• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of geneticsGeneticsCurrent IssueInformation for AuthorsEditorial BoardSubscribeSubmit a Manuscript
Genetics. Apr 2006; 172(4): 2449–2463.
PMCID: PMC1456379

Maize Adaptation to Temperate Climate: Relationship Between Population Structure and Polymorphism in the Dwarf8 Gene

Abstract

To investigate the genetic basis of maize adaptation to temperate climate, collections of 375 inbred lines and 275 landraces, representative of American and European diversity, were evaluated for flowering time under short- and long-day conditions. The inbred line collection was genotyped for 55 genomewide simple sequence repeat (SSR) markers. Comparison of inbred line population structure with that of landraces, as determined with 24 SSR loci, underlined strong effects of both historical and modern selection on population structure and a clear relationship with geographical origins. The late tropical groups and the early “Northern Flint” group from the northern United States and northern Europe exhibited different flowering times. Both collections were genotyped for a 6-bp insertion/deletion in the Dwarf8 (D8idp) gene, previously reported to be potentially involved in flowering time variation in a 102 American inbred panel. Among-group D8idp differentiation was much higher than that for any SSR marker, suggesting diversifying selection. Correcting for population structure, D8idp was associated with flowering time under long-day conditions, the deletion allele showing an average earlier flowering of 29 degree days for inbreds and 145 degree days for landraces. Additionally, the deletion allele occurred at a high frequency (>80%) in Northern Flint while being almost absent (<5%) in tropical materials. Altogether, these results indicate that Dwarf8 could be involved in maize climatic adaptation through diversifying selection for flowering time.

MAIZE arose from a single domestication that occurred in southwestern Mexico ~9000 years ago from a strain of teosinte Zea mays ssp. parviglumis (Matsuoka et al. 2002b). Native Americans spread maize over North and South America, generating a wide diversity of landraces adapted to local environmental conditions, from tropical to cool temperate (nowadays Canada and southern Chile). One of the prominent maize types in this adaptation history is the Northern Flint race (Brown and Anderson 1947) cultivated in northeastern America during pre-Colombian times. This early flowering type has adapted to cold temperate regions and was reported to have an exceptional genetic divergence compared to other pre-Colombian maize tropical or subtropical landraces that were cultivated in present-day Mexico, the southwestern United States, and the Caribbean islands (Doebley et al. 1986). In North America, primary genetic pools (Northern Flint, tropical, and subtropical) cultivated by Native Americans were then used by colonists to create new landrace varieties. Both the historical records and genetic studies show that the subtropical Southern Dent type was crossed with Northern Flint ~200 years ago, leading to the Corn Belt Dent type adapted to the temperate midwestern United States region (Doebley et al. 1988).

With respect to maize adaptation in Europe, the work of Rebourg et al. (2003) based on both molecular and historical data revealed that at least two introductions occurred in the old continent: first, Colombus brought Caribbean material to southern Spain in 1493, and then Northern Flint material was introduced by either Spanish or French explorers from the eastern coast of North America during the first half of the sixteenth century. This second material was already cultivated on a significant scale in northern Europe by 1539 (Rebourg et al. 2003; Dubreuil et al. 2006). On the basis of molecular marker records, these studies also showed that landraces cultivated at intermediate latitudes in Europe resulted from the hybridization of these two parental types. Northern Flint material therefore has played a unique role in the adaptation of maize to temperate climates (i) in North America, southern Chile, and northern Europe as well as (ii) in independent hybridization processes in the Corn Belt and in Europe. These traditional Corn Belt Dent and European Flint landraces then played a key role in the development of hybrid breeding for the United States (Duvick et al. 2004; W. Tracy, personal communication) and Europe.

Regarding the genetic basis of this adaptation, numerous quantitative trait loci (QTL) have been detected for flowering time in maize and a recent meta-analysis suggests a total of >60 QTL, some of which have large effects in several genetic backgrounds (Chardon et al. 2004). Association studies now offer a promising avenue to go beyond QTL mapping by identifying genes that contribute to trait variation. This approach was first applied to plants by Thornsberry et al. (2001) who reported that polymorphism in Dwarf8, a gene involved in the maize gibberellin pathway (Peng et al. 1999), was associated with flowering time variation in a 102 maize inbred line panel representing North American temperate and subtropical modern origins. A key factor for a relevant application of association genetics is to have a good understanding of population structure, to avoid false associations due to linkage disequilibrium between physically distant loci. Association studies were originally carried out to investigate disease susceptibility traits in humans (Pritchard et al. 2000b), where differentiation of susceptibility allele frequency among genetic groups can be considered as mainly random. On the contrary, spatial adaptation in plants leads to a relationship between genetic groups and adaptative traits, as described in Oryza glaberrima (Semon et al. 2005). Nonrandom differentiation among genetic groups for causal polymorphisms is expected in this context. Diversifying selection can ultimately lead to the fixation of different alleles in phenotypically contrasted populations. In such an extreme situation, association studies, which integrate population structure as covariate(s), have no power to reveal statistical associations between phenotypic and genetic variations. Thus, efficient identification of genes involved in adaptation requires the consideration of both association between phenotypes and polymorphisms and differentiation among groups.

The aim of our study was first to investigate population structure in maize with special attention to its relationship to flowering time and second to evaluate the contribution of Dwarf8 to maize adaptation to temperate climate. To do so, we defined a collection of 375 inbred lines covering a wide range of flowering time from late photoperiod sensitive to extremely early types with a special emphasis on typical Northern Flint and its derivatives. This collection was characterized for flowering time under both long- and short-day conditions, genotyped for genomewide simple sequence repeat (SSR) markers and an indel polymorphism in Dwarf8, previously reported to be potentially involved in flowering time variation (Thornsberry et al. 2001). The indel polymorphism in Dwarf8 was also characterized in a collection of 275 traditional landraces from both American and European origins, previously analyzed for SSR markers (Dubreuil et al. 2006) and for which flowering time data were available (Rebourg 2000; Gouesnard et al. 2002). For these different genetic materials, we investigated (i) population structure based on SSR markers using the Structure software (Pritchard et al. 2000a); (ii) the relationship between population structure and flowering time variation, to assess the specificities of the different genetic groups with respect to this trait; (iii) Dwarf8 differentiation among groups; and (iv) the association between flowering time variation, population structure, and Dwarf8 polymorphism by combining association tests with the study of among-group differentiation.

MATERIALS AND METHODS

Plant material:

A 375 inbred line collection was defined with the objective of representing European and American diversity and covering a wide range of flowering times. The collection includes 153 inbreds obtained directly by selfing from traditional landraces (open-pollinated varieties). These 153 lines, referred to as the first-cycle inbred panel, were used to assess genetic structure of the ancestral inbred gene pool used for modern selection. We added to this panel 220 inbreds either of most advanced selection cycles or from synthetic populations like CYMMIT or Stiff Stalk populations. These 220 inbreds, representative of more recent material, should present new phenotypic characteristics such as those associated with high performance. The 375 inbred panel, referred to as the whole inbred panel, includes the 102 inbred line panel studied by Remington et al. (2001) and Thornsberry et al. (2001) and additional dent lines studied by Liu et al. (2003). A detailed list of genotypes and origins is available as supplemental information at http://www.genetics.org/supplemental/ or from the corresponding author, with information regarding reference stock centers for seed request.

