Chapter 17 marker-assisted selection in fish and shellfish ... · marker information in...

Chapter 17

marker-assisted selection in fish and shellfish breeding schemes

Victor Martinez

Marker-assisted selection – Current status and future perspectives in crops, livestock, forestry and fish330

SummaryThe main goals of breeding programmes for fish and shellfish are to increase the prof-itability and sustainability of aquaculture. Traditionally, these have been carried outsuccessfullyusingpedigreeinformationbyselectingindividualsbasedonbreedingvaluespredictedfortraitsmeasuredoncandidatesusingan“animalmodel”.Thismethodologyassumes that phenotypes are explained by a large number of genes with small effectsand random environmental deviations. However, information on individual genes withmediumorlargeeffectscannotbeusedinthismanner.Inselectivebreedingprogrammesusing pedigree information, molecular markers have been used primarily for parentageassignmentwhentaggingindividualfishisdifficultandtoavoidcausingcommonenviron-mentaleffectsfromrearingfamiliesinseparatetanks.Theuseofthesetechniquesinsuchconventionalbreedingprogrammesisdiscussedindetail.

Exploiting the great biological diversity of many fish and shellfish species, differentexperimental designs may use either chromosomal manipulations or large family sizesto increase the likelihood of finding the loci affecting quantitative traits, the so-calledQTL,byscreeningthesegregationofmolecularmarkers.Usinginformationonidentifiedloci in breeding schemes in aquaculture is expected to be cost-effective compared withtraditional breeding methods only when the accuracy of predicting breeding values isratherlow,e.g.fortraitswithlowheritabilitysuchasdiseaseresistanceorcarcassquality.Oneoftheproblemsfacingaquaculture is thatsomeoftheresourcesrequiredto locateQTLaccurately,suchasdense linkagemaps,arenotyetavailable for themanyspecies.Recently, however, information from expressed sequence tag (EST) databases has beenused for developing molecular markers such as microsatellites and single nucleotidepolymorphisms(SNPs).Marker-assistedselection(MAS)orgenome-widemarker-assistedselection(G-MAS)usinglinkagedisequilibriumwithinfamiliesoracrosspopulationsarenot widely used in aquaculture, but their application in actual breeding programmes isexpectedtobeafertileareaofresearch.Thischapterdescribeshowgenomictoolscanbeused jointly with pedigree-based breeding strategies and the potential and real value ofmolecularmarkersinfishandshellfishbreedingschemes.

Chapter 17 – Marker-assisted selection in fish and shellfish breeding schemes 331

introduCtionThemaingoalsoffishandshellfishbreedingprogrammes are to increase the profit-ability and sustainability of productionenterprises,whilemaintaininggeneticvari-ability in theculturedstock.Traditionally,selective breeding has targeted traits suchasbodyweightthatcanbeeasilyimprovedusingmassselection.Relativelyfewstudieshaveanalysedothertraitsthatareeconomi-callyimportant.However,diseaseresistanceandcarcassqualityaretraitsthatarediffi-culttomeasureoncandidatesforselection,but have major effects on the productionefficiencyandprofitabilityofmanyspeciesinaquaculture.

When developing efficient breedingprogrammes, pedigree information isrequired to maximize effective populationsizes and to use information from rela-tivestoincreasetheaccuracyofpredictingbreeding values for all traits included inthe breeding objective. In most commer-cial applications, pedigree information islacking; therefore, markers can be used toinferrelatednessbetweenindividuals,withor without parental information. Severalissuesneedtobeconsideredonacase-by-case basis when applying such molecularinformationforincreasingtheprofitabilityofbreedingprogrammesinpractice.

For traits that are difficult to measureon candidates for selection, prediction ofbreedingvaluehastorelyonmeasurementson relatives. Under these circumstances,the accuracy of predicted breeding values(and thus, response) is lower than whenrecords are obtained directly on candi-datesforselection.Inaddition,there isanincreased probability of co-selecting rela-tives. It is especially for these traits thatmolecular markers that directly affect orare linked to quantitative trait loci (QTL)have been regarded as useful for marker-

assisted selection (MAS) or gene-assistedselection(GAS)programmes.

This chapter begins by discussing thestatus of “conventional” breeding pro-grammes, the challenges involved whenstarting such programmes for new spe-cies and the possibilities of incorporatingmarkerinformationin“conventional”pro-grammes. An outline is then provided ofthemolecularmarkersdevelopedforaqua-culturespeciesandoftheiruseforgeneticanalysis. The main features of designs forQTL mapping, including the use of chro-mosomal manipulations, are described,followed by a discussion of the prospectsandchallengesofGASorMASfordiseaseorcarcasstraits.Finally,newgenomictoolsareconsideredbriefly.

Breeding programmeS and reSponSe to SeleCtionManagement of modern breeding pro-grammes in aquaculture requires usingpedigree information to carry out soundand efficient statistical evaluations (usingbest linear unbiased prediction [BLUP]methodology). This approach enablesbreeders to maximize genetic gain whilelimiting rates of inbreeding to acceptablelevels (Meuwissen, 1997; Toro and Mäki-Tanila,1999).

Most of the genetic improvement infish and shellfish species to date has beenmade through the use of traditional selec-tive breeding (reviewed by Hulata, 2001).Well-designed breeding programmes haveshown substantial response to selectionforbodyweight,e.g.Atlanticsalmon,10–14percent (Gjøen and Bentsen, 1997). Inrainbowtrout, ratesofgeneticgainvariedfrom 8percent for indirect selection forbody weight at sea (Kause et al., 2005)to 13percent for direct selection (Gjerde,1986).Theresponsetoselectionwasabout


10percent for body weight in a breedingprogrammeforcohosalmon(CMG-IFOP)initially funded by FAO (Martinez andHidalgo, unpublished data), and a similarresponse was obtained for this species intheUnitedStatesofAmerica(Hershbergeret al., 1990). Estimates for tilapia followlargely the same trend,witha responseofabout10percent(Ponzoniet al.,2005).Incommon carp, responses to selection forbodyweightwereinconsistentbetweenup-selectedanddown-selected lines,althoughexhaustion of additive genetic variationfor increased growth rate, genotype-by-environment interaction, or competitioneffects could not be ruled out (Moav andWohlfarth,1976).Inoysters,asymmetricalresponsetoselectionforbodyweightwasfound(Toroet al.,1995;Ward,EnglishandMcGoldrick,2000).

Although responses to selection havenotbeenwelldocumented,significantesti-mates of genetic parameters have beenobtained for carcass traits (Gjerde andSchaeffer,1989;Kauseet al.,2002;Quinton,McMillan and Glebe, 2005) and diseaseresistance (Gjøen et al., 1997; Henryon et al., 2002, 2005). Rates of genetic gain areexpected to be lower for these traits thanfor body weight because breeding valuepredictions rely solely on measurementsfromrelatives.

Severalbreedingprogrammeshavebeeninitiatedrecentlyfornewaquaculturespe-cies,suchasmussels,scallops,Artemiaandshrimp.Thebiologyofthesespeciesposesinteresting avenues for the design of con-ventionalbreedingprogrammes,takingintoaccount factors such as self-fertilization,intrafamily competition, cannibalism, lackofmethodsforphysicaltagging,andmatingpreferences.Forexample,competitioncanaffect the expression of quantitative traitsdue to co-variances among members of a

groupmanagedtogetherinapondortankand, ifnotconsideredproperly, this effectcanseriouslyaffecttheratesofresponsetoselection(Muir,2005).However,thiseffectcan be included explicitly in the model ofanalysisusingtheco-varianceamongmem-bers of a group, the so-called “associativeeffects”fromothergenotypesinthegroup.The theory of Griffing (1967) for BLUPevaluationwasdevelopedinthecontextoftreebreeding,butdeservesfurtherinvesti-gation in the analysis of fish and shellfishbreeding. This may be especially true forspecies taken recently from the wild orthosethatshowcannibalisticbehaviour.

Another recent example is the devel-opment of scallop breeding programmes.Argopecten purpuratus is a simultaneouslyhermaphroditicspecies.Inthefirstbreedingphase, the scallop liberates sperm, afterwhich the eggs are expelled. To decreasethelevelofself-fertilization,itiscustomaryto use only the last pulses of eggs. Thissystem reduces rates of self-fertilizationto 20 percent (A. Vergara, personal com-munication), but a residual proportion ofeggs are still already fertilized with spermfrom the same individual. As this processoccurs within the reproductive tract, itis not possible to detect which individ-uals are selfed or outcrossed, althoughthe rate of residual self-fertilization varieswidelyamongfamiliesandproducesbiasedestimates of heritability (Martinez and diGiovanni,2006).Informationfrommolec-ularmarkerscanbeofbenefitunderthesecircumstances(seebelow).

dna markerS uSed in aquaCulture Mutations in the genome create geneticvariability (or polymorphism), whichis reflected as allelic diversity of molec-ular markers. While genomic sequencingwould greatly facilitate the development


ofmolecularmarkers, themanyspecies inaquaculturewouldmake thisacostly task(Liu and Cordes, 2004). Hence, a varietyof approaches have been taken to developgeneticmarkersforaquaculturespecies.

Dominantly-expressed markers havebeenusedextensivelyinaquaculturestudies.Amplified fragment length polymorphism(AFLP) markers (Vos et al., 1995) pro-vide a cost-effective alternative for specieswhereDNAsequencing isnotunderwayor when there are restricted resources forQTL mapping. Dominant AFLP markersarepreferredoverrandomamplifiedpoly-morphic DNA (RAPD) markers becausethey are more reproducible both in otherlines or populations and in other labora-tories (e.g. Nichols et al., 2003), and theycangeneratehundredsofmarkers(asinglepolymerasechainreactioncommonlygen-erates over ten markers). Furthermore,heterozygotes can often be distinguishedfrom homozygotes using the fluorescentband intensity (Piepho and Koch, 2000;Jansenet al.,2001).

Microsatellite markers are simplesequencerepeats(SSRs)arrangedintandemarrays scattered throughout the genome,both within known genes and in anony-mous regions. Microsatellite markers areused increasingly in aquaculture species(reviewed by Liu and Cordes, 2004), dueto their elevated polymorphic informa-tion content (PIC), co-dominant mode ofexpression, Mendelian inheritance, abun-dance and broad distribution throughoutthe genome (Wright and Bentzen, 1994).Microsatellites are generally Type IImarkers,whichareassociatedwithgenomicregions that have not been annotatedto known genes (O’Brien, 1991). OthermolecularmarkerscanbedistinguishedasTypeImarkers,whichare linkedtogenes(of known function). TypeI markers are

more desirable because they are gener-ally more conserved across evolutionarilydistant organisms, enabling comparativegenomics,assessmentofgenomeevolutionandcandidategeneanalysis.

Two procedures are used to generatemicrosatellite markers. The first uses agenomic library enriched with microsatel-lite-bearing sequences to generate clonesthat bear specific SSRs. These clones arethen sequenced to identify microsatel-lite-bearing sequences and then to designprimers to amplify the regions with spe-cific SSR. Validation is required to studythelevelofpolymorphismandthenumberofnullalleles,andtoidentifyanylocithatare duplicates due to any recent evolu-tionary genome duplication event givingrisetomultiplecopiesoflociinthehaploidgenome (Coulibaly et al., 2005). This isdone by screening a sample of individualsfromthetargetpopulation.

Many laboratories have been workingon developing expressed sequence tags(ESTs)derivedfromcomplementaryDNA(cDNA) libraries for a variety of fish andshellfish species (Panitz et al., 2002; Riseet al.,2004a;Hayeset al.,2004;Rexroadet al.,2005;A.Alcivar-Warren,personalcom-munication). EST sequences can be usedfor marker development in species wherethe full genome is not currently beingsequenced. The cDNA libraries are con-structed using messenger RNA (mRNA)thatwasexpressedindifferenttissues,suchaskidneyandgills.Theexpressedfragmentsofsequencedataarenotthefullsequenceofaknowngene,butwhatwas incorporatedintoamaturemRNAmolecule.

In addition to the library-basedmethodofmarkerdevelopmentpreviouslydescribed,microsatellitescanbedevelopedfrom EST databases or from known genesequences.As it ispossible toconnect the


functionofthetranscriptofgenes(fromanESTsequence)withthepresenceofamic-rosatellite,thesemarkersareTypeImarkers(O’Brien, 1991; Serapion et al., 2004; Nget al., 2005). This strategy of developingmicrosatellite markers from known genesandESTshasbeenused forcommoncarp(Yue,HoandOrban,2004),rainbowtrout(Rexroad et al., 2005; Coulibaly et al.,2005)andAtlanticsalmon(Nget al.,2005;Vasemägi,NilssonandPrimmer,2005).

In all these analyses, high levels oftransferability between populations andspeciescanbeexpectedifthemicrosatellites

are included in coding regions. Suchtransferability has been observed e.g.betweenAtlanticsalmonandrainbowtrout(Vasemägiet al.,2005;Rexroadet al.,2005),makingthesemarkersidealforanalysesofpopulationgeneticsandcomparativemaps.For example, microsatellites derived fromEST sequences have been used to studydivergenceofAtlanticsalmonpopulationsin salt, brackish and freshwater habitats(Vasemägi,NilssonandPrimmer,2005).

Bioinformatic tools can be used forpotential discovery of SNPs using DNAsequencealignment“in silico”(Marthet al.,

table 1recently published linkage maps for various fish and shellfish species used in aquaculture

Species number of markers

marker type

map length female/male

male female reference

cm (kosambi)

cm (kosambi)

atlantic salmon

473 aFlP 8.26:1 103 901 Moen et al., 2004a

54 Microsatellites

65 Microsatellites 3.92 np np Gilbey et al., 2004

rainbow trout

226 Microsatellites - 4 590 nichols et al., 2003

973 aFlP

4 allozymes

72 VntR

29 Known genes

12 Minisatellites

5 RaPDs

38 Sine*

oysters 115 Microsatellites 1.31:1 776 1 020 Houbert and Hedgecock, 2004

Sea bass 174 Microsatellites 1.6:1 567.4 905.9 chistiakov et al., 2005

kuruma prawn

195 aFlP 1 780 1 026 li et al., 2003

tilapia 525 Microsatellites 1:1 1 300 lee et al., 2005

21 Genes

Scallops 503 aFlP 1.27:1 2 468 3 130 Wang et al., 2005

Common carp

110 Microsatellites - 4 111 Sun and liang, 2004105 Known genes

57 RaPDs

Japanese flounder

111 Microsatellites 7.4:1 741.1 670.4 coimbra et al., 2003

352 aFlP

Channel catfish

313 Microsatellites 3.18:1 1 958b Waldbieser et al., 2001

b Sex-averaged

* Short interspersed elementsnp = not published


1999). Although it is possible to use basequality values to discern true allelic vari-ations from sequencing errors, validationisakeystep for truepositivedetectionofSNPs(Marthet al.,1999).ThisisgenerallycarriedoutusingaproportionoftheSNPsdetectedinasampleofindividualsfromthetarget population. This strategy has beenusedrecentlyforSNPdetectionusingESTsequences fromAtlanticsalmon(Panitzet al.,2002;Hayeset al.,2004).

linkage mapsA linkage map is an ordered collection ofthe genes and genetic markers occurringalong the lengths of the chromosomes ofa species, with distances between themestimated on the basis of the number ofrecombinationeventsobservedinthedata.Genetic linkagemapshavebeenpublishedfor rainbow trout (Young et al., 1998;Sakamotoet al.,2000;Nicholset al.,2003),channel catfish (Waldbieser et al., 2001),tilapias(Kocheret al.,1998;Leeet al.,2005)and Japanese flounder (Coimbra et al.,2003).Referencestoupdatedlinkagemapsof themajoraquaculture species aregivenin Table 1. Dense linkage maps includinga relatively large number of markers areunderdevelopment.

Different patterns of recombinationappearamongregionsoflinkagegroupsincertainmalemaps,withmarkers clusteredin centromeric regions, an extremeexample being Atlantic salmon whererecombination in males is greatly reduced(Moen et al., 2004b). The molecularmechanismsresponsibleforthedifferencesin recombination rates between sexes arenot well understood, although studies onmodel organisms such as zebrafish, wheregenomic sequencing is currently underway, may help to clarify this (Singer et al.,2002).

uSing markerS to aid Conventional fiSh and ShellfiSh Breeding programmeS Molecularmarkersmaybeusedinanumberofwaystoaidconventionalbreedingoffishandshellfishspecies,andsomeofthesearedescribedandexemplifiedbelow.

parentage analysisOneofthemainconstraintsfacingeffectivebreedingprogrammesforfishandshellfishisthatnewbornindividualsaretoosmalltobe tagged individually. Application of theanimalmodelapproach (i.e.usinga statis-tical genetic model to predict individualbreedingvalues)requirestaggingaconstantnumber of individuals from each familywith passive integrated transponders (PITtags) when they become sufficiently largeafteraperiodof individual familyrearing.However, this system of early manage-ment creates common environmental (i.e.tank)effectsforfull-sibfamilies(Martinez,Neira and Gall, 1999). To address thisissue,mixturesofequal-agedprogenyfromdifferent families can be reared commu-nallytoprecludethedevelopmentofsuchfamily-specific environmental effects, andgenetic markers can be used subsequentlytoassignindividualstofamiliesafterevalu-ation of individual performance (Doyleand Herbinger, 1994). Thus, the impactof early common environmental effectsis considerably reduced if markers areusedforparentageanalysiswhenselectingindividuals for early growth rate traits(Herbingeret al.,1999;Norris,BradleyandCunningham,2000).

The amount of marker data needed toachieve acceptable levels of correct par-entageassignmentdependsonthenumberofloci,thenumberofallelesandthenumberof parent-pairs (sires and dams) availablefor reconstructing the pedigree (Jamieson


andTaylor,1997;Villanueva,VerspoorandVisscher, 2002).The information from themarker data available for each species canbe studied using exclusion probabilities,whicharethenusedtocalculatetheprob-ability(PC)ofcorrectlyassigningthetrueparent-pair(sireanddam)tooffspringthatare genotyped (Villanueva, Verspoor andVisscher,2002).

