From plant genomics to breeding practice

From plant genomics to breeding practice Michele MorganteÃy and Francesco Salamini z§New alleles are constantly accumulated during intentional crop selection.The molecular understanding of these alleles has stimulated new genomic approaches to mapping quantitative trait loci(QTL)and haplotype multiplicity of the genes concerned.A limited number of quantitative trait nucleotides responsible for QTL variation have been described,but an acceleration in their rate of discovery is expected with the adoption of linkage disequilibrium and candidate gene strategies for QTLfine mapping and cloning.Additional layers of regulatory variation have been studied that could also contribute to the molecular basis of quantitative genetics of crop traits.Despite this progress,the role of marker-assisted selection in plant breeding will ultimately depend on the genetic model underlying quantitative variation.AddressesÃDipartimento di Produzione Vegetale e Tecnologie Agrarie,Universita’di Udine,Via delle Scienze208,33100Udine,Italyy e-mail:morgante@dpvta.uniud.itz Max-Planck-Institut fuer Zuechtungsforschung,Carl-von-Linne-Weg 10,50829Koeln Germany§e-mail:salamini@mpiz-koeln.mpg.deCurrent Opinion in Biotechnology2003,14:214–219This review comes from a themed section onPlant biotechnologyEdited by Csaba Koncz0958-1669/03/$–see front matterß2003Elsevier Science Ltd.All rights reserved.DOI10.1016/S0958-1669(03)00028-4AbbreviationsLD linkage disequilibriumMAS marker-assisted selectionQTL quantitative trait lociQTN quantitative trait nucleotidesSNP single nucleotide polymorphismIntroductionSpecies relevant to agriculture have a long breeding history.During this period,two bottlenecks have restricted the genetic base of breeding populations:spe-cies domestication,which for a large part of crops has been monophyletic[1–3],and the post-Mendelian adop-tion of breeding procedures separating environmental from genetic effects(i.e.only genetically superior vari-eties have entered intentional crossing programs). Despite the narrow base of current breeding materials, substantial genetic progress has been achieved for crop plants[4].In almost all cases the rate of genetic gain appears linear over time(sting for hundreds of generations[5 ]).This type of response calls for the existence of very many genes affecting each relevant trait,a situation known as the Fisher’s infinitesimal model [6].Recent major contributions to the study of quantita-tive variation in natural and experimental populations [5 ,7]indicate,however,that quantitative trait loci (QTL)can support large phenotypic effects.When super-ior alleles at relevant QTL arefixed,intentional selection should have exhausted a large part of the genetic variance present in populations,thus reducing the rate of genetic advance.It is true that the number of QTLs supporting a specific trait may be under-estimated[5 ],an evidence in favour of the infinitesimal model,but it is also known that nucleotide diversity is not random but structured in haplotypes[8 ,9,10]having significant allelic differences in terms of phenotype[7].Barton and Keightley[5 ] have discussed a genetic model that explains long-term selection responses,while accepting the existence of relatively few and major QTL for a trait.The model includes the possibility that during intentional selection new QTL alleles appear and reinforce trait variances. Besides recombination,the model assigns a central role to mutation as supportive of genetic advances[11],a situa-tion explaining the rate of progress reported for gene pools with a narrow genetic base.The model calls for a better understanding of the origin,nature,number,allele multiplicity and phenotypic value of QTL genes support-ing quantitative variation.This review covers recent work to address the molecular nature of quantitative variation. In addition,the article summarizes valuable information on linkage disequilibrium(LD)-based QTL mapping and on the difficulty of understanding and using gene reg-ulatory variation.From quantitative trait loci to quantitative trait nucleotidesThe genomics revolution of the past10years has improved our understanding of the genetic make up of living organisms.Together with the achievements repre-sented