To compare inbred lines with traditional maize accessions, we conducted new statistical analyses for SSR data from the landrace panel defined and genotyped by Dubreuil et al. (2006). It includes 131 European landrace accessions from eight geographical areas of primary introduction or traditional cultivation (Spain, Portugal, Italy, France, Germany, former Yugoslavia, Switzerland, Austria, Poland, Romania, Ukraine, Bulgaria, Czech Republic, and Slovakia) and 144 accessions among American racial groups including various types from the southwestern and northeastern United States, highland Mexico, core Andes, southern Chile and Argentina, lowland Mexico, Guatemala, the Caribbean, and Central America.

Field data:

For inbred lines, flowering time under long-day conditions (~15 hr) was evaluated in 2002 at two locations, St. Martin de Hinx (southwestern France) and Gif-sur-Yvette (Paris region, France). Two replicates were planted at each location for each line, each replicate consisting of 15 plant rows planted at a density of approximately six plants per square meter. Days to pollen expressed in thermal time following Ritchie and Nesmith (1991) (with parameter values Tb = 6° and To = 30°) was selected as a measure of flowering time because of its high heritability. A global ANOVA of the data was performed using the GLM procedure in SAS (1989) to test the significance of genotype and location effects and genotype-by-location interaction. Considering the result that genotype-by-location interaction was low compared to genotype effect, we used the adjusted mean (estimated using the LSMEANS statement in the GLM procedure of SAS) of each genotype for further analyses. Heritability was calculated as described in Gallais (1990).

Flowering time under short-day conditions (13 hr 10 min) for inbred lines was evaluated in Petit Bourg (Guadeloupe) in 2003 and 2004 with a two-repetition trial and the same sowing conditions as in Gif-sur-Yvette and Saint Martin de Hinx. As for long-day conditions, days to pollen expressed in thermal time was selected as a measure of flowering time. The 2004 trial was lost because of extreme climatic conditions and insect pressure.

We used data from Rebourg (2000) and Rebourg et al. (2001) for flowering time of temperate landraces under long-day conditions and those from Gouesnard et al. (2002) for flowering time of tropical landraces under both short- (13 hr 10 min) and long-day (15 hr) conditions. Both temperate and tropical landraces used in these studies were evaluated in Montpellier and Gif-sur-Yvette in 1997.

Photoperiod sensitivity was taken from Gouesnard et al. (2002) for tropical landraces and evaluated for inbred lines using flowering time data from Gif-sur-Yvette (15 hr 33 min) and Petit Bourg (13 hr 10 min), with the same method as that of Gouesnard et al. (2002).

Genotyping:

Fifty-five SSR were used to analyze genetic diversity and population structure of both inbred panels. Primer pairs were chosen on the basis of their ability to detect a single locus, their broad genome coverage, and their good reproducibility in allele size determination. We avoided dinucleotide SSR because of their possible high mutation rate (Vigouroux et al. 2002), which may cause departure from the infinite-allele model (Excoffier and Hamilton 2003). Table 1 summarizes the main characteristics of these loci in terms of position and motif. Primer sequences are available from the maize database project, at the University of Missouri (http://www.maizegdb.org/). Eighteen of these SSR loci are in common with the 24 used by Dubreuil et al. (2006) to characterize maize landraces using a bulk method.

TABLE 1
Diversity of SSR markers in three maize panels: (i) 275 landraces, (ii) 153 first-cycle inbred lines issued from landraces, and (iii) 375 inbred lines

Leaf tissue samples for the whole inbred panel were obtained from 10 plants per inbred line. DNA was extracted following Tai and Tanksley (1991) with minor modifications. PCR reactions were performed in 20-μl volumes containing 15–30 ng DNA template, 1× PCR buffer, 0.2 mm dNTPs, 3 mm MgCl2, 1 unit of Taq polymerase, 0.1 μmol of each forward (with M13 tail) and reverse primer, and 0.03 μmol of IRD-labeled M13 primer. Thermocycling consisted of initial denaturation of DNA template at 95° for 5 min followed by 30 cycles of 95° for 20 sec, 56° for 20 sec, and 72° for 30 sec, and a final extension of 72° for 3 min. SSR were multiplexed by two according to their expected size, with loading buffer (IR2 loading solution, Li-Cor), heated at 95° for 5 min, and then placed on ice. Denatured samples (0.5 μl) were loaded on 6.5% KB+ (Li-Cor) gels in 1× TBE buffer (89 mm Tris, 89 mm Borate, 2 mm EDTA, pH 8.2) and electrophoresed at 2000 V for 3 hr on an automated DNA sequencer (model IR2, Li-Cor). Gels were run in a 96-well format. Size standards (50–350 bp, Li-Cor) were regularly spaced in gels (every 19 lanes). Fragment sizes were determined on the basis of their migration relative to that of size standards using One-Dscan v. 2.05 software (Scanalytics). Alleles displaying very close sizes were pooled. SSR data for inbred lines are available as supplemental information at http://www.genetics.org/supplemental/ or from the corresponding author.

For the Dwarf8 gene (located on chromosome 1, in bin 1.09, see Gardiner et al.1993 for bin definition), we genotyped, for both inbred panels and the landrace panel analyzed by Dubreuil et al. (2006), the polymorphism that showed the stronger association with flowering time according to Thornsberry et al. (2001). This polymorphism is a 6-bp insertion/deletion at position 3472 and is referred to as D8idp. The deletion allele is further noted as D8-deletion. This polymorphism was characterized through the size difference of PCR products. PCR reactions were performed in 20-μl volumes containing 37 ng of template DNA, 1× PCR buffer, 0.1 mm of each dNTP, 1× QIAGEN (Valencia, CA) buffer, 0.6 unit of Taq polymerase, 0.04 μmol of forward primer (5′-CGTTCCTCGACCGCTTCACC-3′ with M13 tail), 0.4 μmol of reverse primer (5′-GGTACACCTCCGACATGACCT-3′), and 0.36 μmol of IRD-labeled M13 primer. A touchdown PCR amplification was carried out as follows: 4 min at 95°; 10 cycles of 20 sec at 95°, 20 sec annealing decreasing 1° per cycle from 64° to 55°, and 30 sec elongation at 72°; followed by 21 cycles of 20 sec at 95°, 20 sec at 54°, and 30 sec at 72°; and a final extension of 5 min at 72°. Fragment migration and gel reading were done with the same protocol as that for SSR. Inbred lines were individually genotyped for D8idp. For landraces, the frequency of the deletion allele was estimated using bulked DNA samples from Dubreuil et al. (2006) for SSR. Reliability of D8idp bulk analysis on landraces was assessed by comparing expected and estimated allele frequencies obtained from 10 controlled DNA pools made from 100 mg lyophilized leaf from the FV75 inbred (homozygote for the deletion allele) and the LAN496 inbred (homozygote for the insertion allele) in various proportions. R2 between expected and estimated allele frequencies was 98.4%.

Diversity and genetic structure analysis:

We calculated Nei's unbiased genetic diversity Ht (Nei 1978) at SSR loci for the panels of inbred lines. Ht estimates heterozygosity expected under the hypothesis of panmixy. To have a measure of allelic richness independent of sample size, we used FSTAT software (Goudet 2001) to calculate the estimator proposed by Petit et al. (1998). These values were compared to those obtained by Dubreuil et al. (2006) on landraces. However, allelic richness values were similar to the average number of alleles and were thus not shown. Genetic structure was assessed using the Structure software package (Pritchard et al. 2000a; Falush et al. 2003) on SSR genotypic data for landrace, whole-inbred, and first-cycle inbred panels. Structure results of individual attribution to genetic groups (percentage of genome of each individual attributed to each group) were graphically displayed using the Distruct software (Rosenberg 2002).

