection (MAS).
breeding program of an integrated pig production enter-
prise. Detection of QTL and MAS on identified QTL
regions for a multitrait breeding goal and associated
genotyping costs and extra returns from the production
phase of the integrated enterprise were considered in
the economic assessment. They concluded that imple-
mentation of LE-MAS was feasible for the assumed
cost and price parameters. They also found that, in
particular if QTL detection was based on small sample
sizes, stringent thresholds should be setduring the QTL
detection phase such that genotyping costs during the
Figure 3. Integration of phenotypic and molecular data on polygenic and monogenic traits, including data on direct(D), linkage disequilibrium (LD), and linkage equilibrium (LE) markers, in a selection program that will meet businessgoals, using analysis tools to estimate breeding values (EBV), molecular scores, and genotypes (or genotype probabili-ties). Solid and broken arrows indicate the flow of information for polygenic and monogenic traits, respectively.
implementation phase arereduced and selection of false
positives is minimized. An economic analysis of intro-
gression of the Booroola gene into dairy sheep breeds
is given in Gootwine et al. (2001) and of MAS prese-
lection in dairy cattle in Brascamp et al. (1993), Mackin-
non and Georges (1998), and Spelman and Garrick
(1998).
Whereas Hayes and Goddard (2003) evaluated eco-
nomic returns from MAS from increased profit at the
production level, which is proportional to extra genetic
gain, most commercial breeding programs derive profit
Figure 4. Effect (%) of 50% preselection of young bulls for entry into a progeny-test program on genetic gain (meanEBV of the top 10% progeny-tested bulls) and market share (number of bulls in top 10% and top 1%). Preselectionis on an index of genotype for a QTL with additive effect a (in genetic standard deviations, σg) and a polygenic EBVthat has a correlation (r) with true polygenic breeding values among young bulls of 0.0 or 0.1.
from increased market share of breeding stock or germ-
plasm. In general, implementation of MAS will have a
greater impact on market share than on genetic gain.
An example is given in Figure 4, which evaluates the
effect of GAS preselection of young dairy bulls in a
competitive market. A deterministic model of a mixture
of two normal distributions to represent sons that re-ceived alternate QTL alleles from their heterozygous
sire was used. Extra response from preselection of sons
from heterozygous sires depends on the variation that
is still present among selection candidates for polygenic
EBV for the overall selection criterion, which is based
on pedigree information only. If stringent selection on
EBV has been applied to bull dams and bull sires, this
variation will be limited and the accuracy of polygenic
EBV to further differentiate selection candidates will
be small (Dekkers, 1992). In Figure 4, correlations esti-
mated with true polygenic breeding values among
young bulls of 0.0 and 0.1 are evaluated and prese-
lection is on an index of the MS and polygenic EBV.
For a QTL with a substitution effect of 0.3 genetic
standard deviations and a polygenic EBV accuracy of
0.0, preselection increased genetic gain of selected (top
10%) progeny-tested bulls by 7%, but the number of
bulls in the top 10 and 1% increased by 20 and 30%,
respectively (Figure 3). Mean EBV and market share
increased with effect of the QTL and decreased with
increasing accuracy of the polygenic EBV at the time
of preselection. Effects on market share were, however,
always greater than effects on mean EBV. This does
notimply that the economic feasibility of MAS is greater
in a competitive market than when returns are derived
from commercial production, as was evaluated by
Hayes and Goddard (2003). Economic feasibility not
only depends on the proportional increase in the objec-
tive, but also on the absolute returns associated with
a percentage increase in genetic gain vs. market share;
in fact, Brascamp et al. (1993) showed that economicreturns from increased market share were less than
from increased production for a preselection situation
similar to that considered here. Nevertheless, it is im-
portant that economic analysis is conducted in relation
to business and market realities and goals. Computa-
tional approaches using genetic algorithms (e.g., King-
horn et al., 2002) can be used to develop selection and
mating strategies based on multiple sources of informa-
tion, including markers that meet multiple business
goals and constraints. Weller (1994) provides further
discussion of alternative criteria to economically evalu-
ate alternative breeding programs.
Other Opportunities
Optimal implementation of MAS involves careful con-
sideration of alternative selection strategies, business
goals, and integration of molecular with other technol-
ogies (e.g., reproductive technologies following Georges
and Massey, 1991). Opportunities also exist to imple-
ment LD-MAS in synthetic lines, capitalizing on the
extensive disequilibrium that exists in crosses and their
power to detect QTL (Zhang and Smith, 1992). In addi-
tion, strategies must be developed to estimate gene ef-