Figure 1 presents the results for threemicrosatellites and combinations of mic-rosatellites to predict the probabilities ofexclusion and PC. The allelic frequenciesofthethreemicrosatelliteswerecalculatedwithasample(n=100)fromacohosalmon(O. kisutch)farminsouthernChilemanagedundercommercialconditions.Theanalysisshowed that the probability of assigningthe true parent-pair depended greatly on

the number of parent pairs available forparentage.Onlyforanunrealisticallysmallnumber of ten sires and dams is there ahigh probability of assigning the correctparent-pair to offspring. For a breedingprogramme of 200 or 300 parent-pairs,PC decreased considerably. Therefore, inthisexample,moremarkersareneededforaccuratepedigreereconstruction.Successfulparentageassignmentexperimentstypicallyhaveusedsixtoeightmicrosatellitemarkers(Herbingeret al., 1995;GarciadeLeonet al.,1998;Norris,BradleyandCunningham,2000; Castro et al., 2004). In practice, thepresenceofgenotypingerrors,null alleles,realized mutations and non-Mendeliansegregation can seriously affect the effi-ciency of parentage assignment (Castro et al.,2004).Parentageassignmentinthecon-

FiGURe 1predicted probability of assigning the correct sire-dam (parent-pair) for a given number of parent-pairs (x axis) using different combinations of three microsatellites (l1, l2 and l3)

amplified in Oncorhynchuskisutch

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 50 100 150 200 250 300 350 400

Number of parent-pairs

P (c

orr

ectl

y as

sig

nin

g t

he

tru

e p

aren

t-p

air)

L1 (0.985) L2 (0.644) L3 (0.646)

L1+L2 (0.995) L1+L3 (0.995) L2+L3 (0.874)

L1+L2+L3 (0.998)

Probabilities of exclusion included between parentheses in the legend are obtained using the deterministic method of Villanueva, Verspoor and Visscher, 2002. Data obtained from Diagnotec Sa.


text of fish breeding is also discussed bySonesson(thisvolume).

For most breeding programmes, phys-ical tagging will prove efficient both ineconomic and biological terms to achieveacceptableratesofgeneticgain,whilemini-mizingratesofinbreeding.Geneticmarkertechnology can still be costly for rou-tine assignment of parentage, althoughthese costs can be reduced using mul-tiplex polymerase chain reaction (PCR)technology (Paterson, Piertney and Knox,2004; Taris, Baron and Sharbel, 2005) inwhichmorethanonemarkercanbegeno-typed simultaneously in a single gel laneor capillary. This is especially the casewhenonlyDNAmarkersareusedwithoutphysicaltagging,asindividualsmustbere-typedwhen records formultiple traits areincluded in the selection criteria (Gjerde,VillaneuvaandBentsen,2002).

It is expected that rates of genetic gainforeconomictraitswillnotbeaffectedsig-nificantly when common environmentaleffectsarepresent.Thisisbecause,inmanyspeciesofculturedsalmonids,thecommonenvironmentaleffectdecreasesconsiderably,fromabout20percentforalevinweightto5percentforbodyweightatharvest,whichis the trait with most impact on profit(Herbinger et al., 1999; Henryon et al.,2002;Kauseet al.,2005).Hence,commonenvironmental effects should not decreasetheratesofgeneticgainfortraitsmeasuredat harvest when physical tagging is used.Furthermore,multistageselectionoffersthepossibilityoffirstselectingindividualsonawithin-familybasisdirectlyfromtanks(fortraitsinfluencedbycommonenvironmentaleffects),andthenselectingatasecondstagefor traits measured at harvest (Martinez,2006a).Thisalternativewouldeithermain-tainratesofgainwhiledecreasingthecostsassociated with tagging, or even increase

rates of gain, when recording from tanks(within families) can be carried out rela-tivelyinexpensively(Martinez,2006a).

The sample size (i.e. the numbers ofindividuals and markers required forreconstructing the pedigree of a popu-lation accurately) is a practical issue, asnot all individuals in a population can begenotyped for all markers available. Suchissues arise in species where physical tag-ging is not possible or not economicallysound, as in nucleus populations withoutelectronic tagging (e.g. when recovering aback-up population for nucleus breedingprogrammes) or when disease challenges(e.g.forinfectiouspancreaticnecroticvirus[IPNV]) are carried out early in the lifecycle(Martinezet al.,inpreparation).Smallsamplesizes,togetherwithspermcompeti-tion (Withler and Beacham, 1994), matingpreference(asinArtemia;G.Gajardo,per-sonalcommunication)andotherbiologicalfactors after fertilization can increase thevarianceoffamilysize, therebydecreasingthe effective population size to unsus-tainable levels (Brown, Woolliams andMcAndrew,2005).

Another problem arises in practicewhenselection iscarriedoutbeforegeno-typing with markers. In this case, BLUPof breeding values is likely to be biasedbecause not all phenotypic informationis used when predicting breeding values.Themagnitudeofre-ranking isdependenton the amount of information from afamily within the selected group. In theseinstances,themixedmodelequationsneedtobemodifiedtoaccountforsuchselecteddata(MortonandHowarth,2005).

establishing breeding programmes using molecular informationThe choices made at the founding of abreeding programme have a critical


bearingonitsultimatesuccess.Criteriaforchoosing individuals thatwillbe foundersshouldbeessentiallythesameasthoseusedwhen the selection response is optimizedunder restricted co-ancestry when pedi-gree information is available (Meuwissen,1997; Toro and Mäki-Tanila, 1999). Thus,it is necessary to avoid matings betweencloserelativesformanagingexistingquan-titativegeneticvariationat thestartof theprogramme. Experiments with the plank-tonic microcrustacean Daphnia spp. haveshown that neutral genetic variation giveslittle indication of the levels of quantita-tivegeneticvariationavailableforselection(Pfrenderet al.,2000).However,increasingthepopulationsizeatthebeginningofthebreedingprogrammewilldiminishthesub-sequenteffectofrandomgeneticdrift,andtherefore larger founding populations willhave an increased likelihood of showingresponse to selection. Lack of adequatebasepopulationsisthemainreasonforthelackofselectionresponseobservedinsomespeciesoffish(Gjedrem,2000).

The effective population size (Ne)required for setting up a breeding pro-gramme depends on the policy regardingrisk management (Brown, Woolliams andMcAndrew, 2005), but to prevent declineinfitness,someauthorshaverecommendedNevaluesrangingfrom31to250,whichintermsofratesofinbreedingshouldbelessthan2percent(MeuwissenandWoolliams,1994). Duetothelargefamilysizespossibleformanyfishandshellfishspecies,breedingprogrammesthatfailtocontrolthegeneticcontributionsofparentsineverygenerationareexpectedtoincurrelativelyhighratesofinbreeding (Meuwissen, 1997). The situa-tionisevenmoreextremewhenselectionisbasedonacomplexbreedingobjectivethatincludes information from relatives andmanytraitsjointly(Martinez,2006b).

Fish within commercial productionpopulations generally are not taggedindividually and pedigree information isthereforelacking.Geneticmarkersallowtheestimationofpairwiserelatednessbetweenindividuals or sib-ship reconstructioneven with unknown ancestors (Toro andMäki-Tanila, 1999; Thomas and Hill,2000; Toro, Barragán and Óvilo, 2002;Wang, 2004; Fernandez and Toro, 2006).There is a plethora of estimators forcalculating pairwise relatedness (Quellerand Goodnight, 1989; Lynch and Ritland,1999). The efficiency of inferring pairwiserelatednessusingmarkerswithoutparentalinformationisaffectedbyassumingknownallele frequencies in the base populationand unlinked loci in Hardy-Weinbergequilibrium. Furthermore, pair-wisemethodscanleadtoinconsistentassignationsbetweentripletsofindividualsbecausetheyuseinformationfromonlytwoindividualsat a time (Fernandez and Toro, 2006). Inaddition,itisdifficulttosetthresholdsforclaiming different types of relatedness inthe data (Thomas and Hill, 2000; Norris,Bradley and Cunningham, 2000). On theotherhand,sib-shipreconstructionmethodsdo not attempt to calculate co-ancestry;rather,theyattempttoreconstructfull-orhalf-sib or other family groups (Thomasand Hill, 2000; Emery, Boyle and Noble,2001; Smith, Herbinger and Merry, 2001).Such reconstructions of full- or half-sibfamilies or even other groups of relativesappearrobusttolackofknowledgeofbasepopulationallele frequencies (ThomasandHill,2000;FernandezandToro,2006).

Markerinformationcanbeusedtoinferrelatednessbetweenindividualsavailableascandidate broodstock to generate the firstgenerationofoffspringinthebreedingpro-gramme,andtherebyavoidmatingamongclose relatives. This approach uses molec-


ular information to infer the genealogicalpedigree. A simulation was conducted toreconstruct the pedigree of 100 potentialcandidates from ten full-sib families (withaPoissonfamilysizeequaltotenusingsixequally-frequent microsatellites, withoutparental genotypes (Martinez, 2006c).Theposterior probability of either full [P(FS)]or half-sib [P(HS)] groups was obtainedusingtheBayesianmodelofEmery,BoyleandNoble(2001).Inthesimulationresults,therewasa tendencytooverestimaterela-tionships,withposteriorprobabilitiesover0.5whenindividualswereinfactunrelated.On the other hand, not all true full-sibswereassignedtothecorrectfull-sibfamilywiththegreatestprobability,andsometruefull-sibfamilymemberswerereconstructedas half-sibs. On average (among ten rep-licates), the probability of mating relatedindividuals was significantly smaller wheninformation from molecular markers wasused, compared with what was expectedbychance(4.7percentversus18.1percent,p=0.002).Thepracticalimplicationisthatinbreedingintheprogenygenerationwouldaverage 5percent when random matingis used and 1percent when optimizationusingmolecularinformationisused.

In practice, to perform mating in thebase population, the relatedness inferredfrommolecular informationdoesnotneedtobeperfectlyaccurate,butitdoesrequirethat relatedness is not underestimatedgreatly. Among the technical issues thatarise when using marker data are that apair of individuals could be misclassifiedas related when they are in fact unrelated(Type I error) or a pair may be wronglyclassified as unrelated when the pair is infact related (Type II error). Type II errorisofgreatestconcernasthiscouldresultinrelated pairs being mated. This is becausemating of individuals (males and females)

as unrelated when in fact they are fullsibs will increase true inbreeding in thepopulation, while misclassification leadingto unrelated individuals being assigned toa full-sib family would not increase theinbreedingintheprogeny.Thepresenceofmutations,nullallelesorgenotypingerrorswillunderestimatethetruerelationshipsinthe population and eventually increase theprobabilityofmatingtruefull-sibs(Butleret al.,2004).Recently,Wang(2004)suggesteda method for inferring relationships formarker data with a high error rate andmutation that can be used to address thisissue. It should also be noted that studiesdealing with estimation of heritability orpredictionofbreedingvalueswithpedigreesreconstructed using molecular markersmay be very inefficient when pedigreesarereconstructedwithanincreasedrateofTypeIerrors(Mosseau,RitlandandHeath,1998;Thomas,PembertonandHill,2000).

detecting self-fertilization in scallopsIn scallops, a main drawback whenimplementingbreedingprogrammes is theoccurrence of self-fertilization, even whengametesfromlaterspawningpulsesareusedfor obtaining family material (Martinezand di Giovanni, 2006), i.e. a mixtureof selfed and outcrossed individuals canbe present even at later stages within asingle family. Bias in estimating geneticparametersisexpectedduetothisresidualself-fertilization, which can occur withconsiderablefrequency(average20percent)withinparticularfamilies.

Asimulationstudywasusedtoinvesti-gatetowhatextentmarkerswithdifferentinformation content can be used todiscriminatebetweenselfedandoutcrossedindividualswithinafamily(Figure2).Theresults showed that microsatellites gavemeanvaluesofposteriorprobabilitiesgreater


than0.95inabout90percentofthefamiliessimulated(100intotal).Similarresultswereobtainedwith30AFLPmarkers,butthesepercentageswereconsiderablyreducedforsmallernumbersofAFLPsorSNPs.

The information from these markerscanbeusedtocullindividuals,toconstructa relationship matrix in which all unusualrelationships are incorporated in analysesused for obtaining unbiased estimates ofheritability and genetic correlations, andfor estimating breeding values from realdatasets(Martinez,2006a).

identifying qtl and maJor geneS influenCing Complex quantitative traitSMolecularbiologycangreatlyhelpthedis-coveryoffactorsinfluencingtheexpressionofquantitativetraits.Thereareanumberof

waysinwhichthisinformationcanbeused,thedifferencebetweenthembeingthelevelof resolutionwithwhich these factorscanbe mapped. For example, loci with majoreffects on quantitative traits (QTL) aremapped by using markers to track inher-itance of chromosomal regions in familiesor in inbred line crosses using the extentof linkage disequilibrium generated in thepopulation. This approach gives a limitedamountofmapping resolution.Finemap-ping requires information from additionalmarkersandindividualssampledacrosstheoutbred population and, while helping tonarrowtheconfidenceintervaloftheposi-tion of the QTL, this is only the startingpointforidentifyingthepolymorphismsinthe performance-determining genes them-selves. In practice, identification of genesinfluencingspecifictraits isachievedusing

FiGURe 2identification of selfed individuals within families of scallops

using different types of marker data1

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

MICRO-5 SNPs-5 AFLP-5 AFLP-15 AFLP-30

Vertical bars represent the proportion of the 100 replicates in which the mean posterior probabilities of being selfed (for true selfed individuals) were greater than 0.95.

1 MicRo-5: five microsatellites with six equally frequent alleles each. SnPs-5: five SnP markers with equal allele frequencies. aFlP-5, -15 or -30: 5, 15 or 30 aFlP markers. the design of the simulations of self-fertilization in scallops: the amount of self-fertilization was modelled using a truncated normal distribution which best fitted the empirical distribution of self-fertilization (Martinez and di Giovanni, 2006). a bayesian model was used to infer mutually exclusive posterior probabilities of being either selfed or outbred (anderson and thompson, 2002). it was assumed that parental information was lacking, with unlinked markers and vague priors. Selfed individuals were regarded as having been detected correctly when the posterior probabilities of being selfed were greater than 0.95 (this criterion was determined empirically for operational reasons).

Source: V. Martinez, in preparation.


acombinationofgeneticmapping(linkageand fine mapping) to localize the QTL toa small region on the chromosome underanalysis, and candidate gene or positionalcloning approaches to identify the geneswithintheQTLregion.

In some cases, sufficient biochemicalor physiological information is availableto investigate the association between thequantitative expression and the level ofmarker polymorphisms within specificgenes.Nevertheless,thisapproachrequiresa great amount of detailed information inorder to choose which gene explains thegreatesteffectandtohavesufficientpowertodetecttheassociation.This informationisstartingtoappearintheaquaculturelit-erature from multinational projects suchas theConsortiumofGenomicResourcesfor All Salmonids Project (cGRASP) (Nget al.,2005).

qtl mapping in fish using linkage dis-equilibrium: theoretical and practical considerations Value of chromosomal manipulationsThe great reproductive flexibility of fishenables different breeding designs to beimplemented relatively easily. Completelyhomozygous fish can be produced inonly one generation using chromosomeset manipulations, without the many gen-erations of inbreeding needed in othervertebrates. These manipulations enabledoubling of the chromosomal comple-ment of a haploid gamete (Young et al.,1996; Corley-Smith, Lim and Bradhorst,1996). Androgenetic double haploid indi-vidualscanbeobtainedby fertilizingeggsthat were inactivated with gamma radia-tion, yielding haploid embryos containingonlypaternalchromosomes.Alternatively,gynogeneticdoublehaploidindividualscanbeobtainedbyactivatingthedevelopment

of eggs with ultraviolet-inactivated sperm,yielding haploid embryos containing onlymaternal chromosomes. In each case,diploidy is restored using methods thatsuppressthefirstmitoticdivision(Figure3;Streisingeret al.,1980;Corley-Smith,LimandBradhorst,1996;Bijma,vanArendonkand Bovenhuis, 1997; Young et al., 1998).The use of these reproductive manipula-tions to provide experimental populationsforgeneticanalysisofcomplexquantitativetraits has been well described (Bongers et al.,1997;Robison,WheelerandThorgaard,2001;Tancket al.,2001).

Double haploids from inbred line crossesAfter a second roundofuniparental repro-duction(Figure3),acollectionofclonallinescan be obtained that collectively is likelyto represent all the genetic variants fromthe base population (Bongers et al., 1997).Crossesofsex-reverseddoublehaploidindi-vidualsfromlinesthatdivergeforthetraitsof interestcanproduceF1linesincompletelinkage disequilibrium. These F1 popula-tionscanbeusedforfurtherexperimentationbased on F2 or backcross designs. AnotherroundofandrogenesisofF1individualswillproduce a population of fully homozygousindividuals. This design will have twice thepowerfordetectingQTLasthestandardF2design(Martinez,2003).Thestandarddevia-tionofQTLpositionestimatesishalvedforthe double haploid design. This is due toan increase in the additive genetic variance,which is doubled for the double haploiddesign due to redistribution of the geno-type frequencies in the progeny generation(FalconerandMackay,1996).

Informativedoublehaploidpopulationsof this sort have been utilized to performQTLanalysis for embryonicdevelopmentrate in rainbow trout (Robison, Wheelerand Sundin, 2001; Martinez et al., 2002;


Martinez et al., 2005). At least four QTLof relatively large effect explain about40percent of the phenotypic variance ofthe mapping population and most of the2.5 standard deviations of the differencebetween the original clonal lines usedto generate the F1 population (Robison,Wheeler and Thorgaard, 1999). Twolinked QTL were in repulsion phase inthe F1 population, and were undetectedin the analysis using composite intervalmapping.Thisresultwasnotsurprisingasevidencewasaccumulatedamongreplicatesof lines that were incubated at differenttemperatures (Robison, Wheeler andSundin, 2001), and the Bayesian multipleQTLmethodincorporatedalltheavailableinformation of environmental co-variatesin the analysis (Martinez et al., 2005).Recently, these double haploid lines havebeenusedformappingQTLrelatedtothenumberofpyloriccaeca(Zimmermanet al.,2005)andforconfirmingQTLinfluencingdevelopmentrate(Sundinet al.,2005).