Results on the whole inbred and first-cycle inbred panels were obtained as described in our companion technical note (L. Camus-Kulandaivelu, J.-B. Veyrieras, B. Gouesnard, A. Charcosset and D. Manicacci, unpublished results, companion technical note available on request from the corresponding author). For both panels, we performed 10 independent runs of Structure for numbers of groups varying from 2 to 10, leading to 90 Structure outputs. For landraces, allele frequencies obtained from the analysis of pools (Dubreuil et al. 2006) were used to simulate sets of five haplotypes per landrace, under the hypothesis of equilibrium. This made it possible to use the haploid option in Structure as for inbred panels and thus to keep the data a manageable size. We calculated the group attribution of each landrace as the mean of its five simulated haplotype group attributions. Because the options of the iteration procedure used for inbreds would be too computer-time consuming for landraces, we set a 105 iteration burn-in period and 105 iteration sampling period and performed only 3 independent runs for each group number, leading to 27 Structure outputs.

For the three panels, we selected the best output for each group number on the basis of goodness-of-fit criteria. We estimated group mean contribution (GMC) using Structure to quantify each group's relative importance. We also estimated the divergence of each group relative to a putative ancestral pool using Fst-statistics analogous to Wright's Fst (Wright 1951). Finally, we determined the best group number for each panel on the basis of goodness-of-fit criteria (see our companion technical note available on request from the corresponding author) and, to a limited extent, on the basis of consistency of group composition based on a priori knowledge of genetic origins. For these three outputs (landrace, first-cycle inbred, and whole inbred panels), further referred to as reference outputs, we calculated Nei's relative genetic differentiation among groups (Gst) as well as Nei's diversity index of each group, using allelic frequencies estimated for the different groups.

Relationships between the first-cycle inbred panel and the landrace panel were also investigated by running Structure on these two data sets pooled together, using the 18 common SSRs. Relationships between genetic groups of the whole inbred panel and the first-cycle inbred panel were identified by calculating Euclidian distances between groups defined for the reference output of each panel, using group predicted allelic frequencies provided by Structure.

Relationship between flowering time, D8idp, and genetic structure:

The relationship between population structure and flowering time was investigated using the linear regression model,

equation M1

(GLM procedure in SAS 1989), where Tj stands for the trait value of genotype j, ao for the intercept, gij for the proportion of the genotype j genome attributed to group i (k groups in total), ai for the effect of group i, and ej for the residual. The effect of population structure on the trait was quantified with the determination coefficient (R2). Average value of group i was estimated as a0 + ai for ik and as a0 for group k.

The effect of population structure on flowering time under long-day conditions was evaluated on the whole inbred panel using the 90 Structure outputs, while only the reference output was considered for the first-cycle inbred panel and landrace panel. Effect of genetic structure on flowering time under short-day conditions and on photoperiod sensitivity was also evaluated for the three panels with Structure reference outputs.

The effect of population structure on D8idp repartition was investigated using the logistic regression model (LOGISTIC procedure in SAS 1989):

equation M2

SNPj stands for the D8-deletion allele frequency in genotype j, i.e., 0 or 1 for inbred lines. Max-rescaled pseudo-R2 of logistic regression was used to quantify the association between population structure and D8idp. Estimates of D8-deletion frequency in Structure groups were calculated as

equation M3

and

equation M4

This was calculated for each of the 90 outputs obtained for the whole inbred panel and for reference output only for the first-cycle inbred panel. Since D8-deletion frequency varies in a continuous way in landrace pooled samples, population structure effect on D8idp and D8-deletion frequency in landrace groups were calculated using linear regression as for phenotypic traits.

Association between D8idp and flowering time was investigated using both linear (GLM procedure) and logistic (LOGISTIC procedure) regressions (SAS 1989), and test significance was evaluated using Student's and chi-square tests for linear and logistic regressions, respectively. In both cases, the proportions of individual genome attributed to each group were included in the statistical model as covariates, accounting for k − 1 d.f. D8idp effect on flowering time was estimated from the corresponding linear regression coefficients. We tested association between D8idp and flowering time under long-day conditions on the 90 outputs obtained for the whole inbred panel (see our companion technical note, available on request from the corresponding author), to examine test stability. This association was also tested on landrace and first-cycle inbred panels for reference outputs only.

Association between D8idp and the two other traits (flowering time under short-day conditions and photoperiod sensitivity) was tested for the whole inbred panel and on the tropical landrace subpanel using reference outputs.

RESULTS

Flowering time variation:

Flowering time variation among inbred lines was measured under long (15 hr 30 min) and short (13 hr 10 min) day conditions, allowing us to assess genotype and location effects and their interaction. For the inbred line experiment, genotype effect on flowering time under long day conditions was highly significant (mean square = 70,430; P < 0.0001). We observed very contrasted flowering times in the whole inbred panel, from 808 degree days (dd) (FV268) to 1647 dd (H16), with a mean value of 1043 dd and a standard deviation of 155 dd. We found no significant effect of repetition, a high repeatability (Pearson's correlation coefficient of 0.95 between observations for the genotypes in the two replicates within a location), and a high heritability (0.97). Genotype-by-location interaction (mean square = 1415, P < 0.0001) was significant but low compared to genotype effect. It was thus not considered in further analyses.

Under short-day conditions, flowering time varied significantly among lines (mean square = 20,787; P < 0.0001), from 782 dd (AP1) to 1329 dd (Ky228), with mean of 1020 dd and standard deviation of 97 dd. We observed a lower repeatability (Pearson's correlation coefficient of 0.79) than that for trials in long-day conditions. This and the fact that a single trial could be considered lead to a lower heritability (0.85). Results for flowering time under short-day conditions and photoperiod sensitivity should therefore be considered with caution.

Photoperiod sensitivity evaluation showed contrasted behaviors of the tested inbred lines: photoperiod-sensitive lines with more temperature requirements under long days than under short days (up to 131.90 dd · hr−1 for inbred EM1197), photoperiod-insensitive materials (e.g., −0.80 dd · hr−1 for inbred FC201), and also lines with less temperature requirements under long days than under short days (up to −74.40 dd · hr−1 for inbred Fv66). This last situation was observed for temperate lines and is likely due to a slowdown of plant development due to a lack of adaptation to environmental conditions (e.g., air dryness) rather than to the effect of day length. The mean value of photoperiod sensitivity was 0.64 dd·hr−1 and the standard deviation was 42.00 dd·hr−1.

Flowering times under long- and short-day conditions were significantly correlated to each other (R2 = 0.45; P < 0.0001). Photoperiod sensitivity was correlated to flowering time under long-day conditions (R2 = 0.45; P < 0.001) but not to flowering time under short-day conditions (R2 = 0.01; P = 0.058).

Diversity at SSR loci and structure analysis:

For the 55 SSR loci used to characterize inbred lines, we scored a total of 358 alleles for the whole inbred panel, of which 330 were present in the first-cycle inbred panel. The average number of detected alleles per locus was 6.5 for the whole inbred panel and 6.0 for the first-cycle inbred panel. Average diversity (Ht) was 0.61 for both the whole inbred panel and the first-cycle inbred panel. For the sake of comparison, note that the landrace panel genotyped for 24 SSR showed on average 7.7 alleles per locus and an average diversity of 0.63 (Dubreuil et al. 2006). Considering only the 18 SSR common to the landrace and inbred panels, average diversities per locus were 0.62, 0.64, and 0.62 for the whole inbred panel, the first-cycle inbred panel, and the landrace panel, respectively. Detailed diversity statistics are presented, for each locus, in Table 1.

