Iowa and Illinois) there was little evidence for G X E interaction for most QTLs. .... North Carolina; 4 and 5 = Plymouth, North Carolina; 6 = Bloomington, Illinois; ...
Copyright 0 1992 by the Genetics Society of America
Identification of Genetic Factors Contributing to Heterosisin a Hybrid From Two Elite Maize Inbred Lines Using Molecular Markers Charles W. Stuber,*?t.'Stephen E. Lincoln,$*§David W. Wolff: Tim Helentjarisnand Eric S. Lander*.§ *United States Department of Agriculture, Agricultural Research Service, Raleigh, North Carolina 27695-761 4, tDepartment of Genetics, North Carolina State University, Raleigh, North Carolina 27695-7614, $Whitehead Institute for Biomedical Research, Cambridge, Massachusetts 02142, §MassachusettsInstitute of Technology, Cambridge, Massachusetts 02139 and 'Department of Biology, University of Arizona, Tucson, Arizona 85721 Manuscript received October 15, 199 1 Accepted for publication July 3 1, 1992
ABSTRACT The use of molecular markers to identify quantitative trait loci (QTLs) affecting agriculturally important traits has become a key approach in plant genetics-both for understanding the genetic basis of these traits and to help design novel plant improvement programs. In the study reported here, we mapped QTLs (and evaluated their phenotypic effects) associated with seven major traits (including grain yield) in a cross between two widely used elite maize inbred lines, B73 and Mo17, in order to explore two important phenomena in maize genetics-heterosis (hybrid vigor) and genotypeby-environment (G X E) interaction. We also compared two analytical approaches for identifying QTLs,the traditional single-marker method andthe more recently described interval-mapping method. Phenotypic evaluations were made on 3168 plots (nearly 100,000 plants) grown in three states. Using 76 markers that represented 90-95% of the maize genome, both analytical methods showed virtually the same results in detecting QTLs affecting grain yield throughout the genome, except on chromosome 6 . Fewer QTLs were detectedforother quantitative traits measured. Whenever a QTL for grain yield was detected, the heterozygote had a higher phenotype than the respective homozygote (with only one exception) suggesting not only overdominance (or pseudooverdominance) but also that these detected QTLs play a significant role in heterosis. This conclusion was reinforced by a high correlation between grain yield and proportion of heterozygous markers. Although plant materials were grown and measured in six diverse environments (North Carolina, Iowa and Illinois) there was little evidence for G X E interaction for most QTLs. mapped genetic markers for identifying, locating and manipulating QTLs have been reviewed in a recent paper on biochemical and molecular markersin plant breeding (STUBER 1992). In principle,thegenetic tern and pigmentation with seed size differences in analysis of QTLs should provide both a molecular and Phaseolus vulgaris. RASMUSSON (1 933) a n d EVERSON a practical understanding of key phenomena in plant a n d SCHALLER (1 955)subsequently reported linkages improvement. between single genetic markers and quantitative trait In this paper, we describe a study using mapped loci (QTLs), and THODAY (1961) greatly elaborated genetic markers to explore two important issues in upon the subject. These classical studiesemployed maizegenetics,heterosis a n d genotype-by-environmorphologicalmutationsasgeneticmarkers-which posed major limitationson such research because only ment (G X E) interaction. Heterosis (or hybrid vigor) is the principal reason for thesuccess of the commera few such markers could be followed in any given cial maize industry. The term, heterosis, was coined cross and because the markers themselves oftenproby G. H . SCHULLa n d firstproposedin 1914 (see duced confounding phenotypic effects (TANKSLEY et HAYES1952) a n d is usually described in terms of the al. 1989; STUBER 1989, 1992). Recently,molecular superiority ofF1 hybrid performance over some measmarkers-particularly DNA polymorphisms-have proure of parental performance. However, the underlyvided geneticists with a n essentially unlimited supply ing genetic basis for the phenomenon has not been of phenotypically neutral markers withwhich to study satisfactorily explained. Possible explanations include the inheritance of quantitative traits and to maniputrue overdominance (i.e., single loci at whichtwo late these traits for plant and animal improvement. alleles have the property that the heterozygote is truly Contributions to the concepts and theory of using superiortoeitherhomozygote),pseudo-overdomi' To whom reprint requests should be sent. nance (i.e., nearby loci at which alleles having domi-
ENETIC markers have been used to study quantitativelyinheritedtraitsfornearly 70 years. SAX(1 923)reported the association of seed coat pat-
G
Genetics 132: 823-839 (November, 1992)
824
C. W. Stuber et al.
nant or partially dominant advantageous effects are in repulsion linkage phase), or even certain types of epistasis. By mapping QTLs contributing to heterosis in a cross between two inbred maize lines, B73 and Mo17, and evaluating the phenotypic effects associated with these QTLs, our goal was t o lay a foundation for understanding the basis for this important phenomenon. G X E interaction is a n essential issue in the assessment of mechanismsofinheritanceas well as the prediction of performance in breeding programs because genotypic values must be inferred from phenotypic responses. Clearly, phenotypic performance depends on both genetic and nongenetic influences on plant development. The relativerankingsofgenotypes may well differ in different environments and the relationship may be quite complex (ALLARD and BRADSHAW 1964). Many quantitative traits in maize, including grainyield, show significant variation attributable to genotype-environmental interactions(MOLL et al. 1978). Classical studiesonquantitativetraits have measured G X E interaction averaged across the entire genome rather than for individual QTLs, while recent studies of Q T L s (e.g., EDWARDS,STUBERand WENDEL 1987; STUBER,EDWARDS and WENDE~ 1987; ABLER,EDWARDS and STUBER1991; EDWARDSet al. 1992)havefocusedonmappingQTLs ina fixed environment (but see PATERSONet al. 1991). Here, we have attempted to discern the degree of G X E interactionatindividualQTLs by first comparing QTL maps generated insix diverse environments. We then contrasted these results with location (environment) by marker interaction variances obtained from traditional analyses of variance. In addition to these scientific objectives, the current work also had the methodological objective of comparing two different analytical methods for identifying QTLs: thetraditionalsingle-markerapproach(in which the chromosomal position of the Q T L is assumed tolie exactly at the marker locus) and the more recentlydescribedintervalmapping(LANDERand BOTSTEIN1989) in which t h e Q T L is taken to lie at its most likely position between two markers flanking an interval. MATERIALS AND METHODS Experimentalprocedures: The experimental materials were developed by first intercrossing two elite maize inbred lines, B73 and Mo17. This cross produces superior hybrid performance and the parental lines, or lines derived from them, have been and are stillwidelyusedin commercial hybrids. From thiscross, 264 Fslineswere developed through two selfinggenerations with each FJ line tracing to a different F2 plant. A single plant from each F I’me was: (1) selfed to produce F4 progeny, of which approximately 10 were bulked and used to infer the genotype of the Fa parent, and(2) backcrossed to each of the two parental lines
B73
x
Mol7
1
Fl
F, 264plants
m
x
Fg
J I
x
Mol7
264 lines
264 BC (to B73) famllies fldd evaluation 6 environments
t
264 BC (to Mo17) famliies M d evaluation 6 environments
264 F4 families RFLP (67) & isozyme (9) scoring to infer F3genotype
FIGURE 1.-Diagram of procedures for developing, genotyping and phenotyping experimental materials.
to produce progeny which were phenotyped in field evaluations. Figure 1 outlines these procedures. Genotypes were determinedfor9 isozymelociin the maize isozyme laboratory at Raleigh, North Carolina, using techniques reported by STUBERet al. (1988), and for 67 restriction fragment length polymorphism (RFLP) loci at Native Plants, Inc., at Salt Lake City, Utah, using methods reported by HELENTJARIS et al. (1985). Identification and distribution of the 76 molecular markers in the maize genome is shown in Figure 2. Phenotypes were evaluated for each of the 528 (= 264 lines X 2 parents) backcrossfamilies by growing 24-35 plants in each ofsix diverse environments (locations) in 1987-four in North Carolina, one in Iowa, and one in Illinois. Therefore, 3 168plots (nearly 100,000 plants) were evaluated and growing conditions were good at all locations. Field plots were two rows; row length varied from 3.66 to 4.57 m and planting density varied from 36,000 to more than 50,000 plants per hectare (14,600 to more than 20,000 plants per acre). For the field design, the 528 backcross families were subdivided into 22 sets of 24 families each, with each set containing the crosses of 12 randomly chosen FS lines to each of the two parental lines, B73 and Mol”. Each family was replicated six times, once in each of the six locations. The quantitative traits measured are listedin Table 1. Linkage analysis of genetic markers: Although isozyme and RFLP markers usedinthis study had been mapped et al. 1988; HELENTJARIS, WEBERand previously (STUBER WRIGHT 1988), these earlier studies involved different, and usually smaller, populations. Accordingly, genetic maps were calculated from the genotypic data de novo and checked for consistency with previouslyreported maps. Pairwise and multipoint linkage analyses wereperformed using a modifiedversionof the MAPMAKER program (LANDER et al. 1987) on a Digital DECStation 5000. MAPMAKER’S linkage analysis algorithms were modified to allow for correct multipoint maximum likelihood calculation of recombination rates in an FJ population, ie., for each locus, probability distributions were calculated over all possible combinations of FPand FS phase-knowngenotypes (16
Genetic Factors Affecting Heterosis
825
TABLE 1 Means of quantitative traits measured ateach location (environment) where data were recorded Trait Ear Plant Ear area Grain yield height ( 4 (bu./acre)
Locationa
1 3 4 5 6 7
77.67 109.17 115.54
0.87
0.86
Mean (over locations) a
77.46 2.74 88.20
1.07 557.1 0.87
62.97
88.50
1 .oo
leaf height (m)
2.59 64.4 2.39 68.8 2.36 2.83 2.59
(
4
17.6 593.8 619.4 559.8 21.6 67.4 582.6
b
Days to tassel (no.)