Whentraitsareassociatedandbytakingintoaccountthecorrelatedstructureofthedata,multivariateestimationofQTLeffectsisexpectedtobemorepowerfulthansingletraitanalysis(JiangandZeng,1995).Also,from a genetic standpoint, joint analysisprovides the means for testing differenthypothesesaboutthemodebywhichgenesexplained the genetic co-variation (Wu et al., 1999). For example, after hypothesistesting(followingKnottandHaley,2000),a single pleiotropic QTL with oppositeeffects for development rate and lengthbest explained the multivariate data (asdetailed earlierbyMartinez et al., 2002b).This finding was also consistent with thenegativecorrelationestimatedwiththedata(Martinezet al.,2002).

Double haploids in outbred populationsMartinez, Hill and Knott (2002) derivedanalyticalformulaetopredictthepoweroflinkageanalysisforintervalmappingunderthree different mating designs in outbred

FiGURe 3Chromosomal manipulations in fish

ANDROGENESIS

Double haploid

Haploid

DNA inactivation (ova)

Normal diploid

Late Schock

Double haploid

Haploid

DNA inactivation (ova)

Normal diploid

Late Schock

OUTBRED MALE POPULATION

RECOMBINANT INBRED PROGENY(DOUBLED HAPLOID)

INBRED CLONAL (SEX REVERSAL)

DOUBLED HAPLOID

androgeneSiS

in androgenesis, fertilization is carried out using radiation-inactivated ova, and a late shock is used to suppress first mitosis and thereby restore diploidy.adapted from thorgaard and allen, 1987.


populations: full-sib mating, hierarchicalmating, or double haploid designs. Thisanalysis suggested that the use of doublehaploids appeared to be of benefit whendetectingQTL,particularlywhenboththevarianceof theQTLandof thepolygeniceffects was small. Furthermore, given therelatively large size of full-sib families infish, there appeared to be little advantageofhierarchicalmatingover full-sibmatingdesigns for detecting QTL, the optimumfamily size depending on the size of theQTL and the population structure usedfor mapping (Martinez, Hill and Knott,2002).Thegaininpowerofthedoublehap-loiddesigncomesfromtheincreaseinthevariance of the Mendelian sampling termwithin families, which is effectively dou-bledfortraitsthatareexplainedbyadditiveeffects(FalconerandMackay,1996).

As experimental settings constrain thetotal number of individuals genotyped,designs aimed at QTL mapping shouldinclude a small number of families of rel-atively large size in order to maximizethe likelihood of detecting the QTL. Thisis because most of the information formapping QTL uses linkage informationthatcomesfromwithin-familysegregation(Muranty, 1996; Xu and Gessler, 1998).However, increasing power comes at theexpense of reducing the accuracy of esti-mating the additive genetic variance forpolygeniceffects.AQTLmappingmethodhas been developed for double haploids,which efficiently accommodates all theuncertainties that pertain to outbred pop-ulations, such as unknown linkage phasesand differing levels of marker informa-tiveness, using the identical-by-descentvariance component method (see below;Martinez,2003).Also,itispossibletocom-binedoublehaploidsandoutbredrelativesinthesamefamily.Simulationsofdiffering

amountsofmarker informationandherit-abilityfortheQTLwereusedtocomparetheempiricalpowerofthedoublehaploidand full-sib designs. While the power ofthefull-sibdesignwas lowerthanthatfordouble haploids, QTL position estimatesfor double haploids had large confidenceintervals(about30cMascomparedwith40cMforfull-sibs;Martinez,2003).

The double haploid design was usedfor mapping QTL for stress response incommoncarpusingsinglemarkeranalysis(Tanck et al., 2001). The authors foundonlysuggestiveevidenceforQTL,whichisnotsurprisingduetolimitedgenomecov-erageformarkersusedintheanalysis.

PublishedresultshaveshownthatdoublehaploidlinesareausefulresourceforQTLdetection studies. However, double hap-loidlinesaredifficulttodevelopduetotheexpression of deleterious recessive alleles(McCuneet al.,2002)andthelowsurvivalfollowingshocksappliedtorestorediploidytothehaploidembryo.Astherateofmalerecombination is depressed, the precisionofmappingQTLinandrogeneticfamiliesislowerthanthatobtainedusingrecombina-tioneventsfromfemales.Anotherpracticalmatteristhelabourneededfordevelopingaclonalline,asatleasttwogenerationsarerequired(Figure3).Thisdelaycanbequiteexpensive and time-consuming for specieswith a long generation interval, such assalmonortrout(twotofouryears).

aspects of qtl mapping in outbred populations of fishInbred line crosses are ideal for mappingQTLbecausetheyareexpectedtobecom-pletely informative for both markers andQTL, providing that the inbred lines arefixedforalternativealleles.Outbredpopu-lations are not completely informative forbothQTLandmarkers;thus,experimental


powerisexpectedtobelowerthanthatforcrosses between clonal lines. The powerfor detecting the QTL depends on allelefrequencies,theprobabilityofsamplinganinformativeparentandfamilysize.

Factors influencing the power of detecting QTLDue to the large family sizes that can beobtained in many fish species, differentmatingdesignsusingfull-sibgroupscanbecarried out for outbred populations. Forexample,fullfactorialdesignsmaybeusedinwhichmanymalesandfemalesarematedto one another, and hierarchical designsmaybeappliedinwhicheachmaleismatedwithmultiplefemales,oreachfemalewithmultiplemales.Foragivensizeofexperi-ment,factorialandhierarchicaldesignshavepotentiallyalowerprobabilityofsamplingaheterozygousparent(becausefewersiresand or dams are sampled overall), com-pared with the full-sib design in whicheachfamilyhaspotentiallytwoinformativeparents.Forthisreason,factorialandhier-archicaldesignscanpotentiallygive lowerpower compared with the simple full-sibdesign(Muranty,1996;Martinez,HillandKnott,2002).

Theoptimumnumberoffull-sibfamiliessampled in the QTL mapping popula-tiondependsontheintrinsicpoweroftheexperiment(i.e.sizeoftheQTLeffectandsizeof thepopulation).Asexpected, largefamilysizesareneededfordetectingQTLofsmalleffects(Martinez,HillandKnott,2002).When theQTLexplains10percentof the phenotypic variance, the optimumfamilysizeappearstobe50individualsperfamily for a reasonably-sized QTL map-ping experiment in outbred populations(Figure4).Furtherincreasesinthenumberof individuals per family provide only amodestincreaseinpower.Further,thesame

results used simulation models showingdominance and additive effects under thevariancecomponentsmethodformappingQTL(Martinezet al.,2006a).

Methods of analysisThemethodofchoicewhenanalysingdatafrom outbred populations is the variancecomponentmethod, inwhichQTLeffectsare included as random effects with a co-varianceproportionaltotheprobabilitythatrelatives(e.g.full-sibs)shareallelesidenticalbydescentconditionalonmarkerdata(Xuand Atchley, 1995). This model is similartotheoneusedmoregenerallyforgeneticevaluation of candidate fish for selection,butincludestherandomQTLeffect.

Aconsiderableproportionofthegeneticvariance for growth-related traits in fishpopulations has been explained by domi-nance (Rye and Mao, 1998; Pante, Gjerdeand McMillan, 2001; Pante et al., 2002).When mapping QTL using the randommodel, it is assumed that only additiveeffects are of importance and thereforeonlymatricesofadditiverelationshipscon-ditional on marker data are fitted in theresidual effect maximum likelihood pro-cedure(George,VisscherandHaley,2000;Pong-Wong et al., 2001). However, thelarge family sizes in fish enable hypoth-eses for different modes of inheritance atthe QTL to be tested using the within-family variance. While some authors havespeculatedthatincludingdominanceinthemodelwillincreasethepowerofdetectingQTL (Liu, Jansen and Lin, 2002), others(Martinez, 2003; Martinez, 2006a) haveshownthatpowertodetectQTLwascom-parable between models including or notincludingdominance.Thiswasparticularlythe case for the larger family sizes simu-lated and it was concluded that for mostscenarios, the additive model was quite


robust for detecting QTL and there waslittlelossofinformationfordetectingQTLwhendominanceispresentbutnotusedintheQTLmappinganalysis.

qtl mapping in practiceTodate,QTLmappinginfishusingoutbredpopulations has been carried out mostlywithsinglemarkeranalysis(microsatellitesand AFLP markers), and using relativelysparse linkage maps when interval map-pingisused.Intilapia,theF2designandafour-waycrossbetweendifferentspeciesofOreochromis havebeenused fordetectingQTL affecting cold tolerance and bodyweight (Cnaani et al., 2003; Moen et al.,2004c). In outbred populations of salmo-nids,QTLthatinfluencebodyweighthavebeen mapped (Reid et al., 2005 and refer-encestherein).

Studies seeking linkage of markers totraits amenable to MAS, such as disease

resistance,havebeguntoappearintheliter-atureoverthepastfewyears.Forexample,QTL for resistance have been mappedfor infectious pancreatic necrosis virus (Ozaki et al., 2001), infectious salmonidanemia (Moen et al., 2004c), infectioushaematopoietic necrosis (Rodriguez et al.,2004; Khoo et al., 2004), and stress andimmune response (Cnaani et al., 2004).Also, Somorjai, Danzmann and Ferguson(2003andreferencestherein)reportedevi-denceofQTLforupperthermaltolerancein salmonids with differing effects in dif-ferentspeciesandgeneticbackgrounds.

from fine mapping to finding genes influencing complex traitsWhen the number of meioses in thegenotypedpedigreeisnotsufficientforthelinkageanalysistoobtainaprecisepositionfor the QTL, there is a wide confidenceintervalaroundanestimatedQTLposition.

FiGURe 4deterministic power calculation (following martinez, hill and knott, 2002) for a qtl explaining

10 percent of the phenotypic variance for a variable total population size (x axis) and family size (25, 50 and 100)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

100 200 300 400 500 600 700 800 900 1 000

Population size

Pow

er o

f d

etec

tin

g Q

TL

25 50 100


Finemappingmethodsattempttoovercomethis problem by quantifying the gameticphase or linkage disequilibrium (LD)presentinanoutbredpopulation,i.e.acrossfamilies. This method makes use of thenumber of generations as the appearanceof a mutation and can produce extremelyprecise estimates of the QTL position(Pérez-Enciso et al., 2003). The rationalebehindusingLDformappingQTListhatwhen the population size is rather small,founders of the population would haveonly a limited number of haplotypes, andwithverytightlylinkedlocitheremaynotbe sufficient time for recombination tobreak up the association between markersandthemutationaffectingthequantitativetrait.

LDmappingiscarriedoutbycalculatingtheprobabilitiesthathaplotypessharedbyindividualsareidenticalbydescentfromacommon ancestor conditional on markerdata(assumingtgenerationsasthecommonancestor and a certain Ne; Meuwissen andGoddard, 2001). The LD in the popula-tion depends on a number of populationparameters such as the degree of admix-tureorstratificationinthepopulationandtheactual levelof associationbetween thecausal mutation and the polymorphisms.The correct determination of phases andof genotypes at the QTL is required forfine mapping purposes (Meuwissen andGoddard, 2001; Pérez-Enciso, 2003). Forthese reasons, apureLDanalysis is likelytoresultinalargenumberoffalsepositives,i.e.falselyinferringassociationwhenthereisnolinkage.

Methods that incorporate the linkageinformation (within families) and LDjointlyarepreferred,becausethelikelihoodof spurious association (i.e. LD withoutlinkage) diminishes, making much betteruseof thewholedata set (Meuwissenand

Goddard,2001,2004;Pérez-Enciso,2004).All of these methods, however, require agreat deal of genotyping of tightly linkedmarkerssuchasSNPs,whichcurrentlyarenot widely available for fine mapping inaquaculturespecies.

Usingfinemappingtechniques,thecon-fidence interval for QTL position can bereducedconsiderably.However,todevelopa direct test for a favourable polymor-phismrequiresuseofcomparativemappingapproaches with model species, such aszebrafish or fugu, to select the candidategenes that most likely affect the trait ofinterest.Otherwise,enrichmentofmarkersin a specific region of the genome (tonarrowfurtherthemost likelypositionofthe polymorphism) following sequencingis needed to compare sequences betweenindividualsthatshowdifferentphenotypesoralternativeQTLalleles.

Candidate gene analysisIt is tempting to invokevariationatgeneswithaknownroleinthephysiologyunder-lying a complex trait such as growth toexplainphenotypicvariabilityforthetrait.Thesegenescanbesearched forpolymor-phisms (e.g. SNPs) and the variants thentestedtodeterminewhethertheyarecorre-latedwiththeexpressionofthequantitativetrait. This approach requires knowledgeof the biology of the species, biochemicalpathways and gene sequences in order totarget variation at those specific genes. Inaquaculture,mostofthisinformationiscur-rentlylacking,butitisexpectedthatmoregenes will be incorporated in databases inthenearfuture.Thepossibilityexiststouti-lizedatafromhighlystudiedmodelspecies,suchaszebrafishorrainbowtrout,incom-parativebioinformaticapproaches.

Todate,thisstrategyhasnotprovenpar-ticularly successful for explaining genetic


variation underlying complex (polygenic)traits.Thisisbecausealthoughthebiologyof the trait and the genes most likelyinvolved in the expression of the pheno-type may be known, in complex traitsmany other genes may be involved in themetabolic pathway that are not obviouscandidates.Forexample,inaquaculturespe-cies,candidategeneshavebeenstudiedforgrowth-related traits using ten conservedgenesequencesknowntoberelatedtothegrowth hormone axis (Tao and Boulding,2003).InthisstudyofArcticcharr,onlyasingleSNP(of ten) fromfiveof tengeneswas found to be associated with growthrate.

Another example for disease resistancetraits isthemajorhistocompatibilitycom-plex (MHC). The genes of this complexencode highly polymorphic cell surfaceglycoproteins involved inspecific immuneresponses and either specific alleles orheterozygotes at this complex were asso-ciatedwithresistanceandsusceptibility toA. salmonicidaorinfectioushaematopoieticnecrosis(IHN)virus(Langefors,LohmandGrahn,2001;Lohmet al.,2002;Arkushet al.,2002;Grimholtet al.,2003;Bernatchezand Landry, 2003). Nevertheless, thebackground genome was quite importantfor explaining the difference in resist-ance between individuals within a family(Kjøglum,GrimholtandLarsen,2005).

microarrays, gene expression and identification of candidate genes for qtl analysisMicroarray technology (Knudsen, 2002)enables the expression of thousands ofgenes to be studied simultaneously. Untilnow, this information has been used pri-marily for following gene expression intreatmentandcontrolexperimentsinmanyfields such as disease exposure and stress

response.This informationcanbeused todiscovernewsetsofcandidategenes,pos-siblywithorwithoutfunctionalassignmentthatmayberelatedtothequantitativetraitof interest (Walsh and Henderson, 2004).Genes whose expression differs betweentreatments are likely to be trans-actinggenes, i.e. their expression is regulatedby other genes. Therefore, it seems likelythat seeking polymorphisms within thesegenes may not yield information aboutfactors that explain the phenotype, andtheremightbeproblemsassigningthecor-rect significance threshold (Pérez-Encisoet al., 2003). Further, because many genesarepartofmetabolicpathwaysanddonotact individually, the expression of a singlegenemaybeinsufficienttoexplainpheno-typicdifferencesbetweenindividuals.Onlythosegenesthatdirectlyaffectphenotypicexpression (i.e. cis-acting genes) can betreated as candidate genes for subsequentuseinMASafterstudyingpolymorphismsin their sequences. In salmonids, a micro-arraymadeavailablefromtheConsortiumfor Genomics Research on all SalmonidsProject (cGRASP)hasbeenused to studygene expression in fish exposed or notexposed to Pisciricketsia salmonis (Rise et al., 2004b), and microarrays in other fishand shellfish species are currently underdevelopment.

A gene expression pattern can itself beregarded as a quantitative trait. Here, theinterest is in finding associations betweendifferent patterns of gene expression andmarker loci. This analysis was coined as“genetical genomics” by Jansen and Nap(2001). As is usual in QTL mapping, theanalysisattemptedtodissectthetranscrip-tionalregulationoftheentiretranscriptomeand to identify the effects of individualQTL affecting gene expression (the so-calledeQTL;e.g.Hubneret al.,2005).To


date, this analysis relies upon the use ofsegregatingpopulations (ofknownorigin)suchasrecombinantinbredlines(Carlborget al., 2005), and the analysis of outbredpopulationsposesgreaterchallenges(Pérez-Enciso,2004).Still,aquaculturespeciescanprovide sufficient information due to thelarge family sizes needed to unravel com-plex regulatory gene networks. How allthis information can be included in MASprogrammesisyetunclear.

inCorporating moleCular markerS into Breeding programmeS for fiSh and ShellfiShgeneral aspects of incorporating molecular information in breeding programmesThe response to selection ∆G is estimatedas: ∆G =iσHr

wherei=theintensityofselection,r=thecorrelationbetweenthebreedingobjectiveandtheselectioncriteria(i.e.accuracy),andσH =theadditivegeneticstandarddeviationfor the breeding objective. As the majorimpact of incorporating information frommolecular markers will be on accuracyestimates, improvementoftheresponsetoselectionwillbehigherfortraitsthathaverelatively small accuracy than for traits ofrelatively large accuracy. Thus, breedingprogrammesfortraitswithlowheritabilityand relatively few records per trait meas-uredsuchascarcassanddiseaseresistanceare those most benefiting from incorpo-rating marker information (Meuwissen,2003).

Therelativeincreaseinaccuracydependson the amount of variation explained bymarkers, which in turn depends on thenumberofQTLidentifiedandusedinMASor GAS schemes (Lande and Thompson,

1990). QTL experiments in other specieshave shown that the effects of markedgeneshavea leptokurticdistribution,withasmallnumberofgeneshavinglargeeffectsandpolygenes(HayesandGoddard,2001),whichislikelytobethecaseinaquaculturespecies(Martinezet al.,2005).Hence,itisexpected that more than a single markedgene will be needed for MAS schemes tobeefficient.

