| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Genetics, Vol. 169, 1659-1672, March 2005, Copyright © 2005
doi:10.1534/genetics.104.035543
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,2
* University of Missouri, Department of Biology, Saint Louis, Missouri 63121
John Innes Centre, Colney, Norwich NR4 7UH, United Kingdom
1 Corresponding author: Department of Biology, 1 University Blvd., University of Missouri, St. Louis, MO 63121.
E-mail: adoust{at}umsl.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT MATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
10,000 species, and variation in inflorescence architecture is important in determining yield of cereal grains. Even among closely related species the structure of the inflorescence may vary in the number of branches, the number of orders of branching, and the extent of elongation of various axes (KELLOGG 2000; DOUST and KELLOGG 2002). To understand the evolution of these phenotypes, we need to characterize them carefully and identify the genes that underlie them. The most complete source of information on genes controlling inflorescence development in grasses comes from studies on the model system, maize. Many genes have been found to affect inflorescence development in maize and these are all possible candidates for the control of inflorescence morphology in other grass species. However, several aspects of the morphology of maize are highly unusual within the grasses, and thus its developmental evolution may not be representative of other grass species. In contrast, other cereal domestication events have involved a greater number of inflorescence characteristics that vary among related wild species. We were interested to investigate morphological changes in a wild-domesticate pair in which the domesticate is morphologically similar to multiple wild species. Study of domestication in such a pair of species might shed light not only on the process of domestication, but also on morphological diversification in general.
We have examined the domesticated crop foxtail millet (Setaria italica) and its close relative and presumed wild progenitor, green millet (S. viridis), which are grasses in the subfamily Panicoideae, tribe Paniceae. They are in the same subfamily as maize, sugar cane, and sorghum, but in a different tribe, having diverged from the maize lineage roughly 28 million years ago (GAUT and DOEBLEY 1997). Foxtail millet is an important crop in China, and green millet is one of the world's worst weeds of arable land (HARLAN 1992). Domestication of foxtail millet led to the production of much larger inflorescences with a more complex branching pattern, as well as a reduction in basal (tillering) and axillary vegetative branching (HARLAN 1992; DOUST and KELLOGG 2002; DOUST et al. 2004).
Some of the same changes during domestication are also seen in the domestication of maize from its wild progenitor teosinte, and quantitative trait locus (QTL) studies of crosses between maize and teosinte have revealed five major regions of the genome controlling differences between the two species (DOEBLEY and STEC 1991, 1993). Later analyses of mutants and gene expression identified teosinte branched1 as causing differences between teosinte and maize (DOEBLEY et al. 1997), with noncoding regions upstream of the gene being under selection (CLARK et al. 2004). Increased expression of tb1 in maize relative to its wild progenitor influences inflorescence architecture and sex expression and represses lateral vegetative branch elongation (HUBBARD et al. 2002). However, in a previous study, we found that tb1 plays only a minor role in controlling vegetative branching in foxtail millet (DOUST et al. 2004).
We have previously characterized the development of both foxtail and green millet (DOUST and KELLOGG 2002). Both species produce sterile inflorescence branches, commonly called bristles. These generally terminate the various inflorescence axes and also appear to be approximately paired with spikelets (Figures 1–3; DOUST and KELLOGG 2002).
|
|
|
When first initiated, spikelets and bristles appear to be paired (Figure 3) and, if all spikelets grew to maturity, the numbers of spikelets and bristles would be approximately equal. However, a number of spikelets cease development and persist as tiny rudiments (Figure 1G). This failure of development may happen to spikelets at various stages and, in some cases, cessation of growth of the spikelets is so early that it is difficult to recognize the structures as spikelets. The result is that there are often many more bristles than spikelets, especially in green millet.
To determine the number—and ultimately the identity—of genes controlling the developmental characters, we have used QTL analysis of differences in inflorescence structure between foxtail and green millet. We have then exploited the colinearity of grass genomes to identify a likely subset of candidate genes for these differences (GALE and DEVOS 1998; HAKE and ROCHEFORD 2004). To localize genes of interest we have used common markers among foxtail millet, maize, and rice genomes. Cross-species comparisons are feasible because the millet genome is broadly colinear with rice and maize (DEVOS et al. 1998; WANG et al. 1998).
Our previous work has identified several QTL of major effect that control vegetative branching (DOUST et al. 2004); for each of these QTL we have hypotheses of candidate genes. Here we describe QTL that control inflorescence branching, compare their location to that of vegetative QTL, and suggest possible candidate genes.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
10-cM intervals. F3 offspring selfed from 120 of the original 127 F2 plants were used to evaluate the number and location of QTL controlling the morphological characters distinguishing the two parents. Additional RFLP markers from maize were added to the original map using 23 probes from University of Missouri-Columbia, a maize tb1 cDNA clone kindly provided by J. Doebley, and probes amplified from foxtail millet using primers for terminal ear1 (te1; WHITE and DOEBLEY 1999), phytochrome B (MATHEWS et al. 2000), knotted1 (kn1; unpublished primers provided by Anthony Verboom, University of Missouri-St. Louis), barren stalk1 (ba1; unpublished primers provided by Andrea Gallavotti, University of California at San Diego), and branched silkless1 (bd1; CHUCK et al. 2002). Southern hybridizations were performed against restricted F2 DNA of the original S. italica acc. B100 x S. viridis acc. A10 population (DEVOS et al. 1998; WANG et al. 1998). Marker data were scored by two people separately and then results were cross-checked. Position of these markers on the genetic map was established using the two- and three-point linkage routines in Mapmaker version 3 (LANDER et al. 1987).
