| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Dep. of Agronomy, Univ. of Wisconsin-Madison, Madison, WI 53706
* Corresponding author (deleonn{at}msu.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Abbreviations: AFLP, amplified fragment length polymorphism • CIM, composite interval mapping • cM, centimorgan • LG, linkage group • PCR, polymerase chain reaction • QTL, quantitative trait locus • SSR, simple sequence repeat • tb1, teosinte branched1
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Prolificacy has been the subject of several selection studies in maize because prolific types of maize may be more stress tolerant (Coors and Mardones, 1989; Maita and Coors, 1996; Carena et al., 1998; Subandi, 1990; Jampatong et al., 2000; de Leon and Coors, 2002). Prolific types tend not to become barren under conditions of inadequate moisture or fertility, and they can produce ear shoots under the intense plant-to-plant competition typical of modern high planting densities, although prolificacy may also be associated with poor stalk strength and standability (Carlone and Russell, 1987; Thomison and Jordan, 1995; Maita and Coors, 1996).
One of the longest-running selection programs for prolificacy is that involving the Golden Glow maize population [GG(MP)] (de Leon and Coors, 2002). After >25 cycles of mass selection for increased number of ears plant–1, there has been a significant change in overall plant morphology in GG(MP). The most extreme plants from advanced cycles of GG(MP) have a highly branched morphology when grown under optimal conditions and low planting densities. Selection studies with GG(MP) comprise appropriate material for studying the effects of selection on morphological and genetic changes.
An intensively studied example of changes in plant architecture involves the domestication of maize from its ancestor teosinte. These two subspecies diverged in several morphological features, due mainly to the effects of cultivation and selection by man (Galinat, 1988). The teosinte plant branches from its base and then generates branches on those branches, which results in a profuse type of plant architecture compared with the single-stalk morphology of modern maize (Kellogg, 1997). The inflorescence terminating the main culm is male in both maize and teosinte plants. In teosinte, the primary lateral branches are normally elongated, and the inflorescences terminating these branches are generally male. Maize typically produces branches at only two or three of the nodes along the main stem. At most other nodes, axillary buds are present, but they are arrested early in development. The primary lateral branches in modern maize are much shorter, and the inflorescences terminating these branches are usually female (Doebley, 1990; Doebley and Wang, 1997). Although the morphological differences between maize and teosinte are dramatic, it has been proposed that as few as five genomic regions may control the major traits differentiating cultivated maize from teosinte (Beadle, 1980; Doebley and Stec, 1991, 1993).
To better characterize the changes in morphology accompanying selection for prolificacy in GG(MP), we developed a mapping population from the cross of nonprolific inbred A679 with a highly prolific S1 plant derived from Cycle 23 of GG(MP). The phenotypic evaluation of this population showed that significant variation was observed among F3 families for all 16 traits analyzed (de Leon et al., 2005). High heritability estimates were observed for most traits. This suggested that a number of these traits had a comparatively simple type of inheritance and/or were only moderately affected by environmental or other random, nongenetic factors. A substantial plasticity, nevertheless, was observed in the expression of a number of these morphological characteristics. In addition, a number of morphologically related traits were highly correlated with each other, indicating that some of these traits may possibly be controlled by similar genetic factors (de Leon et al., 2005).
The first objective of our study was to identify molecular genetic markers linked to chromosomal regions that influenced the number of ears plant–1 and other traits related to plant architecture. The second objective was to determine whether genes previously shown to influence branching patterns in maize were associated with QTLs influencing prolificacy and related traits. The third objective was to determine whether there were epistatic interactions among putative QTLs influencing the number of ears plant–1 and correlated morphological traits.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Molecular Markers
The PCR amplification of diluted genomic DNA samples with SSR primers and electrophoresis of these samples of amplified DNA was performed according to Senior and Heun (1993) with some slight modifications. The primer sequences used for the PCR were obtained from the MaizeDB web site (http://www.maizegdb.org/; verified 22 Feb. 2005).