SSR polymorphisms were used to assess the genetic structure of the three panels using the Structure software. A detailed analysis of the stability of the results obtained for different runs of Structure for the same group number was conducted for the inbred panels (see our companion technical note, available from the corresponding author on request). For the three panels, Table 2 describes the best output for each group number based on goodness-of-fit criteria (see our companion technical note). The designation of a given group (j) was based on the origins of the materials with a high genome proportion attributed to this group (gij > 0.80). Two-group models discriminated between a “Flint” group (early materials from both European and American origins) and a “non-Flint” group, for each of the three panels. The three-group models subdivided the Flint group into a “Northern Flint” group and a “European Flint” group for both landrace and first-cycle inbred panels, whereas it subdivided the non-Flint group into “tropical” and “Corn Belt Dent” groups for the total panel. Similarly, a further increase in group number led mostly to group subdivisions. Interestingly, Northern Flint materials from European or American origins were never split in any panel and group number, confirming their very high similarity. Similarly, some southern Spanish materials always remained clustered with tropical material.

TABLE 2
Group composition in Structure outputs with best goodness-of-fit, depending on the number of groups (Groups) for the three panels (landraces and first-cycle inbred panel and whole inbred panel)

Among the best outputs for each number of groups, we determined the best model for each panel using the goodness-of-fit criterion. As discussed in our companion technical note (available on request from the corresponding author), this led unambiguously to five groups for first-cycle inbreds, the origin of which appeared highly consistent with former knowledge (see discussion). A similar pattern was observed for landraces. However, in the latter panel, while highest goodness-of-fit was observed for eight groups, seven- and eight-group models differed only by an additional group of four obviously unrelated landraces. We therefore considered the seven-group output more relevant. Model choice was less straightforward for the whole inbred panel due to the lack of a clear stabilization of the statistics with the increase in group number (see our companion technical note). The two possible best outputs (five and seven groups) showed a high consistency with knowledge on material pedigrees. The output with the smallest group number (five groups) was chosen as advised in the Structure documentation. Beyond the optimal group number, Structure tended to generate small groups of heteroclite material in landrace and first-cycle inbred panels (Table 2). On the contrary, for the whole inbred panel, additional groups corresponded to line families [groups of related inbreds with a high contribution of one or few major progenitor(s)]. This was observed in the seven-group output, for instance, for lines related to the two major progenitors F2 and F7 issued from the Lacaune population.

For the reference outputs, Fst showed contrasted values between groups, indicating variable differentiation levels from the ancestral pool (Table 2). Tropical groups for both inbred panels as well as “Mexican” and “Caribbean” groups for the landrace panel showed very low Fst-values, indicating that allelic frequencies of these tropical groups are very close to those of the maize ancestral genetic pool that have generated the materials analyzed here. On the contrary, groups shaped by recent breeding such as that of inbreds related to Stiff Stalk materials showed high Fst-values, indicating strong differentiation. The Northern Flint group also displayed high Fst-values in the three panels, consistent with its established high divergence from other materials (Doebley et al. 1986).

To give a synthetic picture of maize population structure in the three panels, we represented jointly the reference outputs in Figure 1. To further investigate the relationships between these models, Structure was run on first-cycle inbreds and landraces together. This yielded an optimum of seven groups, with the same origins (Northern Flint, “Pyrenees–Galicia Flint,” “Italian Flint,” Corn Belt Dent, Mexican, Caribbean, and “Andean”) as those defined for the landrace panel only. As expected, inbred lines from Northern Flint origin and Northern Flint landraces were attributed to the same group. A similar pattern was observed for Corn Belt Dent lines and landraces. European Flint inbreds were mostly grouped with Pyrenees–Galicia Flint landraces. On the other hand, inbred lines related to Italian landraces were attributed to the European Flint group in the inbred analysis. In the joint analysis of landraces and inbreds, some of them were attributed to an Italian Flint group whereas others were considered as the result of admixture between Italian Flint and Pyrenees–Galicia Flint groups. Spanish inbred lines attributed to the Tropical group in the inbred line analysis were attributed mainly to the Caribbean group, while South American inbreds like ARGL256 and ZN6 showed an admixture between Andean and Italian Flint groups. A single inbred (P9COS6) had >50% of its genome originating from the Mexican group. These relationships permit inferences of filiations from landrace to first-cycle inbred groups (Figure 1, arrows). Afterward, relationships between population structure of the first-cycle inbred panel and the whole inbred panel were determined from the genetic distances between the corresponding groups (Table 3). Adding lines from advanced breeding generations to the first-cycle inbred lines led to the identification of a new dent subgroup (“Stiff Stalk”) in addition to the four main groups determined for the first-cycle inbred panel (Northern Flint, European Flint, Corn Belt Dent, and Tropical), the allelic frequencies of which remained stable. However, it can be noted that the small “Popcorn” inbred group identified from the first-cycle inbred panel clustered with the Corn Belt Dent group in the whole inbred panel.

Figure 1.
Models for population structure at three steps of maize selection history: landraces (seven groups), first-cycle inbreds (five groups), and whole inbred panel (five groups). Groups for each panel are represented by colors as indicated at the bottom. For ...
TABLE 3
Genetic distances between groups defined from the first-cycle panel and the whole inbred panel reference Structure outputs (Figure 1)

Relationship between population structure, flowering time, and D8idp:

In the whole inbred panel, population structure showed a highly significant (P < 0.0001) and very strong association with flowering time under long-day conditions from 3- to 10-group outputs (Table 4). It explained on average (over runs for the same group number) from 42.9% (3 groups) to 54.7% (10 groups) of flowering time variation. For 3–10 groups, different runs of Structure with the same group number led to stable results (standard errors from 0 to 3%). For 2 groups, genetic structure showed a contrasting effect on flowering time variation depending on group composition. Two-group outputs can be classified into two categories (see our companion technical note available on request from the corresponding author) that correspond to two very different genetic structures. For some outputs, the two groups correspond to Flint and non-Flint materials. Population structure explained in this case 24% of flowering time variation. Conversely, for the other outputs, Flint and tropical materials clustered into a single group. The genetic structure described by these outputs failed to explain flowering time variation (R2 = 0.2%).

TABLE 4
Relationship between population structure and flowering time (FT) under long-day conditions (R2), association between population structure and D8idp (pseudo R2), and association between D8idp and flowering time taking population structure into account ...

For the first-cycle inbred and landrace panels, structure effect on flowering time under long-day conditions was calculated only for reference outputs. The five-group model for the first-cycle inbred panel and the seven group model for the landrace panel explained, respectively, 39.4 and 66.8% of flowering time variation under long-day conditions. The estimation of average group flowering time (Table 5) showed the same trend in the three panels. Northern Flint, Pyrenees–Galicia Flint, and, to a lesser extent, Italian Flint groups in the landrace panel displayed the earliest flowering time. Tropical groups, including Mexican, Caribbean, and Andean landrace groups, displayed the latest flowering time, while Dent groups (Corn Belt Dent and Stiff Stalk in the whole inbred panel) exhibited an intermediate flowering time. The largest differences between groups were 387 dd between Tropical and Northern Flint groups in the whole inbred panel and 706 dd between Caribbean and Pyrenees–Galicia Flint in the landrace panel.