by complete genomic sequences(Arabidopsis[12] and rice[13,14]),high-throughput and parallel approa-ches are available for the analysis of transcripts,proteins, insertional mutants and chemically induced mutants.All this information allows us to understand the function of genes in terms of their relationship to the phenotype. Projects such as‘Arabidopsis2010’(/ pubs/2002/bio0202/start.htm)have the goal of defining the function of each of the genes identified in Arabidop-sis.The end result of such projects will be the under-standing of how all genes act and interact with the environment.Nevertheless,despite its greatrelevance,214such an understanding could be of little value to plant breeding.In fact,the genomics revolution will be at the core of plant breeding only when it can provide the relationship between genetic variation in gene sequences and phenotypic variation in traits,rather than just between a gene and a mutant phenotype.Although traits of plant breeding interest are mainly quantitative in nature,the relationships between quantitative trait varia-tion and molecular diversity of genes can be studied based on a genomic approach(Figure1).QTL detection has been successful in species where inbred strains are available,owing to the simplicity of the genetics of the system and because tests associating significant phenotypic effects to chromosomal intervals are straightforward to carry out.A simple architecture of crop trait complexity is emerging,with only few loci controlling most of the variation[15].The resolution of these experiments is,however,poor(10–30cM).This represents a major drawback in successive attempts to identify genes responsible for QTLs.QTL experiments can be carried forward to a sufficient resolution to attempt a positional cloning effort(1cM or less).In such cases, mapping is based on near isogenic lines(NILs)differing for the QTL region and on the analysis of thousands of progenies from their cross[16].Fine mapping of QTLs has sometimes revealed the presence of tightly linked loci affecting the same trait[17 ,18].Eight genes responsible for QTLs have been identified in plants so far[19–21,22 ,23 ,24,25 ,26](Table1).With the exception of the maize tb1gene cloned by transposon tagging,the others were identified by positional cloning.Some of the genes encode for transcription factors,whereas others encode proteins involved in metabolism or active in signalFigure1Genomic approaches used to decode the molecular basis of trait quantitative variation.EST,expressed sequence tag;SNP,single nucleotide polymorphism.Table1Genes identified as components of trait quantitative variation(QTL genes)in plants.Species Trait QTL Gene Biochemical function Mutations Variationexplained(%)ÃTomato Fruit size fw2.2OFRX Unknown,possibly regulatory Promoter variants5–30Tomato Sugar content Brix9-2-5Lin5Apoplastic invertase Coding and/or intronic variants naTomato Fruit shape ovate ovate Unknown Nonsense coding48–67Rice Flowering time Hd1Se1Transcription factor,Constans-like Coding deletion,intron insertion67Rice Flowering time Hd6CK2Protein kinase Nonsense coding naRice Flowering time Hd3a Hd3a Unknown na naMaize Apical dominance Tb1Tb1Transcription factor Regulatory region variants20–25 Arabidopsis Flowering time ED1CRY2Cryptochrome Amino acid substitution27–56ÃThe fraction reported corresponds to the part of total genetic variation explained by the QTL.na,not available.From plant genomics to breeding practice Morgante and Salamini215perception/transduction pathways.Two of the three genes identified as responsible for QTLs forflowering times in rice were homologs of Arabidopsis genes affect-ingflowering time.The examples show that Arabidopsis data can help in predicting the molecular nature of traits shared between species.Work has been carried out to characterize sequence variation and evolution in two genes—tb1of maize[27,28]and fw2.2[29 ,30]of tomato —where the causative mutation seems to be located in the promoter/regulatory region and affects transcription level.Mutations in fw2.2have been detected in the2.6kb upstream of the coding region and appear to cause het-erochronic expression of the gene.The causative muta-tion of the tb1allele appears to be located more than50kb upstream of the transcription