QTL trials:
Multiple replicates of the 120 F2:3 families were grown in four separate trials in a climate-controlled greenhouse at UM-St. Louis. Trials 1 and 2 were at a high density of 5 plants of each family per pot while trials 3 and 4 had 15 plants per family at a lower density of 1 plant per pot. The position of the pots was randomized to minimize the effect of differences in light intensity and other environmental variables within the greenhouse. Trials 1 and 3 were grown in May–June (early summer) of 2000 and 2001, respectively, and trials 2 and 4 were grown in July–August (late summer) of 2000 and 2002, respectively. Soil, fertilizer, and water conditions were standardized, and a 16-hr day length was maintained by artificial lighting when necessary. Trials 2 and 4 had both a higher natural light intensity and higher average temperatures than trials 1 and 3, owing to the later growing time, although temperatures in both trials were kept between 25° and 35°.
Measurement of phenotypic traits:
Plants were harvested after the seeds had ripened, and measurements were taken of the inflorescence that terminated the primary culm. All plants in trials 1 and 2 and five randomly chosen plants for each family in trials 3 and 4 were measured. Primary branch density was estimated by counting the number of primary inflorescence branches in a 1-cm region in the middle of the inflorescence axis. Multiplication of this number by the length of the inflorescence (measured from the inflorescence tip to the most basal branch that produces fertile spikelets) gives an estimate of the number of primary branches per inflorescence. We verified the reliability of this approach by establishing a significant correlation (P < 0.01; R2 = 0.34) between the estimated number of primary branches and the actual number; the actual number was determined by counting all primary branches for 184 randomly selected inflorescences of various sizes. The numbers of spikelets and bristles were measured for one primary branch from each inflorescence (taken from the middle of the inflorescence). Because approximately one bristle and one spikelet per order of branching are produced (Figure 3; DOUST and KELLOGG 2002), the number of bristles can be used as a surrogate for the number of orders of branching of the primary branch (Figures 1D, 2D, and 3).
Data management:
Means were calculated for each trait for each of the families in each trial. To construct mean high- and low-density trait values, means per family were calculated for combined high-density (trials 1 and 2) and combined low-density (3 and 4) trials. The means of both the individual and the two density trial averages were tested for normality (Kolmogorov-Smirnov test, P < 0.05; SPSS 2001) and transformed where necessary.
QTL detection:
We investigated QTL (1) from each trial individually (referred to here as individual QTL), (2) with the trials combined by density (high-density QTL vs. low-density QTL), and (3) with all four individual trials analyzed together (joint QTL). Most results presented here reflect analyses of the high- vs. low-density trials, but other analyses are presented where appropriate. QTL were detected using composite interval mapping (CIM), as implemented in QTLCartographer (BASTEN et al. 2002). CIM tests the hypothesis that a QTL is present in an interval between two adjacent markers, while at the same time controlling for the effects of segregating QTL elsewhere in the genome (RIESEBERG et al. 2003a). Background markers were selected at P = 0.05, and five background parameters were included as cofactors in each CIM model. Tests were made at 2.0-cM intervals and a window size of 10 cM was used to exclude flanking markers from the search for new QTL. Data were also analyzed by multiple interval mapping, using the recommended information criterion 1 with a threshold of 0.0 for acceptance of new QTL (BASTEN et al. 2002). This gave qualitatively similar results, so only the CIM results are presented here. Individual and high- and low-density averages were analyzed for CIM using the program module Zmapqtl (BASTEN et al. 2002). Where multiple QTL were detected in close proximity, they were accepted only when the LOD score dip between the QTL peaks was >1 LOD interval (BASTEN et al. 2002). QTL in high- and low-density trials were considered to be identical if their 1-LOD support intervals overlapped and the sign of their additive effects was the same. In addition, a joint analysis of each trait for all four trials taken together was analyzed using the module JZmapqtl (BASTEN et al. 2002). The joint analysis allows an estimate of the amount of genetic by environmental (G x E) interaction between the trial values for each trait and provides a measure of both the main and the interaction effects of detected QTL.
Significance thresholds for QTL for individual and joint CIM analyses were calculated by 1000 permutations of the original data, using the same parameter settings as for the original analysis (CHURCHILL and DOERGE 1994; DOERGE and CHURCHILL 1996). To strike a balance between detection of true QTL and a high rate of acceptance of false positive results, chromosome-wide significance levels at P < 0.05 were used to declare the presence of QTL. The chromosomes differed in the number and density of markers, and thus the critical LOD score values for P < 0.05 differed among them. Permutation tests gave critical LOD score values per chromosome similar to those of a Bonferroni correction based on the total number of independent comparisons (after correcting for linkage between markers on a particular chromosome; CHEVERUD 2001). A more stringent genome-wide significance level of P < 0.05 was also calculated. The identification of QTL based on multiple chromosome-wide significance levels will increase type I error compared to the genome-wide level because nine different tests are being performed (one for each chromosome), but will also increase the probability of identifying more true QTL (CHEVERUD 2001).