Due to the biology of many fish andshellfish species, multistage selectionwill likely prove useful in MAS or GASschemes. Basically, a first stage of selec-tion can be applied for traits expressedearly in the life cycle (e.g. body weight),anda second stageof selectionwill incor-porate information from relatives plusmarkedQTL.Optimizationwillbeneededtodeterminetheintensityofselectionthatshouldbeappliedateachstagetomaximizeprofit while reducing the cost and labourof keeping individuals until later stages(Martinezet al.,2006b).

Health and carcass traits are difficultto select for in fish and shellfish becausephenotypicrecordsareobtainedfromrela-tivesandnotfromcandidatesforselection(Gjoen and Bentsen, 1997). Sib or pedi-greeevaluationhasmanydisadvantages inrelation to theamountofgeneticprogressthatcanberealizedwithinaselectionpro-gramme using only pedigree informationtopredictbreedingvaluesusingananimalmodel. First, selection accuracy using sibinformationislowerthanwhenpredictingbreeding values based on an individual’sown information (Falconer and Mackay,1996). Second, there is no variation ofestimated breeding value for polygeniceffects.Thus,variationofMendelian sam-plingeffectswithinafamilycannotbeusedand consequently there may be a limitedscope for constraining rates of inbreeding


to acceptable levels when the number offamiliesisrelativelylow.

To date, little has been publishedregardingtheeconomicprofitsarisingfromthe extra genetic gain obtained by MASor GAS schemes in aquaculture or ter-restrial species. Information of this natureis essential because the additional gainsare dependent on the magnitude of thealleliceffectsandthusthemarginalincreaseshould offset the costs of applying thetechnology. This trade-off may be moreimportant when a single marked QTL,rather than multiple marked QTL (andmultipletraits),istargetedbyselection.

Pleiotropic effects can be important ifthe polymorphisms under MAS or GASalso have negative effects on fitness orother traits of economic importance. Forexample,negativegeneticcorrelationshavebeenfoundforresistancetoviralandbac-terialdiseases(Gjøenet al.,1997;Henryonet al.,2002,2005),whichmaybeaproblemin practical breeding when the goal is toselect fish resistant to a range of patho-gens. For example, in natural and selectedpopulations,MHCpolymorphismislikelyto be maintained by frequency-dependentselection (Langefors, Lohm and Grahn,2001; Lohm et al., 2002; Bernatchez andLandry, 2003), suggesting that selectionfavours rare alleles, but works against thesame alleles at high frequency. Therefore,it seems likely that a MAS scheme usingMHCinformationorQTLinLDwithdis-easeresistanceshouldfocusonmaintainingpolymorphismratherthanonselectingforaparticularcombinationofalleles.

maS in populations in linkage equilibriumWhen populations are in LE betweenmarkers and QTL, the information usedfor selection purposes is given by the

Mendelian co-segregation of markers andQTL within each of the full-sib familiesin the population under selection. Inpracticalterms,thismeansthatco-ancestryconditional on marker information needsto be computed within a family for agiven segment in the genome. In effect,the segregation of regions that individualsshare as identical-by-descent (“more” or“less” than average) is being traced and,under suchcircumstances, the accuracyofpredicting breeding values using markerinformation is mainly dependent on theproportion of the within-family varianceduetotheQTL(Ollivier,1998).

The effect of family size on the rela-tiveaccuracyofpredictingbreedingvalues(comparingMASandBLUP)usingmarkerinformationwasstudiedindetailusingsim-ulations(Table2;V.Martinez,unpublisheddata). Compared with the GAS schemespresented below, for LE-MAS to be effi-cient, large full-sib families are requiredforpredictingbreedingvaluesfortheQTLaccurately. This is because breeding valuepredictioniscarriedoutonawithin-familybasis; thus, large families are required toobtainbreedingvaluesforpredictingQTLeffects with reasonable accuracy. Whenindividuals do not have records for thequantitative trait, the extra accuracy ofMASwashighestforthelargestfamilysizesimulated (50individuals, 25 with recordsand 25 without records; the difference isequal to 7percent). The accuracy of pre-dictingbreedingvalueswasverysimilarinBLUP or MAS for individuals that haverecordsforthetraitinmostofthescenariossimulated,suggestingthatMASisexpectedtobeoflittleuseunderthesecircumstances(Villanueva, Pong-Wong and Woolliams,2002).

The advantage of MAS will comeboth from increased accuracy and from


increasing the realized selection inten-sity in sustainable breeding schemes withrestrictedratesofinbreeding.Insib-testingschemes, candidates without records canonly be selected randomly within familiesbecauseanestimateoftheMendeliansam-pling terms cannot be obtained. MarkersprovideanestimateoftheQTLeffectsthatsegregate within a family, and thereforethe realized selection differential (at thesameratesofinbreeding)isexpectedtobegreater than that obtained using standardsib/familytesting.

Allthebenefitsoutlinedabovecomeatan expense. MAS using LD within fami-liesrequiresagreatdealofgenotypingandrecording of phenotypes on relatives, dueto the fact that the linkage phase betweenmarkersandQTLneedstobere-estimatedin each generation. This is because LDbetweenmarkersandtheQTLisestablishedonlywithinfamiliesineachgenerationandnotacrossthepopulation.Forthisreason,itisnotpossibletopredictbreedingvaluesfor the QTL using molecular marker datawithout recordswhenexploiting informa-tion from a single generation. Therefore,pre-selection using this approach is more

difficult to apply in practice. This meansthatfordiseaseresistanceorcarcassqualitytraits,challenge(measurement)willhavetobecarriedoutateverygeneration,inallthefamiliesavailablewithintheprogramme,asisalwaysthecaseforconventionalbreedingprogrammes.

Due to the low resolution when map-ping the QTL, it is likely that inaccurateestimates of position will lead to over-optimistic estimates of rates of geneticgain. In the simulations, it was assumedthat the QTL position was known withinthe interval and the markers surroundingthe QTL were completely informative.Thus, the increase in accuracy presentedin Table2 represents the upper bounds ofaccuracyestimates.

utilizing direct test of genes in gaS schemesThe mean phenotype of the populationfor a quantitative trait can be modifiedby increasing the frequency of favour-able alleles of genes influencing the trait.In the literature, greater genetic gain hasbeen predicted for GAS schemes than forMAS schemes (using LE populations) at

table 2empirical correlation between predicted breeding values using molecular* and pedigree information (m+Blup) or pedigree information (Blup) and true breeding values

Scenario individuals with records

family size (number of families)

10 (100) 20 (50) 50 (20)

M+ blUP

blUP M+ blUP

blUP M+ blUP

blUP

i no 0.47 0.45 0.55 0.52 0.64 0.57

YeS 0.60 0.60 0.65 0.64 0.70 0.65

ii no 0.41 0.41 0.49 0.47 0.56 0.52

YeS 0.58 0.58 0.62 0.61 0.64 0.63

* Molecular information comprises a completely informative marker bracket of 10 cM around a Qtl and all individuals genotyped for the markers. the matrix of identity-by-descent values was calculated using the deterministic method of Martinez (2003). the estimated values of h2 using residual effect maximum likelihood for the polygenic and Qtl effects were, on average, 0.13 and 0.09, respectively. the results are presented for different nuclear family sizes (number of families, between parentheses) and for candidates with or without phenotypic records. the population size was equal to 1 000, where 50 percent (Scenario i) or 25 percent (Scenario ii) of the individuals within each full-sib family had records for the trait.Source: Martinez, unpublished data.


the same rate of inbreeding (Pong-Wonget al., 2002). This is because the accuracyof predicting QTL effects using markersis always smaller than when the QTLeffects are known, as in GAS schemes. Inreality, it is likely that MAS will be car-ried out using information from manymarkerstopredictthealleliceffectsofmorethanoneQTLsimultaneouslywhereas, inGAS schemes, only a limited number ofpolymorphisms are likely to be available.Therefore, on the whole, MAS schemesmayyieldgreatergeneticresponsebecausea greater proportion of the genetic varia-tionismarkedandused.Still,moremarkergenotyping is required for MAS schemes,whichmeansthattheadditionalproportionof the variance typed should pay for theincreaseinthecostofmanymarkerstypedsimultaneously.

Due to the biology of many speciesin aquaculture, large family sizes can beused in a breeding programme. Followingthe deterministic model of Lande andThompson (1990), Figure 5 describes theeffect of family size and amount of poly-genicvariationontherelativeefficiencyofaccuracy estimates for an index using dif-ferent numbers of full-sibs measured forthe trait, versus an index also includinginformationoncandidatesforselectiongen-otypedatlocitargetedforGASschemes(V.Martinez,unpublishedresults).ForasingleQTL explaining 10percent of the geneticvariance,whentheheritabilityisrelativelylarge,familysizehasasmallimpactontheaccuracy.Ontheotherhand,whentheher-itabilityofthetraitissmall,selectionforaknownQTLhasamajorimpactonrelativeefficiency,particularlywhenthefamilysize

FiGURe 5relative efficiency of combined gaS (for different family sizes [full-sibs]) and a known qtl,

explaining 10 percent of the genetic variance) versus an index using information from full-sibs only for different values of the overall heritability (h²)

0

10

20

30

40

50

60

70

80

0 10 20 30 40 50

Family size

Rel

ativ

e ef

fici

ency

of

com

bin

ed M

AS

(fu

ll-si

bs

plu

s a

QTL

) ve

rsu

s fu

ll-si

b in

form

atio

n o

nly

(per

cen

tag

e)

0.1 0.2 0.3 0.4

the data were obtained from the ratio of square root of the variance of the indices. Selection index formulae were derived from lande and thompson (1990).


isrelativelysmall.Hence,thisapproachcanbeimportantfortraitsthatareexpensiveordifficulttomeasuresuchascarcassquality,diseaseresistanceorantibodyresponse.

Giventheresearcheffortscarriedoutatdiverse laboratoriesworldwide, it is likelythatdirecttestswillbeavailableinthenearfutureforGASschemesfordifferenttraits.WithanincreasingamountofdataonESTs,together with a greater understanding ofthe function of known genes in aquacul-turespeciesandnewgenediscovery,thereisapossibilityofmorerapidlyidentifyingand subsequently using polymorphismsthat are within coding regions. However,theresearcheffortrequiredtodeveloptestsforpolymorphismsexplainingalleliceffectscannot be underestimated, and the factorsinfluencing the profitability of GAS willinclude:• theamountofvariationexplainedbythe

testandthenumberoftests(genes)avail-ableforexplainingthephenotype;

• thefrequencyofthefavourableallelein,and the presence of the direct test (e.g.SNPs),forthetargetpopulation;

• the interaction between the polymor-phism and the background genome andpossiblepleiotropiceffectsonfitness;

• thetrade-offbetweenthemarginalreturngiven by the additional genetic gainobtainedthroughthenon-linearchangesin theallele frequencyof the favourablealleleuntilfixation;

• fixed costs of implementing genotypingandpatenting.

maS in populations in ldUsing information from dense markermaps, it is possible to make use of LDbetween the markers and the beneficialmutationsinfluencingthequantitativetraitsacrossthepopulation.Underthisscenario,therearetwopossiblewaystousetheLD

inMASprogrammesi.e.usinginformationonasinglehaplotypeeffectinLDwiththebeneficialpolymorphismacross thepopu-lation,orpredictingthetotalgeneticvalueusing genome-wide, dense marker maps(genome-wide marker-assisted selection,or G-MAS) (Lande and Thompson, 1990;Meuwissen,HayesandGoddard,2001).

The effectiveness of each scenario islargelydependentontheactualmagnitudeoftheeffectsassociatedwiththepolymor-phism, either across the whole genome orat specific genes. It is likely that, in thenear future, high-throughput SNP tech-nologywillmakedensemarkermapscosteffectiveforselectivebreedingpurposesinaquaculture. Thus, it can be expected thatLD-MAS will be implemented over thewhole genome, basically using markers tounravelthegeneticarchitectureofquantita-tivetraits.Informationfrommultipletraitsjointly and for multiple genes (and theirinteractions within and between loci) willbe used, rather than first relying on map-pingQTLinexperimentalpopulationsandthenimplementingthisinformationinMASprogrammes. A profit analysis includingmultiple traits (e.g. to study undesirablepleiotropic effects on the breeding goal)will be needed on a case-by-case basis todetermine whether the use of a single ormultiplehaplotypessimultaneouslyismostprofitableandwhichmethodofLD-MASbettersuitsthepopulationunderselection.

Specific genes are not being evaluatedwhen LD is used across the population;rather,haplotypeeffectsonthephenotypeare being estimated. As this is done on asinglegenerationacrossthewholegenome,itwouldbepossibletousethesehaplotypeeffects for selecting candidates some gen-erationsaftertheinitialestimationwithoutrelyingonphenotypes(Meuwissen,Hayesand Goddard, 2001). Recombination will


erode the initial LD and therefore it isexpected that accuracy of estimating thebreeding value of many haplotypes willdecay(ZhangandSmith,1992), theextentof the erosion being dependent on sev-eral population parameters (Meuwissen,HayesandGoddard,2001).Inpractice,theresponsetoselectionobtainedneedstobeverifiedineachgeneration;thus,re-estima-tioncanbeusedbasedonarandomsampleofindividualsfromthepopulation.

Onepossiblecaveatisthatbyassuminga certain mode of gene action (i.e. onlyadditiveeffects),theremayinfactbeamorecomplicatedgeneticarchitectureinfluencingquantitative traits. For example, whenestimatingdominanceandepistasiswiththesamedata,morehaplotypeeffectsneed tobeestimated.Therefore,itislikelythattheaccuracyofindividualeffectswilldecrease.Anotherpotential complication that ariseswhenthetruemodelinvolvesnon-additiveeffectsisthatassignmentofpotentialmatesneedstobeoptimizedtoincreasethemeanphenotypeofthepopulationsimultaneouslythroughheterosisarisingfromcombinationofdifferentQTLalleles.Inthe longterm,the frequency of homozygotes that areidentical-by-descent will increase withinthe population as a whole; consequently,methodsarerequiredtoconstraintheratesof inbreeding to obtain similar changes ofthe population mean across generations.Furthermore, expression of differentcombinations of alleles after selection will

requirere-estimationofbetween-haplotypeeffectsineachgeneration.

ConCluSionQTL mapping and MAS are not as welladvanced in aquaculture species as interrestrial plants and animals. However,themergerbetweengeneticsandgenomicsis expected to be a fertile area of researchinthecomingyearsduetotheplethoraofinformationthatiscurrentlybeinggatheredby many laboratories around the world.It is through these research efforts thatvariations affecting complex traits in fishand shellfish species may be detected andused for increasing theusefulnessofMASschemes.Inthefinalanalysis,however,allthese techniques must be cost-effective iftheyaretobeprofitableinactualbreedingprogrammes.

aCknowledgementSIamindebtedforcommentsfromMrEricHallerman who greatly helped clarify themanuscript. This chapter has been fundedpartially by: INNOVA-Corporación deFomento de la Producción (CORFO)(05CT6PP-10) on “Using functionalgenomics forunderstandingdisease resist-ance in salmon” from the Governmentof Chile; Vicerrectoria de Investigaciony Creación, Universidad de Chile hatchproject03/03312;andFondodeCienciiasyTecnologia(FONDECYT)1061190.

referenCeSAnderson, E.C. & Thompson, E.A. 2002. A model-based method for identifying species

hybridsusingmultilocusgeneticdata. Genetics160:1217–1229.Arkush, K.D., Giese, A.R., Mendonca, H.L., McBride, A.M., Marty, G.D. & Hedrick, P.W. 2002.

Resistance to three pathogens in the endangered winter-run Chinook salmon (Oncorhynchus tshawytscha): effects of inbreeding and major histocompatibility complex genotypes. Can. J. Fisheries and Aquatic Sci.59:966–975.


Bernatchez, L. & Landry, C.2003.MHCstudies innonmodelvertebrates:whathavewe learnedaboutnaturalselectionin15years?J. Evolut. Biol.16:363–377.

Bijma, P., van Arendonk, J.A.M. & Bovenhuis, J.H. 1997.Breedingvalueandvariancecomponentestimationfromdatacontaininginbredindividuals:applicationtogynogeneticfamiliesincommoncarp.(Cyprinus carpioL.).Genet.145:1243–1249.

Bongers, A.B., Bovenhuis, J.H., van Stokkom, A.C., Wiegertjes, G.F., Zandieh-Doulabi, B., Komen, J. & Ritcher, C.J.J.1997.Distributionofgeneticvarianceingynogeneticorandrogeneticfamilies.Aquaculture153:225–238.

Brown, C.R., Woolliams, J.A. & McAndrew, B.J.2005.Factorsinfluencingeffectivepopulationsizeincommercialpopulationsofgiltheadseabream,Sparus aurata.Aquaculture247:219–225.

Butler, K., Field, C., Herbinger, C.M. & Smith, B.R. 2004. Accuracy, efficiency and robustnessof four algorithms allowing full sibship reconstruction from DNA marker data. Mol. Ecol. 13:1589–1600.

Carlborg, Ö., De Koning, D.J., Manly, K.F., Chesler, E., Williams, R.W. & Haley, C.S. 2005.Methodologicalaspectsofthegeneticdissectionofgeneexpression.Bioinformatics21:2383–2393.

Castro, J., Bouza, C., Presa, P., Pino-Querido, A., Riaza, A., Ferreiro, I., Sanchez, L. & Martinez, P.2004.Potentialsourcesoferrorinparentageassessmentofturbot(Scophthalmus maximus)usingmicrosatelliteloci.Aquaculture242:119–135.

Christiakov, D.A., Hellemans, B., Haley, C.S., Law, A.S., Tsigenopoulos, C.S., Kotoulas, G., Bertotto, D., Libertini, A. & Volckaert, F.A.M. 2005. A microsatellite linkage map of theEuropeanseabass,Dicentrarchus labrax L.Genetics170:1821–1826.

Cnaani, A., Hallerman, E.M., Ron, M., Weller, J.I., Indelman, M., Kashi, Y., Gall, G.A.E. & Hulata, G.2003.Detectionofachromosomalregionwithtwoquantitativetraitloci,affectingcoldtoleranceandfishsize,inanF2tilapiahybrid.Aquaculture223:117–128.

