Quantitative trait loci and metabolic pathways

Abstract

The interpretation of quantitative trait locus (QTL) studies is limited by the lack of information on metabolic pathways leading to most economic traits. Inferences about the roles of the underlying genes with a pathway or the nature of their interaction with other loci are generally not possible. An exception is resistance to the corn earworm Helicoverpa zea (Boddie) in maize (Zea mays L.) because of maysin, a C-glycosyl flavone synthesized in silks via a branch of the well characterized flavonoid pathway. Our results using flavone synthesis as a model QTL system indicate: (i) the importance of regulatory loci as QTLs, (ii) the importance of interconnecting biochemical pathways on product levels, (iii) evidence for “channeling” of intermediates, allowing independent synthesis of related compounds, (iv) the utility of QTL analysis in clarifying the role of specific genes in a biochemical pathway, and (v) identification of a previously unknown locus on chromosome 9S affecting flavone level. A greater understanding of the genetic basis of maysin synthesis and associated corn earworm resistance should lead to improved breeding strategies. More broadly, the insights gained in relating a defined genetic and biochemical pathway affecting a quantitative trait should enhance interpretation of the biological basis of variation for other quantitative traits.

The past decade has seen an explosion of information on the structure, organization, and functions of the maize (Zea mays L.) genome, including the development of high density molecular marker maps. One application of new mapping technologies has been the genetic dissection of quantitative traits with much greater precision than was previously possible (1, 2). Still, the quantitative trait loci (QTLs) detected are generally rather poorly defined regions, and the size of a QTL’s phenotypic effect is sometimes confounded with its location relative to the nearest marker or to a nearby QTL. For most traits, genetic and biochemical information on metabolic pathways is extremely limited, and, therefore, it is difficult to interpret QTL results in terms of regulatory and structural genes, duplicate function loci, feedback inhibition, branched pathways, or other phenomena affecting trait expression. Our goal in this research project is to analyze the genetic control of a quantitative trait of economic importance [antibiosis to the corn earworm (CEW)] and to interpret the results in terms of the well characterized flavonoid pathway.

The CEW Helicoverpa zea (Boddie) is a major insect pest of maize and other crops (cotton, soybeans, peanuts) in the United States and elsewhere in the Western Hemisphere (3, 4). Corn earworm eggs are laid on the silks, and the larvae access the ear by feeding through the silk channel. Host–plant resistance to CEW by antibiosis is caused by the presence of the C-glycosyl flavones maysin, apimaysin, and methoxymaysin and related compounds (Fig. 1) in maize silks (5, 6). Although these flavones occur in various proportions in different maize lines and populations, maysin is the predominant compound in most genotypes (M.E.S., unpublished data). Upon ingestion by CEW, the flavones are oxidized to quinones, which bind amino acids, making them unavailable and thus inhibiting larval growth (7).

Figure 1.

The branch of the flavonoid pathway for the synthesis of flavones in maize. Enzyme abbreviations: CHS, chalchone synthase; CHI, chalcone isomerase; F3H, flavanone-3-hydroxylase; DFR, dihydroflavanone reductase; F3′H, flavanone-3′-hydroxylase; FNS, flavone synthase; RT, rhamnosyl transferase.

The branches of the flavonoid pathway leading to the synthesis of C-glycosyl flavones (including maysin and related CEW resistance factors), phlobaphenes (responsible for red cob and pericarp pigments), and 3-deoxyanthocyanins are regulated by the p1 locus. The p1 locus encodes a myb-like transcription factor. That locus also affects the silk-browning trait, whereby silks of some genotypes turn brown after wounding (8, 9). Based on our current understanding, maysin synthesis requires appropriate alleles at p1, c2, and/or whp1 [encoding chalcone synthases (10, 11)], chi1 [encoding chalcone isomerase (12)], pr1, [controlling the 3′-hydroxylation of the flavonoid B-ring to convert monohydroxy to dihydroxy compounds (13)], and unidentified additional loci encoding flavone synthase, C-glycosyl transferase, glucose oxidase, rhamnosyl transferase, and an enzyme such as glutathione S-transferase for transport to the vacuole (14, 15) (Fig. 1).