In this study, we focused primarily on QTL for a particular phenotypic trait that were found in both high- and low-density trials. This is because our interest is in QTL that may represent genes explaining species differences, which we expect to produce the same phenotype under multiple environmental conditions. By this approach we are ignoring QTL and, by extension, genes that are important under certain environmental conditions (as represented by the different densities and growing times).
It was not possible to directly measure transgressive segregation (the extent to which the range of values of the hybrid populations exceeded that of the two parents) in the high- and low-density trials, because the parents of the cross were grown in only trials 2 and 4. However, the parental and hybrid ranges were examined for evidence of transgressive segregation in those individual trials.
Additive and dominance effects:
The additive effect of each QTL changes the phenotypic value away from the overall mean of the parental values (the midpoint of the distance between the mean for foxtail millet and that for green millet) and toward the mean of one parent or the other. Dominance effects are detected by examining the phenotypes of individuals heterozygous for markers at a particular QTL. Heterozygotes are expected to have phenotypic values corresponding to the mean of the parental values; deviation from the mean indicates a dominant effect at or near that marker. The sign of the effect indicates the direction of change in the phenotype; in this study, a positive sign indicates an increase and a negative sign a decrease in the trait value.
Additive effects are presented for each of the replicated QTL in the high- and low-density trials. It was not possible to measure the extent to which the additive effects for the replicated QTL explained the parental differences, because parents were grown in only trials 2 and 4. The range of values of the F2:3 population means in the high- and low-density trials was compared with the sum of the absolute values of the additive effects (
|A|) of the replicated QTL. If combinations of all positive or all negative additive effects are sufficient to explain the observed phenotypic extremes, then the range of values of the hybrids should be equal to twice |A|. Transgressive segregation for a trait can be inferred when the range of the hybrid values exceeds twice |A|.
Epistasis:
The program Epistacy (HOLLAND 1998) was used to identify digenic epistatic interactions by examining each pair of markers for each trait and testing for a significant interaction term. Because of the large number of comparisons that must be made for each trait, a Bonferroni correction to the P < 0.05 significance level was applied so that overall significant interactions were reported at P < 1.0 x 10–5 (RIESEBERG et al. 2003a).
Comparative mapping:
To identify possible candidate genes from maize, we used markers mapped on both maize and foxtail millet to define intervals on the maize map that correspond to QTL regions in foxtail millet. Where regions between common markers were larger than the 1-LOD confidence intervals of the QTL, the entire region on the maize map was searched for candidate genes. MaizeGDB (LAWRENCE et al. 2004) and GRAMENE (WARE et al. 2002a,b) were used to identify genes that had mutant phenotypes in which inflorescence branching was affected. Several of these genes, including ba1, bd1, zea floricaula leafy1 (zfl1), and luminidependens (a marker closely linked to tassel seed4, ts4) were screened for RFLP polymorphisms between the parents and, in the case of ba1 and bd1, were mapped.
Identification of candidate genes from the rice genome:
Common markers mapped on both the foxtail millet and rice genomes (DEVOS et al. 1998) were used to delimit regions that contained the identified QTL. These regions were in most cases covered by a number of unassembled contigs. Each of these contigs was scanned using FgeneSH (SALAMOV and SOLOVYEV 2000), and open reading frames (ORFs) were translated and BLASTed against ORFs from other contigs from the same QTL region to reduce redundancies (ALTSCHUL et al. 1997). The final data set of translated proteins for each region was BLASTed against the NCBI database (ALTSCHUL et al. 1997). Hits with eV < 10–7 were evaluated, and possible candidate genes were identified. Where annotated chromosomes for the rice genome were available, these were searched directly for candidate genes, using the intervals defined by the common markers between foxtail millet and rice.
A dual approach was also used, identifying the location of orthologs of developmentally important maize genes on the rice genome and then using common markers between the rice and millet genomes to identify whether the presumed orthologous regions on the millet genome would contain the orthologs. This enabled greater precision in placement of genes relative to QTL because there is greater coverage of the millet genome with rice markers than with maize markers.
|
RESULTS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
|
|
|
QTL detected:
The CIM approach, using a chromosome significance level of P < 0.05, found 33 QTL in the high-density and 38 QTL in the low-density trials. Of these, 5 in the high density and 12 in the low density were also significant at the more stringent genome level of P < 0.05. The 71 QTL found at the chromosome significance level of P < 0.05 were compared between high- and low-density trials to identify those that were overlapping and could represent the same gene or genes. We found 14 replicated QTL, 3 each for primary branch number and primary branch density, 6 for spikelet number, and 2 for bristle number (Table 1). Because we were looking for QTL that accounted for differences between the two species, rather than differences between environments, we focused on these replicated QTL and did not consider further the QTL found in only one trial. The QTL that were replicated in position and direction of additive effect at high and low densities can be interpreted as identifying genomic regions responsible for differences between the parents independent of environment. Some of the QTL for different traits overlap, suggesting that individual QTL may have effects on more than one trait (Table 1; Figure 7).
|
|
|
The hybrid range was compared to double the sum of the absolute values of the additive effects (|A|) (the sum of the absolute values of the additive effects is doubled because the range spans both sides of the overall mean). |A| explains only between 38 and 81% of the hybrid range in the high-density trial and between 50 and 88% in the low-density trial (Table 1). This suggests that, as in trials 2 and 4, significant transgressive segregation occurs in the high- and low-density trials.