Cnaani, A., Zilberman, N., Tinman, S., Hulata, G. & Ron, M.2004Genome-scananalysisforquan-titativetraitlociinanF2tilapiahybrid.Mol. Genet. Genom.272:62–172.

Coimbra, M.R., Kobayashi, M.K., Koretsugu, S., Hasegawa, O., Ohara, E., Ozaki, A., Sakamoto, T., Naruse, K. & Okamoto. N.2003.AgeneticlinkagemapoftheJapaneseflounder,Paralichthys olivaceus.Aquaculture220:203–218.

Corley-Smith, G.E., Lim, C.H.J. & Brandhorst, B.P. 1996.Productionofandrogeneticzebrafish(Danio rerio).Genetics142:1265–1276.

Coulibaly, I., Gharbi, K., Danzmann, R., Yao, J. & Rexroad, C.E. 2005. Characterization andcomparisonofmicrosatellitesderivedfromrepeat-enrichedlibrariesandexpressedsequencetags.Anim. Genet.36:309–315.

Doyle, R.W. & Herbinger, C.M.1994.TheuseofDNAfingerprinting forhigh intensity,within-familyselectioninfishbreeding.Proc. 5th Wrld. Congr. Genetics Appl. Livestock Prodn.,Vol.19,pp.364–371.Guelph,Ontario.

Emery, W., Boyle, C. & Noble, F.2001.Assignmentofpaternitygroupswithoutaccesstoparentalgenotypes:multiplematinganddevelopmentplasticityinsquid.Molec. Ecol.10:1265–1278.

Falconer, D.S. & Mackay, T.F.C. 1996. Introduction to quantitative genetics, 4th edition. Harrow,Essex,U.K,Longman.

Fernandez, J. & Toro. M. 2006. A new method to estimate relatedness using molecular markers.Molec. Ecol.(inpress).


Garcia de Leon, F.J., Canonne, M., Quillet, E., Bonhomme, F. & Chatain, B.1998.Theapplicationofmicrosatellitemarkerstobreedingprogramsintheseabass,Dicentrarchus labrax.Aquaculture159:303–316.

George, A., Visscher, P. & Haley, C.S.2000.Mappingquantitativetraitlociincomplexpedigrees:atwostepvariancecomponentapproach.Genetics156:2081–2092.

Gilbey, J., Verspoor, E., McLay, A. & Houlihan, D.2004.AmicrosatellitelinkagemapforAtlanticsalmon(Salmo salar).Anim. Genet.5:98–105.

Gjedrem, T.2000.Geneticimprovementofcold-waterfishspecies.Aquaculture Res.31:25–33.Gjerde, B.1986.Growthandreproductioninfishandshellfish.Aquaculture57:37–55.Gjerde, B. & Schaeffer, L.R.1989.Bodytraitsinrainbowtrout:II.Estimatesofheritabilitiesandof

phenotypicandgeneticcorrelations.Aquaculture80:25–44.Gjerde, B., Villanueva, B. & Bentsen, H.B.2002.Opportunitiesandchallengesindesigningsustain-

ablefishbreedingprograms.Proc. 7th Wrld. Congr. Genetics Appl. Livestock Prodn.,CD-ROM06-01.Montpellier,France.

Gjøen, H. & Bentsen, H.1997.Pastpresentandfutureofgeneticimprovementinsalmonaquacul-ture.ICESJ. Marine Sci.54:1009–1014.

Gjøen, H.M., Refstie, T., Ulla, O. & Gjerde, B. 1997. Genetic correlations between survival ofAtlanticsalmoninchallengeandfieldtests.Aquaculture158:277–288.

Griffing, B.1967.Selectioninreferencetobiologicalgroups.I.Individualandgroupselectionappliedtopopulationsofunorderedgroups.Austr. J. Biol. Sci.20:127–139.

Grimholt, U., Larsen, S., Nordmo, R., Midtlyng, P., Kjoeglum, S., Storset, A., Saebo, S. & Stet, R.J.2003.MHCpolymorphismanddiseaseresistanceinAtlanticsalmon(Salmo salar);facingpathogenswithsingleexpressedmajorhistocompatibilityclassIandclassIIloci.Immunogen.55:210–219.

Hayes, B. & Goddard, M.E. 2001.Thedistributionoftheeffectsofgenesaffectingquantitativetraitainlivestock.Genet. Sel. Evol.33:209–229.

Hayes, B., Lærdahl, J., Lien, S., Adzhubei, A. & Høyheim, B.2004.Largescalediscoveryofsinglenucleotide polymorphism (SNP) markers in Atlantic salmon (Salmo salar). AKVAFORSK,InstituteofAquacultureResearch(availableatwww.akvaforsk.no/norsk/HentFil.asp?fid=41).

Henryon, M., Jokumsen, A., Berg, P., Lund, I., Pedersen, P.B., Olesen, N.J. & Slierendrecht, W.J.2002. Genetic variation for growth rate, feed conversion efficiency and disease resistance existswithinafarmedpopulationofrainbowtrout.Aquaculture209:59–76.

Henryon, M., Berg, P., Olesen, N.J., Kjaer, T.E., Slierendrecht, W.J., Jokumsen, A. & Lund. I.2005.Selectivebreedingprovidesanapproachtoincreaseresistanceofrainbowtrout(Onchorhynchus mykiss)tothediseases,entericredmouthdisease,rainbowtroutfrysyndromeandviralhaemor-rhagicsepticaemia.Aquaculture250:621–636.

Hershberger, W.K., Myers, J.M., Iwamoto, R.N., Mcauley, W.C. & Saxton, A.M. 1990. Geneticchangesinthegrowthofcohosalmon(Oncorhynchus kisutch) inmarinenet-pens,producedbytenyearsofselection.Aquaculture85:187–197.

Herbinger, C.M., Doyle, R.W., Pitman, E.R., Paquet, D., Mesa, K.A., Morris, D.B., Wright, J.M. & Cook, D.1995.DNAfingerprint-basedanalysisofpaternalandmaternaleffectsofoffspringgrowthandsurvivalincommunallyrearedrainbowtrout.Aquaculture137:245–256.

Herbinger, C.M., O’Reilly, P.T., Doyle, R.W., Wright, J.M. & O’Flynn, F.1999.Earlygrowthper-formanceofAtlanticsalmonfull-sibfamiliesrearedinsinglefamilytanksversusinmixedfamilytanks.Aquaculture173:105–116


Houbert, S. & Hedgecock, D.2004.LinkagemapsofmicrosatelliteDNAmarkers for thePacificoyster(Crassostrea gigas)Genetics 168:351–362.

Hubner, N., Wallace, C.A., Zimdahl, H., Petretto, E., Schulz, H., Maciver, F., Mueller, M., Hummel, O., Monti, J., Zidek, V., Musilova, A., Kren, V., Causton, H., Game, L., Born, G., Schmidt. S., Muller, A., Cook, S.A., Kurtz, T.W., Whittaker, J., Pravenec, M. & Aitman, T.J.2005.Integratedtranscriptionalprofilingandlinkageanalysisforidentificationofgenesunderlyingdisease.Nature Genet.37:243–253.

Hulata, G.2001.Geneticmanipulationsinaquaculture:areviewofstockimprovementbyclassicalandmoderntechnologies.Genetica111:155–173.

Jamieson, A. & Taylor, S.T.1997.Comparisonofthreeprobabilityformulaeforparentageexclusion.Anim. Genet.28:397–400.

Jansen, R.C., Geerlings, H., Van Oeveren, A.J. & Van Schaik R.C.2001.Acommentoncodomi-nantscoringofAFLPmarkers.Genetics158:925–926.

Jansen, R.C. & Nap, J.P.2001.Geneticalgenomics:theaddedvaluefromsegregation.Trends Genet.17:388–391.

Jiang, C. & Zeng, Z.B. 1995.Multiple trait analysisofgeneticmapping forquantitative trait loci.Genetics140:1111–1127.

Kause, A., Ritola, O., Paananen, T., Ritola, O., Paananen, T., Mäntysaari, E. & Eskelinen, U.2002.Couplingbodyweightanditscomposition:aquantitativegeneticanalysisinrainbowtrout.Aquaculture211:65–79.

Kause, A., Ritola, O., Paananen, T., Wahlroos, H. & Mantysaari, E.A. 2005. Genetic trends ingrowth,sexualmaturityandskeletaldeformationsandrateofinbreedinginabreedingprogrammeforrainbowtrout(Oncorhynchus mykiss).Aquaculture247:177–187.

Khoo, S.K., Ozaki, A., Nakamura, F., Arakawa, T., Ishimoto, S., Nickolov, R., Sakamoto, T., Akutsu, T., Mochizuki, M., Denda, I. & Okamoto, N.2004.Identificationofanovelchromo-somalregionassociatedwithinfectioushematopoieticnecrosis(IHN)resistanceinrainbowtrout.Fish Path.39:95–102.

Kjøglum, S., Grimholt, U. & Larsen S. 2005.Non-MHCgeneticandtankeffectsinfluencediseasechallengetestsinAtlanticsalmon(Salmo salar).Aquaculture250:102–10.

Knott, S.A. & Haley, C.S.2000.Multitraitleastsquaresforquantitativetraitlocidetection.Genetics156:899–911.

Knudsen, S.2002.A biologist’s guide to analysis of DNA microarray data.NewYork,NY,USA,JohnWileyandSons.

Kocher, T. D., Lee, W. J., Sobolewska, H., Penman, D. & McAndrew, B.1998.Ageneticlinkagemapofacichlidfish,thetilapia(Oreochromis niloticus).Genetics 148:1225–1232.

Lande, R. & Thompson, R. 1990. Efficiency of marker-assisted selection in the improvement ofquantitativetraits.Genetics124:743–756.

Langefors, J., Lohm, M. & Grahn, O.2001.AssociationbetweenmajorhistocompatibilitycomplexclassIIBallelesandresistancetoAeromonas salmonicidainAtlanticsalmon.Proc. R. Soc. Lond. B268:479–485.

Lee, B.Y., Lee, W.J., Streelman, J.T., Carleton, K.L., Howe, A.E., Hulata, G., Slettan, A., Stern, J.E., Terai, Y. & Kocher, T.D. 2005.Asecond-generationgeneticlinkagemapoftilapia(Oreochromisspp).Genetics170:237–244.


Li, Y., Byrne, K., Miggiano E., Whan, V., Moore, S., Keys, S., Crocos, P. & Lehnert, S. 2003.GeneticmappingofthekurumaprawnPenaeus japonicususingAFLPmarkers.Aquaculture219:143–156.

Liu, Y., Jansen, G.B. & Lin, C.Y. 2002. The covariance between relatives conditional on geneticmarkers.Genet. Select. Evol.34:657–667.

Liu, Z.J. & Cordes, J.F. 2004. DNA marker technologies and their applications in aquaculturegenetics.Aquaculture238:1–37.

Lohm, J., Grahn, M., Langefors, A., Andersen, O., Storset, A. & von Schantz, T. 2002.Experimentalevidenceformajorhistocompatibilitycomplexallele-specificresistancetoabacterialinfection.Proc.R. Soc. Lond. B269:2029–2033.

Lynch, M. & Ritland, K.1999.Estimationofpairwiserelatednesswithmolecularmarkers.Genetics 152:1753–1766.

Marth, G.T., Korf, I., Yandell, M.D., Yeh, R.T., Gu, Z., Zakeri, H., Stitziel, N.O., Hillier, L., Kwok, P.Y. & Gish, W.R.1999.Ageneralapproachtosingle-nucleotidepolymorphismdiscovery.Nature Genet.23:452–456.

Martinez, V. 2003. Quantitative genetic analysis of fish populations using molecular markers.UniversityofEdinburgh,UK(Ph.D.thesis).

Martinez, V. 2006a. Using marker data in conventional breeding programmes: a case of study inselective breeding programs of scallops. In Proc. Internat. Assocn. Genet. Aquaculture, p 66.Montpellier,France.

Martinez, V. 2006b. Further insights of the variance component method to detect QTL in fishpopulations:relaxingtheassumptionofadditivegeneticeffects.InProc. Internat. Assocn. Genet. Aquaculture, p.66.Montpellier,France.

Martinez, V.2006c.Importanceandimplementationofmoleculartechnologiesinselectivebreedingprograms for aquaculture species. In Proc. 8th Wrld. Congr. Genet. Appl. Livest. Prodn.,Communication09-01.MinasGerais,Brazil.

Martinez, V. & di Giovanni. S. 2006. Breeding programmes of scallops: effect of self-fertiliza-tion when estimating genetic parameters. In Proc. Internat. Assocn. Genet. Aquaculture, p.65. Montpellier, France.

Martinez, V., Hill, W.G. & Knott, S.A.2002.OntheuseofdoublehaploidsfordetectingQTLinoutbredpopulations.Hered.88:423–431.

Martinez, V., Neira, R. & Gall, G.A.E. 1999. Estimation of genetic parameters from pedigreedpopulations:lessonsfromanalysisofalevinweightincohosalmon(O. kisutch).Aquaculture330:22–30.

Martinez, V., Sillanpaa, M., Thorgaard, G., Robison, B., Woolliams, J. & Knott, S.2002.Evidenceof a pleiotropic QTL influencing components of early development in double haploid lines ofrainbowtrout.Proc. 7th Wrld. Congr. Genet. Appl. Livestock Prodn.,CDROMCommunication06-08.Montpellier,France.

Martinez, V., Thorgaard, G., Robison, B. & Sillanpää, M. 2005. Simultaneous consideration ofenvironmentalandgeneticeffectsinQTLmapping:anapplicationtoearlydevelopmentindoublehaploidlinesofrainbowtrout.Genet. Res.86:209–221.

Martinez, V., Kause, A., Mäntysaari, E. & Mäki-Tanila, A.2006a.Theuseofalternativebreedingschemes to enhance genetic improvement in rainbow trout: I. One-stage selection. Aquaculture254:182–194.


Martinez, V., Kause, A., Mäntysaari, E. & Mäki-Tanila, A.2006b.Theuseofalternativebreedingschemestoenhancegenetic improvement inrainbowtrout:II.Two-stageselection.Aquaculture254:195–202.

McCune, A.R., Fuller, R.C., Aquilina, A.A., Dawley, R.M., Fadool, J.M., Houle, D., Travis, J. & Kondrashov, A.S.2002.Alowgenomicnumberofrecessivelethalsinnaturalpopulationsofbluefinkillifishandzebrafish.Science296:2398–2401.

Meuwissen, T.H.E.1997.Maximizingtheresponseofselectionwithpredefinedrateofinbreeding.J. Anim. Sci.75:934–940.

Meuwissen, T.H.E. 2003. Genomic selection the future of marker-assisted selection and animalbreeding. In Marker-assisted selection: a fast track to increase genetic gain in plant and animal breeding?(availableatwww.fao.org/biotech/torino.htm).

Meuwissen, T.H.E. & Goddard, M.E. 2001. Prediction of identity by descent probabilities frommarker-haplotypes.Genet. Select. Evol.33:605–634.

Meuwissen, T.H.E. & Goddard, M.E.2004.MappingmultipleQTLusinglinkagedisequilibriumandlinkageanalysisinformationandmultitraitdata.Genet. Select. Evol.36:261–279.

Meuwissen, T.H.E. & Woolliams, J.A. 1994. Effective sizes of livestock populations to prevent adeclineinfitness.Theor. Appl. Genet.89:1019–1026.

Meuwissen, T.H.E., Hayes, B. & Goddard, M.2001.Predictionoftotalgeneticvalueusinggenome-widedensemarkermaps.Genet.157:1819–1829.

Moav, R. & Wohlfarth, G.1976.Two-wayselectionforgrowthrateinthecommoncarp(Cyprinus carpio L.).Genet.82:83–101.

Moen, T., Hoyheim, B., Munck, H. & Gomez-Raya, L.2004a.AlinkagemapofAtlanticsalmon(Salmo salar) reveals an uncommonly large difference in recombination rate between the sexes.Anim. Genet.35:81–92.

Moen, T., Kjersti, K.T., Fjalestad, H. & Gomez-Raya, L.2004b.Amultistagetestingstrategyfordetection of quantitative trait loci affecting disease resistance in Atlantic salmon. Genet. 167:851–858.

Moen, T., Agresti, J.J., Cnaani, A., Moses, H., Famula, T.R., Hulata, G., Gall, G.A.E. & May, B.2004c.Agenomescanofafour-waytilapiacrosssupportstheexistenceofaquantitativetraitlocusforcoldtoleranceonlinkagegroup23.Aquacult. Res.35:893–904.

Morton, R. & Howarth, J.M.2005.Estimationofbreedingvaluesusingselectedpedigreerecords.Biomet.61:422–432.

Mousseau, T.A., Ritland, K. & Heath, D.D.1998.Anovelmethodforestimatingheritabilityusingmolecularmarkers.Hered.80:218–224.

Muir, W.M.2005. Incorporationofcompetitiveeffects in forest treeoranimalbreedingprograms.Genet.170:1247–1259.

Muranty, H.1996.Poweroftestsforquantitativetrait locidetectionusingfull-sibfamilies indif-ferentschemes.Hered.76:156–165.

Ng, S.-H., Chang, S., Brown, A., Koop, B. & Davidson, W.S.2005.TypeImicrosatellitemarkersfromAtlanticsalmon(Salmo salar)expressedsequencetags.Mol. Ecol. Notes5:762–766.

Nichols, K.M., Young, W.P., Danzmann, R.G., Robison, B.D., Rexroad, C., Noakes, M., Phillips, R.B., Bentzen, P., Spies I., Knudsen, K., Allendorf, F.W., Cunningham, B.M., Brunelli, J., Zhang, H., Ristow, S., Drew, R., Brown, K.H., Wheeler, P.A. & Thorgaard, G.H.2003.Aconsolidatedgeneticlinkagemapforrainbowtrout(Oncorhynchus mykiss).Anim. Genet.34:102–115.


Norris, A.T., Bradley, D.G. & Cunningham, E.P.2000.ParentageandrelatednessdeterminationinfarmedAtlanticsalmon(Salmo salar)usingmicrosatellitemarkers.Aquaculture182:73–83.