start site(J Doebley,per-sonal communication).Although the reported results point to QTL variation as residing within gene transcrip-tion units,they do not negate that molecular events taking place distantly from the affected genes(s)can contribute to a phenotype.In this sense,thefinding that transposable element insertion can modulate gene tran-scription[5 ]has relevance to the origin of allelic diver-sity.This suggests that Mendelian genes affecting a given trait have an additional role in the quantitative control of the same character.A new paradigmWith the recent advances in DNA sequencing and single nucleotide polymorphism(SNP)genotyping,new approaches to QTL mapping and quantitative trait nucleotide(QTN)identification are available.The emerg-ing concept is to exploit the possibility of looking at variation directly in genes and not at anonymous markers (candidate gene association studies),as well as to saturate the genome with markers(whole genome scan)[31,32]. Both approaches rely on the detection of LD(i.e.non-random association between alleles at linked loci)and take advantage of recombination events accumulated over many generations to restrict a mapping interval. Association studies are based on existing populations/ germplasm collections,which is a major advance for species where experimental populations are difficult to access.Although the whole scan approach has not yet been used,the candidate gene approach is being applied to crops.Both procedures benefit,in different ways,from the level/amount of LD present in a species or in a population(LD is a function of history and recombination and seems population-rather than species-specific). When LD is low,high-resolution mapping is achievable.A low LD,however,has a drawback:a high number of SNPs is required to detect an association.Conversely, if LD is high,a lower mapping resolution becomes evi-dent but less markers are necessary to apply the genome scan approach.The level and distribution of LD in plants are yet to be determined.Maize studies have shown that LD decays over1500bp distances[9,31],but other studies have also shown that LD values may vary considerably.Sizeable regions of LD have been detected for the human genome in which regions with only a few common haplotypes are observed[10].In the laboratory mouse—the result of a breeding experiment having more relevance to modern crops—linkage blocks extend over1Mb with two hap-lotypes usually observed,thus reflecting the large impact of the breeding history on LD[33 ].Simulation studies have shown that when SNPs critical to distinguish haplo-types are used[34],haplotype blocks allow for a reduction in the genotyping efforts without loosing power in the search for associations.The influence of demographic processes on LD block structure has been stressed and the emerging concept is that,alone,the average extent of LD is not very useful in light of the possible differences among populations and genomic regions[35].Additional work is needed to determine the extent of LD in crop species and to provide guidelines on the marker density needed to reveal significant associations.The choice of proper genes is the key issue if the candidate approach to QTL mapping is taken.Genes qualified by mutant analysis to be involved in the deter-mination of the trait of interest have proven to be good candidates[36 ,37].Candidates can also be identified from expression profiling experiments,under the assump-tion that genes that show genotype-specific differences in their level of expression could be the causative agents for the variation in a trait.This approach has to be taken with caution,however,and considered only when a QTL for a trait of interest is shown to co-map with a QTL control-ling the expression of a candidate gene.When SNP markers derived from QTL genes are avail-able,they can be effectively employed in marker-assisted selection(MAS)without the need for further testing their association in populations.In MAS,however,what still is largely limiting is our capacity to produce,for all genetic determinants contributing to quantitative variation,both a genetic map and a precise molecular interpretation. This heavily depends on the size and precision of QTL mapping experiments and on knowledge of the