Our approach has been to develop F₂ or F_2:3 populations from crosses of inbred lines chosen to address the contribution of specific genes of the flavonoid pathway on maysin synthesis and antibiotic earworm resistance as measured in the laboratory by larval weight bioassay (16, 17). F₂ plants or F_2:3 family rows are grown in the field, and silks are collected 2–3 days after silk emergence. The silks are weighed, frozen, and freeze-dried (for recent populations), and concentrations of maysin, apimaysin, and other analogs are determined by reversed-phase HPLC (18). Whorl tissue is collected from F₂ or bulked from F_2:3 plants, DNA is extracted (19), and genotypes of the individuals in the population are determined by restriction fragment length polymorphism (RFLP) analysis by using primarily the University of Missouri-Columbia core RFLP markers (20, 21) and probes for genes of the flavonoid pathway (22). Analysis of the phenotypic data for silk chemical constituents and bioassay results is carried out primarily with ANOVA (analysis of variance), general linear model, and correlation procedures of SAS statistical analysis software (23). Linkage maps for the populations are constructed with the computer program mapmaker/exp (24), which uses a maximum likelihood procedure. QTL, chromosomal position, significance, and the contribution to trait variance are determined through interval mapping with mapmaker/qtl (25) and composite interval mapping by using qtl cartographer software (World Wide Web address: http://statgen.ncsu.edu).

In this paper, we will review the results of our QTL analyses to identify loci controlling flavone synthesis and will address the implication of our research on understanding the basis of quantitative variation for agronomic traits.

RESULTS

Population 1: (GT114 × GT119)F₂.

Contribution of the major regulatory locus p1. The main purpose of the first population was to determine the magnitude of effect of functional vs. nonfunctional alleles at the p1 locus on the quantitative expression of silk maysin concentration. The inbred GT114, which has a silk maysin concentration of 0.5–0.6% of fresh silk weight, has a functional p1-wrb allele (white pericarp, red cob, browning silks) whereas the inbred GT119, which has negligible levels of silk maysin, has a nonfunctional p1-www allele (white pericarp, white cob, nonbrowning silks). Maysin concentration was determined from 3-day-old silks of 285 F₂ plants and genotypes determined by RFLP analysis. As expected, the p1 locus had the largest effect on maysin concentration, accounting for 58% of the variation (22). The effect of the positive allele from GT114 was strikingly additive, with mean maysin concentrations of 0.023, 0.280, and 0.630% of fresh silk weight for the GT119 homozygous, GT114/GT119 heterozygous, and GT114 homozygous classes, respectively. A second, unexpected result was the detection of a major QTL on chromosome 9S (Table 1). A substantial portion of the contribution from the 9S locus comes from epistatic interaction with p1. The 9S region increases maysin only when the 9S region is homozygous for the GT119 allele and there is a functional p1 (GT114) allele. In another study (22), we put forward the hypothesis that the 9S locus may function in a pathway intersecting the maysin synthetic pathway. To extend these results to understanding CEW resistance, silks from a subset of 76 F_2:3 lines, corresponding to the high and low tails of the F₂ maysin distribution, were evaluated for CEW antibiosis in a dried-silk bioassay. The same chromosome regions significant for maysin concentration also were detected in the ANOVA for CEW larval weight (17).

Table 1.

Genotype class means for silk maysin concentration at chromosome 9S loci in three populations

GT114/GT119 (Population 1)
	Percent maysin (umc105a)*
GT114	0.25  a
F1	0.25  a
GT119	0.46  b
GE37/FF8 (Population 2)
	Percent maysin, MO (wx1)	Percent maysin, GA (wx1)
FF8	0.35  a	0.24  a
F1	0.39  a	0.29  a
GE37	0.52  b	0.42  b
GT114/NC7A (Population 3)
	Percent maysin (wx1)
GT114	0.27  a
F1	0.31  a
NC7A	0.53  b

Locus in parentheses showed highest significance level. Within each population, means followed by the same letter are not significantly different at the 0.05 level. All indicated loci are on chromosome 9S in the order sh1-bz1-umc105a-wx1. Distance between sh1 and wx1 ranged from 25 to 35 cM for the different populations. 