Dominance effects were generally much less than additive effects, and the differences observed are in several cases below the power of this experimental design to unambiguously detect (LANDER and BOTSTEIN 1989). Larger dominance effects were detected for several of the spikelet QTL at both planting densities. The sign of dominance effects was not always consistent between high- and low-density trials.
Epistasis:
An analysis of digenic epistasis revealed 17 significant epistatic interactions between markers, occurring exclusively in the high-density trials (Table 3). In 9 cases one of the markers in the interaction was associated with a QTL; in 2 of these the QTL were found in both the high- and low-density trials, and in the other 7 cases the QTL were found in only one or the other of the density trials. In the remaining 7 cases neither marker in the interaction was associated with a detected QTL. In no case was there a significant interaction where both markers were associated with QTL. It is noteworthy that the interactions were identified in only the high-density trial, where additive effects explained less phenotypic variation than in the low-density trial.
|
We hybridized clones of two potential candidate genes for inflorescence architecture that were polymorphic between the two parents. A clone of bd1 was placed at the base of linkage group II, between markers rgr1789 and psf360, and a clone of ba1 was placed on the long arm of linkage group V, between markers psm768 and rgc385, both as predicted by synteny with the maize and rice genomes (Figure 7). The clone of luminidependens (closely linked to ts4 in maize) did not identify polymorphisms between the parents and so could not be mapped. The same was true for zfl1, but a closely linked marker identified through MAIZEGDB, umc44, was mapped to the long arm of linkage group VII. The position of this marker in rice was checked via Gramene, where it was found to be closely linked to the rice ortholog of zfl1.
The position of a number of other candidate genes for inflorescence variation was inferred using synteny between the millet, maize, and rice genomes. Of these, good matches between QTL and potential gene position were obtained for barren inflorescence2 (bif2) and ts4 (Figure 5).
We also examined regions of the sequenced rice genome corresponding to the QTL regions on foxtail millet, using common markers mapped on both the rice and the foxtail millet genomes to delimit appropriate rice genomic regions. We identified several hormone biosynthesis pathway genes and many transcription factors in the putatively colinear regions on rice. The rice genome was also used to map the position of candidate genes from maize more precisely onto the millet genome.
|
DISCUSSION |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
There was a significant effect due to planting density for all traits, with larger mean trait values at low density. Together with the QTL results, this suggests that, as expected, variation between foxtail and green millet is a combination of fixed species differences and variable environmental responses.
Declared QTL:
This study identifies 14 QTL that control the major portion of variation in the four measured phenotypic traits. In several cases, replicated QTL for separate traits overlap, raising the possibility that they actually indicate a single locus with pleiotropic effects. Taking all instances where this could be the case reduces the set of QTL differentiating the species to 11. Thus, it is possible that only a small number of QTL may underlie the differences in inflorescence architecture between foxtail and green millet.
The number of QTL controlling differences between the species is expected to be an underestimate of the number of genomic regions responsible for the control of all variation in inflorescence traits. This is because only QTL that were replicated between high- and low-density trials were considered, thus excluding the 43 QTL that were unique to one or the other trial. These unique QTL were excluded because they indicate gene effects that occur only in certain environmental situations, but their presence suggests that genotype by environment interactions are important in determining the absolute values of each trait. The results from the joint analysis of individual trials support this conclusion as two of the six joint QTL detected had significant G x E interactions and were in positions that differed from the replicated high- and low-density QTL.
Spikelet and bristle number show a mixture of overlapping and separate QTL. Some overlapping QTL were expected because both spikelet and bristle number reflect the number of orders of branching within each primary branch, and developmental analyses indicate that spikelets and bristles initiate in approximately equal numbers. However, a proportion of the spikelets fail to develop to maturity, with green millet having more undeveloped spikelets than foxtail millet. Cessation of spikelet growth occurs later in development than initiation of branching and differentiation of primordia into spikelets or bristles. This suggests that QTL for spikelet number that overlap with QTL for bristle number represent regions for genetic control of branching within each primary branch, whereas nonoverlapping QTL for spikelet number may control spikelet development or suppression.
It is possible that not all QTL affecting these traits have been detected, due to the relatively small population of F2 plants on which the study is based (BEAVIS 1998). The effect of the small population size may also be responsible for the relatively broad 1-LOD intervals for several of the detected QTL. The small population size potentially decreases the power to detect small but significant changes in dominant effects (LANDER and BOTSTEIN 1989). However, the presence of such effects in replicated trials may support our inference of small but real dominance effects for these QTL.