TABLE 5
Average flowering time under long-day conditions (FT) and average D8-deletion frequency for each group of the three panels (landrace, first-cycle inbred, and whole inbred)

Population structure effect on flowering time under short-day conditions and on photoperiod sensitivity was calculated for the whole inbred panel and for a tropical landrace subpanel using reference outputs. Structure effect on flowering time under short-day conditions was significant (P < 0.0001) for both inbred lines (R2 = 0.26) and tropical landraces (R2 = 0.25). For the sake of comparison, genetic structure effect on flowering time under long-day conditions for the tropical landrace subpanel was found significant (R2 = 0.33; P < 0.0001). In the whole inbred panel, the Northern Flint group remained the earliest (926 dd) and the Tropical group the latest (1081 dd) flowering material under short-day conditions. However, the difference in flowering time between these two extreme groups was reduced (155 dd) in short-day conditions, when compared to long-day conditions (387 dd). Structure effect on photoperiod sensitivity was significant for both tropical landraces (R2 = 0.41; P < 0.0001) and inbreds (R2 = 0.08; P = 0.0001). In the inbred panel, the only photosensitive group (+32 dd · hr−1) is the tropical one, while other groups are photoperiod insensitive (−17 to −4 dd · hr−1).

Population structure showed a significant and very strong effect on D8idp polymorphism distribution in the whole inbred panel using 3- to 10-group Structure outputs (Table 4), with max-rescaled pseudo-R2 averaging 0.37 (3 groups) to 0.47 (9–10 groups). Different runs of Structure for the same group number exhibited stable results (standard errors varying from 0.013 to 2.2%). As observed for flowering time under long-day conditions, two-group outputs showed contrasted max-rescaled pseudo-R2 values, depending on their group composition. Outputs discriminating Flint vs. non-Flint showed very high max-rescaled pseudo-R2 values (0.35) while outputs gathering Flint and Tropical material exhibited lower max-rescaled pseudo-R2 values (0.12). The first-cycle inbred reference output showed a max-rescaled pseudo-R2 of 0.51 and the landrace reference output showed a max-rescaled pseudo-R2 of 0.55. Estimated frequencies of D8-deletion in the landrace panel, in the first-cycle inbred panel, and in the whole inbred panel for reference outputs are presented in Table 5. Northern Flint and Tropical groups displayed contrasted frequencies for D8idp. In the whole inbred panel, the deletion was rare in the Tropical group (1.9%) while predominant in the Northern Flint group (82%). In the landrace panel, while the contrasted frequency of D8-deletion between Northern Flint (83.0%) and tropical groups Caribbean (2.0%) and Mexican (4.0%) remained strong, we also observed a high frequency of the D8-deletion in the Andean group (58.0%) that gathers material originating from elevated places (on average 2200 m) of Andean tropical regions (Peru, Ecuador, and Bolivia). To illustrate these results, the deletion frequency in landraces was represented according to the population geographical coordinates (Figure 2). The D8-deletion prevails in the landraces originating from northern Europe, northern North America, southern South America, and the Andean tropical region. D8idp Gst was 0.388 in the whole inbred panel and 0.407 in the landrace panel, which is higher than that observed for any SSR (Table 1).

Figure 2.
Geographical distribution of D8-deletion frequency among landraces.

Association between D8idp and flowering time under long-day conditions was highly significant for the whole inbred panel, when ignoring genetic structure, with both logistic and linear regressions (P < 0.0001). The D8-deletion was associated in this case to a 130-dd earlier flowering (relative to the D8-insertion). This association was also significant when considering two-group Structure outputs as covariates (P < 0.01). However, two contrasting patterns were observed among two-group outputs. The D8-deletion allele was associated on average with 148 dd earlier flowering when Flint material was clustered with Tropical material and only with 54 dd earlier flowering when Flint material was separated from Tropical material. Note that the deletion was associated with early flowering time. Using logistic regression with three or more groups yielded 79% significant association tests (63 of 80) while 4% of these tests (3 of 80) exhibited P-values > 0.1. Linear regression always showed larger P-values than logistic regression. Linear regression yielded 19% significant tests (15 of 80) and 36% tests (29 of 80) with P-values > 0.1. Mean P-values obtained for each group number are presented in Table 4. For models with three or more groups, the D8-deletion was associated with a 29- to 37-dd earlier flowering (Table 4).

Association between D8idp and flowering time under long-day conditions was also tested for the first-cycle inbred line and landrace panels using reference outputs. No significant association was found on the first-cycle inbred panel, potentially due to the smaller size of this sample. However, when not taking structure into account, significant association was found in the first-cycle inbred panel (P = 0.01 for both linear and logistic regressions). Conversely, D8idp and flowering time under long-day conditions were very strongly associated (P = 0.0007) in the landrace panel. In this case, the D8-deletion was associated with a 145-dd earlier flowering time. No association between D8idp and flowering time under long-day conditions was detected on the tropical landrace subpanel.

Still taking Structure reference outputs into account, no association between D8idp and flowering time under short-day conditions was detected either on the tropical landrace panel or on the whole inbred panel. Association between D8idp and photoperiod sensitivity was close to significance (logistic regression: P = 0.052) in the whole inbred panel using reference Structure output. When considering all the outputs from 3–10 groups as covariates, 25% of the association tests showed a P-value <0.05 with logistical regression and 28.9% with linear regression. No association was found between photoperiod sensitivity and D8idp in the tropical landrace subpanel.

DISCUSSION

Effect of history and recent selection on population structure:

The average amount of genetic diversity at SSR markers showed low variation among the three panels that were analyzed: (i) the total collection of 375 inbred lines (whole inbred panel), (ii) the subcollection of 153 first-generation inbred lines issued from landraces (first-cycle inbred panel), and (iii) a collection of 275 traditional landraces (landrace panel). Average allele number per locus varied from 6.0 to 7.7 and genetic diversity from 0.61 to 0.64. These numbers are comparable to the ones from previous studies in maize (Senior et al. 1996; Taramino and Tingey 1996; Lu and Bernardo 2001; Matsuoka et al. 2002a). However, diversity in our study was lower than that reported by Liu et al. (2003) on a collection of 260 diverse maize inbred lines: 21.7 alleles per locus on average and an average genetic diversity of 0.82. This discrepancy is probably due to the presence in Liu et al.'s (2003) article of dinucleotide SSR that have high mutation rates (Vigouroux et al. 2002).

The most clear-cut trend observed in the population structure of all three panels that we analyzed is the splitting of Northern Flint (and related material such as European Flint or Pyrenees–Galicia Flint) from the rest of the collection. This is in accordance with previous work showing the striking divergence of Northern Flint from Corn Belt Dent and some tropical material on the basis of isozyme loci (Doebley et al. 1986). This genetic feature of Northern Flint was not reported by Thornsberry et al. (2001) and Liu et al. (2003) probably because they focused mostly on Corn Belt Dent and tropical diversity. A second important result is the organization of tropical landraces into three groups that can be referred to as Mexican, Caribbean, and Andean. Mexican and Caribbean groups showed a low differentiation relative to the assumed ancestral gene pool (Table 2), which is consistent with their geographical proximity with the domestication center. The Andean group appeared clearly isolated and was more differentiated than other tropical material, as reported by Matsuoka et al. (2002a). This is consistent with the ancient introduction of maize in core Andes (Freitas et al. 2003). Contrary to tropical landraces, tropical inbred lines did not display any genetic structure, for both inbred panels. This may be due to the heterogeneous origins of these lines (southern Spain, southern America, Mexican Tuxpeno, and highlands) and to the fact that they were derived mostly from synthetic populations with a broad genetic basis (Reif et al. 2003).