QTN basis of phenotypic variation.The dark sideWhen considering the molecular basis of phenotypic variation,besides a simple understanding of the point mutations affecting gene transcription units,additional layers of complexity should be considered.For example, the role of epistasis in QTL variation is still poorly understood when based on current QTL mapping designs and population sizes.Nevertheless,in silico experiments suggest that epistasis has a key role in the long-term evolution of adaptive traits and in the dynamics of popu-lation divergence[38].Epistasis refers to allele-specific gene to gene,or protein to protein,interactions.Some216Plant biotechnologyproteins,like prions and chaperones(e.g.heat-shock proteins)[39,40],seem more prone than others to interact generating epigenetic bufferings or releasing hidden genetic variation.The additional role of other epigenetic phenomena and the relationships between gene silenc-ing,DNA methylation,RNA interference and hetero-chromatic DNA have been the subject of aflurry of recent papers[41,42],which have highlighted the complexity of RNA regulation operating through small non-coding RNAs.The presence of microRNAs,a class of small RNAs,and their role in development and morphogenesis has been confirmed in plants[43].The number of genes encoding microRNAs is yet to be determined and stra-tegies for their large-scale in silico identification are still largely missing.Thus,such genetic determinants could go entirely undetected in a positional cloning effort. The relevance for the phenotype of emerging new levels of gene regulation is difficult to evaluate.This type of regulatory variation,in fact,frequently concerns gene or genomic regions—like promoters,introns,silencers and other non-coding sequences mapping away from tran-scriptional units—all known to be more variable among individuals than protein-coding DNA sequences.Such regulatory variation affecting both the level and pattern of expression of the genes concerned appears to be fairly common[44,45 ,46].However,the effect of mutations in these regions is difficult to describe for three reasons:they could be distant from the regulated gene[47 ];their differentiation from neutral base pair substitutions is labour-intensive;and the prediction of their phenotypic effects,unlike mutations affecting protein sequence,is still extremely difficult.In plants,an additional role in gene expression variation is played by long terminal repeat retrotransposons,which are capable of altering the expression of adjacent genes [48].Given that maize inbreds can be polymorphic for the presence of specific retrotransposons[49 ],a new dimen-sion for allelic variation among genotypes appears to consist of differences in large blocks of intergenic sequences or of large indels even involving genes.In this respect,the gene complement of different maize lines may be variable,both locally as well as globally.This may have a bearing on the molecular basis of hybrid vigor,a phenomenon that,while widely exploited in crop breed-ing,is still missing a mechanistic molecular explanation. It seems logical to predict,in closing this section,that regulatory variation,although genetic in nature and,as such,fully inheritable,will be difficult both to understand and to assign to a measurable fraction of the phenotypic variation of a trait.ConclusionsThe positional cloning approach to the identification of QTL genes has proven successful.Two components seem to be needed:genomic technologies and biological resources.Although in the near future genomic resources will no longer be limiting,biological resources are already insufficient.Thus,an effort is required to standardize methods and provide models for the analysis of crop trait variability.In this sense,and due to the fact thatfine QTL mapping via LD can be carried out in germplasm collec-tions,the use of such an approach allows the collection to be genotyped only once,while monitoring several traits simultaneously.In addition,unlike in human genetics, plant breeders have a wide choice of crossing schemes to choose from.Long-term efforts may be required to adapt experimental populations to the desired level of resolu-tion,but they will definitely pay back.Among the pos-sible approaches in