O’Brien, S.J.1991.Moleculargenomemapping:lessonsandprospects.Curr. Opin. Genet. Devel.1:105–111.

Ollivier, L.1998.Theaccuracyofmarker-assistedselectionforquantitativetraitswithinpopulationsinlinkageequilibrium.Genet.148:1367–1372.

Ozaki, A., Sakamoto, T., Khoo, S., Nakamura, K., Coimbra, M.R., Akutsu, T. & Okamoto, N.2001. Quantitative trait loci (QTLs) associated with resistance/susceptibility to infectious pan-creaticnecrosisvirus(IPNV)inrainbowtrout(Oncorhynchus mykiss).Mol. Genet. Genom.265:23–31.

Panitz, F., Holm, L.-E., Høyheim, B., Hagen-Larsen, H., Taggart, J., Cairney, M., Davey, G.C., Powell, R. & Bendicen, C.2002.ESTsequencinginAtlanticsalmon.Proc. 7th Wrld. Congr. Genet. Appl. Livestock Prodn.CD-ROMCommunicationNo06-14.Montpellier,France.

Pante, M.J.R, Gjerde, B. & McMillan, I. 2001.Effectof inbreedingonbodyweight atharvest inrainbowtrout,Oncorhynchus mykiss.Aquaculture192:201–211.

Pante, M.J.R., Gjerde, B., McMillan, I. & Mistzal, I.2002.Estimationofadditiveanddominancegeneticvariancesforbodyweightatharvestinrainbowtrout,Oncorhychus mykiss.Aquaculture204:383–392.

Paterson, S., Piertney, S.B. & Knox, D. 2004. Characterization and PCR multiplexing of novelhighlyvariabletetranucleotideAtlanticsalmon(Salmo salar L.)microsatellites.Mol. Ecol. Notes4:160–162.

Pérez-Enciso, M.2003.Finemappingofcomplextraitgenescombiningpedigreeandlinkagedis-equilibriuminformation:aBayesianunifiedframework.Genet.163:1497–1510.

Pérez-Enciso, M. 2004. In silico studyof transcriptomegeneticvariation inoutbredpopulations.Genet.166:547–554.

Pérez-Enciso, M., Toro, M.A., Tenenhaus, M. & Gianola. D.2003.Combininggeneexpressionandmolecularmarkerinformationformappingcomplextraitgenes:asimulationstudy.Genet.164:1597–1606.

Pfrender, M.E., Spitze, K., Hicks, J., Morgan, K.L. & Lynch, L.M. 2000. Lack of concordancebetweengeneticdiversityestimatesatthemolecularandquantitative-traitlevels.Conserv. Genet.1:263–269.

Piepho, H-P. & Koch, G. 2000. Codominant analysis of banding data from a dominant markersystembynormalmixtures.Genet.155:1459–1468.

Pong-Wong, R., Villanueva, B. & Woolliams, J.A.2002.Comparisonofdirectandmarker-assistedselectionwithoptimisedcontributions.InProc. 7th Wrld. Congr. Genet. Appl. Livestock Prodn.CD-ROMCommunicationNo.33-02.Montpellier,France.

Pong-Wong, R., George, A.W., Woolliams, J.A. & Haley C.S. 2001Asimpleand rapidmethodfor calculating identity-by-descent matrices using multiple markers. Genet. Select. Evol. 33:453–471.

Ponzoni, R.W., Hamzah, A., Tan, S. & Kamaruzzaman, N.2005.GeneticparametersandresponsetoselectionforliveweightintheGIFTstrainofNiletilapia(Oreochromis niloticus).Aquaculture247:203–210.

Queller, D.C. & Goodnight, K.F. 1989. Estimating relatedness using genetic markers. Evol. 43:258–275.


Quinton, C., McMillan, I. & Glebe, B.D.2005.DevelopmentofanAtlanticsalmon(Salmo salar)genetic improvement program: Genetic parameters of harvest body weight and carcass qualitytraitsestimatedwithanimalmodels.Aquaculture247:211–217.

Reid, D.P., Szanto, A., Glebe, B., Danzmann, R.G. & Ferguson, M.M.2005.QTLforbodyweightandconditionfactor inAtlanticsalmon(Salmo salar):comparativeanalysiswithrainbowtrout(Oncorhynchus mykiss)andArcticcharr(Salvelinus alpinus).Hered.94:166–172.

Rexroad, C.E., Rodriguez, M.F., Gharbi, K., Danzmann, R.G., DeKoning, J., Phillips, R. & Palti, Y.2005.Comparativemappingofexpressedsequence tagscontainingmicrosatellites inrainbowtrout(Oncorhynchus mykiss).BMC Genom.6:54.

Rise, M.L., Von Schalburg, K.R., Brown, G.D., Mawer, M.A., Devlin, R.H., Kuipers, N., Busby, M., Beetz- Sargent, M., Alberto, R., Gibbs, A.R., Hunt, P., Shukin, R., Zeznik, J.A., Nelson, C., Jones, S.R., Smailus, D.E., Jones, S.J., Schein, J.E., Marra, M.A., Butterfield, Y.S., Stott, J.M., Ng, S.H., Davidson, W.S. & Koop, B.F.2004a.DevelopmentandapplicationofasalmonidESTdatabase andcDNAmicroarray:dataminingand interspecifichybridizationcharacteristics.Genome Res.14:478–490.

Rise, M.L., Jones, S.R., Brown G.D., von Schalburg, K.R., Davidson, W.S. & Koop, B.F. 2004b.Microarray analyses identify molecular biomarkers of Atlantic salmon macrophage and hemat-opoietickidneyresponsetoPiscirickettsia salmonisinfection.Physiol. Genom.20:21–35.

Robison, B., Wheeler, P. & Sundin, K.2001.CompositeintervalmappingrevealsaQTLofmajoreffectonembryonicdevelopmentrateinrainbowtrout(Onchorhynchus mykiss).J. Hered.92:16–22.

Robison, B., Wheeler, P. & Thorgaard, G.1999.Variationindevelopmentrateamongclonallinesofrainbowtrout(Oncorhynchus mykiss).Aquaculture173:131–141.

Rodriguez, F.M., LaPatra, S., Williams, S., Famula, T. & May, B. 2004.Geneticmarkersassociatedwithresistancetoinfectioushematopoieticnecrosisinrainbowandsteelheadtrout(Oncorhynchus mykiss)backcrosses.Aquaculture241:93–115.

Rye, M. & Mao, I.L.1998.NonadditivegeneticeffectsandinbreedingdepressionforbodyweightinAtlanticsalmon(Salmo salar L.).Livest. Prodn. Sci.57:23–32.

Sakamoto, T., Danzmann, R.G., Gharbi, K., Howard, P., Ozaki, A., Khoo, S.K., Woram, R.A., Okamoto, N., Ferguson, M.M., Holm, L-E., Guyomard, R. & Hoyheim, B. 2000.Amicros-atellite linkagemapof rainbow trout (Oncorhynchus mykiss) characterizedby large sex-specificdifferencesinrecombinationrates.Genet.155:1331–1345.

Serapion, J., Waldbieser, G.C., Wolters, W. & Wolters, Z.J.2004.DevelopmentofTypeImarkersinchannelcatfishthroughintronsequencing.Anim. Genet.35:463–466.

Singer, A., Perlman, H., Yan, Y., Walker, C., Corley-Smith, G., Brandhorst, B. & Postlethwait, J.2002.Sex-specificrecombinationratesinzebrafish(Danio rerio).Genet.160:649–657.

Smith, B.R., Herbinger, C.M. & Merry, H.R. 2001.Accuratepartitionofindividualsintofull-sibfamiliesfromgeneticdatawithoutparentalinformation.Genet.158:1329–1338.

Somorjai, I.M.L., Danzmann, R.G.& Ferguson, M.M.2003.DistributionoftemperaturetolerancequantitativetraitlociinArcticcharr(Salvelinus alpinus)andinferredhomologiesinrainbowtrout(Oncorhynchus mykiss).Genetics165:1443–1456.

Streisinger, G.,Walker, C.H., Dower, N., Knauber, D. & Singer, F.1980.Productionofclonesofhomozygousdiploidzebrafish(Brachydanio rerio).Nature291:293–296.

Sun, X. & Liang, L.2004.Ageneticlinkagemapofcommoncarp(Cyprinus carpio L.)andmappingofalocusassociatedwithcoldtolerance.Aquaculture238:165–172.


Sundin, K., Brown, K.H., Drew, R.E., Nichols, K.M., Wheeler, P.A. & Thorgaard, G.H. 2005.GeneticanalysisofadevelopmentrateQTLinbackcrossesofclonalrainbowtrout,Oncorhynchus mykiss.Aquaculture247:75–83.

Tanck, M.W.T., Palstra, A.P., van de Weerd, M., Leffering, C.P., van der Poel, J.J., Bovenhuis, H. & Komen, J.2001.Segregationofmicrosatelliteallelesandresidualheterozygosityatsinglelociinhomozygousandrogeneticcommoncarp(Cyprinus carpio L.).Genome44:743–751.

Tao, W.J. & Boulding, E.G.2003.Associationsbetweensinglenucleotidepolymorphismsincandi-dategenesandgrowthrateinArcticcharr(Salvelinus alpinus L.).Hered.91:60–69.

Taris, N., Baron, S. & Sharbel, T.F.2005.AcombinedmicrosatellitemultiplexingandboilingDNAextractionmethodforhigh-throughputparentageanalysesinthePacificoyster(Crassostrea gigas).Aquaculture Res.36:516–518.

Thomas, S. C., Pemberton, J.M. & Hill, W.G.2000.Estimatingvariancecomponentsinnaturalpop-ulationsusinginferredrelationships.Hered.84:427–436.

Thomas, S.C. & Hill, W.G.2000.Estimatingquantitativegeneticparametersusingsib-shipsrecon-structedfrommarkerdata.Genet.155:1961–1972.

Thorgaard, G.H. & Allen S.K. 1987.Chromosomemanipulationandmarkers in fisherymanage-ment.InN.Ryman&F.M.Utter,eds.Population genetics and fishery management,pp.319–331.Seattle,WA,USA,UniversityofWashingtonPress.

Toro, M. & Mäki-Tanila, A.1999.Establishingaconservationscheme.InJ.K.Oldenbroek,ed.Gene banks and the conservation of farm animal genetic resources, pp.75–90. Lelystad, Netherlands,DLOInstituteofAnimalHealth.

Toro, M., Barragán, C. & Óvilo, C.2002.EstimationofcoancestryinIberianpigsusingmolecularmarkers.Conserv. Genet.3:309–320.

Toro, J.E., Sanhueza, M.A., Winter, J.E., Aguila, P. & Vergara, A.M.1995SelectionresponseandheritabilityestimatesforgrowthintheChileanoysterOstrea chilensis(Philippi,1845).J. Shellfish Res.14:87–92.

Vasemägi A., Nilsson, J. & Primmer, C.R.2005.Expressedsequencetag(EST)-linkedmicrosatel-litesasasourceofgeneassociatedpolymorphismsfordetectingsignaturesofdivergentselectioninAtlanticsalmon(Salmo salar L.).Mol. Biol. & Evol.22:1067–1076.

Villanueva, B.,Verspoor, E. & Visscher, P.M.2002.Parentalassignmentinfishusingmicrosatellitegeneticmarkerswithfinitenumbersofparentsandoffspring.Anim. Genet.33:33–41.

Villanueva, B., Pong-Wong, R. & Woolliams, J.A.2002.Markerassistedselectionwithoptimisedcontributionsofthecandidatestoselection.Genet. Select. & Evol.34:679–703.

Vos, P., Hogers, R., Bleeker, M., Reijans, M., Van De Lee, T., Hornes, M., Friters, A., Pot, J., Paleman, J., Kuiper, M. & Zabeau, M.1995.AFLP:AnewtechniqueforDNAfingerprinting.Nucl. Acids Res.23:4407–4414

Waldbieser, G.C., Bosworth, B.G., Nonneman, D.J. & Wolters, W.R.2001.Amicrosatellite-basedgeneticlinkagemapforchannelcatfish,Ictalurus punctatus.Genet.158:727–734.

Walsh, B. & Henderson, D.J.2004.Microarraysandbeyond:Whatpotentialdocurrentandfuturegenomicstoolshaveforbreeders?Anim. Sci.82:E292–E299.

Wang, J.2004.Sibshipreconstructionfromgeneticdatawithtypingerrors.Genet.166:1963–1979.Wang, L., Song, L., Chang, Y. & Xu, W. 2005. A preliminary genetic map of Zhikong scallop

(Chlamys farreri, JonesetPreston1904).Aquaculture Res.36:643–653.


Ward, R D., English, L. & McGoldrick, J. 2000. Genetic improvement of the Pacific oyster, Crassostrea gigas(Thunberg)inAustralia.Aquaculture Res.31:35–44.

Withler, R. & Beacham, T.1984.Geneticconsequencesofthesimultaneousorsequentialadditionofsemenfrommultiplemalesduringhatcheryspawningofchinooksalmon(Oncorhynchus tshawyt-scha).Aquaculture126:11–20.

Wright, J.M. & Bentzen, P.1994.Microsatellites–geneticmarkersforthefuture.Rev. Fish Biol. & Fisheries4:384–388.

Wu, W.-R., Li, W.-M., Tang, D.-Z., Lu, H.-R. & Worland, A.J.1999.Time-relatedmappingofquan-titativetraitlociunderlyingtillernumberinrice.Genet.151:297–303.

Xu, S. & Atchley, W. 1995.Arandommodelapproachtointervalmappingofquantitativetraitloci.Genet.141:1189–1197.

Xu, S. & Gessler, D. 1998. Multipoint genetic mapping of quantitative trait loci using a variablenumberofsibsperfamily.Genet. Res.71:73–83.

Young, W.P., Wheeler, P.A., Fields, R.D. & Thorgaard, G.H.1996.DNAfingerprintingconfirmsisogenicityofandrogenetically-derivedrainbowtroutlines.J. Hered.87:77–81.

Young, W.P., Wheeler, P.A., Coryell, V.H., Keim, P. & Thorgaard, G.H.1998Adetailedlinkagemapofrainbowtroutproducedusingdoubledhaploids.Genet.148:839–850.

Yue, G.H., Ho., M.Y. & Orban, L.2004.MicrosatelliteswithingenesandESTsofcommoncarpandtheirapplicabilityinsilvercruciancarp.Aquaculture234:85–98.

Zhang, W. & Smith, C.1992.Computersimulationofmarkerassistedselectionutilizinglinkagedis-equilibrium.Theor. Appl. Genet.83:813–820.

Zimmerman, A.M., Wheeler, P.A., Ristow, S.S. & Thorgaard, G.H.2005.Compositeintervalmap-pingrevealsthreeQTLassociatedwithpyloriccaecanumberinrainbowtrout.Aquaculture247:85–95.

Section Vi

Selected issues relevant to applications of marker-assisted

selection in developing countries

Chapter 18

marker-assisted selection in crop and livestock improvement: how

to strengthen national research capacity and international

partnerships

Maurício Antônio Lopes


SummaryItisgenerallyrecognizedthatmarker-assistedselection(MAS)isatoolthatbreederscanuse to accelerate the speed and precision of crop and livestock breeding in developingcountries. However, its practical application has been more difficult than previouslyexpected. Although advances in molecular marker technology have uncovered manypossibilities for transferringgenes intodesiredcrops and livestock throughMAS,moremethodological development and better planning and implementation strategies willbe needed for its successful and expeditious application to breeding programmes. Also,this technology shouldnotbe regardedas anend in itself,but as an interactingpartofcomplex strategies and decision-making processes. An appropriate mix of technologiesand capabilities together with effective approaches to networking must be viewed askey ingredients for its correct development and application to breeding programmes.Thischapterdescribessomestrategies toguidedecisionsaboutstructures,methodsandcapacities that may contribute to enhancing the access and successful use of MAS indevelopingcountries.

Chapter 18 – Marker-assisted selection in crop and livestock improvement 367

introduCtionThe tremendous advancesmade inmolec-ular marker techniques in the past twodecadeshaveledtoincreasedunderstandingof the genetic basis of many agriculturaltraits inavarietyofplantandanimalspe-cies. The use of these techniques has alsomade it possible to accelerate the transferof desirable traits among varieties and tointrogress novel genes from related wildspecies.

DNA markers have many advantagesover conventional approaches available tobreeders.Theyareespeciallyadvantageousfortraitsthatareotherwisedifficulttotag,suchasresistancetopathogens,insectsandnematodes,tolerancetoabioticstressesandqualityparameters.Theyoffergreatscopefor improving the efficiency of conven-tional breeding by carrying out selectionnotdirectlyon the traitof interestbutonlinkedgenomicregions.Additionallythesemarkers are unaffected by environmentalconditions and are detectable during allstagesofgrowth(Mohanet al.,1997).

Molecularmarkertechniqueshavethere-fore moved beyond their early projectedrole as tools for identifying chromosomalsegments and genes to uncovering manypossibilitiesforeasingthetransferofgenesintodesiredcultivarsandlines.MASgen-eratedgreatenthusiasmas itwasseenasamajor breakthrough, promising to over-come many limitations of conventionalbreedingprocesses(FAO,2003).However,despite advances in the theory of MAS,directutilizationoftheinformationitpro-videsforselectingsuperiorindividualswithcomplex traits is stillvery limited (Young,1999; Ferreira, 2003). Nevertheless, thereis still optimism about the contributionsof MAS, which is now balanced by therealization that genetic improvement ofquantitative traits using this tool may be

more difficult than previously considered(FAO,2003).In1999,Youngreviewedthedevelopment of MAS, analysing in detailitsmaindrawbacks,manyofwhichremaintoday. He concluded that because MAStechnologywassochallengingitshouldnotbeareasonfordiscouragementbut,instead,reasonformoreingenuityandbetterplan-ningandexecution.