More than 300 SSR primer sets were screened against the parents of the population [GG(C23) and A679] to resolve polymorphism. Briefly, amplification reactions were performed in 15-µL volumes containing 50 ng of genomic DNA, 0.2 µL of AmpliTaq Gold DNA Polymerase (Applied Biosystems, Bedford, MA), 0.06 µL of dNTPs (25 mM), 0.5 µL of each primer (50 ng µL–1), 1.5 µL of nonacetylated BSA (10 mg mL–1), 0.9 µL of MgCl2 (25 mM), 1.5 µL 10x AmpliTaq Gold buffer, and 0.1 µL of 
-33-phosphorus. Cycling conditions were as follows: an initial denaturation step of 10 min at 95°C was followed by 20 amplification cycles of 1 min denaturation at 95°C, 1 min at 94°C, 1 min annealing at 65°C, 1 min extension at 72°C, for the first cycle, and then a 0.2°C decrement for the annealing temperature, each repeated cycle, until the annealing temperature reached 55°C. Another 20 amplification cycles were done of 1 min denaturation at 94°C, then 1 min of annealing at 55°C, and 2 min extension at 72°C. Once the polymorphism and the band size were determined, primer pairs were multiplexed in the same PCR reaction based on band sizing by adding 0.5 µL of each primer and reducing the same amount of sterile water in the mix. Up to three primer pairs were used per reaction. Sixty-one polymorphic SSR primer pairs were used in the initial construction of the linkage map.
Amplified fragment length polymorphism analysis was performed according to the manufacturer's instruction (Instruction Manual AFLP Analysis System I, AFLP Starter Primer Kit, Gibco BRL, Life Technologies). The combination of two restriction endonucleases, EcoRI and MseI was used for the genomic DNA digestion. Enzyme EcoRI was 
-33-Phosphorus-labeled before amplification. Six AFLP primer pairs were assayed (AGC-CAC, AGC-CAG, AGG-CTC, AGGCAA, AGG-CTA, AGG-CAC). An aliquot of 2.5 µL of each PCR product from the AFLP selective amplification as well as for the SSR reactions was run on polyacrylamide gels [5% polyacrylamide, 8 M urea, 5 x TBE (1,1,2,2-tetrabromoethylene), TEMED (N,N,N',N'-temex,tetramethylethylenediamine), and ammonium persulfate]. Following, the gel was fixed, dried, and exposed to x-ray film at room temperature. The auto-radio-grams were visually scored and reviewed by a second reader. The AFLP markers were visually assessed as presence or absence of a band in a genotype; therefore, only dominant type scoring was possible in this case. The polymorphic bands were named A, B, or RA (depending on whether one parent, the other, or none of them had the absent band) followed by a number corresponding to the chronological order by which they were scored. Averages of 15 to 20 polymorphic scorable bands were found per primer combination. Eighty AFLP markers were used in the initial construction of the linkage map.
Data Analysis
Linkage analyses were performed using JoinMap Version 3.0 (Van Ooijen and Voorrips, 2001) with the Kosambi mapping function (Kosambi, 1944), a minimum LOD score of 4.0, and a maximum recombination fraction of 0.50. Linkage group numbers were assigned based on information available in the MaizeDB web site (http://www.maizegdb.org). Quantitative trait loci analyses were performed using the marker linkage information given in Fig. 1. Means and standard deviations were calculated for each phenotypic trait as described in de Leon et al. (2005), and the Shapiro-Wilks statistics from the CAPABILITY procedure of SAS (SAS Institute, 2000) was used to study the empirical distribution of each trait at a 0.05 level of significance. For traits that deviated from normality, a data transformation approach was considered using the Box–Cox power transformation family (Box and Cox, 1964).
|
Quantitative trait loci mapping was performed using the composite interval mapping (CIM) technique of PlabQTL Version 1.1 software (Utz and Melchinger, 1995). This approach uses a multiple regression procedure that adjusts for background effects of markers other than those in the interval being tested. All the markers located outside of the interval, which were chosen in the preselection, were used as cofactors in the analysis. For these analyses, intervals of two centimorgans (cM) between markers and putative QTLs were tested. The threshold of the LOD score for declaring a putative QTL significant was chosen to be 3.77, corresponding to a 0.0006 comparison-wise significance level, which was approximately equivalent to applying a significance level of 0.05 for the experimental-wise error rate. The position of a QTL was always described in relationship to the nearest marker to the left of the LOD-score peak. The additive and dominant effects of a single QTL at its estimated position were obtained from the output of the QTL analysis using PlabQTL (Utz and Melchinger, 1995). Dominance and two-loci epistatic effects were included in the analysis. Initially, the detection of QTLs was conducted without epistatic effects in the model. Subsequently, digenic epistatic effects were estimated for the detected set of QTLs. The epistatic effects were chosen in a stepwise regression procedure (Utz and Melchinger, 1995). The phenotypic variance accounted for by a single QTL was calculated as the square of each partial correlation coefficient. This value was the coefficient of determination of the specified QTL and the phenotypic observations, keeping all other QTL effects fixed. The total phenotypic variance accounted for by all QTLs affecting a single trait was calculated by simultaneously fitting all QTLs detected for the specific trait by a multiple regression model (Utz and Melchinger, 1995).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Sixteen morphological traits were evaluated, and 10 of them presented some deviation from normality (de Leon et al., 2005). Even after transformation, five of the 10 transformed traits did not present satisfactory approximation to a normal distribution; therefore, they were analyzed in their original scales (Table 1). Traits that required transformation were also analyzed in their original scale to compare the results (data not provided). Magnitudes of R2 values and LOD scores for the transformed data did not differ appreciably from those for the raw data. This was expected because the QTL analyses were performed using the regression interval mapping method, which is relatively robust with respect to nonnormality (Utz and Melchinger, 1995). Nonetheless, the QTL analyses presented in this study used transformed data whenever appropriate (Table 1).