plants,segmental introgression lines are already in use[50].Populations from several founder lines that have been intercrossed for many generations have been developed in mouse and have shown superior attributes forfine mapping of QTLs[51,52].Some of the existing long-term selection populations in crops could be employed in a similar fashion,but new ad hoc populations should also be created.Breeding lines for which the pedigree relationships are known could also be adapted to LD approaches:phenotypic data for traits of interest are often already available for them,from multiple years and locations,giving a direct access to LD mapping.An accelerated rate of QTL discovery should also be possible with the transfer across species of results and knowledge on candidate genes.Once genes responsible for quantitative variation become available,information can be passed onto plant breeding in the form of MAS.This will only be adopted if it improves the efficiency of the conventional breeding process,which so far has been a success.Additional opportunities will arise from knowledge of the molecular basis of variation,from the identification of new trait supporting alleles in wild germplasm,and from targeted mutagenesis.However,the success of molecular breeding will,in the large part,be dictated by the genetic model underlying quantitative variation.It has been correctly argued that if Fisher’s infinitesimal model(i.e.many genes with small effects)applies,the efficiency of MAS will always be inferior to that of phenotypic selec-tion,even if all genes are known and mapped[53].Time will tell how general this prediction is.References and recommended readingPapers of particular interest,published within the annual period of review,have been highlighted as:of special interestof outstanding interest1.Salamini F,O¨zkan H,Brandolini A,Scha¨fer-Pregl R,Martin W:Genetics and geography of wild cereal domestication in thenear East.Nat Rev Genet2002,3:429-441.2.Zohary D:The mode of domestication of the founder crops ofSouthwest Asian agriculture.In The Origins and Spread ofAgriculture and Pastoralism in Eurasia.Edited by Harris DR.University College Press,London;1996:142-157.From plant genomics to breeding practice Morgante and Salamini2173.Matsuoka Y,Vigouroux Y,Goodman MM,Sanchez GJ,Buckler E,Doebley J:A single domestication for maize shown bymultilocus microsatellite genotyping.Proc Natl Acad Sci USA 2002,99:6080-6084.4.Slafer GA:Genetic Improvement of Field Crops.Marcel Dekker Inc.,New York,1994.5. Barton NH,Keightley PD:Understanding quantitative genetic variation.Nat Rev Genet2002,3:11-21.A critical and timely review of the area of quantitative genetics as it relates to breeding,selection and the evolutionary forces that maintain genetic variability.6.Fisher RA:The Genetics of Natural Selection.Oxford UniversityPress,Oxford;1930.7.Glazier AM,Nadeau JH,Aitman TJ:Finding genes that underliecomplex traits.Science2002,298:2345-2349.8. Schneider K,Weisshaar B,Borchardt DC,Salamini F:SNP frequency and allelic haplotype structure of Beta vulgaris expressed genes.Mol Breed2001,8:63-74.A report on a species with a clear haplotype structure of expressed genes (see also[9]for a similar case in maize).9.Ching A,Caldwell K,Jung M,Dolan M,Smith H,Tingey S,MorganteM,Rafalski A:SNP frequency and haplotype structure of18maize genes.BMC Genetics2002,3:19.10.Gabriel SB,Schaffner SF,Nguyen H,Moore JM,Roy J,Blumenstiel B,Higgins J,DeFelice M,Lochner A,Faggart M et al.: The structure of haplotype blocks in the human genome.Science2002,21:2225-2229.11.Hill WG:Rates of change in quantitative traits fromfixation ofnew mutations.Proc Natl Acad Sci USA1982,79:142-145.12.The Arabidopsis Genome Initiative:Analysis of the genomesequence of theflowering plant Arabidopsis thaliana.Nature2000,408:796-815.13.Goff SA,Ricke D,Lan TH,Presting G,Wang R,Dunn M,GlazebrookJ,Sessions A,Oeller P,Varma H et al.:A draft sequence of the rice genome(Oryza sativa L.ssp.japonica).Science2002,296:92-100.14.Yu J,Hu S,Wang J,Wong GK,Li S,Liu B,Deng Y,Dai L,Zhou Y,Zhang X et al.:A draft sequence of the rice genome(Oryza sativa L.ssp.indica).Science2002,296:79-92.15.Kearsey MJ,Farquhar AG:QTL analysis in plants;where are wenow?Heredity1998,80:137-142.16.Salvi S,Tuberosa R,Chiapparino