Abstract
Secondary metabolites are a diverse set of plant compounds believed to have numerous functions in plant-environment interactions. Despite this importance, little is known about the regulation of secondary metabolite accumulation. We are studying the regulation of glucosinolates, a large group of secondary metabolites, in Arabidopsis to investigate how secondary metabolism is controlled. We utilized Ler and Cvi, two ecotypes of Arabidopsis that have striking differences in both the types and amounts of glucosinolates that accumulate in the seeds and leaves. QTL analysis identified six loci determining total aliphatic glucosinolate accumulation, six loci controlling total indolic glucosinolate concentration, and three loci regulating benzylic glucosinolate levels. Our results show that two of the loci controlling total aliphatic glucosinolates map to biosynthetic loci that interact epistatically to regulate aliphatic glucosinolate accumulation. In addition to the six loci regulating total indolic glucosinolate concentration, mapping of QTL for the individual indolic glucosinolates identified five additional loci that were specific to subsets of the indolic glucosinolates. These data show that there are a large number of variable loci controlling glucosinolate accumulation in Arabidopsis thaliana.
THE sessile nature of plants forces them to cope directly with environmental changes rather than escape to more favorable sites. Secondary metabolites are a large and diverse set of molecules that allow plants to respond to these environmental challenges, which include insect herbivory, pathogen attack, UV radiation, and drought. There are currently >100,000 known secondary metabolites, which probably represent <10% of the total in nature (Wink 1988). This large chemical diversity presumably arises from a vast set of biosynthetic enzymes. Although numerous studies have investigated the enzymes and regulatory factors controlling biosynthesis of secondary metabolites, little is known about the genetics controlling quantitative and qualitative natural variation in secondary chemistry. To address this question, we are studying glucosinolate production in Arabidopsis thaliana.
Glucosinolates are amino acid-derived thioglycosides. Glucosinolates and their hydrolyzing agent, myrosinase, are spatially separated within plant cells (Bones and Rossiter 1996). When the cell is disrupted, myrosinase cleaves the sugar from the glucosinolate, and a series of toxic compounds are released. These toxins include nitriles, isothiocyanates, oxozaladines, and epithioalkanes. Glucosinolates and their breakdown products have been shown to alter insect herbivory (Siemens and Mitchell-Olds 1996; Stowe 1998; Liet al. 2000), nematode survival (Donkinet al. 1995), and fungal resistance (Mithenet al. 1986; Maniciet al. 1997). Additionally, certain glucosinolates have anticarcinogenic activities in mammals (Verhoevenet al. 1997).
Arabidopsis utilizes a three-part biosynthetic pathway to produce glucosinolates from methionine, tryptophan, and phenylalanine (Hoggeet al. 1988; Halkier and Du 1997). The first step in this pathway is the elongation of a protein amino acid (Figure 1). This is accomplished by the deamination of the amino acid to the 2-oxo acid and subsequent methylene group addition (de Quiroset al. 2000; Graseret al. 2000; Figure 1). The 2-oxo acid can then be reaminated to the elongated amino acid or undergo further elongation (Figure 1). The amino acid enters the general pathway of glucosinolate biosynthesis by conversion to an aldoxime via one of a series of cytochrome P450 oxygenases (Kahnet al. 1999; Wittstock and Halkier 2000). The resulting amino acid is then converted to a glucosinolate by three additional steps. Finally, a number of amino acid-specific side chain modifications can occur (Kliebensteinet al. 2001b; Figure 1). Several of the genes controlling these steps have been mapped and/or cloned in A. thaliana, generating a detailed model of glucosinolate biosynthesis. This provides a solid base from which to interpret natural variation in glucosinolates.
More than 35 different glucosinolates have been documented in the species A. thaliana, with 25 of these coming from differential elongation and side-chain modification of methionine (Mithen and Campos 1996; Kliebensteinet al. 2001b). In contrast to the large number of different glucosinolates present in the species as a whole, the leaves of individual Arabidopsis plants contain only 7–14 different glucosinolates, with 4–11 of these coming from methionine. This qualitative variation in methionine-derived glucosinolates can be explained by natural variation at four biosynthetic loci (Kliebensteinet al. 2001b). In contrast to qualitative variation in Arabidopsis, little is known about the factors controlling quantitative variation in aliphatic, indolic, or benzylic glucosinolates (Kliebensteinet al. 2001b).