|
|
|
Two other QTLs were found for total number of ears plant–1 on LGs 4 and 7, which coincide with QTLs found to be associated with the number of ears on main stalk. The directions of the additive effects in the two QTLs were the same for both traits. As expected, the parent GG(C23) conferred the dominant allele that increased the total number of ears plant–1 and the number of ears on main stalk for the QTL associated with marker umc1366. On the other hand, A679 conferred the dominant allele that increased both traits for the QTL associated with marker A22 (Table 1). Contributions such as this coming from parent A679 would explain some of the transgressive segregation observed in the phenotypic evaluation of these traits (de Leon et al., 2005). Two more QTLs were found for number of ears on main stalk on LGs 2 and 6 that accounted for 15.2 and 11.2% of the phenotypic variation, respectively (Table 1).
The position of the lowermost ear on the main stalk was affected by three similarly sized QTLs located on LGs 1, 2, and 6, which accounted for 10.1, 10.9, and 11.0% of the phenotypic variation, respectively (Table 1). Although there was a high phenotypic correlation between position of the lowermost ear on the main stalk with total number of ears plant–1 and with number of ears on main stalk (de Leon et al., 2005), only the QTL located on LG 2 was shown to affect another trait involved in the formation of ears (i.e., number of ears on main stalk). Contrary to expectation, for two of the three QTLs, parent GG(C23) conferred the allele that increased the node number of lowermost ear on main stalk, and in all cases the allele from parent A679 was dominant (Table 1).
A single QTL was found to influence number of aboveground nodes on the main stalk, and it was located on the short arm of LG 1. Parent GG(C23), the parent with the fewest nodes, conferred the dominant allele that increased number of aboveground nodes for this QTL (Table 1). The largest association in this study was found between marker phi055 on LG 1 and frequency of terminal male inflorescences on tillers. This QTL accounted for 36.5% of the phenotypic variation for this trait. This same marker was also associated with number of tillers plant–1. This QTL accounted for 16.3% of the phenotypic variation. As expected, parent GG(C23) contributed the dominant allele that increased the value of both traits. A second marker, A5, was associated with number of tillers plant–1. In this case, parent GG(C23) also conferred the allele that increased number of tillers plant–1, but this allele was recessive to the A679 allele. As mentioned previously, these two markers, phi055 and A5, were also associated with total number of ears plant–1 and number of ears on tillers, two traits associated with axillary meristem activation (Table 1).
A single QTL located on LG 1 was found to affect frequency of plants with completely active regions on main stalk. This QTL accounted for 10% of the phenotypic variation for this trait. Contrary to expectations, none of the other traits involved in the development of lateral branches on the main stalk were affected by this region of the genome.
Main stalk total length was the trait with the largest number of detected QTLs (Table 1). This set of QTLs accounted for the largest percentage of phenotypic variation of all traits analyzed (Table 2). Medium to large effect QTLs were found for this trait, and they were distributed in five different LGs. Two of those regions, located on LGs 5 and 9, were also found to influence main stalk internode length. The QTL located on LG 5 was also shown to affect the uppermost lateral branch length, a trait also involved in the elongation of the intercalary meristem. In all three cases, parent A679 conferred the allele that increased length, and this allele was dominant. Another three QTLs were found to affect the uppermost lateral branch length. These were located on LGs 1, 6, and 9 and accounted for 13.7, 9.9, and 10.7% of the phenotypic variation, respectively (Table 1).
Single QTL in different regions of LG 1 were found for mid-silk and mid-pollen dates. Parent GG(C23) conferred the allele that increased the value of each trait for both QTLs; however, the A679 allele was dominant for mid-silk date. Two QTLs that accounted for 10.2 and 9.0% of the phenotypic variation affected anthesis–silk interval. They were located on LGs 5 and 7. Interestingly, neither of these QTLs influenced mid-pollen date or mid-silk date, two traits intimately related to anthesis–silk interval. As expected, the prolific parent conferred the alleles that increased the interval, but these alleles were recessive to the A679 allele (Table 1).