65.7 0.99 0.97 66.6 71.6 -
Grain moisture (%)
Ears/plant (no.)
1.02 13.0 13.7 17.7 16.7
0.92 0.85 0.95
1 = Clayton, North Carolina; 3 = Lewiston, North Carolina; 4 and 5 = Plymouth, North Carolina; 6 = Bloomington, Illinois; 7 =
Johnston, Iowa.
.
b,,-,,.
Indicates that no data were recorded.
per locus). The hidden Markov chain approach was adapted for the expectation step of the E-M algorithm (LANDER and GREEN1987). Linkage groups were determined using pairwise analyses with a LOD threshold of 4.0.From each linkage group, a subset of well spaced and highly informative markers was chosen, and multipoint analyses comparing thousands of candidate orders were used to determine the mostlikely framework map (with genetic orders accepted when they had a 1000-fold higher likelihood than alternative orders). Consistency of the data for the framework markers was checked by successively removing markers one at a time from the linkage group andreanalyzing the group's datade nouo. Significant changes in the map distances or orders were used to detect potential scoring errors. In addition, when final map orders were determined, isolated obligate double crossover events were used to detect further potential data errors. The remaining markers from each linkage group were placed into intervals in the framework using three point analysis and relative orders were determined and confirmed using extensive multipoint analysis involving permutations of loci (at 100:1 likelihood ratio). QTL mapping using traditional (one-marker-at-a-time) approach The traditional approach for detecting a QTL near a marker (SOLLER and BRODY1976; TANKSLEY, MED INA-FILHO and RICK1982; EDWARDS, STUBER and WENDEL 1987) involves comparisons among the phenotypic means of appropriate marker classesof progeny. In this study, progeny were generated from backcrossing Fs lines to the parents, B73 and Mol 7. If B, designates the allele at the ith locus originating from B73 and M , designates the allele at the ith locus originating from Mo17, then the expected genotypic ratio of the Fs lines would be 3/8 B,Bz: 1/4 B,MZ: 3/8 M,MZfor the ith locus. T o analyze the effect of the ith locus in the backcross to B73 (for example), we compared the phenotypic means of the backcross progeny from FS parents having genotype B$, (whose backcrossprogeny had genotype B&) to the phenotypic meanof the backcross progeny from F3 parents having genotype MSZI, (whose backcross progeny had genotype B&). Because those Fs parents with the B,M, genotype would produce backcross progeny segregating at thislocus,theywould not correspond to either cell in a traditional single-marker analysis for the ith locus and so were omitted. [The omission of these progeny represents a limitation of the traditional single-marker approach for the type of experimental materials used in this study, which becomes serious if one wishes
to construct multilocus regression models. Witha fourlocus model, one would expect to retain only about 32% (= 0.754) of the data. For other types of experiments, such as those using F2 individuals or recombinant inbred lines, this limitation would not exist for the single-marker analysis.] For this study, about 600 observations were available for computing each of the backcross progeny means usedto estimate the effects associated with each marker locus. Therefore, the means were estimated with a high level of precisionand the omission of progeny originating from Fs parents with the B,M, genotype was not considered to be a serious limitation for the single marker locus comparisons. Again, using the backcross to B73 as an example, the difference between the two marker class means provided an estimate of the phenotypic effect of substituting an M allele for a B allele at the QTL associated with the marker. To test the significance of the difference, we performed two different analyses of variance and evaluated the results with an appropriate F-test. First, we performed a simple onefactor ANOVA ignoring location and set effects: each Fs parent was assigned the mean phenotypic value of its backcross progeny averaged over sets and locations (environments). (It should be noted that F- tests for the one-factor ANOVA were conducted both with and without the B,M, genotypic class. Significance levelstended to be lower with the inclusion of the B,Mi class, however, overall interpretations of the results did not differ.) This analysis was appropriate for comparison withthe LOD score analysis (discussed in the next section), inasmuch as the latter analysis also did not account for set and location effects. Second, we performed a three-factor ANOVA that accounted for set and location effects. Because the location (environment) by marker ( L X M ) and the location by set by marker ( L x S x M ) components of variance usually were not significant, the mean square associated with the set by marker (S X M ) source of variation was used as the error variance for this F-test (see Table 2a). This is expected to provide a conservative testbecause both sampling variation (only 12 lines were included in each set) and environmental variation would contribute to thesize of this error term. For each marker, we calculated the following two quantities according to the ANOVA in Table 2: ,R' = SS[Markers]/SS[Families], and sR*2= SS[Markers in Sets]/(SS[Markersin Sets] + SS[Families in Markers in Sets]).
C. W. Stuber et al.
826 TABLE 2 ANOVA calculations SO”rCea
d.t
a) Form of analysis of variance for partitioning variance to test significance of phenotypic effects associated with markersb Locations ( L ) Sets (S) Markers ( M ) LXS LXM S X M L X S X M Residual
5 21 1 105
5 21 105
I,) Form of analysis for partitioning variance due to families for
additive: F3 individuals heterozygous at any particular Q T L necessarily produce backcross families which resemble, on average, a 50-50 mixture of backcrossed families produced by the two types of homozygotesfor the QTL.Accordingly, the analysis has 1 d.f. Also, a LOD score threshold of 2.6 corresponds to an approximate nominal significance level of P = 0.001 per test, or P = 0.05 for the entire maize genome (LANDER and BOTSTEIN1989). In this work, we used a more stringent threshold of 3.0 for declaring the existence of a QTL, andwe considered LOD scores between 2.0 and 3.0 as “suggestive.” For each LOD peak, we determined 1 .0 LOD and 2.0 LOD support intervals (that is, the region in which the LOD score remains within 1.0 or 2.0 units of the peak). For any point in the genome, we calculated the quantity: ,R2 = 1 - (r2/ulot2.
computing ,x’ (see text) Locations Families Locations X families
5 263 1315
c) Fornt of :unalysis of variance for assessing proportion of genetic varianceascribed toQTLs associated with specific markers Locations ( L ) Sets (S) Markers ( M )in S Families ( F ) in M in S Residual
5 21 22 Varies
a Locations, sets and lines were assumed to be randomvariables; 11urkers were assumed to be fixed. Significance test for markers:F,,, = M/(S X M). Significance test fiw locations X markers: F,. x = ( L X M ) / ( L X S X M ) .