The finding of southern Spanish landraces clustering with tropical material (Caribbean and Mexican) as well as the clustering of northeastern European material with U.S. Northern Flints is consistent with the hypothesis of a double introduction of maize in Europe proposed by Rebourg et al. (2003) and Dubreuil et al. (2006). This remains true whatever the group number considered in Structure runs, confirming the close proximity between these European and American materials (Rebourg et al. 2003). The rest of European material, spread mainly in central and northern Spain, France, and Italy, appeared as a distinct genetic cluster that does not have any counterpart in America. This material appeared as a single European Flint group for inbred lines and as two distinct Pyrenees–Galicia Flint and Italian Flint groups for landraces. This supports the hypothesis of Rebourg et al. (2003) that this material results from the hybridization of populations derived from Tropical and Northern Flint introductions in Europe. Similarly, Corn Belt Dent is known to result from the hybridization between Northern Flint and Southern Dent (Doebley et al. 1988). Interestingly, populations derived from these two independent hybridizations between Northern Flint material and late materials from tropical origins are identified by Structure as individualized groups (European Flint, Pyrenees–Galicia Flint, and Corn Belt Dent, see Table 2). The signature of admixture between parental populations is thus not detected using the Structure software. This may be due to the accumulation of numerous recombination events and/or genetic drift, resulting in a strong differentiation of the derived populations.

Results obtained from the whole inbred panel, as compared to those observed for the first-cycle inbred and landrace panels, show that population structure has been further shaped by modern breeding achieved during the last half-century. Indeed, inbred lines related to the Iowa Stiff Stalk synthetic group cluster together in the whole inbred panel (Table 2, Stiff Stalk group), whereas initial Corn Belt Dent material appears as a single homogeneous group within the landrace panel. This is consistent with the current structuring of U.S. hybrid programs into Stiff Stalk and non-Stiff Stalk materials (Duvick et al. 2004). In addition, the development of new breeding populations by crossing existing inbred lines together (Gerdes and Tracy 1993) also led to some highly related “families” of inbred lines that are detected as individual groups when running the Structure software for high group numbers. As a consequence, the Structure software shows no clear stabilization of the goodness-of-fit criteria as group number increases for the whole inbred panel (see our companion technical note, available from the corresponding author on request), making it difficult to conclude for any group number and composition. This effect of relatedness among inbred lines on the stability of Structure outputs is similar to that described by Liu et al. (2003).

Diversifying selection on flowering time:

The strong variation for flowering time and photoperiod sensitivity in maize has been known for long time (Bonhomme et al. 1994; Gouesnard et al. 2002) and is illustrated by the range of variation found in this study (more than a twofold variation in degree days needed to reach flowering time in long-day conditions, from the earliest to the latest genotype). Strong correlations were observed in the whole inbred panel between population structure (five groups reference output) and (i) flowering time under long-day conditions (R2 = 0.47), (ii) flowering time under short-day conditions (R2 = 0.26), and (iii) photoperiod sensitivity (R2 = 0.08), although to a lesser extent. This effect of population structure was even larger for landrace flowering time under long-day conditions (R2 = 0.67), which may be due to the effect of recent selection that tended to eliminate very early types from European and Corn Belt genetic pools (A. Charcosset, personal communication). These results show that groups established on the basis of neutral markers (SSR) are strongly differentiated for their flowering time determination.

Geographical variation of a phenotypic trait such as flowering time may be the result of adaptation and/or genetic drift. The very high correlation of population structure and flowering time under long-day conditions (R2 = 0.47) and the clear consistency between group average flowering time and local climatic characteristics preclude that this differentiation is the effect of genetic drift only. For instance, in the whole inbred panel, the largest between-group difference in precocity is that observed between Northern Flint and Tropical materials, Northern Flints flowering 387 dd earlier. Northern Flint is a genetic pool created by Native Americans and cultivated in eastern North America up to cool regions of the Saint Laurent bay (Dubreuil et al. 2006) at the time of its discovery by Europeans. Following the discovery, the Northern Flint group was crossed with tropical or subtropical materials by North American colonists and Europeans, leading to new temperate material, Corn Belt Dent and European Flint/Pyrenees–Galicia Flint, respectively, that are adapted to intermediate climates. In the landrace panel, the Pyrenees–Galicia Flint group that includes materials originating from the Pyrenees Mountains appears on average as early as the Northern Flint group, whereas Corn Belt Dent material is later than its Northern Flint parental pool. This is consistent with local climatic characteristics, the Corn Belt being warmer on average than the Pyrenees Mountains. These results call for a detailed investigation of the growing season length, i.e., the duration of maize cultivation possible given the local environmental conditions, at the geographical origins of the landraces. This investigation should ideally consider possible variation in climate through history (see, for instance, Haug et al. 2003). The investigation of the relationship between growing season length, flowering time, population structure, and the frequency of polymorphism potentially involved in flowering time variation should bring more precise elements into the maize adaptation process.

Dwarf8 association with flowering time and its possible role in maize adaptation to temperate climate:

In this study, we confirmed the association between D8idp and flowering time in panels of inbred lines and landraces representing American, European, and tropical maize diversity. These are wider maize samples than the panel used by Thornsberry et al. (2001), who mainly worked on Corn Belt Dent and tropical inbred material. More recently, this association was reinvestigated in a 71 elite European inbred line panel by Andersen et al. (2005), who found a very strong association (P < 0.0001) between D8idp and flowering time. This association, however, was no longer significant after correction for population structure. We confirmed in the present study a strong association between D8idp and flowering time in the panels when not correcting for population structure. After correcting for population structure, this association remained highly significant for the landrace panel but was no longer significant for the first-cycle inbred panel. For the whole inbred line panel, association was significant with logistic regression and close to significance with linear regression. Considering that the direction of the effect was known a priori (Thornsberry et al. 2001), unilateral tests indeed would have led to average P-values from 0.03 to 0.05 for 4–10 groups. Altogether, these results illustrate that the power of association test while correcting for population structure is a burning issue.

Maize flowering time variation among panels from diverse origins certainly represents an extreme situation in this respect, because the genes involved in this trait have played a key role in defining population structure. When correcting for population structure, population structure estimates consume part of the effect of the candidate gene that is correlated to it. When considering linear regression, the power of the test is directly related to the fraction of the variation explained by the candidate polymorphism after correction for population structure, i.e., the partial R2. Given that the remaining variation contains environmental variation, extreme situations may occur where there is no longer any genetic variation to be explained. In this study, flowering time measured for inbred lines under long-day conditions has a high heritability (97%) and population structure explains approximately half of genetic variation (38 and 49% for the first-cycle and whole inbred panels, respectively). In such a situation, taking into account population structure will increase the power of the test for genes in which the polymorphism is loosely linked to population structure and conversely decrease it in the case of a strong association with population structure. The association between D8idp and flowering time is clearly that of the latter case (see below). In this context, the repeatability of this association in three diverse studies strongly supports the relevancy of association studies for other traits more loosely related to population structure.