Epistatic Effects
Tests for digenic epistasis were performed only for combinations of chromosomal regions that had an association with a particular trait. No additive-by-additive interactions were found for any trait; however, additive-by-dominance interactions were detected for total number of ears plant–1, number of ears on the main stalk, main stalk total length, uppermost lateral branch length, and position of the lowermost ear on the main stalk. The magnitude of these interaction effects was relatively small, with values that ranged from 4.50 to 12.23% of the additive variance (Table 3).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
In general, QTL analyses have some limitations in estimating the number of important genomic regions controlling any particular trait. First, several small, closely linked QTLs cannot be distinguished from a single QTL of large effect (Lynch and Walsh, 1998), even when using the CIM. In addition, recombination between molecular markers and the genomic region of interest may restrict the ability to detect QTLs, particularly in a situation like in this study where portions of the genome remain unmapped. Lastly, the ability to detect QTLs with small effects may be limited by the number of progeny evaluated. If the progeny number is low, false negatives are likely (Beavis, 1998). Ordinarily, these limitations would result in an underestimation of the number of genomic regions affecting any given trait. Therefore, it is adequate to assume that the number of QTLs detected in this study represents a minimal estimate of the true number of factors affecting any of the 16 traits evaluated.
The vast majority of the QTLs found on LGs 1, 2, 4, 6, and 7 involved traits related to the axillary meristems, such as total number of ears plant–1, number of ears on the main stalk, number of ears on tillers, number of aboveground nodes on the main stalk, number of tillers plant–1, frequency of male inflorescences at the end of tillers, and position of the uppermost and lowermost ears on the main stalk. On the other hand, QTLs found on LGs 3, 5, and 9, for the most part, involved traits related to elongation associated with activation of intercalary meristems.
Fourteen of the 33 associations involved LG 1, and a large percentage of them had considerably large effects. Many of those were traits that affected the activation of axillary meristems such as total number of ears plant–1, number of ears on tillers, frequency of male inflorescences at the end of tillers, and number of tillers plant–1, and were also associated with marker phi055. Parent GG(C23) was the contributor of the partially dominant allele that increased values for all traits associated with this marker. The largest QTL in this region was found for frequency of male inflorescences at the end of tillers and accounted for 36.5% of the phenotypic variance for this trait.
Marker phi055 is located in bin 1.09 of the maize genome (http://www.maizegdb.org), the same bin where tb1 is located (Doebley et al., 1997). The tb1 mutant phenotype is conditioned by a recessive allele of maize that affects plant architecture (Burnham, 1959). This region has been shown to be one of the five major genomic regions responsible for the morphological differentiation between maize and teosinte (Doebley and Stec, 1991, 1993). When the segment of chromosome carrying the maize allele was introgressed into a teosinte genetic background, it was found that this single QTL completely suppressed the elongation of lateral branches, and it converted the inflorescence terminating these branches from male to female (Doebley et al., 1995). Molecular cloning and analysis of tb1 suggested that this mutant's protein functions as a repressor of organ growth (Doebley et al., 1997).
Although it was not possible to precisely characterize the genetic mode of action of this specific region of LG 1 with our mapping population, this important region linked to marker phi055 shares many similarities with the pattern of development suggested for the tb1 mutant (Doebley and Stec, 1991, 1993). We conclude that this segment of the genome was very important in distinguishing the morphology of the tillering, highly prolific parent GG(C23) from parent A679. It is intriguing to ponder whether a similar genetic factor or factors could have been involved in both selection for prolificacy in GG(MP) and the domestication of maize from its ancestor teosinte.
To test whether tb1 could be considered as the single factor influencing the change in morphology seen in GG(MP), D.V. Butruille and co-workers (1999, unpublished data) genotyped 82 and 103 individuals from Cycles 0 and 21 of this population, respectively, using a SSR primer pair flanking the location of tb1. No significant differences were observed for the frequency of the five alleles found for this locus (D.V. Butruille, 1999, unpublished data). This indicated that the frequency of tb1 alleles present in GG(MP) did not change appreciably across cycles of selection.