T h e quantity $R2is the proportion of the total phenotypic variance among families (without controlling for effects of locations and sets) that could be ascribed to QTLsassociated with an individual marker, while the quantity ,R*‘ reflects the proportion of the variance explained when controlling for theeffects of both locations and sets, which more nearly reflects the proportion of genetic variance explained. QTL analysis using maximum likelihood methods based on interval mapping: Maximum likelihood methods have recently been described (LANDER and BOTSTEIN1989) that generalize the traditional single-marker analysis to the situation in which the Q T L does notlie exactly at the marker locus but rather between two flanking markers andin which some data may be missing. T h e strength of evidence for linkage is reflected in a LOD score, or logarithm to thebase 10 of the likelihood ratio. In this work, interval mapping of QTLs was performed using the basic approach described previously for backcrosses (PATERSON et al., 1988), butusing the version of the MAPMAKER-QTL modified for FB progeny (with changes as described above). For all traits, analyses were first performed on the untransformed phenotypic data. For those traits that were not approximately normally distributed, the datawere also transformed to more closely fit a normal distribution and the analyses were repeated. Such analyses did not substantially alter any of the conclusions and so are not reported. For the main analysis, each F3 parent was assigned the mean phenotypic value of its progeny averaged over environments-with set and location effects being ignored (inasmuch as MAPMAKER-QTL does not include an option to account for such effects). Because each FX individual’s phenotype was defined asthe average phenotype of its backcross progeny,the measured Q T L effects must necessarily be
where utotis the total phenotypic varianceand u2 is the unex plained variance for the model in which the phenotype is explained by a single Q T L at the oint having the maximum likelihood effect. T h e quantity ,R 1s analogous to $R2defined for the single-marker analysis. In contrast to the traditional single-marker analysis (see above), the maximum likelihood approach does not omit F3 individuals with heterozygous genotypes; this makes it possible to directly fit simultaneous multiple QTL models without omitting large fractions of the data. Accordingly multiple-QTL models were fit to the data to search for QTLs with effects that might be detected only after controlling for larger QTLs. In addition, multiple-QTL models were used to discern whether chromosomes with multiple LOD peaks contained single- or multiple-segregating QTLs (LANDER and BOTSTEIN1989; LINCOLN and LANDER 1989;PATERSON et al. 1990). When adding QTLs toa model, LOD score increases of 1.0wereconsidered suggestive of an additional Q T L and 2.0 were considered indicative. Finally, in order to evaluate the effects of environments, we repeated the complete analyses separately for each environment. Effects of sets were again ignored in these analyses. Analyses of genotype by environmental interaction: Two different procedures were used to assess G X E interaction. T h e first was based on the single-marker analysis; to studyoverall G X E interaction,the significance of the location by marker interactionvariance was examined using an F-test (see Table 2a). T h e second was based on the LOD score analysis; to test whether the apparent position of a QTL differed significantly between two environments, we compared the LOD score obtained under the hypothesis of two distinct QTLs in the two environments (each located at its maximum likelihood position) to the LODscore obtained under the hypothesis of a single Q T L in both environments (located at the maximum likelihood position). T h e test has one degree of freedom. Relationship of heterozygosity withtrait expression: T h e relationships of heterozygosity with trait expression were evaluated by regressing mean trait values on the percent heterozygous marker loci in the 264 BCI families for the two backcross populations. Analysis of epistasis: After apparent QTLs were located, we tested each pair for possible epistasis. For example, let m l , m‘, m B and m4 denote the expectedphenotypic effect of individuals in the backcross to B73 having, respectively genotypes BB;BB,BB;BM,BM;BB and BM;BM. We compared the LOD score for the maximum likelihood model allowing for epistasis (i.e., allowing m l ,m2, ma and m4 to have their maximum likelihood values) to the LOD score for the
P.
Genetic Factors Affecting Heterosis 7
1 6
5
2
827 9
8
4
3
10
npll14' npl4 17 np125A
-
np1409
Ort8'-
npl2531
24.9 np1422 I
34.4
-
npi287-
20.8
!.O
nplB1-
-
npi249
np1297- '4.6 npl456-
1.9
np1446'
4D 20.4
15.2
-
np15'
bnll5.45' 4.4 np1267. np1449
1.5
bn18.21" 9.8 npl465 n p d 6.1
22.6
-
np1392"
np1453
6.0
npi212" 25.7
np1292
6.4
el--
6.7
npi208' 0 ' 5.6 2 .. 26
1 1.2 np1444
npl429
7.9
np145F 1 .-a npl425
npi 104
np1317 amp 1
>,; np125
np1256
np1296"
np1428 26.8
-18.8
%.l
np1354
14.1
7.1
24.0
npl4l 1
np1366+
'
npl458
np1288
bz
3.4
lt
nplllO25.4 np1283
-
5.0 1
bnl5.10
bn11.45!1.4
npi426np130 36.3
npi350
tp15-
31 .9
npilO;
1 .8
npl447
7.0
np1255
5.5
bn17.25 np1238 gdh 1
9.4
3.3
1;IGURt: '?.-M;tise clwotlmsome map showing locations of the 76 isozyme and KFLP markers as calculated from genetic data in this study using MAPMAKEK (LANDERet al. 1987) as modified for F:,family data. lsozyme loci are: a m p l , g d h l , est8, pgm2, a m p j , mdh2, tpij,sad1 and glul The other nlarkers are RFLPs. All markers are linked to the mapwith LOD scores exceeding 3.0, except where indicated by dashed lines. Centrotnere positions are approximate. Distances between markers are given in Kosambi centiMorgans.
maximum likelihoodmodel subject to the constraint of no epistasis (ie., m l+ m4 = my + mB).This test has 1 d.f.
markerswerelinkedtothemap with LODscores exceeding 3.0. Weestimatethatthemarkersare detectably linked with 90-95% of the maize genome. RESULTS agreed order Map previously with published orders [isozymes: STUBERet al. (1988); BNL RFLP probes: Genetic map: T h e geneticmap is shown in FigureBURR et al. (1988); NPI RFLPprobes:WEBERand 2, with distancesindicated in centiMorgans. All 76 HELENTJARIS (1989)l with theexceptions: (1) thetwo
828
C. W. Stuber et al.
linked probes NPI104 and NPI317 on the long arm of chromosome 4 were reversed in order on our map, (2) the isozyme marker Mdh2 onthelongarm of chromosome 6 and thetwo linked probes NPI366 and NPI422 on the short arm of chromosome 10 could not be assigned to any of the 10 linkage groups in our mapping analyses, and (3) the three linked probesBNL8.21,NPI59 and NPI392-which unequivocally mapped to chromosome2 on our map had previously been assigned to chromosome 7 (B. BURR,personal communication). It is known that there is extensive duplication between chromosomes 2 and 7 (HEI~ENTJARIS, WEBERand WRIGHT1988), with the result that many probes detect fragments on both chromosomes. We presumethat the probes revealed polymorphic fragments on chromosome 7 in the crosses used in the previous studies, while they revealed polymorphic fragments on chromosome 2 (and monomorphic, and thus unmappable, fragments on chromosome 7 ) in the population used in our study. QTLs for grain yield: Because most metabolic processes in the maize plant ultimately affect reproduction, it seems very likely that the inheritance of grain yield must involve multiple genetic factors. Consequently, this trait probably was the most complex of the traits evaluated in this study. It also is the most importanttraitfor plant breeders.Consequently, analyses of grain yield will be discussed in more detail than the other traits evaluated. The results for this trait,averagedover the six environments are shown in Figure 3. Three types of analyses were performed for mapping and ascertainingthe significance of QTL effects: single-marker one-factor analysis of variance, interval mapping using LOD scores, and single-marker three-factor analysis of variance. The first two methods are directly comparable, with each using phenotypic trait means averaged over the six environments, thus ignoring set and location effects. These two methods gave broadly similar results with the following differences: (1) interval mapping tended to show higher overall significance levels; (2) interval mapping detected a significant effect for one markerlying at some distance from a QTL (BNL8.21 on chromosome 2 in the backcross to B73) which the single-marker one-factor analysis of variance did not (but was detected by the three-factor analysis); and (3) single-marker one-factor analysis of variance detected one barely significant result (NPI306 on chromosome 10 in the backcross to Mol 7) not detected by interval mapping but this could be interpreted as a false positive. The third method (single-marker three-factor)of analysis accounted for set and location effects, thereby eliminating variance due to these causes. This approach gave results similar to the first two methods, with the exception that it was able to detect small effects on chromosome 4 in the
backcross to B73 and on chromosome 6 in the backcross to Mo 17 (see Figure 3). Table 3 shows the effects of allelic substitution at the marker locus nearest the apparent QTL, computed both by the single-marker three-factor analysis of variance and by the interval mapping (see columns titled “Phenotypic effect”). For example, at the Amp3 marker in the backcross to B73, the effect of substituting the allele from Mol 7 is about 1 1 bushels per acre (0.69 Mg ha”). This is more than 12% of the 88.5 bushels per acre (5.55 Mg ha”) mean grain yield over the six environments. Because the three methods of analyses provided similar results, we will focus on the LODscore analysis because it providesa simple picture of the likely location of the QTLs. Markers showing significant association with yield in at least one backcross were found on all 10 chromosomes. In the B73 backcross population, the long arm of chromosome 1 and the centromeric regionsof chromosomes 2 and 5 showed highly significant effects with LOD scores greater than 6.0, the centromeric regionsof chromosomes 7 and 9 showed LOD scores greater that4.0, and the centromeric region of chromosome 10 showed a LOD score greater than 3.0. In the Mol7 backcross, the long arms of chromosomes 3 and 4 and the centromeric region of chromosome 5 showed highly significant effects with LOD scores greater than 6.0, the proximal region of the long arm of chromosome 1 showed a LOD score greater than 4.0, and the centromeric regions of chromosomes 7 , 8 , 9 and 10 showed LOD scores exceeding 2.77. Thus, the backcross to B73 showed at least six QTLs and the backcross to Mol 7 showed at least eight QTLs for grain yield. Separate analyses for each of the six individual environments based on interval mapping using LOD scores are shown in Figure 4. In spite of the large variation in mean grain yields among environments (Table l), results for each environment are remarkably consistent, particularly in those regions for which the LOD scores (for means over environments) were greater than 3.0. (Deviations from this consistency in the backcross to B73 were found on chromosome 1 in the region between markersNPI453 and Amp1 and on chromosome 10 between markers Glul and NPI461. Similar exceptions occurred in the backcross to Mol 7 on chromosome 2 in the region between markers NPIBl and BNL8.21, on chromosome 4 between markers NPI267 and NPI292, and on chromosome 10 between Sad1 and NPI264.) Statistical tests based bothontraditional analysesof variance and LOD-score analyses confirmedthegeneral lack of evidence for significant QTL-by-environment interaction. Finally, no convincing evidence for epistasis was found: about 1% of the pairwise tests were significant
2
7
4
3
5
9
I I
t
tl =2.0 LOD = 5.0 bu/ac FIGURES.-Maize chronwsome map showing location of QTLs affecting grain yield in the blckcrosses to inbred lines B T S and Ylo I7 cv;~luatedover six environtnents. l'he m;~psummarizes the results of three analyses: interval mapping using LOD scores, single-marker onclilctor analysis ofwriance, and single-marker three-factor analysis of variance [accounting for set and location (environnwllt) effects]. Intcrval m;lpping is repres~ntedby QIL likelihood plots showing LOD score curves exceeding the threshold of 2.0. Singlemarker one-factor analysis is not explicitly represented because it agreed extremely well with the interval mapping analysis; the few discrepancies are indicated by an asterisk at the one IOCIIS a t which the LO11 score was significant but t h e one-factor analysis of variance w a s not, and by :I triangle a t the one locus a t which the opposite H'BS found. Single-marker three-factor analysis is shown by b a r s protrilding from the chrotnosome. whose length indicates the estinmted phenotypic effect of substituting an allele a t the QTl. in the vicinity of the marker. Bars ;Ire shaded light and dark gray t o indicate significant association exceeding the 0.01 and 0.00 I levels, respectively. For ;dl analyses, results from backcrosses to RTS are shown on the left ; ~ n dresults from backcrosses to MolT are shown on the right of the vertical lines representing each chro1nosome.