E,Maccaferri M,Veillet S,van Beuningen L,Isaac P,Edwards K,Phillips RL:Toward positional cloning of Vgt1,a QTL controlling thetransition from the vegetative to the reproductive phase inmaize.Plant Mol Biol2002,48:601-613.17. Fridman E,Liu YS,Carmel-Goren L,Gur A,Shoresh M,Pleban T, Eshed Y,Zamir D:Two tightly linked QTLs modify tomato sugar content via different physiological pathways.Mol Genet Genomics2002,266:821-826.This paper introduces the possibility that QTL mapping may have com-plex facets.The case discussed here concerns the possibility that QTL genes can be tightly linked.18.Steinmetz LM,Sinha H,Richards DR,Spiegelman JI,Oefner PJ,McCusker JH,Davis RW:Dissecting the architecture of aquantitative trait locus in yeast.Nature2002,416:326-330. 19.Doebley J,Stec A,Hubbard L:The evolution of apical dominancein maize.Nature1997,386:485-488.20.Fridman E,Pleban T,Zamir D:A recombination hotspot delimitsa wild-species quantitative trait locus for tomato sugar contentto484bp within an invertase gene.Proc Natl Acad Sci USA2000, 97:4718-4723.21.Frary A,Nesbitt TC,Grandillo S,Knaap E,Cong B,Liu J,Meller J,Elber R,Alpert KB,Tanksley SD:fw2.2:a quantitative trait locus key to the evolution of tomato fruit size.Science2000,289:85-88.22. Yano M,Katayose Y,Ashikari M,Yamanouchi U,Monna L,Fuse T,Baba T,Yamamoto K,Umehara Y,Nagamura Y,Sasaki T:Hd1,amajor photoperiod sensitivity quantitative trait locus in rice,isclosely related to the Arabidopsisflowering time geneCONSTANS.Plant Cell2000,12:2473-2484.Model species can contribute to crop molecular genetics.This contribu-tion supports this conclusion across the monocot-dicot divide.23.Takahashi Y,Shomura A,Sasaki T,Yano M:Hd6,a ricequantitative trait locus involved in photoperiod sensitivity,encodes the a subunit of protein kinase CK2.Proc Natl Acad SciUSA2001,98:7922-7927.The success in positional cloning of a minor QTL gene using the advancedbackcross strategy is described.The gene identified did not correspondto any of the genes identified by mutant analysis in Arabidopsis with a rolein controllingflowering time.24.El-Din El-Assal S,Alonso-Blanco C,Peeters AJ,Raz V,KoornneefM:A QTL forflowering time in Arabidopsis reveals a novel alleleof CRY2.Nat Genet2001,29:435-440.25.Liu J,Van Eck J,Cong B,Tanksley SD:A new class of regulatorygenes underlying the cause of pear-shaped tomato fruit.Proc Natl Acad Sci USA2002,99:13302-13306.Interspecific QTL dissection and cloning contribute to the understandingof crop morphology traits.A QTL assigned to a chromosome interval isnow restricted to a gene with a specific function.26.Kojima S,Takahashi Y,Kobayashi Y,Monna L,Sasaki T,Araki T,Yano M:Hd3a,a rice ortholog of the Arabidopsis FT gene,promotes transition toflowering downstream of Hd1undershort-day conditions.Plant Cell Physiol2002,43:1096-1105.27.Lukens L,Doebley J:Molecular evolution of the teosintebranched gene among maize and related grasses.Mol Biol Evol2001,18:627-638.28.Hubbard L,McSteen P,Doebley J,Hake S:Expression patternsand mutant phenotype of teosinte branched1correlate withgrowth suppression in maize and teosinte.Genetics2002,162:1927-1935.29.Cong B,Liu J,Tanksley SD:Natural alleles at a tomato fruit sizequantitative trait locus differ by heterochronic regulatorymutations.Proc Natl Acad Sci USA2002,99:13606-13611.A detailed analysis of the expression levels and patterns of a QTL genehas been performed to explain how natural variation in the promoter ofthis gene can affect fruit size.The changes in timing of gene expressionbetween alleles are identified as the cause of the observed phenotypicdifferences.30.Nesbitt TC,Tanksley SD:Comparative sequencing in the genusLycopersicon.Implications for the evolution of fruit size in thedomestication of cultivated tomatoes.Genetics2002,162:365-379.31.Remington DL,Thornsberry JM,Matsuoka Y,Wilson LM,Whitt SR,Doebley J,Kresovich S,Goodman MM,Buckler ES iv:Structure of linkage disequilibrium and phenotypicassociations in the maize genome.Proc Natl Acad Sci USA2001,98:11479-11484.32.Rafalski A:Applications of single nucleotide polymorphisms incrop genetics.Curr Opin Plant Biol2002,5:94-100.33.Wade CM,Kulbokas EJ III,Kirby AW,Zody MC,Mullikin JC,Lander ES,Lindblad-Toh K,Daly MJ:The mosaic structure ofvariation in the laboratory