Studies about the mode of action of tb1 in the differentiation of maize and teosinte have shown that even though the expression of the maize allele was about twice the level of the teosinte allele in the inflorescence primordia terminating the primary lateral branches, no differences in the amino acid composition of the protein have been detected between them. This suggests that the tb1 protein function has not really changed (Doebley and Lukens, 1998). Changes in expression are the most likely explanation for the evolutionary differences observed between maize and teosinte (Doebley and Lukens, 1998). Therefore, the absence of allelic frequency variation between cycles at the GG(MP) tb1 locus might be expected.
Further research is necessary to confirm the association of the region linked to marker phi055 on LG 1 and the maize mutant tb1. However, on the basis of the phenotypic observations and the results found on the regulatory nature of tb1(Doebley and Lukens, 1998), it is reasonable to believe that a tb1-like gene and/or its regulation are important factors in the morphological changes seen in GG(MP) across cycles of selection.
Selection for the number of ears in GG(MP) was for a trait opposite in direction to that which occurred during domestication. Of particular interest, not only did the number of ears plant–1 increase across cycles of selection in GG(MP), but also the morphology of many plants in the advanced cycles began to resemble teosinte. The question now becomes whether the genetic factors that triggered changes seen in GG(MP) were similar to the ones shown to be important in the domestication of maize from teosinte. Two hypotheses may be drawn in this regard. The first is that the selection procedure affected the same genetic factors that humans modified during the transformation of teosinte into modern maize. The second is that the highly branched GG(MP) phenotype was achieved by genetic factors independent from those involved with domestication. Determining the validity of the two hypotheses is of considerable consequence given the ever-more powerful genetic tools available to modern geneticists and breeders. If the same genes, genomic regions, or regulatory networks used when selecting for prolificacy in GG(MP) were also used during the initial steps of domestication, this would indicate that ancestral genetic features remain important for current cultivars and that future modifications of plant architecture will likely build on these same factors. However, domestication occurred about 7000 yr ago, and the manner in which maize is cultivated and harvested has changed considerably, as has the genetic form of cultivars (i.e., hybrids vs. open-pollinated varieties). A number of unrelated novel genetic factors may now be as important for determining morphology as those initially involved in domestication.
Several authors have recently suggested that genetic variation that is unexpressed as phenotypic variation (cryptic variation) is a relatively common phenomenon in some species (Rutherford, 2000; Lauter and Doebley, 2002). They suggested that genetic modifiers, which have no effect by themselves, can, however, influence the degree of expression of a phenotype and, therefore, be eventually revealed (Gottlieb et al., 2002). This process can be illustrated as the basic principle of epistatic interactions (Phillips, 1998). Through the effect of continued selection, and as long as no biological ceiling is reached, those phenotypically unexpressed genetic variants that stabilize the optimum phenotype continue to accumulate, and what was considered cryptic variation originally could later become neutral genetic variation affecting the phenotype (Waddington, 1957; Schmalhausen, 1986; Dudley and Lambert, 1992). The interaction of all these forces would provide a significant degree of morphological plasticity.
Even though no digenic additive epistatic interactions were found in this study, most likely because of low statistical power of the experimental setting used, it remains possible that interactions between chromosomal regions played an important role in the morphological changes seen in GG(MP). By assuming that a common factor has affected selection of prolificacy in GG(MP) and the domestication of maize, at least regarding the elongation of lateral branches and inflorescence at the end of tillers, one might speculate that some phenomenon related to the transformation of cryptic variants into neutral phenotypic variants was the cause of some portion of the variation seen for most morphological traits in the GG(MP). After all, the advanced cycles of GG(MP) originated from Cycle 0 plants having the normal morphology of modern maize. Under conventional quantitative genetic theory, loci that control traits that differentiate modern maize from their ancestor teosinte should have become fixed long ago. Cryptic variation has been suggested to affect a number of other traits involved in the domestication of maize as well (Lauter and Doebley, 2002).
The mapping population used in this study was derived from a cross involving a single prolific plant from a highly heterogeneous, cross-pollinated maize population. In other words, the QTLs detected in this study using one particular GG(C23) parent may or may not be similar to the QTLs that would be detected using a mapping population derived from another prolific plant from GG(MP). One would have to analyze a large number of mapping populations using many different parental plants from GG(MP) to develop an overall understanding of how prolificacy and related traits are inherited in this single maize population.
The number of genetic factors detected in our study that underlie any particular trait is likely to be a minimal estimate because only one allele was sampled from GG(MP) at each locus. It is clear that, even though most of the traits analyzed in this research were highly heritable and are likely influenced by some of the most thoroughly studied genetic factors in maize, their genetic underpinnings are far from fully understood.
|
ACKNOWLEDGMENTS |
|---|
Received for publication October 3, 2003.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Vadose Zone Journal | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