Additive effects:
QTL for primary branch, spikelet, and bristle number in the high- and low-density trials had total additive effects consistent with higher values in foxtail millet, whereas primary branch density had additive effects consistent with lower values in foxtail relative to green millet. We did not expect that the declared additive effects should explain all of the observed variation in each of the trials, as QTL that occur in only a single trial were excluded. However, the large dominance effects observed for some of the identified QTL, particularly in the high-density trials, suggest that additive effects are unlikely to explain all of the variation observed.
Transgressive segregation was observed for all traits in the trials in which the parents were grown. Transgressive segregation is commonly found in QTL studies and has been interpreted as the result of combination of favorable alleles in the hybrids from both of the parents (RIESEBERG et al. 2003b). Transgressive segregation was also indicated in the high- and low-density trials, and in many cases the hybrid range exceeded that obtainable by combination of favorable QTL alleles. The increase in hybrid extreme values may be due to the exclusion of QTL that were not replicated between the high- and low-density trials or to the epistatic and dominance effects that were detected.
Epistasis:
All of the significant digenic epistatic effects between markers occurred at high density, where additive effects explained less of the phenotypic variation than at low planting density. It was striking that in no case were both markers associated with QTL and that in just under half of the cases neither marker was associated with a QTL. This suggests that epistatic control can be exerted by genes that do not by themselves have a detectable phenotypic effect, indicating that a more complex model of interacting gene networks must exist to mediate these responses (LARK et al. 1995).
Inflorescence vs. vegetative QTL:
In a previous article we discussed the position and possible candidate genes underlying QTL for vegetative branching in the same F2:3 cross (DOUST et al. 2004). That article discussed QTL found in the two low-density trials and used a chromosome-wide significance level for declaring QTL of P < 0.01 (a more stringent cutoff criterion than the P < 0.05 level used in this article). Three QTL appear to be in similar positions in both the vegetative and the inflorescence analysis. Two QTL for vegetative branching are in positions similar to replicated QTL for inflorescence traits; one for tillering at the base of chromosome III matches the broad position of a replicated QTL for spikelet number, and one for axillary branching at the base of chromosome IX matches inflorescence QTL for primary branch density and spikelet number. Another QTL for axillary branching near the centromere on chromosome V is in the same general position as a QTL identified from the joint analysis. It is possible that these overlapping QTL represent the same genes, and that therefore the same mechanisms govern vegetative and inflorescence branching, but a rigorous test of this hypothesis awaits fine mapping studies.
Candidate genes:
For genes to be considered candidate genes in this study they had to have a mutant phenotype in a model organism that was similar to some aspect of the phenotypic differences between foxtail and green millet, and they had to be localized to a region of the genome (QTL) implicated in control of that phenotype. An example of such a gene is zea floricaula leafy1 (zfl1), a gene that is duplicated in maize and affects variation in long branch number in the tassel inflorescence (BOMBLIES et al. 2003). In addition, zfl1 has been linked to a QTL in maize that affects inflorescence phyllotaxy and may act to increase the number of primary branches that are initiated, presumably by increasing the size of the apical meristem (BOMBLIES et al. 2003). It also affects the transition to flowering and, in rice, may affect panicle branching (KYOZUKA et al. 1998). It was not possible to map zfl1 directly, because of the lack of polymorphisms between the parents. However, in maize and rice it is closely linked to the RFLP probe umc44, which was hybridized to the mapping filters and localized to foxtail millet chromosome VII. This places the gene in close proximity to QTL for primary branch number and density on millet chromosome VII. This gene appears to be a good candidate for control of primary branch number, as it confers a similar phenotypic effect in maize and in foxtail millet and apparently maps to two of the replicated QTL detected in foxtail millet .
Tasselseed4 is another candidate gene for which the mutant phenotype of increased inflorescence branching (IRISH 1997) is similar to phenotypic variation in foxtail and green millet and where the inferred position of the gene matches an identified QTL. It was not possible to map ts4 to the millet genome directly because the gene has not yet been cloned, and a closely linked marker, luminidependens, was monomorphic between the parents of the cross. However, on the basis of comparative mapping, using the common markers umc60 and umc321, ts4 is near a joint QTL for bristle number near the centromere on chromosome V. In this case, the joint QTL was not matched by replicated high- and low-density QTL, presumably because of a significant genotype by environment interaction. Ts4 affects determinacy of the spikelet pair meristem in maize, leading to proliferation of branches on which spikelet production is greatly delayed. We interpret the spikelet pair meristem as simply another order of branching and thus regard the ts4 mutant as having increased numbers of orders of branching compared to wild type. Bristle number in foxtail millet is a measure of the number of orders of branching; thus the QTL effect and the phenotype of ts4 match. More bristles (and thus more orders of branching) are in foxtail than in green millet.