829
C. W. Stuber et al.
830
TABLE 3 Percentage of variance (R') and phenotypic effect attributable to individual QTLs affecting grain yield in backcrosses to B73 and Mo17, for single-markerand interval mapping analyses Interval-mapping analyses Single-marker analyses N-QTLs Phenotypic
Phenotypic 'R' Chr. marker
Backcross to B73: I NPI255 24.0 NPIB 2 1 5 Amp? NP1216 6.60 7 15.8 Y NP1427 10 NP1264 3.54
(%)
12.8 15.1 21.3 8.4 10.7 8.1
All 6 QTLs: ,R' = 60.9% Backcross to Mol 7: 1 NP1429 312 NP12 NP1444 4 5 Amp? NP1216 7 8 BNL1.45 9 NP1427 10 Glul
9.7 9.3 15.4 13.0 8.7 5.9 7.6 7.7
All 8 Q T L s : iR' = 59.1%
,R*' (%)
6.91 20.0 6.63 34.99.73 23.0 6.52 17.8
Phenotypic effect (bu./ acre)
8.07 8.80 9.76 10.85 7.65 6.2 6.65
Single-QTL
,R' LOD
9.72 4.44 4.80 8.70 3.16
(%)
15.18.72 13.3 18.0 8.8 6.74 10.3
effect (bu.1 acre)
effect (bu.1 acre)
10.40 11.30 7.98
10.40 7.56
Total effects = 46.72 bu./acre
16.8 19.2 26.3 29.5 25.1 15.7 15.0 21.0
8.41 8.01 10.86 9.71 7.86 6.33 7.23 7.60
4.78 6.53 8.01 6.86 3.31 2.73 2.97 3.56
9.5 14.4 13.9 12.9 6.4 5.8 5.6 6.5
9.50 12.38 11.34 13.72 8.02 7.68 7.52 7.06
7.71 1 1.57
10.55 8.42 5.38 6.70 7.60 4.68
Total effects = 62.61 bu./acre
Descriptions of R' values are given in MATERIAIS AND METHODS. Under interval mapping, N - Q T L s column refers to a multiple QTL model i n which phenotypic effects were estimated simultaneously for all Q T L s .
at approximately the 0.01 significance level. QTLs for other traits: The results of LOD score analyses for each chromosome are shown for plant height, ear leaf area, days to tassel, grain moisture, and ears per plant in Figure 5. These were computed on the mean trait values over all environments (see Table 1) in which the trait was evaluated. A minimum of three QTLs, on chromosomes I , 9 and 10,were detected forplant height in the backcross to B73. For this trait, at least five QTLs were found on chromosomes 2, 3, 4 , 7 and 8 in the backcross to Mol 7. For ear leaf area, QTLswere detectedonly on chromosomes I , 2 and 9 in the backcross to B73. However, at least six QTLs were detected on chromosomes 2 , 3 , 4 , 5 , 8 a n d 10 in the backcross to Mol 7. A minimum of four QTLs associated with days to tassel were noted in the B73 backcross on chromosomes I , 3, 6 and 8; in the Mol 7 backcross at least three QTLs were detected on chromosomes 1, 8 and 9 . Grain moisture showed associations with a minimum of four QTLs on chromosomes 1, 2, 8 and 10 in the backcross to B73; also at least four QTLs were notedfor this trait in the backcross to Mol 7 on chromosomes 1, 2, 5 and 8. Only two QTLs were detected for ears perplant in each of the backcrosses: on chromosomes 3 and 6 in B73, on 7 and 8 in Mo17. Proportions (R2) of variances and phenotypic effects attributed to grain yield QTLs: For the singlemarker three-factor analyses, ,R2 and sR*2were com-
puted only for individual markers nearest the QTL identified using LOD scores. However, for the interval mapping analyses, ;R2 values were computed for the presumed location of each QTL as determined by the peak LOD score. All R2 values were converted to percent (by multiplying by 100) for discussion purposes (Table 3). The ,R2 values represent the proportions of the phenotypic variance among backcross families accounted for by the respective QTLs and ranged from 5.9 for marker B N L l . 45 in the Mol 7 backcross to 2 1.3 for Amp3 in the B73 backcross. The ,R2 values were 5.8 and 18.0 for those two markers, respectively. Table 3 shows that the magnitudes of the ,R2 and the iR2values were quite similar for all of the markers. Multiple Q T L analyses could be done only using interval mapping and these results showed that the six QTLs (associated with grain yield) accounted for about6 1% of the phenotypic variance in the B73 backcrosses and the eight QTLs (associated with grain yield) accounted for 59% of the phenotypic variance in the Mol7 backcrosses. The sR*2values, which presumably reflect the proportion of genetic variance accountedfor by the QTLs in the vicinity of the markers ranged from 15% to nearly 35%. Although each marker listed is on a different chromosomeand would segregate independently, obviously the ,R*' values are not independent because their sums were greater than 100% in each backcross. However, when the relative contributions
Genetic Factors Affecting Heterosis
83 1
Chromosomo 7 Environment
acto 873 mean
Peak
......................................................................... w.. .............................................................. .." .................................................... " 1 ... . ........................................................................ ..... .......................... ................................................................ 4 " .. ...................................... 22.0
30.5
16.9
I
26.4
11.8 7.0
27.9
1
I
I
I
I
I
11
bn17. npi238
3
BC t M o ol7
4 5 6 7
mean
9.4 3.3
25.5 1
npi4 npi255 1 amp1 1npi429 npi453 npi428 npi354
1
7
npi44
.... ................................... ................................ .+.+ .... ................................................................................. ................................................................................... ....................................... 1 - 1 ............................ ................................................................. ................................... + + .-. . . . . . . . . . . .. .. . . . . . . .. . .
Chromosomo 2
Environment
BC t o 873
5
. .. . . . . . . . { A I . . . . . . . . . . . . . . . . . . " . . ..................-& .. . . . . . . . ................ .. . .. . . . . . . . ~"""..'........... . . . .. . . . .. . ... ..
6 7 mean
.................................... ........... ........... ... . . .