Recent developments in high-through-put genotyping, single nucleotide polymor-phism(SNP)andtheintegrationofgenomictechnologies are advances that will play animportantroleinthedevelopmentofMASasaneffectivetoolforsustainableconservationandincreaseduseofcropgeneticresources(Ferreira,2006). However, research teams,funding agencies, commodity groups andtheprivatesectorwillneedtoworktogetherto develop MAS technology further andensurethatbreedershavethebestavailabletools. Also, the tools and strategies willneed to go beyond markers themselves toinclude genome-based knowledge derivedfrommodelsystems,high-throughputcosteffectivetechnology,aswellasbettertech-nologies and strategies for handling largevolumesofinformation.

The purpose of this chapter is to dis-cuss the access to and utilization of MAStechnology by breeding programmes,withspecialemphasisonstrategiestohelpstrengthen research capacity and partner-ships in developing countries. Wheneverpossible, recommendations are presentedtohelpguidedecisionsthatmaycontributetoenhancing theaccessandsuccessfuluseofMASbynationalprogrammes.

perCeptionS aBout the uSe of maS in Crop and liveStoCk improvement As MAS is still an evolving technology,there are not many detailed studies avail-able describing the state-of-the-art of its


applicationtobreedingprogrammes.Also,thereareveryfewprospectivestudiesindi-cating future trends in the application ofthis technology. The FAO BiotechnologyForum hosted an e-mail conference on“Molecular marker-assisted selection asa potential tool for genetic improvementof crops, forest trees, livestock and fishin developing countries”. This provideda comprehensive overview of the percep-tions of scientists from different parts ofthe world about key aspects of the appli-cation of MAS to genetic improvementin developing countries (www.fao.org/biotech/logs/c10logs.htm).

As described in Chapter 21, this FAOconferencewasveryinclusive,withatotalof627peoplesubscribing.Eightpercentofthese (52 people) submitted 85messages,whichwerereceivedfromallmajorregionsof the world, including Asia (33percent),Europe(26percent),LatinAmericaandtheCaribbean (14percent), Africa (9percent)Oceania (9percent) and North America(8percent).Peoplefrom26differentcoun-triesparticipated,withatotalof50messages(59percent)fromdevelopingcountriesand35 messages (41percent) from developedcountries. Institutional representation wasalsoample,includingnationalresearchinsti-tutes,centresbelongingtotheConsultativeGroup on International AgriculturalResearch (CGIAR), universities, consult-ants, farmer organizations, governmentagencies, non-governmental organizations(NGOs),etc.Althoughonly52peopleoutof 627 subscribers participated directly inthe conference, the number is significantconsidering the broad representation, thehigh level of the (moderated) discussionsandthenumberofrelevantissuesdiscussed(www.fao.org/biotech/logs/c10logs.htm).

Topreparethischapter,adetailedreviewwascarriedoutoftheconferenceresultsin

anattempttocapturethemainperceptionsand concerns related to access to and uti-lization of MAS in developing countries.Thisanalysisrevealedavarietyofideasandcreativesuggestionstoovercometheprob-lems of MAS utilization. Although thereis a riskofnarrowingviewson importantissuesdiscussedduringtheconference,fourmajorperceptionswereclearfromtherichcontentofthediscussions:

Perception 1.Thereisaneedfordevelop-ment of priority-setting mechanisms andcost benefit analysis to guide informeddecisions on how best to apply MAS andothertechnologicalinnovationstocropandlivestockbreedingindevelopingcountries.

Perception 2. MAS has to be under-stood as part of a complex process.Complementarities, mix of technologies,integration of capabilities and networkingmust always be viewed as key ingredi-ents for its correct application in breedingprogrammes.

Perception 3.Thereisaneedforanobjec-tive definition of public-private functionsand responsibilities in relation to fundingand development of technological innova-tionindevelopingcountries.Public-privateand north-south partnerships are essen-tial to accelerate progress and effectiveapplicationofMASandother innovationsto breeding programmes in developingcountries.

Perception 4. Developing countries mustfocus on capacity building and humanresource development oriented to shapeeffective strategies of technologicalinnovation.

Inthefollowingsections,possiblestrat-egies and alternatives to deal with thechallenges and opportunities indicatedabove are outlined, including the need


forobjectivepriority-setting,developmentof partnerships, complementarities andcapacity building for compatible humanresourceformation.

maS aS part of a Complex proCeSS – Setting prioritieS and taking aCtionBeforediscussionofMASasatechnologicalalternativetoincreasethecapacityofbreedingprogrammes it is importanttodiscussandconsiderthefutureofthebreedingprocessitself.Untilrecently,selectionwasbasedonobservablephenotypes,withoutknowledgeof the genetic architecture of the selectedcharacteristics (Dekkers and Hospital,2002). However, advances in molecularmarker techniques and rapid advances inlarge-scale sequencing are creating newperspectives for exploiting the immensereservoir of polymorphism in genomes.Moleculargeneticanalysisoftraitsinplantandanimalpopulationsisleadingtoabetterunderstandingofquantitativetraitgenetics.Morerecently,thediscoveryandscoringofsingle nucleotide polymorphisms (SNPs)using automated and high-throughputinstrumentation are already providing theincreasedresolutionneededtoanalysesetsof genes involved in complex quantitativetraits(Altshuleret al.,2000;DeLaVegaet al.,2002;Rafalsky,2002,LörzandWenzel,2005;Ferreira,2006).

What impacts will all these develop-ments have on breeding programmes? Asanticipated by Stuber, Polacco and Senior,in1999,“whengenomicsisaddedtofuturestrategies for plant and animal breeders,theprojectedoutcomesaremind-boggling.There is every reason to believe that thesynergy of empirical breeding, MAS andgenomicswilltrulyproduceagreatereffectthan the sum of the various individualactions.”Despitethepositiveviewofmany

who find technological development anopen venue for enhancement or completeredesign of traditional breeding, there aremanyuncertaintiesaboutitsfuture.Theriseofgeneticengineeringandthebio-industry,and the widespread granting of intellec-tualpropertyrights,followedbyprofoundchangesintherelationshipbetweenpublicand private science make it very difficultto anticipate future developments in bothpublicly fundedbreedingresearchand thecommercialbiotechnologyindustry.

Unfortunately,verylittleefforthasbeendirected to thinking about the future ofbreeding,especiallyindevelopingcountries(Castroet al.,2002,2006).Manypastandcurrenteventsarechangingtheperformance,therelationshipsand thespace thatpublicandprivate researchorganizationshave inthe market, raising the need for a deeperunderstandingoftheirunfoldingimpactsonthepublicactivityofresearch(Price,1999;Graffet al.,2003).Thecurrentscenarioofchanges and uncertainties has generatedthe necessity for strategic re-alignment ofpublicresearchinmanypartsoftheworld.Therefore, research organizations needinformation that is not currently availableabout such changes and influences andtheirimpactonthefutureofkeyactivities,such as crop and livestock breeding. Toobtain and to organize this information,prospective studies need to be developedon the present and future performanceof breeding programmes and their relatedproductionsystems.

The future configuration of breedingprogrammes depends on knowledgeto guide strategic decisions about struc-tures, methods and capacities in order totake advantage of new opportunities andtechnological niches. Foresight method-ologies have been applied to this end,using systemic analysis of the past and


present performance of a research field,determining critical factors of perform-ance(LinstoneandTuroff,1975;Castro,deCobbeandGoedert,1995;Castro,deLimaandFreitasFilho,1998;Castroet al.2002,2006;Limaet al.,2000).

An innovative model of a prospectivestudy was proposed and tested by Castroet al. (2002, 2006), based on the Braziliannational system of genetic resources andbreeding. The effort started with thedistinction between two component sub-systems– public and private. The authorsconsidered that the two subsystems admittwo possible states or situations, currentand future, after the effect of current andemerging events (Figure 1). Prospectiveeffortsbasedonthisframeworkcanbeveryuseful to guide diagnosis of national pro-grammes,identifyingthemaindeterminantsof current and past system performancethat can be used to guide decisions about

the configuration of genetic resources,breeding programmes and the associatedseedindustry.

This type of study can help identifychanges in the system and in the corres-ponding technology market, analysingtheir current and future impacts, deter-mining future opportunities and threatsto the strategic positioning of researchorganizations in the technology market.Thereisalsotheperspectiveofdevelopingpossible alternative scenarios for therelationships between public and privateresearch, and of these with the market, toguide the strategic positioning of publicresearch. Results of this effort couldindicate new opportunities and niches forpublic breeding programmes, as well asareas of extreme value where the publicsector would have to acquire capacity inthe future. Key decisions on investmentsinnewtechnologiesandprocessesapplied

FiGURe 1Conceptual framework of a prospective study on genetic resources and breeding r&d in Brazil

Public subsystem of genetic resources and breeding R&D

(current state)

Public subsystem of genetic resources and breeding R&D

(future state)

Private subsystem of genetic resources and breeding R&D

(current state)

Private subsystem of genetic resources and breeding R&D

(future state)

Current and Emerging Events Development of biotechnology, widespread granting of intellectual property, changes

in the relationship between public and private science, the rise of geneticengineering, development of bio-industry, concentration in the seed market, etc.

Dark arrows indicate the impact of current and emerging events on both public and private subsystems of R&D in genetic resources and breeding, at present and in the future, considering several alternative scenarios. Vertical arrows indicate the state of the relationship between the public and the private R&D subsystems as it is affected by current and emerging events.Source: castro et al., 2002, 2006.


togeneticresourcesandbreedingresearch,such as MAS, genomic tools, transgenictechnology and others, are better taken iftheseresultsareavailable.

The results of this forward-lookingapproach developed in Brazil allowed theidentification of some important trendsthat must be considered by managers inthe process of adapting breeding effortsfor the future (Castro et al., 2002, 2005,2006). Current and emerging events iden-tified in the process will certainly affectthe performance, methods, technologicalprocesses, portfolio of products and insti-tutionalrelations inthepublicandprivateR&DsectorsdedicatedtoplantbreedinginBrazil. This complexity indicates that it isquite dangerous for developing countries,pressured by market evolution and rapidexpansionofmethodsandtechnologies,toface the challenge of identifying priorityareas for investment without a minimumprospectiveeffort.

In summary, the ability to predictchanges thatmight affect theperformanceof public and private R&D organizationsis essential for decision-makers and man-agers toguide adjustments in the focusofthese sectors in a timelymanner, avoidingthreatsandpromotingaccesstonewtoolsandopportunities.Althoughthesamepro-spectivemethodologymaybeappliedtoawiderangeofcountries, it is important topoint out that situations differ drasticallyfromcountrytocountry,therebyrequiringexamination of future configuration of asectoronacase-by-casebasis.

maS aS part of a Complex proCeSS – Building CapaCitieS, ComplementaritieS and enhanCing networkingMAScannotbeconsideredanendinitselfor a tool detached from the complexi-

ties of breeding strategies. It has to beunderstoodandanalysed inthecontextofan interacting mix of tools and strategiesthathave tobe targeted towardscropandlivestock improvement in a coordinatedmanner. Independently of the outcome ofanypriority-settingeffort, theneedforanexpandednetworkingapproachtobreedingand biotechnological research will alwaysbe an objective to be pursued. This needarises because networking and partner-ships are essential to enable organizationstoattainotherwiseunattainablegoals,addvalue to their products and processes andreducecosts.Also,thecontinuousdemandfor efficiency and relevance presses R&Dprogrammes to move in the direction ofcooperationandalignmentofefforts.

One of the key problems limiting theuse of MAS and other advanced technol-ogies in developing countries is exactlythe difficulty of building effective teamsand networks. Unfortunately, very fewdevelopingcountrieshavetrainedscientistsandadvancedfacilitiesconcentratedinoneplace or institution. Usually, these scarceresourcesarescatteredoverdifferentplacesand institutions, and many times away ordisconnected from the relevant breedingprogrammes.Thisisaseriousdrawbackastheincreasinginterdependenceoftraditionaland upstream disciplines makes it neces-sarytobuildandmanagemultidisciplinaryteamsconsistingofbreeders, agronomists,molecular biologists, biochemists, pathol-ogists, entomologists, physiologists, soilscientists, statisticians,etc.–agoalalwaysdifficulttoachieve.Inadditiontothechal-lenge of working within team alignmentsandcooperation,thereisthepressingneedtodevelopways to share capacities, infra-structure,materialsandinformationamongresearch teams located across a country, aregion,orevencontinents.


Themainprobleminfosteringcollabo-rationandeffectivecooperationtoachievecommongoalsseemstobethedifficultyofrecognizingthatdifferentteamsandorgan-izations have different general interestsand norms. For this reason, competitionusually prevails. While it has been wellacceptedthatcompetitionisoneofthekeyforces that keep industry competitive anddynamic, this view is being challenged bythe concept that many activities can ben-efit from a rational mix of competitionandcooperationthat leads tocomplemen-taryproductsandexpansionofpossibilitiesthroughtheformationofnewrelationshipsorevennewmodesofoperationandman-agement.Increasingly,thesameisalsotrueforR&Dorganizations,whichcanbenefitfromworkingwithpartners (competitors)whoseabilitiesmaketheirownmoreattrac-tive in the eyes of clients (Brandenburgerand Nalebuff, 1997). Also, faced withgrowing competition from industry andincreasing pressures and demands, publicR&Dinstitutionsmustlookatwaystodomore with fewer resources. Collaborationthrough team nets and other networkingstrategieshavethepotentialtoreducecosts,addvalueandpromotecapacitytorespondquickly to changes. Besides, with the newtoolsofinformationtechnology,collabora-tionwithanypartoftheworldispossibleas this promotes information and otherresourcesharingwithouttheneedforgeo-graphical proximity (Lipnack and Stamps,1993).

HowshouldaR&Dorganizationbehavein a multifaceted relationship, when part-nerscanbealsocompetitors?Organizationsthat enter competitive collaboration mustbeawarethattheirpartnersmaybeouttodisablethem.Thisdilemmahasbeenfacedby a growing number of organizations,which rapidly understand that effective-

ness will be more and more a productof recognizingandusing interdependence.With networks and interdependent teams,cooperationmustbedesignedinthenameof mutual needs and with a clear sense ofsharing risks to reach objectives that arecommontoallpartners(Lopes,2000).

In many parts of the world, includingin developing countries like Brazil, com-petitive funding systems for agriculturalR&D are assuming growing importanceas new sources of funding and as driversfor cooperation among universities, R&Dinstitutes and the private sector, in manycases allowing collaboration even amonginstitutionsthataretraditionalcompetitors(Lopes,2000).Althoughtherulesandpro-ceduresgoverningthecompetitivegrantingsystemindicatetheneedforpartnershipandthegeneralmodeofinteraction,experiencehas shown that industry/university/R&Dinstitutescooperationssucceedonlyiftheyarefoundedontrustandunderstandingandpromise mutual benefits. Also, successfulexperienceshavecome from theclear rec-ognition of objectives and well structuredmanagementwithintensecommunication.

Two experiences are described belowthat rely heavily on cooperation and net-working directed to effective applicationof advanced technologies, including MAS,to genetics and breeding. Both are excel-lent examples of strategies that promoteeffectivepartnershipsandcollaborationbyresearchersfromdifferentinstitutions,dis-ciplines or countries working on specifichigh-priorityprojects.

the case of the Cgiar generation Challenge programme: an internation-al r&d network in genetic resources, genomics and breedingAsthenumberofstakeholdersintheagricul-turaldecision-makingprocessincreasesand


the agricultural research agenda expands,organizationsmustbeabletorespondtoanincreasinglydiverseandcomplexportfolioof priorities by strengthening interactionswithin the system and developing linksandpartnershipswithgroups traditionallyoutside the system. Towards this end, theCGIARhasdesignedastrategytonurturethe definition of objective R&D agendasin key themes and to guide scientists andteams worldwide towards integrated, syn-ergistic involvement and operation. Thisstrategy became known as the “GlobalChallenge Programmes” (www.cgiar.org/impact/challenge/index.html).

The strategy of the Global ChallengeProgrammesrecognizesboth that thecostof conducting research is escalating andthat the complexity of the science neededfor agricultural research is increasing.Research in most fields requires not onlyspecialized equipment and facilities butalso highly trained technical support indiversedisciplines.Increasingly,multidisci-plinaryteamsofscientistswillberequiredto address the complex issues facing agri-cultureand,inmanycases,theprofessionalexpertiseneededmayhavetobeaccessedindifferentpartsoftheworld.

One such Challenge Programme, enti-tled“UnlockingGeneticDiversityinCropsfortheResource-Poor”,alsoknownasthe“GenerationChallengeProgramme(GCP)”(CGIAR, 2003) is an international, multi-institutional, cross-disciplinary publicplatformforaccessinganddevelopingnewgeneticresourcesusingadvancedmoleculartechnologies associated with conventionalmethods. Founded in July2003 by theExecutive Council of the CGIAR withstart-upfundingfromtheWorldBankandtheEuropeanCommission,theGCPhasamembershipoftwenty-twopublicresearchinstitutions around the world, including

nine CGIAR centres, four advancedresearch institutes and nine national agri-cultural research system institutions. Itsbudget in 2005 totalled at US$14 million(GCP,2005).

This platform was designed to ensurethattheadvancesofcropscienceandtech-nologyareappliedtothespecificproblemsand needs of resource-poor people whorelyonagricultureforsubsistenceandtheirlivelihoods.TheGCPaimsto“bridgethatgapbyusingadvancesinmolecularbiologyand harnessing the rich global stocks ofcrop genetic resources to create and pro-vide a new generation of plants that meetthesefarmers’needs”.

The concrete objective of the GCP istoaccessanddevelopgenomicandgeneticresources as enabling technologies andintermediate products for crop improve-mentprogrammes.Itwillnotproduceandrelease finished crop varieties for farmers,but develop new genetic resources andmake the initial gene transfers to locallyadapted germplasm, and then transfer thederived materials to crop improvementprogrammes, particularly those conductedin national agricultural research systemsof developing countries, and to any otherentitiesthathavecropimprovementgoals,especiallythosededicatedtoresource-poorfarmers.

TheGCPis,todate,themostcompre-hensiveefforttocover,inawellstructuredand feasible manner, the complex interac-tions between genetic resources, genomicsandbreeding(Figure2)inordertocapturethe benefits of the revolutions in biologyand direct them to help solve some ofthe agricultural problems in the world’smost difficult and marginal environments.Itaddressesitsthreekeycomponentpartsin a separate but interconnected manner:(1)geneticresourcecollectionsprovidethe


rawmaterials;(2)genomicscienceprovidesthemeanstoexploitgeneticresources(i.e.identifynewalleles);and(3)cropimprove-ment applies traditional and modernmethods of gene/allele transfer into func-tionalcropvarieties(CGIAR,2003).