Population 2: (GE37 × FF8)F_2:3.

Defining loci for quantitative variation in maysin and CEW antibiosis. In contrast to Population 1, this population was developed to investigate the inheritance of maysin concentration when both parents have functional p1 alleles. Both GE37 and FF8 have p1-wrb alleles that are indistinguishable by Southern hybridization analysis (M.D.M. and S. Flint-Garcia, unpublished data); the structure of the p1-wrb allele appears to be identical to that previously described (26). Silks of GE37 have very high maysin levels whereas FF8 silks have low to intermediate values. Two hundred and fifty F_2:3 lines were grown in each of three replicated trials and evaluated for silk maysin concentration (Columbia, MO and Tifton, GA) and CEW antibiosis (Tifton, GA). The F_2:3 lines were evaluated for genotypes at 109 RFLP loci distributed throughout the genome. Composite interval mapping results for the three traits revealed large effects due to loci located on chromosomes 1, 2, 6, and 9 (Table 2). A major discrepancy among traits was that the large effect detected on chromosome 9S for maysin concentration in both locations was not matched by a significant effect for CEW larval weight. Two of the major QTLs corresponded to locations of known flavonoid loci: whp1 on chromosome 2L and the pl1-sm1 region on chromosome 6L. Because chalcone synthase, encoded by whp1, is considered a key enzyme of flavonoid synthesis, an effect on maysin synthesis due to variation at this locus is plausible. On chromosome 6L, the QTL mapped near pl1 and sm1. The pl1 locus, like p1, encodes a myb-homologous transcription activator, and their protein products may compete for binding sites in promoters of flavonoid structural genes. On the other hand, sm1 is known to interact with p1 (27), and sm1 apparently controls the addition of a rhamnose molecule to the C-glycosyl group (unpublished data; Fig. 1). No known flavonoid pathway loci are found in the region of the effect from chromosome 1. However, a QTL for maysin synthesis in Population 1 also was present in this same region (22). Multiple-locus models explained 38–50% of the variation for the three traits.

Table 2.

QTLs for silk maysin concentration in two locations and weight of corn earworm larvae grown on silks of the population (GE37 × FF8)F_2:3

Trait (location)	Chromosome	Nearest RFLP locus	Likelihood ratio*	R2†
Maysin concentration (Missouri)	1S	umc167a	17.65	4.5
	2L	asg20	33.20	11.0
	6S	csu70	11.61	7.9
	6L	pl1	23.14	6.8
	8L	npi268	12.83	5.3
	9S	wx1	33.69	13.6
Multiple-QTL model				40.1
Maysin concentration (Georgia)	1S	umc76a	17.71	5.3
	2L	asg20	34.69	9.7
	6L	umc38a	13.72	3.9
	8L	bnl2.369	21.78	4.4
	9S	wx1	25.15	12.30
Multiple-QTL model				38.0
Larval weight (Georgia)	1S	npi262	21.38	3.7
	2S	umc53a	11.63	2.8
	2L	whp1	78.78	18.7
	4L	gln5	12.11	2.2
	6S	csu70	37.75	8.9
	6L	umc38a	21.25	6.3
	10C	umc64	11.73	1.7
Multiple-QTL model				46.9

Results were obtained with composite interval mapping by using qtl cartographer. S, short arm; L, long arm; C, centromere region. 

^*The likelihood ratio can be converted to the more familiar logarithm of odds score by dividing by 4.6052. 

^†Percent phenotypic variance explained at nearest RFLP marker by using the SAS GLM procedure. 

Population 3: (GT114 × NC7A)F₂.