1 3 4
...................... I
34.4
20.8
15.1 26.84 6 I
I 1
I
npi254
n pniphb71
.................... A
1
4.04 3.31 0.57 1.99 2.37 4.13 4.77
Peak LOD 2.36 4.30 6.10 3.16
....................
I
BC t o Mol7
7.43 4.29 6.77
I
I
I
LOD
2.44 4.16 5.1 1 3.76
2.03 6.67 6.63
9.8 6.1 I I I
nLi456 bnl8.!?l npi392 npi297 npiS9
t ..............
2.80
. . . . . . . . ... . . . .
2.49
................A........................... ......................................... A . 0.89 0.45 ........................................ A... 0.62 ........................................... 0.88 ...............W.".. 3.61 3.35
3 4 5 6 7
mean
....._......."4.
Chromosome 3
LOD
Environment
............................................................ ............................................................. ................................... ...........................
1
B C t o 873
3 4 5 6 7
1.18 0.41 0.31 0.52 0.13 1.30
...............................................................
............................................................. ...............................................................
0.56
mean.".....,........"."."........... .........................
24.0 I I
BC to Mol7
3 4
5 6 7
mean
20.4 22.6
15.2 I
ea18 1
11.21.9 18 8 25.7 I
I
8
I
I
npi212 npi240 npi446 npi296 npi52
.................................. ..........................1 " .................... L + . . . . . . . . . .. ... ......
...............................
............................. ......................
I
1
+-I.....
npi425
npi457
.. ..... . . . . . . . . . . .. . . . . . .. . .....
...........................
A..
+ .+ -
...... . .....
2.85 3.38 3.77 5.12 1 .a4 5.18
6.49
FIGURE4,"Likelihood plots of LOD scores for grain yield in the backcrosses of F j lines to B73 and Mo17. Results depicted are from evaluations in each of the six environments (environments 1, 3, 4, 5 , in North Carolina; 6 in Illinois, and 7 in Iowa) and are shown for each chronlosolne. T h e horizontal line in the center of each plot shows the markers and map distances for that chromosome. Dark shaded bars represent LOD scores greater than 3.0 with the extensions representing LODscores >2.0 and €3.0 T h e values on the right of each plot are the maximum LOD scores and the A's designate the maplocations of the maximumscores which are themost likely locations for the putative QTLs. Asterisks designate markers which showed significant marker (QTL) by interaction effects from analyses of variance.
C . W . Stuber et a1.
832
Chromosome 4
Environment
BC to 873
....................... p. ......... 0.92 ........... 1 - 1 .. 2.07 ................................ A . 1.49
................................ A . ......................... A.......
mean
-
npi267 bn115.45 npilO4
1 3 4 5 6 7 mean
npi208
...H " ......................
........................
... ........................
..
23.1
12.127.4
28.738.7
I
I
npi282 npi400
I
t*
....
3.27 2.73 5.37 8.02
Peak LOD
............... ....................
........ ........ ................... ........
L
......................
I
I
I
I
-
6.30 5.06
5.36
12.2 5.2
I
I
"IL II' 1 m l L
.........................
Chromosome 5 Environment 1 ............... 3 .................................... 4 5 ..6 .............................. 7 ................ mean ........... I
............
........................
............................
BC to 873
0.59 1.32
6.715.8 10.2 2.8
36.0
4.4 I 1 I
BC to Mol7
Peak LOO
........................... A ..... 0.30 .................................. 0.27
3 4 5 6 7
7.20 3.02 6.08 6.57 4.83 5.07 10.37
I
I
pgm2 npi288 npi458 npi440 npi256
amp3
1
BC to Mol7
3 4 5 6 7 mean
........................... 4 "
M.+ ....................... + " ................... ..................... ................................ .................................................................. .................. .......................
Peak LOCI 1 ..................... A ........ 0.16 3 ........................ A ... 4 ........................ A ... 0.59 1.23 5 ........................ .A"" 0.98 6 ......... A ................... 0.52 ........................ A . . . . 0.27 7 A .... 0.75 mean ........................ ~
!11.9 17.2
:
25.4
npi377 npi303npi223 npi262
BC to Mol7
1 3 4 5 6 7 mean
.........................
4.49 6.73 2.62 4.44 3.45
5.50
2.05 7.20
Chromosome 7
Chromosome 6
Environment
BC to B73
....................... ...................... ........................
...................... 4-b
II 3
+
......
bn15.47 m d h l
A...1.69
......................... A ... 0.77 ........ .................... 1.37 ............................ A 0.57 ................ .. 2.25 ..........A .................. 0.88 .................... A....... 1.63
*
Environment
4 .......
1
....-
3 4
5 6 7 mean
.-.3.54 ..'..'.. 2.45 ... .-. ..... 2.41 .....A ........ 1.86 '
,-.
4.44 7.1
I
15.0
13.4
1
npi277 npi304 npi283 npi2 16
5 6 7 mean
Peak LOD 2.81 2.75
I
-
........&..... .A ............. ....... A ....... 1 -.1 .... ..... h ........ .!-.
.
0.96 0.69 1.06 2.67 1.97 4.59 3.29
FIGURE4.
of the QTLs were considered. they were similar for sR2and sR'2. The phenotypic effects attributed to the QTLs in
the vicinities ofthe markers were very similar for both the single-marker and the interval-mapping analyses. The range for the interval-mapping analyses was from
107
Genetic Factors Affecting Heterosis Chromosome S
Environment
3
BCto873
4 5 6 7 mean
.................... A......... ........................................ ............................................. ........ 1-1 ...................................................................... ... p .................................................................. ..................A .................................................... .......... 4 - 1 ..........................................
*
34.1
npi426
bnll.45
BCt0
npill0
-,
Peak LOD 1.09 2.11 0.62 1.08 1.40
2.23 2.17
~....................................... t
............ t
I I
Ipi5
833
25.3
21.4
31.9
36.3
I
I
I
I
I
I
npill4
4
6 7 mean
..................................................................... ..................................................................... ....................................... 1 - 1 ......... ...................................................................... ..... ........................... .....................................................................
.................... 1 -
......................
1
Chromosome g
Environment
.........................
BC to 873
7
I
npi253
5.83.9
brl
I
Mol7
5 6 7 mean
I
W X ~npi427 npi25 *bni10
............................ BC to
19.1
I
.......................... 4
3.30 4.79
25.0
24.9 I
0.93
2.68
Peak LOD 2.41 2.00 3.72 . . A . 1.78 2.88
..............- 4
mean
2.96 1.73 3.41
A ... A .
.......................... ......................... ........................ ..................... .. ..................m........
4 5 6
1.52 1.39
*
Ca
2.10 1.30
..............A .............
0.42 ............ 5.61 ............................ A 1.02 t--”--l..... t”e 2.74 2.75 ................ .. - t p t 2.682.97
.......
w.
Chromosome 1 0 Environment Peak LOD 1 ............................ A .................. 1.03 3 ....................... A ...................... 1.02 4 A..................... 1.70 5 ......................... A .................... 1.52 B73 6 ............... ................ 6.58 5.60 7 ............... 4 ............................... 132 mean ......... 1 3.16 *
.........................
Bcto
......1
10.5
npi366
1
3
BCto Mol7
4 5 6
7
FIGURE4.
mean
.. 10.3
25.1
-
4.0 11.7 ”
16 1
I
1
m d f plul npi264
................... A .......................... ................ A.............................
............................................. ............. .................... .......... 1-1 ......................... ........ 1 - 1 ..................... .............” + .....................
6.5 bu./acre at the NP1264 marker on chromosome 10 in the backcross to B73 to 13.7 bu./acre at the Amp3 marker on chromosome 5 in the backcross to Mol7 . The total phenotypic effects from the multiple
1.70 1.75 1.30
3.39 2.04
2.51
3.56
Q T L analyses were 46.7 bu./acre for the six QTLs in the backcross to B73 and 62.6 bu./acre for the eight QTLs in the backcross to Mol 7 . Relationship of heterozygosityto trait expression:
C . W . Stuber et a1.
x34
Chromosome 1
BC to 873
Plantheight ....................... .I Earleafarm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Days lo tassel . + A t Moisture ................................................. Earsperplant
.....................................................................
t 6.9
22.0
.. 27.9
26.4 30.5
I
I
I
I
I
I
I
I
. .--I.
11.8 7.0
I
I
I
s npi4ZO
np1354 npl453 npi428
npi4 1 1
BCto
-
Peak LOD
D 43.3
25.5
I
.
amp1 np1255 npi44 7
bn17
......................................................
..............................................................................
Chromosome 2
BC to 873
................................. .A. ........ ..................... ......... . . . . . . . . . . . . . h... .............................