Another consequence of the strong correlation between population structure and D8idp is to make it difficult to estimate D8idp effect on flowering time. We propose that the lower bound of the D8-deletion effect on flowering time under long-day conditions is −29 dd (estimated when accounting for population structure for the whole inbred panel) and the upper bound is −130 dd (estimated when not accounting for population structure). It can be noted that the D8-deletion effect on flowering time calculated on landraces is stronger (−145 dd accounting for population structure) than that on inbred lines. This could be due to epistatic interactions between Dwarf8 and other loci that differ in allelic frequencies and/or genetic effects between the inbred and landrace panels.

In addition to investigating association, this study provides us with two other elements supporting that Dwarf8 is a QTL for flowering time. First, Dwarf8 seems to have been submitted to diversifying selection. Diversifying selection on DNA markers is usually detected by comparing differentiation levels (estimates of Fst) obtained in large data sets (Beaumont and Nichols 1996; Vitalis et al. 2001). The differentiation level in the whole inbred panel for D8idp (Gst = 0.388) is higher than that for all neutral SSR markers (maximal Gst-value: 0.282 for marker phi427913). A strong differentiation of D8idp is also found in the two other panels, thus making it very unlikely that the higher Gst-value of D8idp is the result of drift alone. The second element is the predominance of the early allele (D8-deletion) in the earlier material (82% in the whole inbred panel Northern Flint group) and the predominance of the late allele in the later material (98.1% in the whole inbred panel tropical group). This is in accordance with the results of Andersen et al. (2005), who found that the early allele (T, corresponding to the D8-deletion) was fixed among a flint group identified by Structure. Altogether, the results obtained here support the implication of Dwarf8 polymorphism in flowering time variation in maize. Further analysis of the contribution of Dwarf8 to the dynamics of maize adaptation should include a comprehensive study of the correlation between D8idp frequencies and the growing season length at the scale of the geographical origins of the landraces. Also, additional investigations should be performed with the other polymorphisms of Dwarf8 to investigate their possible role on the flowering time phenotype. In particular, it is important to discriminate between the effect of the polymorphisms located in the ORF region (including D8idp) and the effect of the MITE element located in the promoter region because these polymorphisms were found to be linked both by Andersen et al. (2005) and by Thornsberry et al. (2001). Finally, showing that maize lines transformed with the different alleles of Dwarf8 exhibit altered flowering time would be the conclusive experimental evidence for the functional role of the Dwarf8 gene.

The high frequency of the D8-deletion allele is shared by Northern Flint and Andean materials, which otherwise appear as very distant. It is worth noting that Andean material is very particular. It has been reported by Gouesnard et al. (2002) that, although very late on average (see Table 5), Andean materials are on average less sensitive to photoperiod than other tropical materials. Archaeological studies (Freitas et al. 2003) revealed that South American maize diversity is subdivided into two main components, highland maize from core Andes and lowland maize, and that these two maize types originate from two different introductions from Mesoamerica. The introduction of maize in core Andes dates back to 4500 before the present (BP), whereas lowland material was introduced after 2000 BP (Freitas et al. 2003). A relevant question would be to know whether the D8-deletion allele was at a high frequency in ancient material when introduced in core Andes, and thus has been inherited from a Mesoamerican ancestral material that no longer exists as suggested by Jaenicke-Desprès et al. (2003), or if the D8-deletion increased in frequency as a result of drift or adaptation to high altitudes from lowland material.

Finally, although Dwarf8 polymorphism may be responsible for more than the 29-dd variation revealed by association genetics corrected for population structure, it explains only part of the extreme difference (387 dd) between early Northern Flint and late tropical inbreds. Numerous independent QTL potentially involved in flowering time variation in maize have been reported and synthesized through meta-analysis into 62 consensus QTL widely distributed on all 10 chromosomes (Chardon et al. 2004). Large efforts are presently underway in several groups to identify candidate genes and/or to clone these QTL (Salvi et al. 2002). The investigation of the association between their molecular polymorphism and flowering time, as well as that of their differentiation between contrasted genetic groups, should prove extremely useful to better understand the genetic control of flowering time and climatic adaptation in maize.

Acknowledgments

We are grateful to M. Dupin, J. Laborde, and colleagues at Saint Martin de Hinx for managing the seed production of the inbred line collection and for their contribution to field experiments. We are grateful to M. Warburton, from International Maize and Wheat Improvement Center, and C. Rebourg for their contribution to the data on maize landraces. We are grateful to P. Bertin, Ph. Jamin, and D. Coubriche for field experiments at Institut National de la Recherche Agronomique (INRA) le Moulon and to J. Félicité and colleagues at INRA Guadeloupe for field trials under short-day conditions. We are grateful to M. M. Goodman, E. Buckler, and T. Rocheford for kindly providing seeds from the Evolutionary Genomics of Maize project. We are grateful to reviewers for helpful comments and to Heather McKhann for careful editing of American English. Work on inbred lines was funded by INRA and Genoplante. Work on populations was funded by Promaïs members (Caussade Semences, Euralis Génétique, Limagrain Verneuil Holding, Maïsadour Semences, Monsanto France SAS, Syngenta Seeds SAS, Pioneer Génétique, and SDME/KWS France) and by a grant from the French Bureau des Ressources Génétiques. L. Camus-Kulandaivelu is funded by a grant from INRA and the Languedoc-Roussillon region.