Effect of the pr1 (flavonoid 3′-hydroxylase) locus. In crossing a high maysin line (GT114) with a high apimaysin line (NC7A), our intention was to determine the loci responsible for the 3′-hydroxylation of the flavone B-ring. Because the pr1 locus controls an analogous step in anthocyanin synthesis and because our testcross results indicated that NC7A has a recessive pr1 allele, we hypothesized that the pr1 locus would account for most of the variation in levels of the two compounds. In addition, these two parental inbred lines differ in the type of functional p1 allele; GT114, as in Population 1, has a p1-wrb allele whereas NC7A has a p1-wwb allele. We evaluated silks of 316 F₂ plants for maysin and apimaysin concentrations, conducted RFLP analysis on leaf tissue of the same plants, and self-pollinated those plants to produce F_2:3 lines. To date, we have determined genotypes at 53 RFLP loci, located on 18 of 20 chromosome arms, and have completed single-factor analysis based on those data.

As expected, the pr1 locus, as indicated by flanking markers on chromosome 5L, appears responsible for the largest QTL detected for apimaysin, accounting for at least 36% of the phenotypic variation (Table 3). Plants homozygous for the NC7A allele in the pr1 region had a mean apimaysin concentration of 0.30% of fresh weight vs. 0.05% of fresh weight for plants with the GT114 allele. This region did not have a significant corresponding effect on maysin (Table 3). A major QTL for maysin concentration was again detected in the bz1-wx1 interval of chromosome 9S, whereas apimaysin was unaffected by variation in that region (Tables 1 and 3).

Table 3.

The two most significant chromosome regions from a single-factor ANOVA for silk maysin and apimaysin concentrations in the population (GT114 × NC7A)F₂

Chromosome	Locus	Percent maysin		Percent apimaysin
		Significance*	R2†	Significance	R2
5L	bnl5.71 (flanking pr1)	ns		∗∗∗∗∗	36.7
5L	umc126 (flanking pr1)	ns		∗∗∗∗∗	34.9
9S	bz1	∗∗∗∗∗	21.8	ns
9S	wx1	∗∗∗∗∗	21.2	ns

^*ns is not significant at the 0.05 probability level. ∗∗∗∗∗, Significant at the 0.00001 probability level. 

^†Percent phenotypic variance explained by using SAS GLM procedure. 

DISCUSSION

The results of the first population, GT114 × GT119, demonstrated the essential role of the p1 locus in controlling synthesis of maysin and the correlated antibiosis to CEW. The p1 locus encodes the regulatory protein required to activate transcription of structural genes for flavone synthesis. It is clear that, in any population in which the parents differ by having functional and nonfunctional p1 alleles, a substantial fraction of the variation for maysin would be expected to map to the p1 region. Additional support for the primary regulatory role for p1 is seen in its additive gene action; p1 function must be limiting because two doses of active p1 direct the synthesis of approximately twice as much maysin, apparently regardless of genotype at other loci. The additive effect of p1 is interpreted to function by enhancing transcription of rate-limiting structural gene(s) in the flavone pathway. These results indicate that any modification of the expression of the p1 locus will result in quantitative variation in maysin synthesis and in expression of CEW resistance.

No significant variation for maysin was associated with the p1 region for either Population 2 or 3. The parents of Population 2 were chosen specifically because they varied in silk maysin content but had similar p1-wrb alleles. The p1 locus is assumed to be just as important for maysin synthesis in this population, but the locus did not have a significant effect by QTL analysis due to lack of variation at p1. By “holding p1 constant,” we were able to search more effectively for other loci that can contribute to maysin variation in the presence of equivalent functional p1 expression. Although we did not expect a p1 effect in Population 2, we did anticipate that p1 would be significant in Population 3 because of contrasting p1 alleles in the parents (p1-wrb I GT114 and p1-wwb in NC7A). The lack of effect associated with the p1-wwb allele indicates that there is no restriction to using maysin as a CEW control strategy in white cob germplasm, a property that should be of interest to the sweet and food corn industry.