TheGCPisthereforeanambitiousini-tiativetoputintoactionacomplexmixoftools,capacities,conceptsandstrategies.Itis organized and managed to direct theseresourcestowardsthepursuitofgoalsthatare not attainable through the discipli-naryand isolatedmodesofoperation thatunfortunately prevail in the internationalagriculturalR&Darena.Assuch,itispos-siblythebeststructuredinternationaleffortfor development, adaptation and promo-tionof effective (and inclusive) access anduseoftoolssuchasMAS.

Aspartofitscomplexstrategy,theGCPwill define protocols for more efficientgene transfer including molecular markers

thatarecloselylinkedtothegenesforthedesiredtrait,rapidtestsforphenotyperec-ognition, and genetic transformation ofnew genes into locally adapted geneticmaterials, such as improved varieties andlandraces. All of these strategies dependon the adaptation and development ofmarker technology and marker-assistedprocedures, hopefully helping to consoli-date a networking approach to breedingand biotechnological research with effec-tive impact, especially on resource-poorcountries.

ResearchactivitiescommencedinJanuary2005 with the first round of competitiveresearch grants awarded for 17three-yearprojects of approximately US$1 millioneach. In early 2005 a new round of com-missionedgrantswasstarted,whichservedas the basis of the GCP platform of toolsand technologies for genetic studies andapplications. In total, the GCP initiated

FiGURe 2Conceptual basis for the generation Challenge programme – unlocking genetic diversity

in Crops for the resource-poor

Reprinted by permission from the proposal for a cGiaR challenge Programme (cGiaR, 2003).


67competitiveandcommissionedresearchprojects andcapacity-buildingactivities in2005. The 2005 Annual Report and 2006Work Plan summarizes research progressandcapacitybuildingachievementsin2005andpresentsanoverviewofthecompetitiveand commissioned research portfolio andthecapacity-buildinganddeliveryactivitiesfor2006(GCP,2005,2006).

the Case of the genolyptus programme in Brazil: a public/private network in genomics and molecular breeding of Eucalyptus“The challenge and the opportunity forpublicly supported agricultural researchare not in duplicating the private sector’sresearch agenda, but in building uniquepublic/privatepartnershipsorperhapsevenjointly supported consortia for agricul-turalresearch”(CAST,1994).Increasingly,agricultural research will be conducted

through partnerships among private com-panies, public research institutes anduniversities(Figure3).Informingsuchalli-ances, these organizations must recognizethat developing productive relationshipsinvolves non-competitive dialogue andunderstandingofeachothers’abilitiesandlimitations.Partnershipswill flourishonlyif founded on trust and understandingandifdifferences indriversandobjectivesare recognized and accommodated in ini-tiatives with a real perspective of mutualbenefits(Lopes,2000).

An example of a successful public/pri-vatepartnershipwithclearunderstandingofpartners’abilitiesandlimitationsandcleardefinition of responsibilities and benefitstobepursuedis theGenolyptusNetworkin Brazil (Grattapaglia, 2003). This R&Dnetworkwascreatedtoestablishafounda-tion for a genome wide understanding ofthe molecular basis of wood formation in

FiGURe 3research priorities and partnerships among private companies, research institutes

and universities

PRIVATE COMPANIESPRIVATE COMPANIES RESEARCH INSTITUTESRESEARCH INSTITUTES UNIVERSITIESUNIVERSITIES

THE

MUTUAL

BENEFIT

ZONE

Usually strategic or applied

Relatively short time horizons

Ultimately commercially beneficial

Usually basic, fundamental

or speculative

No specific time horizons

or limits

Generation and dissemination

of new knowledge

RESEARCH PRIORITIESRESEARCH PRIORITIESRESEARCH PRIORITIES RESEARCH PRIORITIESRESEARCH PRIORITIESRESEARCH PRIORITIES

PRIVATE COMPANIESPRIVATE COMPANIES RESEARCH INSTITUTESRESEARCH INSTITUTES UNIVERSITIESUNIVERSITIES

THE

MUTUAL

BENEFIT

ZONE

Usually strategic or applied

Relatively short time horizons

Ultimately commercially beneficial

Usually basic, fundamentalor speculative

No specific time horizonsor limits

Generation and disseminationof new knowledge

RESEARCH PRIORITIESRESEARCH PRIORITIESRESEARCH PRIORITIES RESEARCH PRIORITIESRESEARCH PRIORITIESRESEARCH PRIORITIES

Source: adapted from laider (1998).


Eucalyptus, coupled with the translationofknowledgeintoimprovedtreebreedingtechnologies.

This programme relies heavily on thedevelopment of aligned R&D efforts ingeneticresources,genomicsandmolecularbreeding(Figure4).Itmobilizescapacitiesand infrastructure in constructing phys-ical maps, developing expressed sequencetag (EST) databases, generating a databaseof expression profiling of genes that con-trolkeytraitsanddevelopingmethodsforMASfortraitsofhighheritabilityinwoodformation. Also, the network develops acapacity-building and training programmefor professionals in universities and for-estry companies, targeting the integrationofgenetics, genomics andbreedingeffortsofEucalyptus.

Therationaleofthenetworkisbasedontheunderstandingthat,evenwiththemorepowerfultoolsallowingamuchmoreglobal

and integrated view of genetic processes,genomicswillonlysucceedincontributingto the development of improved eucalyptif it is deeply interconnected with inten-sive fieldwork and creative breeding. TheGenolyptus project therefore differs fromotherplantgenomeinitiativesintheinten-sity, refinements and scope of the effortdevoted to field experiments to generatethe diversity of phenotypes necessary tostudygenefunction.Quantitativetraitloci(QTL)detection,thedevelopmentofSNPhaplotypes for association mapping andphysicalmappingwilllinkthephenotypesto genes that control processes of woodformationthatdefineindustrial leveltraits(Grattapaglia,2003).

A key feature of the Genolyptus net-work is its pre-competitive nature. Theresearch programme was designed collab-oratively with no immediate intention ofmarketing its results, even although its

FiGURe 4the genolyptus project – from phenotypes to genotypes in an integrated way

PHENOTYPES

QTL & candidate genes

GENES & LINKED

MARKERS

LINKAGE AND ASSOCIATION

MAPPING

Growth , floweringWOOD

PROPERTIESDisease resistance

FIELD EXPERIMENTS

24 connected full-sib families

among divergent trees

Large progeny sizes (> 1200

individuals)

Partial clonal replication

Several genetic backgrounds and

locations

QTL MAPPING

Genetic map construction in

multiple families

Genome scan with 200+

microsats in multiplexes

Outbred QTL models

PHYSICAL MAPPING

BAC library

Anchoring with genetic map

Complete physical map

BAC end sequencing

EST mapping

Full gene and promoter

identification

GENE DISCOVERY

Several cDNA libraries

EST sequencing

Candidate genes

SSRs and SNPs in cDNAs

Full length cDNA database

Bioinformatics: data mining

Microarrays and SAGE

SNP and association mapping

√

√

√

√

√

√

√

√

√

√

√√

√

√

√

√

√

√

√

√

Source: Grattapaglia, 2003.


planned outputs will eventually lead tothe creation of many new products andprocesses of commercial value. Thus, theactivitiesduringitsfirstphasearedesignedto resolve basic, common technologicalproblems – a sufficient reason to mobi-lize several private companies (that arecompetitors in the market) and publicorganizations. After the first phase of sixyears, the network will have generated aconsolidated base of knowledge and toolsthatwillpromotethedevelopmentofspe-cificinterestprojects,eitherinpartnershipor individually, according to specificbusi-nessstrategiesandmarkettargets.

Also, team organization and man-agement are based on modern tools andconcepts, involving a competent, highlyrespectedscientistwithtalenttoleadnet-workoperations,asteeringcommitteeanda technical committee for adequate plan-ning, decision-making and follow-up, aswell as contract models and negotiationstrategiesappropriatetothecomplexityofthe network. Intellectual property rightsprovisions are based on access limited toparticipants,withall geneticmaterials andpatents produced being co-owned by the20participatinginstitutions.Scientificpub-licationsarehighlyencouraged.

As in the Generation ChallengeProgramme,theGenolyptusnetworkisanoriginal initiative to integrate and align acomplex mix of tools, capacities, conceptsandstrategies.Theabilitytomobilizesucha wide range of organizations, including12privatecompaniesoperatinginahighlycompetitivemarketspace,indicatesthatthenetwork design was successful, while itspre-competitive nature, organization andmanagementstrategyallowedthedefinitionof a “zone of mutual benefits” (Figure3),facilitatingthepursuitofgoalsthatarenotattainable through isolated research. The

Genolyptusnetworkisthereforeanexcel-lentexampleofthefeasibilityofdevelopinga structured public/private effort for inte-grated and effective use of advanced toolssuchasMAS.

ConCluSionS• Although advances in molecular marker

technology have uncovered many pos-sibilities for easing the transfer of genesintodesiredcropsandlivestockthroughMAS, there is still limited recordedimpactofthesetechnologiesinbreedingprogrammes.

• It is generally recognized that geneticimprovement of complex traits usingMAS is more difficult than previouslyconsidered. Therefore, more methodol-ogy development, better planning andimplementationstrategieswillbeneededforitssuccessfulandrapidapplicationtobreedingprogrammes.

• The future configuration of breedingprogrammes is dependent on knowl-edge to guide strategic decisions aboutstructures, methods and capacities thattake advantage of new opportunitiesandtechnologicalniches.Unfortunately,there are very few efforts directed atthinking about the future of breedingprogrammes,especiallyinlessdevelopedcountries. Research organizations needinformation,whichisnotcurrentlyavail-able, about changes and influences andtheir impact in the futureonkeyactivi-tiessuchascropandlivestockbreeding.To acquire and organize this informa-tion, prospective studies on the presentandfutureperformanceofbreedingpro-grammes and their related activities willhavetobedeveloped.

• Priority-setting strategies, together withcost–benefit analysis are necessary toguide informed decisions on how best


toapplyMASandotheradvancedtech-nologies to crop and livestock breedingindevelopingcountries.

• MAS has to be understood and under-takenaspartofacomplexprocess.Com-plementarities and a mix of technolo-giesandcapabilities,togetherwitheffec-tive approaches to networking, must beviewedaskey ingredients for its appro-priate development and application tobreedingprogrammes.

• OneofthekeyproblemslimitingtheuseofMASandotheradvancedtechnologiesin developing countries is the difficultyofbuildingeffectiveteamsandnetworks.Approaches to networking and partner-ships are key to enabling organizationsto attain new goals with less cost andto adding more value to their productsand processes. Also, the demand forefficiency and relevance presses R&D

programmestomoveinthedirectionofcooperationandalignmentofefforts.

• The present and future challenges andopportunities for agricultural researchorganizationsaretobuildpublic/privatepartnerships or new types of consortiadedicatedtoinnovation.Informingsuchalliances,theseorganizationsmustrecog-nizethatdevelopingproductiverelation-ships involves non-competitive dialogueand understanding of each others’ abili-ties and limitations. In order to surviveand flourish, partnerships have to besustainedontrustandunderstanding.

• Developing countries must focus ontrainingtobuildandshapecapacitiesandeffective strategies to support researchin advanced biology applied to breed-ing.Also,newmanagementstrategiesareneeded to deal with the complex natureofmodernbreedingprogrammes.

referenCeSAltshuler, D., Pollara, V.J., Cowles, C.R., Van Etten, W.J., Baldwin, J., Linton, L. & Lander, E.S.

2000.AnSNPmapofthehumangenomegeneratedbyreducedrepresentationshotgunsequencing. Nature407:513–516.

Brandenburger, A.J. & Nalebuff, B.J.1997.Co-opetition,NewYork,USA,Doubleday.CAST (Council for Agricultural Science and Technology).1994.Challengesconfrontingagricul-

turalresearchatlandgrantuniversities.IssuePaper5.Ames,IA,USA.Castro, A.M.G., de Cobbe, R.V. & Goedert, W.J. 1995.Manualdeprospecçãodedemandasparao

SNPA.Brasília,Brazil,Embrapa.Castro, A.M.G., Lima, S.M.V. & Freitas Filho, A. 1998. Manual de Capacitación en Análisis de

CadenasProductivas.Brasília,Brazil,Embrapa.Castro, A.M.G., Lima, S.M.V., Lopes, M.A. & Martins, M.A.A.G.2002.EstratégiadeP&Dparao

MelhoramentoGenéticoemumaÉpocadeTurbulência.Anais do XXII simpósio de gestão da ino-vação tecnológica, p.17.Salvador,BA,Brasil.

Castro, A.M.G., Lopes, M.A., Lima, S.M.V. & Bresciani, J.C. 2005. Trends and technologicalstrategy in the Brazilian seed sector (Cenários do Setor de Sementes e Estratégia Tecnológica). Revista de Política Agrícola Brasília3:58–72.

Castro, A.M.G., Lima, S.M.V., Lopes, M.A., Machado, M.S. & Martins, M.A.G. 2006.OFuturodoMelhoramentoGenéticoVegetalnoBrasil–ImpactosdaBiotecnologiaedasLeisdeProteçãodeConhecimento(ThefutureofplantbreedinginBrazil–impactsofbiotechnologyandintellectualpropertylegislation).Brasília,DF,EmbrapaInformaçãoTecnológica.


CGIAR (Consultative Group on International Agricultural Research).2003.Unlocking Genetic Diversity in Crops for the Resource-poor: A Proposal for a CGIAR Challenge Program(availableatwww.generationcp.org/sccv10/sccv10_upload/CP_proposal.pdf).

Dekkers, J.C.M. & Hospital, F.2002.Theuseofmoleculargeneticsintheimprovementofagricul-turalpopulations.Nature Revs. Genet.3:22–32.

De La Vega F.M., Dailey D., Ziegle J., Williams J., Madden D. & Gilbert D.A. 2002.Newgenera-tion pharmacogenomic tools: a SNP linkage disequilibrium map, validated SNP assay resource,andhigh-throughputinstrumentationsystemforlarge-scalegeneticstudies.Biotechniques Suppl32:S48–S54.

FAO.2003.Molecularmarker-assistedselectionasapotentialtoolforgeneticimprovementofcrops,foresttrees,livestockandfishindevelopingcountries.Background document to Conference 9 of the FAO Biotechnology Forum(availableatwww.fao.org/biotech/C10doc.htm).

Ferreira, M.E.2003.MelhoramentoGenéticodoArroz:ImpactosdaGenômica.InA.Borém,M.Giúdice & T. Sediyama, eds. Melhoramento genômico. Universidade Federal de Viçosa,Viçosa,Brazil.

Ferreira, M.E.2006.Molecularanalysisofgenebanksforsustainableconservationandincreaseduseofcropgeneticresources.InJ.Ruane&A.Sonnino,eds.The role of biotechnology in exploring and protecting agricultural genetic resources.FAO,Rome(availableatwww.fao.org/docrep/009/a0399e/a0399e00.htm).

GCP (Generation Challenge Programme). 2005. Generation Challenge Programme research highlights 2005.Mexico,DF.CIMMYT(availableathttp://www.generationcp.org/sccv10/sccv10_upload/Highlightslittle.pdf).

GCP.2006.Generation Challenge Program Annual Report and Year Three Workplan.Mexico,DF,(availableatwww.generationcp.org/comm/manual/doc_100.pdf).

Graff, G., Cullen, S., Bradford, K., Zilberman, D. & Bennet, A.2003.Thepublic-privatestructureofintellectualpropertyownershipinagriculturalbiotechnology.Nature Biotech.21:989–995.

Grattapaglia, D.2003.Genolyptus.In:A.Borém,M.Giúdice,&T.Sediyama,eds.Melhoramento genômico,pp.51–71.UniversidadeFederaldeViçosa,Viçosa,Brazil.

Laider, D.A.1998.Research partnerships between industry & universities. A guide to better practice.Runcorn,UK,ICITechnology.

Lima, S.M.V., Castro, A.M.G., de Mengo, O. & Medina, M.2000.Entorno y Enfoque Estratégico en I & D: Proyecto Nuevo Paradigma. ArtigosubmetidoaoXXISimpósioInternacionaldeGestãoTecnológica,FIA/USP/PACTO.

Linstone, H.A. & Turoff, M.1975.The Delphi method: techniques and applications.Readings,MA,USA,AddisonWesley.

Lipnack, J. & Stamps, J. 1993. The age of the network. New York, USA, Oliver WrightPublications.

Lopes, M.A. 2000. Cooperation and competition in competitive grants: is “coopetition” beingachieved? In Proc. of the Competitive Grants in the New Millenium: A Global Workshop for Designers and Practitioners, pp.219–230.Brasília,Brazil,Embrapa/IDB/TheWorldBank.

Lörz, H. & Wenzel, G.2005.Molecularmarkersystemsinplantbreedingandcropimprovement.Biotech. in Agric. & Forestry55:XXII.Berlin,Germany,Springer.

Mohan, M., Nair, S., Bhagwat, A., Krishna, T.G. & Yano, M.1997.Genomemapping,molecularmarkersandmarker-assistedselectionincropplants.Mol. Breeding3:87–103.


Price, S.C.1999.Publicandprivateplantbreeding.Nature Biotech.17:938.Rafalski, J.A.2002.Novelgeneticmappingtools inplants:SNPsandLD-basedapproaches.Plant

Sci.162:329–333,Stuber, C.W., Polacco, M. & Senior, M.L. 1999.Synergyofempiricalbreeding,marker-assistedselec-

tion,andgenomicstoIncreasecropyieldpotential.Crop Sci.39:1571–1583.Young, N.D. 1999. A cautiously optimistic vision for marker-assisted breeding. Mol. Breeding 5:

505–510.

Date post:	20-Jul-2020
Category:	Documents
Upload:	others
View:	7 times
Download:	0 times

Chapter 17 marker-assisted selection in fish and shellfish ... · marker information in...

Documents