References

  • Andersen, J. R., T. Schrag, A. E. Melchinger, I. Zein and T. Lübberstedt, 2005. Validation of Dwarf8 polymorphisms associated with flowering time in elite European inbred lines of maize (Zea mays L.). Theor. Appl. Genet. 111: 206–217. [PubMed]
  • Beaumont, M. A., and R. A. Nichols, 1996. Evaluating loci for use in the genetic analysis of population structure. Proc. R. Soc. Lond. Ser. B 263: 1619–1626.
  • Bonhomme, R., M. Derieux and G. O. Edmeades, 1994. Flowering of diverse maize cultivars in relation to temperature and photoperiod in multilocation field trials. Crop Sci. 34: 156–164.
  • Brown, W. L., and E. Anderson, 1947. The northern flint corn. Ann. MO Bot. Gard. 34: 1–28.
  • Chardon, F., B. Virlon, L. Moreau, M. Falque, J. Joets et al., 2004. Genetic architecture of flowering time in maize as inferred from quantitative trait loci meta-analysis and synteny conservation with the rice genome. Genetics 168: 2169–2185. [PMC free article] [PubMed]
  • Doebley, J., M. Goodman and C. W. Stuber, 1986. Exceptional genetic divergence of Northern Flint corn. Am. J. Bot. 73: 64–69.
  • Doebley, J., J. D. Wendel, J. S. Smith, C. W. Stuber and M. M. Goodman, 1988. The origin of cornbelt maize: the isozyme evidence. Econ. Bot. 42: 120–131.
  • Dubreuil, P., M. Warburton, M. Chastanet, D. Hoisington and A. Charcosset, 2006. The origin of maize (Zea mays L.) in Europe as evidenced by microsatellite diversity. Maydica (in press).
  • Duvick, D. N., J. S. C. Smith and M. Cooper, 2004. Long term selection in a commercial hybrid maize breeding program. Plant Breed. Rev. 24: 109–151.
  • Excoffier, L., and G. Hamilton, 2003. Comment on “genetic structure of human populations”. Science 300: 1877. [PubMed]
  • Falush, D., M. Stephens and J. K. Pritchard, 2003. Inference of population structure using multilocus genotype data: linked loci and correlated allele frequencies. Genetics 164: 1567–1587. [PMC free article] [PubMed]
  • Freitas, F. O., G. Bendel, R. G. Allaby and T. A. Brown, 2003. DNA from primitive maize landraces and archaeological remains: implications for the domestication of maize and its expansion into South America. J. Archaeol. Sci. 30: 901–908.
  • Gallais, A., 1990. Théorie de la Sélection en Amélioration des Plantes. Masson, Paris.
  • Gardiner, J. M., E. H. Coe, S. Melia-Hancock, D. A. Hoisington and S. Chao, 1993. Development of a core RFLP map using an immortalized F2 population. Genetics 134: 917–930. [PMC free article] [PubMed]
  • Gerdes, J. T., and W. F. Tracy, 1993. Pedigree diversity within the Lancaster Surecrop heterotic group of maize. Crop Sci. 33: 334–337.
  • Goudet, J., 2001. FSTAT, a program to estimate and test gene diversities and fixation indices (version 2.9.3) (http://www.unil.ch/izea/softwares/fstat.html).
  • Gouesnard, B., C. Rebourg, C. Welcker and A. Charcosset, 2002. Analysis of photoperiod sensitivity within a collection of tropical maize populations. Genet. Res. Crop Evol. 49: 471–481.
  • Haug, G. H., D. Gunther, L. Peterson, D. M. Sigman, K. A. Hughen et al., 2003. Climate and the collapse of Maya civilization. Science 299: 1731–1735. [PubMed]
  • Jaenicke-Desprès, V., E. S. Buckler, B. D. Smith, M. T. P. Gilbert, A. Cooper et al., 2003. Early allelic selection in maize as revealed by ancient DNA. Science 302: 1206–1208. [PubMed]
  • Liu, K. J., M. Goodman, S. Muse, J. S. Smith, E. Buckler et al., 2003. Genetic structure and diversity among maize inbred lines as inferred from DNA microsatellites. Genetics 165: 2117–2128. [PMC free article] [PubMed]
  • Lu, H., and R. Bernardo, 2001. Molecular marker diversity among current and historical maize inbreds. Theor. Appl. Genet. 103: 613–617.
  • Matsuoka, Y., S. E. Mitchell, S. Kresovich, M. Goodman and J. Doebley, 2002. a Microsatellites in Zea—variability, patterns of mutations, and use for evolutionary studies. Theor. Appl. Genet. 104: 436–450. [PubMed]
  • Matsuoka, Y., Y. Vigouroux, M. M. Goodman, J. G. Sanchez, E. S. Buckler et al., 2002. b A single domestication for maize shown by multilocus microsatellite genotyping. Proc. Natl. Acad. Sci. USA 99: 6080–6084. [PMC free article] [PubMed]
  • Nei, M., 1978. Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 89: 583–590. [PMC free article] [PubMed]
  • Peng, J. R., D. E. Richards, N. M. Hartley, G. P. Murphy, K. M. Devos et al., 1999. ‘Green revolution’ genes encode mutant gibberellin response modulators. Nature 400: 256–261. [PubMed]
  • Petit, R. J., A. El Mousadik and O. Pons, 1998. Identifying populations for conservation on the basis of genetic markers. Conserv. Biol. 12: 844–855.
  • Pritchard, J. K., M. Stephens and P. Donnelly, 2000. a Inference of population structure using multilocus genotype data. Genetics 155: 945–959. [PMC free article] [PubMed]
  • Pritchard, J. K., M. Stephens, N. A. Rosenberg and P. Donnelly, 2000. b Association mapping in structured populations. Am. J. Hum. Genet. 67: 170–181. [PMC free article] [PubMed]
  • Rebourg, C., 2000. Diversité génétique de populations européennes et américaines de maïs: analyse moléculaire et morphologique. Ph.D. Thesis, Université Paris XI, Orsay, France.
  • Rebourg, C., B. Gouesnard and A. Charcosset, 2001. Large scale molecular analysis of traditional European maize populations. Relationships with morphological variation. Heredity 86: 574–587. [PubMed]
  • Rebourg, C., M. Chastanet, B. Gouesnard, C. Welcker, P. Dubreuil et al., 2003. Maize introduction into Europe: the history reviewed in the light of molecular data. Theor. Appl. Genet. 106: 895–903. [PubMed]
  • Reif, J. C., A. E. Melchinger, X. C. Xia, M. L. Warburton, D. A. Hoisington et al., 2003. Use of SSRs for establishing heterotic groups in subtropical maize. Theor. Appl. Genet. 107: 947–957. [PubMed]
  • Remington, D. L., J. M. Thornsberry, Y. Matsuoka, L. M. Wilson, S. R. Whitt et al., 2001. Structure of linkage disequilibrium and phenotypic associations in the maize genome. Proc. Natl. Acad. Sci. USA 98: 11479–11484. [PMC free article] [PubMed]
  • Ritchie, J. T., and D. S. Nesmith, 1991. Temperature and crop development, pp. 5–29 in Modeling Plant and Soil Systems, edited by J. Hanks and J. T. Ritchie. ASA, CSSA and SSSA, Madison, WI.
  • Rosenberg, N. A., 2002. Distruct: A Program for the Graphical Display of Structure Results, pp. 1–11 (http://rosenberglab.bioinformatics.med.umich.edu/distruct.html).
  • SAS, 1989. SAS/STAT User's Guide. SAS Institute, Cary, NC.
  • Semon, M., R. Nielsen, M. P. Jones and S. R. McCouch, 2005. The population structure of African cultivated rice Oryza glaberrima (Steud.): evidence for elevated levels of linkage disequilibrium caused by admixture with O. sativa and ecological adaptation. Genetics 169: 1639–1647. [PMC free article] [PubMed]
  • Salvi, S., R. Tuberosa, E. Chiapparino, M. Maccaferri, S. Veillet et al., 2002. Toward positional cloning of Vgt1, a QTL controlling the transition from the vegetative to the reproductive phase in maize. Plant Mol. Biol. 48: 601–613. [PubMed]
  • Senior, M. L., E. C. L. Chin, M. Lee, J. S. C. Smith and C. W. Stuber, 1996. Simple sequence repeat markers developed from maize sequences found in the GENBANK database: map construction. Crop Sci. 36: 1676–1683.
  • Tai, T., and S. Tanksley, 1991. A rapid and inexpensive method for isolation of total DNA from dehydrated plant tissue. Plant Mol. Biol. 8: 297–303.
  • Taramino, G., and S. Tingey, 1996. Simple sequence repeats for germplasm analysis and mapping in maize. Genome 39: 277–287. [PubMed]
  • Thornsberry, J. M., M. M. Goodman, J. Doebley, S. Kresovich, D. Nielsen et al., 2001. Dwarf8 polymorphisms associate with variation in flowering time. Nat. Genet. 28: 286–289. [PubMed]
  • Vigouroux, Y., J. S. Jaqueth, Y. Matsuoka, O. S. Smith, W. F. Beavis et al., 2002. Rate and pattern of mutation at microsatellite loci in maize. Mol. Biol. Evol. 19: 1251–1260. [PubMed]
  • Vitalis, R., K. Dawson and P. Boursot, 2001. Interpretation of variation across marker loci as evidence of selection. Genetics 158: 1811–1823. [PMC free article] [PubMed]
  • Wright, S., 1951. The genetical structure of population. Ann. Eugen. 15: 323–354. [PubMed]

Articles from Genetics are provided here courtesy of Genetics Society of America
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...