There is generally good correspondence between the loci significant for maysin concentration and for antibiosis to CEW. In Population 1, all loci significant for antibiosis to CEW were also significant for larval weight (17, 22). However, the quantitative contribution of the specific loci to antibiosis is difficult to determine because only a subset of families representing the high and low maysin tails was tested. For Population 2, all 250 F_2:3 families were analyzed for silk maysin concentration and antibiotic effect. Chromosome regions 1S, 2L, 6S, and 6L were significant for both traits in agreement with the primary role of maysin in antibiosis. Additional regions significant for CEW antibiosis, but not for maysin per se, were detected on chromosomes 2, 4, and 10. These regions most likely contain loci controlling synthesis of compounds other than maysin that are toxic to CEW. A large discrepancy between Populations 1 and 2 involves chromosome 9S in the region between bz1 and wx1. In both populations, highly significant differences in silk maysin content were associated with variation at 9S loci, but the 9S region was only significant for antibiosis in Population 1. We don’t understand how this region can have such a large affect on maysin in Population 2 without a corresponding affect on antibiosis.

The most consistent result across all three populations was the presence of a major QTL with recessive gene action for high maysin on chromosome 9S. We currently have three hypotheses to explain the nature of the gene on 9S. (i) As we have previously proposed (22), the locus we detected on 9S is brown pericarp1 and is a structural gene in the 3-deoxyanthocyanin pathway; the homozygous recessive genotype results in a shunting of intermediates into the flavone pathway. (ii) The product of the gene from 9S is a negative regulator (inhibitor) of p1 expression. The homozygous recessive class releases this inhibition and results in enhanced p1 expression and consequently increased transcription of all p1-controlled structural genes. (iii) The product of the 9S locus competes or inhibits P1 protein from binding or activating the structural genes required for flavone synthesis. An analogy for this may be the IN1 protein, which has sequence similarity to MYC class transcription activators and down-regulates many of the structural genes of the anthocyanin pathway (28). Additionally, the MYB class gene zm38 has been shown to inhibit transcription activation of the a1 promoter by C1/R1 in transient expression assays (29). We are designing and conducting gene expression studies to attempt to distinguish among these possible models. Two features of the 9S effect that must be understood are its recessive gene action for high maysin concentration and its epistatic interaction with the p1 locus. We believe that understanding the genetic basis of the 9S effect will provide substantial insight into the nature of epistatic interactions between QTL loci.

Population 3 was designed to study a specific, presumably single, enzyme step in the synthesis of maysin to determine how simple differences in related compounds are manifest as QTLs. Maysin differs from apimaysin only in the presence of a hydroxyl group in the 3′-position of the B ring that is absent in apimaysin (Fig. 1). The region of chromosome 5 containing the pr1 locus had a very large effect on apimaysin concentration but no effect on maysin levels (Table 3). Conversely, we detected a large QTL for maysin on chromosome 9S, but this region had no significant effect on apimaysin concentration. Taken together, these results indicate that the synthesis of maysin and apimaysin are controlled independently. Apimaysin is not made at the expense of maysin, or maysin from apimaysin. How is this possible? The simplest way to view this is the existence of separate pathways for the flow of intermediates to maysin and apimaysin synthesis. This view is consistent with the concept of metabolic channeling via membrane-bound enzyme complexes (30). Once chalcone is formed by chalcone synthase (Fig. 1), it is sequestered in a complex with all of the enzymes required for a specific end product, and the intermediates are unavailable to the action of enzymes in complexes with other end product specificities. In this case, there would be two types of complexes differing in the presence or absence of the hydroxylase encoded by the pr1 locus. These complexes must not be competing for limited substrates for the synthesis of maysin and apimaysin to appear independent. Population 3 provides us with a clear example of how the synthesis of two closely related compounds, whose synthesis must share many of the same enzyme steps, nonetheless can behave genetically independently according to QTL analysis.

Our results suggest that the genetic control of quantitative variation for maysin involves the interaction of at least four factors. The first factor is the level of transcription activation of the pathway by the transcription regulator p1. The second is variation in the activity of enzyme-encoding genes (whp1 and sm1 in Population 2). The third factor is the flow of shared intermediates between distinct but connected pathways. The fourth factor is separation of shared enzyme steps into independently regulated complexes allowing independent synthesis of very similar products. The interaction of these factors allows for fine control of related biochemical products and subtle variation in the corresponding agronomic traits.

ABBREVIATIONS

QTL

quantitative trait locus

CEW

corn earworm

RFLP

restriction fragment length polymorphism