20 8
34.4
15.1
I
I
1.8
26 8
3.50 1.09
0.8 6 1
I
I
I
I
bn18.21 nD12.97
BC to Mol7
Peak LOO 1.57
.A......... 1.29
................................. I
Plant Ear leaf area Days totassel Moisture Ears p e r plant
npi392
naidl "r."
.................................. AI.. ...................................
5.36 2.60
.A..........................................
+-
...... ....................
n 73 ".I'.
1 1.37 1.13
.A......................
+-.
Peak LOD
Chromosome 3
BC to 873
height Plant ....................... ............................... Ear leal area 1-1 Days lo tassel ............... ....................... Moisture . ............................................................ Ears per plant ...................." b ..............
........................ ...........................
18 8
24.0 I
eat8
B C t o Mol7
Plant height Ear leaf area Daystotassel Moisture Earnperplant
I
20.4 I I
15.2
25.7
22.6 I
I
I
npi24S npi212 npi446 npi296 npi52
a1
Chromosome 4
Plantheight Eararea leaf Days lo tassel Moisture Ears per plant
............................. .h... ..... A ........................... ......................... ...... .A. ................... A ............. ........... .A. ....................
44
I
npi425
6.7152.8 10.2
4.00 6.06 1.61
0.53 1.50
Peak LOD 0.27 1.59 0.39 1.22 0.75
8
1 I
npi267 bn115.45
BC to Mol7
36.0
2.23 2.54 4.72 0.57 5.82
11 2l-O
I
npi457 ......................... I " ............ ..................... " 1 ........... ............................................................. .............................................................. ....................................... h .....................
BC to B73
1.64 1.88 3.62 1.84 (2.80)' 1.49
. . . . . . . . . . . . . .I 4.934.957 L . .
npi254 npi287 npiBl npl456
height
2.78 3.09 1.96
1
I
npi238 phntheight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Earleafarea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mol7 . .~ -~ ~. ~ t ~ t ~ ~ ~ l Moisture Earsperphnt . .................................................................................
Plantheight E a l e a f area Daystotassel Moisture per plant
9.89
5.30
npiZO8
npl104
Phnt height 4-b .....+ & -IEarleaf area ......... ..A .. .Days totassel P ...... Moisture .A Ears per phnt ........................ W. .-
..........................
.............. ..................
5.35 3.04t 3.322.032.16 0.41 0.38 2.202.497
FIGURE 5.-Likelihood plots of LOD scores for plant height, ear leaf area. days to tassel. grain moisture. and ears per plant in the Ixlrkcrosses of F.,lines to B73 and Mol7 . Results depicted are fromevaluations over environments(see Table 1 for number of environments fiw each trait). I'resentation of results is in the same format as for Figure 4 . Asterisk indicates LOD score after substractingvariation caused b y one o r r ~ ~ o rntajor e QTLs elsewhere i n the g.enotne . t indicates cases where multiple-QTL models did not rule o u t the possible existence o f multiple-linked Q T 1 . s .
Genetic Factors Affecting Heterosis
835 LOD
Chromosome 5
BCto 873
..................................
plantheight EarleafOaystolassel Moisture ~ ~
28.7 23.1
(2.2 5.2
12.1 1
I
I
BCto Mol7
Daystotassel Moisture ~ ~
311.7
I
I
27.4
I
I
npi44O amp3
npi4 58
npi288
.. h ................................................................ ................. ... A ................................................................... ...................... .I ............................................. ~ ~ ~ l
....... A ...................... ~ t
Earleafarea
BCto 873
Days lotassel Moisture Ears per 4.........................
............................ 17.2
!11.9
:
npi377 npi393npi223 npiZ52
...........................
Chromosome 7
Peak LoD
. . .CI ..........
0.70
1.09
2.83 0.87 0.09 0.21
......... .A. ...
-
.
..... .A. ....... 7.1
13.4
I
BC t0 Mol7
1.46 1.09 3.35 0.23 2.91
bn15.47 mdh2
........................... ..................
Plantheight Ear leaf area
1 .00
15.0
I
I
npi304 npi283 npi277 npi216 Plantheight 3.17 Ear leaf area ..... Is. 0.87 Days lo tassel . . .A........... 1.17 Moisture . A 1.09
........
............
....... 3.19
Ears p e r plant
Chromosome 8 Plant height
Peak LOD
..............................................
1.22 0.64
Earleafarea ..................................................... Days to t a w 1 Moisture ....... .... Earsperplant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.......
p -
5.07 3.30 0.23
~
34.1
I I
BC to Mol7
Plant Ear leal area Daystotasd Moisture Earsperplant
6
25.3
I
I
I
npi110 npi426
bnll.45
.............................
.... -H
..........{
A
31 9 I
I tpi5
I I
npilO7
........................
.........I
...................- 1
36.3
21.4
I I
npill4
height
1.04
4.24 0.69
.......
11 2 .534, .
A Plantheight Ear lealarea 1-1 Days to tassel ....................... A ... Moisture A ............................ Earsperplant h
BC to Mol 7
1.59 13.86
LOD Peak
Chromosome 6
.................. .A. ........ ......A ...................... ..................
Plantheight
BC to 873
t
I
1
t
npi256 pgm2 npi282 npi400
plantheight
1.68
.................................................. A ................. 1.00 ............................ A ........................................ 0.21 ................................ A................................... 0.65 .................................................. ~ ~ ~ ~ h A~................. t 1.37 s
Earleafarea
Peak
.................................
A
.. 4 -
I ........ " I
........... .................
............................
..................
3.72 7.55 9.32 3.70 2.23 3.24 t
FIGURE5 .
T h e importance of heterozygosity to trait expression was obtained by examining the mean level of trait performance observed for a given percent of hetero-
zygous loci . For grain yield. there was a highly significant relationship with heterozygosity as evidenced by the r value of 0.68 in the backcrosses to both B73 and
836
C . W. Stuber et al. Chromosome S
BC to 873
~
........................... ~ ~
~
25.0 24.9
5.03.9 I
I
npi253
Plant height Earleafarea h y sto tassel Moisture Ears per plant
BC to Mol7
Peak LOD
. . . .,.-+ ". . . . . . . . . . . . . . . . . . .. .. . .. . . . . . . . .A. .................. ..A ~. . ... ..
Plant height Ear leafIarea Days to ta8sel Moisture
,
bzl
11.64 11.85 1.67 1.OS
~A. 1.63~
I
~
t
I
w11 lnPi25 nPi427 bnl5.10
. . . . . . . . . . . . . . ....... A .. . . . . ........A ................... ........... i A t . . . . . . . . . . . . . . . . . . . . .h .. . ... . . . . . . . . . .A. . . . . . . .. . .. . . . ....
0.53 0.18 2.48 2.83
t
0.99 0.12
Chromosome T O
Peak LOD
. .. . . . . . . . . . . . . . . . . . .A... .. . . . . .
Plant height Ear leaf area . . . . . Days to tm& . Moisture , . . .b . -
.
k
19.1
3.61 0.26
. . . . . . .. ... ... ... ... ... ... . . . . . .A 0.06 . . . . . . . . . . . . . . . 4.93 . . . . . . . . . . . . 0.50 Earsper plant . . . . . . . . . . . . . . . . . .A..
BC to 873
10.5
25.1
10.3
4.0 11.7
16.1
I
1
#*dl glut npi264
npi366
Plant height Ear leaf area Days to tassel Moisture Earsper plant
BC to Mol7
. . . . ... . . . . . . . . . . . . . . . . . . . . . . . . .A 0.56 . . w. . . . . . . . . . . . . . . . . . . . . 3.31 . . . . . . . . . . . . . . . . . . A . . . . . . . . . . . . . 0.55 . . . . . . . . . . . P . .. . . . . . . . . . .. ... .. . . 0.83 . . . . . . . . . . . . . . . . . . A .. . . . . . . . . . . . 0.52
FIGURE5.
I 120 100
06
-
60
E j;
Trait
-
40
y = 50.25 + 0.81X R = 0.68"'
2 o
~
"
~
. BCtOMO17
120
-
100
-
80
-
60
'
40
Correlation coefficients ( 7 ) between several quantitative traits and the percent heterozygous marker lociin 264 BC, families for the two backcross populations
0,".