mouse genome.Nature2002,420:574-578.An example of how genome diversity structure is linked to the history of apopulation.Although concerning the mouse genome,the example caneasily illustrate similar situations known for crop plants.34.Zhang K,Calabrese P,Nordborg M,Sun F:Haplotype blockstructure and its applications to association studies:powerand study designs.Am J Hum Genet2002,71:1386-1394.35.Stumpf MP,Goldstein DB:Demography,recombination hotspotintensity,and the block structure of linkage disequilibrium.Curr Biol2003,13:1-8.36.Thornsberry JM,Goodman MM,Doebley J,Kresovich S,Nielsen D,Buckler ES IV:Dwarf8polymorphisms associate with variationinflowering time.Nat Genet2001,28:286-289.Thefirst example of a success in using the candidate gene approach toidentify determinants of quantitative variation by association mapping.Methods to account for population structure and avoid spurious associa-tions are applied to quantitative variation analysis.218Plant biotechnology37.Osterberg MK,Shavorskaya O,Lascoux M,Lagercrantz U:Naturally occurring indel variation in the Brassica nigra COL1 gene is associated with variation inflowering time.Genetics2002,161:299-306.38.Yedid G,Bell G:Macroevolution simulated with autonomouslyreplicating computer programs.Nature2002,420:810-812. 39.True HL,Lindquist SL:A yeast prion provides a mechanism forgenetic variation and phenotypic diversity.Nature2000,407:477-483.40.Queitsch C,Sangster TA,Lindquist S:Hsp90as a capacitor ofphenotypic variation.Nature2002,417:618-624.41.Volpe TA,Kidner C,Hall IM,Teng G,Grewal SI,Martienssen RA:Regulation of heterochromatic silencing and histone H3lysine-9methylation by RNAi.Science2002,297:1833-1837.42.Zilberman D,Cao X,Jacobsen SE:ARGONAUTE4control oflocus-specific siRNA accumulation and DNA and histonemethylation.Science2003,10.1126/science.1079695.43.Kidner CA,Martienssen RA:Macro effects of microRNAs inplants.Trends Genet2003,19:13-16.44.Cowles CR,Joel NH,Altshuler D,Lander ES:Detection ofregulatory variation in mouse genes.Nat Genet2002,32:432-437.45. Rockman MV,Wray GA:Abundant raw material for cis-regulatory evolution in humans.Mol Biol Evol2002, 19:1991-2004.Sequence variation in cis-regulatory regions was found to be widespread within the human genome and estimated to be more abundant than variation in amino acid encoding sequences.The consequences of variation in cis-regulatory sequences for the genotype to phenotype relationship and for determining the differences between individuals are carefully outlined.46.Oleksiak MF,Churchill GA,Crawford DL:Variation in geneexpression within and among natural populations.Nat Genet2002,32:261-266.47.Stam M,Belele C,Ramakrishna W,Dorweiler JE,Bennetzen JL, Chandler VL:The regulatory regions required for B0paramutation and expression are located far upstream of the maize b1transcribed sequences.Genetics2002,162:917-930. Fine structure recombination mapping and genomic sequencing upstream of the b1locus were employed to show for thefirst time in plants that regulatory elements can be located as far as100kb upstream of the transcription unit.48.Kashkush K,Feldman M,Levy AA:Transcriptional activation ofretrotransposons alters the expression of adjacent genes inwheat.Nat Genet2003,33:102-106.49.Fu H,Dooner HK:Intraspecific violation of genetic colinearity and its implications in maize.Proc Natl Acad Sci USA2002,99:9573-9578.A comparison of sequence variation between two maize inbreds in a large region around the bronze1locus reveals surprising and unexpected levels and types of polymorphisms within a species:intergenic regions that contain completely different retrotransposon arrays and a block of four genes that is found in one line only.50.Zamir D:Improving plant breeding with exotic genetic libraries.Nat Rev Genet2001,2:983-989.51.Mott R,Talbot CJ,Turri MG,Collins AC,Flint J:A method forfinemapping quantitative trait loci in outbred animal stocks.Proc Natl Acad Sci USA2000,97:12649-12654.52.Mott R,Flint J:Simultaneous detection andfine mapping ofquantitative trait loci in mice using heterogeneous stocks.Genetics2002,160:1609-1618.53.Bernardo R:What if we knew all the genes for a quantitative traitin hybrid crops?Crop Sci2001,41:1-4.From plant genomics to breeding practice Morgante and Salamini219。