Glucosinolate biosynthetic pathway. I, 2-alkylmalic acid; II, 3-alkylmalic acid; III, 2-oxo acid. (A) Basic glucosinolate structure. (B) Outline of the pathway, which can be divided into three parts: (1) elongation of the amino acid side chain; (2) formation of the basic glucosinolate skeleton; and (3) further side chain modification. Each chain elongation cycle adds an additional methylene group (Graseret al. 2000).
To identify loci controlling quantitative variation in Arabidopsis glucosinolates we initiated a quantitative trait locus (QTL) mapping experiment using the Landsberg erecta (Ler) × Cape Verdi Islands (Cvi) recombinant inbred lines (RILs; Alonso-Blancoet al. 1998). Previous studies examining glucosinolates QTL focused on the accumulation of seed glucosinolates in Brassica napus. These studies identified four to six QTL that controlled aliphatic glucosinolate concentration (Toroseret al. 1995; Uzunovaet al. 1995). However, little is known about the genes underlying these QTL or about QTL that control nonaliphatic glucosinolates or leaf glucosinolates. In this article, we report on QTL mapping experiments studying aliphatic, indolic, and benzylic glucosinolate concentration in both the leaves and seeds of the A. thaliana Ler × Cvi RILs (Alonso-Blancoet al. 1998). These QTL are discussed in relation to previously described Arabidopsis glucosinolate biosynthetic loci (de Quiroset al. 2000; Kliebensteinet al. 2001a). We also investigate the coregulation of different glucosinolate classes in the two different tissues. Finally, we describe the identification of QTL that control the accumulation of specific glucosinolates.
MATERIALS AND METHODS
Plant growth: All seed stocks were obtained from the Arabidopsis Biological Resource Center (http://aims.cps.msu.edu/aims/). Five plants per Ler × Cvi recombinant inbred lines were planted in a 6-cm pot containing a standard vermiculite/soil mix. The pots were placed in flats of 48 in a randomized design with one pot of Ler and one pot of Cvi per flat. This was repeated three times with independently randomized designs. The plants were allowed to grow at 26° for 3 weeks postgermination. They were placed 10 in. from four 60-W Cool White Delux GE lightbulbs in a 12-hr light/12-hr dark photoperiod. At this time, 10 leaves were harvested for glucosinolate extraction. The same plants were allowed to continue to complete seed maturation. The seeds were then combined from all the plants in each pot and kept separate from the seeds harvested from the other plantings. Seed glucosinolates were analyzed independently from each of the three replicates per recombinant inbred line and from the 10 replicates of Ler and Cvi parental lines.
Sample preparation and HPLC: Samples were extracted and analyzed by HPLC as previously described (Kliebensteinet al. 2001a). Specific glucosinolates were identified by comparison of retention times and UV absorption spectra with purified standards. All glucosinolate absorption data (measured at 229 nm) were converted to micromoles per gram dry weight (DWT) using response factors determined from the purified standards for each of the glucosinolates (Kliebensteinet al. 2001a).
Glucosinolate traits: Glucosinolate traits were generated using data on individual compounds. TALIPH is the sum of all methionine-derived (aliphatic) glucosinolates. TIND is the sum of all tryptophan-derived or indolic glucosinolates. Benzyl is the sum of all phenylalanine-derived or benzyl glucosinolates. MO to I3M = (1-methoxyI3M + 4-methoxyI3M)/I3M.
QTL mapping and statistics: The least-squares mean for each recombinant inbred line for the Ler × Cvi leaf and seed samples was obtained from Systat, utilizing each of the three replicates in an analysis of variance using flat as a categorical factor to control for possible heterogeneous environmental conditions. Relative glucosinolate concentrations for all 162 lines were then used for QTL mapping, utilizing QTLcartographer version 1.13 with both linear interval mapping and composite interval mapping (Bastenet al. 1999). We determined significance thresholds for each trait by doing 500 permutations to estimate the 0.05 significance level. Marker data employed 99 amplified fragment length polymorphism (AFLP) markers from Alonso-Blanco et al. (1999) and 5 microsatellite markers that were previously described (Kliebensteinet al. 2001a). Mapping of the markers for both RI lines employed Mapmaker version 3 (Landeret al. 1987).
For Table 2, heritability was estimated by ANOVA analysis.