'
9
0 n
D O
-
80
e
TABLE 4
1
BCtO 873
-
-
'
"
"
"
'
a
S
b
y = 39.45 + 0.94~ R = 0.68"'
0
$
'
0 30
"
"
30
"
"
40
"
50
60
70
80
Heterozygous Loci (%)
FIGURE6.-Plots
showing associations of grain yield with percent 264 backcrosses toB73and 264 backcrosses to Mo17. Regression andcorrelation statistics are
Grain yield Ears per plant Ear height Plant height Leaf area Days to tassel Grain moisture
Backcross to B73
0.68*** 0.06 0.26*** 0.50*** 0.37*** -0.14* 0.19**
Backcross to Mol 7
0.68*** 0.21***
0.22*** 0.31*** 0.47*** -0.1 1 -0.08
*, ** and *** indicate a significant correlation at the 0.05, 0.01 and 0.00 1 levels, respectively. plant to 0.50 for plant height in the backcross to B73 (Table 4). DISCUSSION
of heterozygous loci forthe
shown
BISO.
Mol 7 (see Figure 6). Other traits also showed a significant relationship between mean trait performance and number of heterozygous loci, but the r values were generally lower, ranging from 0.06 for ears per
Comparison of methodologies: The present study provided an opportunity to compare different methodologies for Q T L mapping. Traditional singlemarker one-factor analysis of variance and the more recently described interval-mapping approach (LANDER and BOTSTEIN1989) are directly comparable, inasmuch as both involved no correction for set and
GeneticHeterosis Factors Affecting location effects. Although the two methods yielded virtually identical results in terms of QTLs identified, the interval mapping approach appeared to offer some advantages, particularly for the type of experimental materials evaluated in this study. For example, interval mapping uses information on flanking markers of putative QTLs, which may provide more power than the single-marker methodfordetectingQTLs.In addition, interval mapping provides information about the likely location of the QTL, even estimating the maximum likelihood position under the assumption that there is a single Q T L in the region (which assumption, of course, may or may not be true for very complex traits such as yield). Interval mapping also allows ambiguous or missing data, whereas (for the experimental design used in this study) the singlemarker analysis excluded Fs individuals that were heterozygous at a locus because their progenies were mixtures of two types (see MATERIALS AND METHODS); thismade itpossiblein this study to use interval mapping to construct multilocus models without cumulative loss of individuals from the dataset. Because interval mapping represents a generalization of the traditional single-marker one-factor analysis, it is not surprisingthat it offered some advantageswithout significant losses. The single-marker three-factor analysis differed by allowing forset and location effects. Although the results were broadly similar to thoseabove, the method provided increased power and allowed detection of QTLs in two additional regions. This third method foundweak but significant evidence for QTLs for yield on chromosome 4 in the B73 backcross and on chromosome 6 in the Mol7 backcross (Figure 3), whereas the interval mapping approachshowed LODs (1.32 and 0.75,respectively) which were below threshold levels and were only suggestive of QTLs on these chromosomes. All major QTLs with LODs of 3.0 or greater were detected by all analytical methods, however. Analogous interval mapping analysis allowing for set and location effects was not conducted, due to the limitations in thecurrent MAPMAKER-QTL software package. However, there is no fundamental limitation to includingthese effects in the likelihood model-which should combine the advantages of the single-factor method with the interval mapping method. (E. S. LANDERhopes to undertakethese modifications in the near future.) QTLs for yield: QTLs affecting grain yield were detected in at least one of the two backcrosses on all 10 maize chromosomes. The finding of a large number of QTLsis not surprising in view of the complex nature of the phenotype. The effects associated with a single QTL ranged as high as 13.7 bu./acre for the Q T L in the vicinity of the Amp3 marker on chromo-
837
some 5 in the backcross to Mol 7. Itshould be noted that in the backcross to B73, the six QTLs associated with grain yield showed an overall phenotypic effect of 46.7 bu./acre, which is more than 50% of the mean for that backcross population. These six QTLs also accounted for more than 60% of the total phenotypic variation. The backcross toMol7 showed a total phenotypic effect of 62.6 bu./acre for eight QTLs which also accounted for nearly 60% of the phenotypic variation. It is intriguing to compare the present findings with our results from 20 other crosses (15 involving elite inbred lines, five involving elite lines with some introgression of exotic material from Latin American maize germplasm) that have been studied in our researchprogram in Raleigh, North Carolina (EDWARDS, STUBERandWENDEL1987; STUBER,EDWARDS and WENDEL 1987; ABLER,EDWARDSand STUBER 1991; EDWARDS et al. 1992; STUBER 1992;our unpublished data).Although some of these crosses have been studied with only a limited number of genetic markers, some consistent patterns are beginning to emerge. For example, of16the 20 populations studied have shown genetic factors significantly associated with yield in the interval flanked by markers Pgm2 and Amp3 on chromosome 5 . Similarly, QTLs for grain yield were found on the long arm of chromosome 1 in 13 of the 18 crosses inwhich linked markers were scored, and onchromosome 9 near Acpl in 13 of 17 crosses inwhich nearbymarkers were scored.Ofcourse,a given quantitativetrait locus-no matter how important its effect on grain yield-can be identified only in crosses in which the parents have different alleles. We will discuss the comparison of these crosses more fully elsewhere. QTLs for other traits: For the other traits studied, yield. fewer QTLs were identifiedthanforgrain However, for traits that are associated with overall plant vigor, such as ear leaf area, plant height and ear height, the QTLsshowed many similarities with those for grain yield. With very few exceptions, QTLs for plantheight andear leaf area were found in the vicinity ofthose associated with grain yield (see Figure 5 ) . The correlation between QTL positions for these traits may resulteitherfrom pleiotropic effects of single QTLs or to linked QTLs. We tend to favor the formernotion,but recognize thatthealternatives cannot be distinguished without finer mapping of the QTLs. Although ears per plant is a component of grain yield, there were few similarities in the location of the QTLs for the two traits. This is not surprising: both of the parental lines are primarily single-eared, there was little variability for number of ears in the progenies, and this limited variation was not associated with the variation in grain yield (data not shown).
838
C. W. Stuber et al.
Heterosis: Our crosses were designed to maximize the ability to detect QTLs contributing to heterosis: we employed F3 individuals so as to increase recombinational segregation of linked QTLsrelative to using FP individuals; and we performed separatebackcrosses to each of the parental strains, so as to more accurately resolve the additive and dominance effects of QTLsrelative to the situation of simply selfing the individuals. The results showed a striking pattern: (1) whenever aQTL for yield was detected, the heterozygote had the higher phenotype (with the sole exception of NPZ253 on chromosome 9 in the back cross to B73). (2)QTLs for yield tended to occur in the same locations in both backcrosses. Significant QTLs were found on chromosomes I , 5, 7 , 8, 9 and 10 in both backcrosses, while chromosomes 2, 3 and 4 showed significant effects only in one backcross (although the Q T L on chromosome 2 in the backcross to B73 was associated with a suggestive, subthreshold effect in the other backcross). Thus, the majority of QTLs were associated with overdominance ( i e . , a higher yield in the heterozygote than in either homozygote),suggesting that these regions may play especially important roles in the phenomenon of heterosis. As CROW(1952) has pointed out, heterosis may result from either true overdominance (single loci at which the heterozygous phenotype exceeds that of either homozygote) or from pseudo-overdominance (linked lociwith advantageous alleles in repulsion phase). Our results cannot distinguish these possibilities. Indeed, they will be difficult to distinguish witho u t extensive recombinationalseparation of linked loci, and perhaps impossible without cloning of the QTLs so as to identify their effects directly. The overall effect of heterosis also can be seen by examining the correlation between a phenotypic trait and theproportion of heterozygousmarkers. This correlation is very high (about0.68 in each backcross) for grain yield, while it is considerably lower for most 6 and Table 4). This obserof the other traits (Figure vation is consistent with our results (and our prior expectations)thatgrain yield is affected by more QTLs than the other traits. A trait controlled by a single locus should show little correlation between phenotype and overall heterozygosity across the genome, atraitcontrolled by two loci should show somewhat higher correlation, and a trait controlled by many loci across the genome should show the high est correlation. (If we assume that a trait is controlled by k loci having equal and purely heterotic effects and having heritability h, it can be shown that the correlation of phenotype with overall proportion of heterozygous genetic markers is proportional to the square root of hk. ) G x E interaction: The limited evidence for interaction of environments with QTLs (see Figure 4) is
surprising for several reasons: (1) when maize traits (such as grain yield) have been evaluated p e r se in several diverse environments, genotype by environment interaction usually has been found to besignificant (MOLL et al. 1978), (2)because this study used six environments in three states (four in North Carolina, one in Illinois, and one in Iowa), the diversity among environmentswas expected to be greater than found in most documented studies of maize, and (3) in the analyses of variance (see Table 2) for this study, the location by set ( L X S) component of variance was usually significant, particularly for grain yield (data not shown). The L X S component of variance should be analogous to the traditional genotype-by-environmentinteraction variance reported inmanymaize studies. Because of the similarity of LOD scores across environments for traitssuch as grain yield (see Figure 4), particularly when the scores are greater than 4.0, we believe that it may be possible to reliably detect major QTLs in relatively few environments, possibly no more than two or three. CONCLUSIONS Identification of QTLs affecting agronomically important traits in maize is a key step in using molecular genetics for plant improvement and in understanding genetic phenomena in plants (such as heterosis and G X E interaction). Here, we have mapped the positions of QTLs, and evaluated the phenotypic effects associated with these QTLs, for several quantitative traits in alarge study designed to shed light on genetic mode of action.Fromthestandpoint of detecting QTLs, we found that two analytical methods, singlemarker and interval mapping, provided virtually identical results in backcross populations developed from the cross between inbred lines B73 and Moly. We also found that QTL alleles causing high grain yield show a strong tendency toward dominance usually and overdominance. QTLsidentified in this cross between B73 and Mol7 may depend on the particular design of the cross. In our study, QTL effects were measured in the genetic background of backcrosses (i.e., 75% recurrent parent).Results for other types of experimental materials (such as F2 plants, FJ families, or testcrosses) may conceivably be quite different. Nonetheless, we find intriguing the high correlations between the regions identified in this study and regions identified in other studies in our research program. Also, the relatively minor evidence in this study for marker (or QTL) by environment interaction was somewhat surprising and may differ from results in other studies and for other traits(PATERSON et al. 199 1; BUBECK et al. 1992). Detailed understanding of Q T L effects will now requirefine-mapping studies such as described by
Heterosis Affecting
Factors
Genetic
PATERSON et al. (1990). Several regions have been identified for such analyses, such as the region in the vicinity of the centromere on chromosome 5 and the long arm of chromosome 4. We thank Pioneer Hi-Bred International and Funks Seeds International (now CIBA-GEIGY Seeds Division) for growing and recording phenotypic data at locations in Iowa and Illinois, respectively. We also thank BRENDAABLER,DIANNEBEATTIE,WAYNE DILLARD, SCOTTWRIGHTand ELIZABETH TERRELL for their technical assistance. The investigations were supported in part by U.S. Department of Agriculture Competitive Grant 86-CRCR-1-2030 (to C.W.S.), National Institutes of Health grants GMI 1546 (to C.W.S.) and HG00198 (to E.S.L.), National Science Foundation grant DCB-8611317 (to E.S.L.), and a Markey Foundation Grant (to E.S.L.).