Teosinte branched1 occurs in QTL for several traits in the comparison between maize and its wild relative, teosinte (DOEBLEY and STEC 1991; DOEBLEY et al. 1995; LUKENS and DOEBLEY 1999), and its promoter region has been shown to have been under selection (WANG et al. 1999; CLARK et al. 2004). It is thus a good candidate gene for involvement in domestication of foxtail millet. Tb1 was found to be a relatively minor player in the control of vegetative branching in foxtail millet (DOUST et al. 2004), but on the basis of map position remains a candidate for control of inflorescence branching. However, the mutant phenotype of tb1 in maize bears little resemblance to anything in foxtail or green millet. In maize, tb1 affects sex expression in the male inflorescence (tassel), but is not involved in inflorescence branching or elongation or in spikelet differentiation or suppression (HUBBARD et al. 2002). The two species of millet investigated here are, however, identical in sex expression, but differ in aspects of branching. Thus the position of tb1 suggests it could underlie the QTL, but its mutant phenotype in maize does not support this hypothesis.
Other possible candidate genes from maize that may be involved in inflorescence variation include branched silkless1 (bd1) and barren stalk1 (ba1). Bd1 and ba1 were mapped directly onto the millet genome; the mutant phenotype of bd1 in maize causes indeterminate branches to form instead of spikelets (CHUCK et al. 2002), while that of ba1 causes a reduction or elimination of branches and spikelets (NEUFFER et al. 1997). Both genes thus affect branching and therefore are candidates for variation in branching in the millets. However, neither gene maps to one of the replicated or joint QTL. Bd1 maps to a position on millet chromosome II that does not have QTL for any of the measured traits, while ba1 maps to a position on chromosome V that is between the 1-LOD intervals for spikelet and bristle number. The lack of close correspondence between bd1 or ba1 and the replicated QTL suggests that, although these genes have mutant phenotypes that mimic the observed phenotypic variation, these genes are not likely candidates for differences in inflorescence branching between the parents.
Several replicated QTL are in regions where no obvious candidate gene from maize is apparent. These include QTL on chromosomes III and IV and the bottom of V. The failure to find candidate genes may be due to the inadequacy of the syntenic comparisons between millet and maize in those regions or may be due to the presence of genes underlying these QTL that are not characterized in maize floral architecture or are not present in the orthologous regions of the maize genome.
To attempt to identify other possible genes we examined sections of the rice genome that were presumed orthologous to regions of the millet genome covered by QTL. Many transcription factors were identified, but, as the phenotypic effect of these genes is unknown, they cannot as yet provide good candidate genes for this study. Rice genomic regions colinear to QTL regions on chromosomes V, VI, and VII also include a variety of auxin and gibberellin pathway mediators, including semidwarf1, the so-called "green revolution" gene in rice (MONNA et al. 2002). Genes in the rice region colinear to the QTL region on millet chromosome IX include a cytochrome P450 gene with high similarity to the lateral meristem proliferating gene SUPERSHOOT1, from Arabidopsis, and genes with high similarity to the Arabidopsis AXR1 gene, all of which are involved in auxin regulation (STIRNBERG et al. 1999; TANTIKANJANA et al. 2001). This approach offers considerable promise in identifying novel candidate genes but would be significantly improved by better annotation of the rice genome. As this happens, we expect further candidate genes from rice and maize to be identified.
The changes in inflorescence morphology seen between green and foxtail millet are the product of domestication, and it is likely that the selective pressure was for increase in seed production (HARLAN 1992). Although such selection must have been the focus of all domestications of cereal grains, the phenotypic traits affected and the genes thus acted upon are likely to differ in the different domestication events. For example, comparisons of foxtail millet with the closely related crop pearl millet (Pennisetum glaucum; DOUST and KELLOGG 2002) show that seed production per primary branch in pearl millet increases only marginally over the presumed wild progenitor (P. mollissimum), but that the inflorescences are much longer, with a concomitant increase in primary branch number (HARLAN 1992; PONCET et al. 1998, 2000). This underscores the importance of careful selection of genes as potential candidates for phenotypic variation and highlights the role that developmental and QTL analysis can have in selecting candidate genes for further analysis.
Conclusions:
This study represents an important first step in understanding the control of inflorescence architecture diversification in millets, and the results are of value because they concentrate on repeatable QTL across high- and low-density trials. Our data show that the inflorescence characteristics that differentiate two millet species are each controlled by a small number of loci. Previous developmental analyses have shown that the characters distinguishing foxtail and green millet are similar to those distinguishing other species of Setaria, Pennisetum, and Cenchrus (DOUST and KELLOGG 2002). Among these species, much of the phenotypic variation involves inflorescence architecture, including the numbers of primary branches, density of primary branches, numbers of orders of branching, and numbers of spikelets that fail to develop. We hypothesize that the loci identified here are the same ones that natural selection has affected during the evolution of the 300+ species of the Setaria-Pennisetum clade. This can be tested by further investigations in mapping populations of closely related species such as pearl millet, as well as by studying the evolution and expression of those candidate genes suggested by QTL analysis and comparative mapping among wild species that exhibit variation in inflorescence morphology.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
|
FOOTNOTES |
|---|
|
LITERATURE CITED |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONACKNOWLEDGEMENTSLITERATURE CITED |
|---|
ALTSCHUL, S. F., T. L. MADDEN, A. A. SCHAFFER, J. ZHANG, Z. ZHANG et al., 1997 Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25: 3389–3402.
BASTEN, C. J., B. S. WEIR and Z-B. ZENG, 2002 QTLCartographer: A Reference Manual and Tutorial for QTL Mapping. North Carolina State University, Raleigh, NC.