LITERATURE CITED ARLER, B. S. B., M. D. EDWARDSand C. W. STUBER, 1991 Isoenzymatic identification of quantitative trait lociin crosses of elite maize hybrids. Crop Sci. 31: 267-274. ALLARD,R. W., and A. D. BRADSHAW,1964 Implications of genotype-environment interactions in applied plant breeding. Crop Sci. 4: 503-508. UUBECK, D.M., M. M. GOODMAN, W. D. BEAVISand D. GRANT, 1992 Quantitative trait loci controlling resistance to gray leaf spot in maize. Crop Sci. (in press). BURR,B., F. A. BURR,K. H. THOMPSON, M. C. ALBERTSENand C. W. STUBER,1988Gene mapping with recombinant inbreds in maize. Genetics 118: 519-526. CROW, J. F., 1952 Dominance and overdominance, pp. 282-297 in Heterosis, edited by J. W. GOWEN.Iowa State College Press, Ames. EDWARDS,M. D., C. W. STUBER and J. F. WENDEL, 1987 Molecular-marker-facilitated investigations of quantitative-trait loci in maize. I . Numbers, genomic distribution, and types of gene action. Genetics 116 113-125. EDWARDS, M. D., T. HELENTJARIS, S. WRIGHT and C. W. STUBER, 1992 Molecular-marker-facilitated investigations of quantitative trait loci in maize. 1V. Analysis based on genome saturation with isozyme and restriction fragment length polymorphism markers. Theor. Appl. Genet. 83: 765-774. EVERSON, E. H., and C. W. SCHALLER, 1955 The genetics of yield differences associated with awn barbing in the barley hybrid (Lion X Atlas”) X Atlas. Agron. J. 47: 276-280. HAYES,H. K., 1952 Development of the heterosis concept, pp. 49-65 in Heterosis, edited by J. GOWEN.Iowa State College Press, Anles. HELENTJARIS, T., D. WEBERand S. WRIGHT, 1988 Identification of the genomic locations of duplicate nucleotide sequences in maize by analysisof restriction fragment length polymorphisms. Genetics 118: 353-363. HELENTJARIS, T., G. KING, M. SLOCUM, C. SIEDENSTRANG and S. WEGMAN, 1985 Restriction fragment polymorphisms as probes for plant diversity and their development as tools for applied plant breeding. Plant Mol. Biol. 5: 109-1 18. LANDER, E. S., and D. BOTSTEIN, 1989 Mapping Mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121: 185-199. LANDER, E. S., and P. GREEN,1987 Construction of multilocus genetic linkage maps in humans. Proc. Natl. Acad. Sci. USA 84: 2363-2367.
839 J. ABRAHAMSON, A. BARLOW, M. J. DALY, LANDER, E. S., P. GREEN, S . E. LINCOLN and L. NEWBURG,1987 MAPMAKER: an interactive computer package for constructingprimary genetic linkage maps of experimental and natural populations. Genomics l: 174-1 8 l . LINCOLN, S . E., and E. LANDER, 1989Mapping Genes Controlling Quantitative Traits with MAPMAKERIQTL.Whitehead Institute for Biomedical Research Technical Report, Cambridge, Mass. MOLL,R. H., C. C. COCKERHAM, C. W. STUBERand W.P. WILLIAMS,1978 Selection responses, genetic-environmental interactions, and heterosis with recurrent selection for yield in maize. Crop Sci. 18:641-645. J. D. HEWITT,S. PETERSON, S. E. PATERSON, A. H., E. S . LANDER, LINCOLN and S. D. TANKSLEY, 1988 Resolution of quantitative traits into Mendelian factors by using a complete linkage map of restriction fragment length polymorphisms. Nature 335: 721-726. PATERSON, A. H., J. W. DEVERNA, B. LANINIand S . D. TANKSLEY, 1990 Fine mapping of quantitative trait loci using selected overlapping recombinant chromosomes, in an interspecies cross of tomato. Genetics 124:735-742. PATERSON, A. H., S. DAMON, J. D. HEWITT,D. ZAMIR,H. D. S. E. LINCOLN, E. S. LANDER and S. D. TANKRABINOWITCH, SLEY,1991 Mendelian factors underlying quantitative traits in tomato: comparison across species, generations, and environments. Genetics 127: 181-197. RASMUSSON, J. M., 1933 A contribution to the theory of quantitative character inheritance. Hereditas 18:245-261. SAX,K., 1923 The association of size differences with seed coat patternand pigmentation in Phaseolusvulgaris. Genetics 8: 552-560. SOLLER, M., and T . BRODY,1976 On the power of experimental designs for the detection of linkage between marker loci and quantitative loci in crosses between inbred lines. Theor. Appl. Genet. 47: 35-39. STUBER, C. W., 1989 Molecular markers in the manipulation of quantitative characters. pp. 334-350 in Plant Population GeM. netics, Breeding, and Genetic Resources,edited by A. BROWN, CLEGG,A. KAHLER and B. WEIR.Sinauer Associates, Sunderland, Mass. STUBER, C. W., 1992 Biochemical and molecular markers in plant breeding. Plant Breed. Rev. 9 37-61. STUBER,W., C. D. M. EDWARDSand J. F. WENDEL, 1987 Molecular marker-facilitated investigations of quantitative trait loci in maize. 11. Factors influencing yield and its component traits. Crop Sci. 27: 639-648. M. M. GOODMAN and J. S. C. SMITH, STUBER, C. W.,J. F. WENDEL, 1988 Techniques and scoring procedures for for starch gel electrophoresis of enzymes from maize (Zea mays L.). NCAgric. Res. Serv. NC State Univ. Tech. Bull. 286, 87 pp. TANKSLEY, S. D., H. MEDINA-FILHO and C. M. RICK,1982 Use of naturally-occurring enzyme variation to detect and map genes controlling quantitative traits in an interspecific backcross of tomato. Heredity 49: 11-25. S. D., N. D. YOUNG, A. H. PATERSON and M. W. BONIER TANKSLEY, BALE,1989 RFLP mapping in plant breeding: new tools for an old science. BioTechnology 7: 257-264. THODAY, J. M., 1961 Location ofpolygenes. Nature 191: 368370. 1989 Mapping RFLP lociin WEBER,D., and T. HELENTJARIS, maize using B-A translocations. Genetics 121: 583-590.
Communicating editor: B. BURR