BEAVIS, W. D., 1998 QTL analyses: power, precision, and accuracy, pp. 145–162 in Molecular Dissection of Complex Traits, edited by A. H. PATERSON. CRC Press, Boca Raton, FL.
BOMBLIES, K., R.-L. WANG, B. A. AMBROSE, R. J. SCHMIDT, R. B. MEELEY et al., 2003 Duplicate FLORICAULA/LEAFY homologues zfl1 and zfl2 control inflorescence architecture and flower patterning in maize. Development 130: 2385–2395.
CHEVERUD, J. M., 2001 A simple correction for multiple comparisons in interval mapping genome scans. Heredity 87: 52–58.[CrossRef][Medline]
CHUCK, G., M. MUSZYNSKI, E. KELLOGG, S. HAKE and R. J. SCHMIDT, 2002 The control of spikelet meristem identity by the branched silkless1 gene in maize. Science 298: 1238–1241.
CHURCHILL, G. A., and R. W. DOERGE, 1994 Empirical threshold values for quantitative trait mapping. Genetics 138: 963–971.[Abstract]
CLARK, R. M., E. LINTON, J. MESSING and J. F. DOEBLEY, 2004 Pattern of diversity in the genomic region near the maize domestication gene tb1. Proc. Natl. Acad. Sci. USA 101: 700–707.
DEVOS, K. M., Z. M. WANG, J. BEALES, T. SASAKI and M. D. GALE, 1998 Comparative genetic maps of foxtail millet (Setaria italica) and rice (Oryza sativa). Theor. Appl. Genet. 96: 63–68.[CrossRef]
DOEBLEY, J., and A. STEC, 1991 Genetic analysis of the morphological differences between maize and teosinte. Genetics 129: 285–295.[Abstract]
DOEBLEY, J., and A. STEC, 1993 Inheritance of the morphological differences between maize and teosinte: comparison of results for two F2 populations. Genetics 134: 559–570.[Abstract]
DOEBLEY, J., A. STEC and C. GUSTUS, 1995 Teosinte branched1 and the origin of maize: evidence for epistasis and the evolution of dominance. Genetics 141: 333–346.[Abstract]
DOEBLEY, J., A. STEC and L. HUBBARD, 1997 The evolution of apical dominance in maize. Nature 386: 485–488.[CrossRef][Medline]
DOERGE, R. W., and G. A. CHURCHILL, 1996 Permutation tests for multiple loci affecting a quantitative character. Genetics 142: 285–294.[Abstract]
DOUST, A. N., and E. A. KELLOGG, 2002 Inflorescence diversification in the panicoid "bristle grass" clade (Paniceae, Poaceae): evidence from molecular phylogenies and developmental morphology. Am. J. Bot. 89: 1203–1222.
DOUST, A. N., K. M. DEVOS, M. D. GADBERRY, M. D. GALE and E. A. KELLOGG, 2004 Genetic control of branching in foxtail millet. Proc. Natl. Acad. Sci. USA 101: 9045–9050.
GALE, M. D., and K. M. DEVOS, 1998 Plant comparative genomics after 10 years. Science 282: 656–659.
GAUT, B. S., and J. F. DOEBLEY, 1997 DNA sequence evidence for the segmental allotetraploid origin of maize. Proc. Natl. Acad. Sci. USA 94: 68090–68094.
HAKE, S., and T. ROCHEFORD, 2004 Exploiting quantitative trait loci in gene discovery. Genes Dev. 18: 597–601.
HARLAN, J. R., 1992 Crops and Man. American Society of Agronomy and Crop Science Society of America, Madison, WI.
HOLLAND, J. B., 1998 EPISTACY: A SAS program for detecting two-locus epistatic interactions using genetic marker information. J. Hered. 89: 374–375.
HUBBARD, L., P. MCSTEEN, J. DOEBLEY and S. HAKE, 2002 Expression patterns and mutant phenotype of teosinte branched1 correlate with growth suppression in maize and teosinte. Genetics 162: 1927–1935.
IRISH, E. E., 1997 Class II tassel seed mutations provide evidence for multiple types of inflorescence meristems in maize (Poaceae). Am. J. Bot. 84: 1502–1515.[Abstract]
KELLOGG, E. A., 2000 Molecular and morphological evolution in Andropogoneae, pp. 149–158 in Proceedings of the Second International Conference on the Comparative Biology of the Monocotyledons, Vol. 2: Symposium on Grass Systematics and Evolution, edited by J. E. EVERETT and S. W. L. JACOBS. CSIRO, Melbourne.
KYOZUKA, J., S. KONISHI, K. NEMOTO, T. IZAWA and K. SHIMAMOTO, 1998 Down-regulation of RFL, the FLO/LFY homolog of rice, accompanied with panicle branch initiation. Proc. Natl. Acad. Sci. USA 95: 1979–1982.
LANDER, E. S., and D. BOTSTEIN, 1989 Mapping Mendelian factors underlying quantitative traits using RFLP linkage maps. Genetics 121: 185–199.
LANDER, E. S., P. GREEN, J. ABRAHAMSON, A. BARLOW, M. J. DALY et al., 1987 MAPMAKER: an interactive computer package for constructing primary genetic linkage maps of experimental and natural populations. Genomics 1: 174–181.[CrossRef][Medline]
LARK, K. G., K. CHASE, F. ADLER, L. M. MANSUR and J. H. ORF, 1995 Interactions between quantitative trait loci in soybean in which trait variation in one locus is conditional upon a specific allele at another. Proc. Natl. Acad. Sci. USA 92: 4656–4660.
LAWRENCE, C. J., Q. DONG, M. L. POLACCO, T. E. SEIGFRIED and V. BRENDEL, 2004 MaizeGDB, the community database for maize genetics and genomics. Nucleic Acids Res. 32: D393–D397.
LUKENS, L. N., and J. DOEBLEY, 1999 Epistatic and environmental interactions for quantitative trait loci involved in maize evolution. Genet. Res. 74: 291–302.[CrossRef]
MATHEWS, S., R. C. TSAI and E. A. KELLOGG, 2000 Phylogenetic structure in the grass family (Poaceae): evidence from the nuclear gene phytochrome B. Am. J. Bot. 87: 96–107.
MONNA, L., N. KITAZAWA, R. YOSHINO, J. SUZUKI, H. MASUDA et al., 2002 Positional cloning of rice semidwarfing gene, sd-1: rice "green revolution gene" encodes a mutant enzyme involved in gibberellin synthesis. DNA Res. 9: 11–17.[Abstract]
NEUFFER, M. G., E. H. COE and S. R. WESSLER (Editors), 1997 Mutants of Maize. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
PONCET, V., F. LAMY, J. ENJALBERT, H. JOLY, A. SARR et al., 1998 Genetic analysis of the domestication syndrome in pearl millet (Pennisetum glaucum L., Poaceae): inheritance of the major characters. Heredity 81: 648–658.[CrossRef]
PONCET, V., F. LAMY, K. M. DEVOS, M. D. GALE, A. SARR et al., 2000 Genetic control of domestication traits in pearl millet (Pennisetum glaucum L., Poaceae). Theor. Appl. Genet. 100: 147–159.[CrossRef]
RIESEBERG, L. H., O. RAYMOND, D. M. ROSENTHAL, Z. LAI, K. LIVINGSTONE et al., 2003a Major ecological transitions in wild sunflowers facilitated by hybridization. Science 301: 1211–1216.
RIESEBERG, L. H., A. WIDMER, A. M. ARNTZ and J. M. BURKE, 2003b The genetic architecture necessary for transgressive segregation is common in both natural and domesticated populations. Philos. Trans. R. Soc. London B Biol. Sci. 358: 1141–1147.[CrossRef][Medline]
SALAMOV, A. A., and V. V. SOLOVYEV, 2000 Ab initio gene finding in Drosophila genomic DNA. Genome Res. 10: 516–522.
SPSS, 2001 Statistical Package for the Social Sciences (Windows), Version 10.1.3. SPSS, Chicago.
STIRNBERG, P., S. P. CHATFIELD and H. M. OTTOLINE LEYSER, 1999 AXR1 acts after lateral bud formation to inhibit lateral bud growth in Arabidopsis. Plant Physiol. 121: 839–847.
TANTIKANJANA, T., J. W. H. YONG, D. S. LETHAM, M. GRIFFITH, M. HUSSAIN et al., 2001 Control of axillary bud initiation and shoot architecture in Arabidopsis through the SUPERSHOOT gene. Genes Dev. 15: 1577–1588.
WANG, R.-L., A. STEC, J. HEY, L. LUKENS and J. DOEBLEY, 1999 The limits of selection during maize domestication. Nature 398: 236–239.[CrossRef][Medline]
WANG, Z. M., K. M. DEVOS, C. J. LIU, R. Q. WANG and M. D. GALE, 1998 Construction of RFLP-based maps of foxtail millet, Setaria italica (L.) P. Beauv. Theor. Appl. Genet. 96: 31–36.[CrossRef]
WARE, D., P. JAISWAL, J. NI, X. PAN, K. CHANG et al., 2002a Gramene: a resource for comparative grass genomics. Nucleic Acids Res. 30: 103–105.
WARE, D. H., P. JAISWAL, J. NI, I. V. YAP, X. PAN et al., 2002b Gramene, a tool for grass genomics. Plant Physiol. 130: 1606–1613.
WHITE, S. E., and J. F. DOEBLEY, 1999 The molecular evolution of terminal ear1, a regulatory gene in the genus Zea. Genetics 153: 1455–1462.
This article has been cited by other articles:
 |
|
 |
|
  T. H. Kebrom and T. P. Brutnell The molecular analysis of the shade avoidance syndrome in the grasses has begun J. Exp. Bot., October 5, 2007; (2007) erm205v1. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Doust Architectural Evolution and its Implications for Domestication in Grasses Ann. Bot., October 1, 2007; 100(5): 941 - 950. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. A. Kellogg Progress and challenges in studies of the evolution of development J. Exp. Bot., October 1, 2006; 57(13): 3505 - 3516. [Abstract] [Full Text] [PDF]
|
|
| ||