The R2 was determined from two different ANOVA models:
Glucosinolate = constant + line + flat
Glucosinolate = constant + flat
The heritability of the trait due to line effects was then estimated as the difference in R2 values obtained between the two models (
For Tables 3 and 4, epistatic interactions among QTL were analyzed by ANOVA using the genetic markers showing the largest statistical control for each QTL. The models used are listed below:
Table 3: Mean leaf aliphatic glucosinolates = constant + GS-AOP + EC198L + GS-ELONG + GS-AOP × GS-ELONG + EC198L × GS-ELONG + GS-AOP × EC198L + GS-AOP × EC198L × GS-ELONG
Table 4: Mean seed benzylic glucosinolate = constant + GS-ELONG + AD182C + GB120C + AD182C × GS-ELONG + AD182C × GB120C + GS-ELONG × GB120C
RESULTS
Qualitative aliphatic glucosinolate variation in the Ler × Cvi RILs: The Ler and Cvi A. thaliana ecotypes accumulate different glucosinolate types. Ler leaves contain predominantly 3-hydroxypropyl and 8-methylsulfinyloctyl glucosinolate while Cvi leaves have principally allyl and 3-butenyl glucosinolate (Table 1 and Figure 2). This difference in glucosinolate profile is controlled by the previously described GS-AOP and GS-Elong loci (Kliebenstein et al. 2001a,b). GS-AOP controls the difference between hydroxyalkyl (i.e., hydroxypropyl) and alkenyl (i.e., allyl and 3-butenyl) glucosinolates while the GS-Elong locus controls C3 (i.e., allyl and hydroxypropyl) vs. C4 (i.e., 3-butenyl) glucosinolate production (Figure 3). These loci are semidominant because the glucosinolate profile of the F1 progeny from a Ler × Cvi cross contains all glucosinolates present in both parents (Figure 2). About one-half of the Ler × Cvi RILs display glucosinolate profiles similar to either the Ler or Cvi parents (Figure 3). However, genetic segregation has generated three new glucosinolate profiles that are not found in either parent (Figure 3). For example, 2-hydroxy-3-butenyl glucosinolate is not found in either parent (Table 1 and Figure 3). This new glucosinolate is produced by the introduction of a hydroxyl group at the C-2 of 3-butenyl glucosinolate (Rossiteret al. 1990). As Cvi is incapable of this reaction, Ler appears to contain an enzyme able to catalyze this reaction, although it does not contain the 3-butenyl precursor (Table 1 and Figure 2). In comparison to leaves, Cvi seeds accumulate low levels of 3-hydroxypropyl and 4-hydroxybutyl glucosinolate in addition to allyl and 3-butenyl glucosinolate (Table 1). Further, seeds of both ecotypes contain high levels of benzoic acid ester derivatives of all hydroxyalkyl and hydroxyalkenyl glucosinolates (Table 1; Kliebensteinet al. 2001b).
Identified glucosinolates
Quantitative glucosinolate variation between Ler and Cvi: In addition to the qualitative variation discussed above, Ler and Cvi exhibit quantitative glucosinolate variation (Figure 2 and Table 2). Cvi leaves accumulate nearly 8-fold more total aliphatic glucosinolates than Ler leaves (Figure 2 and Table 2). Leaves of Ler × Cvi F1 progeny contain intermediate glucosinolate levels (Figure 2). In addition to the large difference between the parents, the RILs displayed a 39-fold difference between the high and low lines for leaf aliphatic glucosinolates (Table 2). Thus, both ecotypes appear to have increasing and decreasing alleles, which allow for transgressive segregation in total quantities of glucosinolates.
Whereas Ler and Cvi had significantly different leaf aliphatic glucosinolate levels, they have nearly identical seed aliphatic glucosinolate accumulation (Figure 2 and Table 2). In contrast to the similarity between the two parents, the high and low Ler × Cvi RILs had a nearly threefold difference in total seed aliphatic glucosinolate concentration (Table 2). This indicates that the similar levels of aliphatic glucosinolates in seeds hide genetic differences between Ler and Cvi for the control of total aliphatic seed glucosinolates.
HPLC chromatograms of compounds present in the leaves of Ler, Cvi, and F1 progeny. Shown are HPLC chromatograms monitored at 229 nm of samples from Ler, Cvi, and Ler × Cvi F1 progeny prepared from equal amounts of 2-week-old rosette tissue. The major glucosinolate peaks are labeled at the top of the chromatogram.
In contrast to methionine-derived aliphatic glucosinolates, where Cvi is higher than Ler, Ler has 2- to 3-fold higher levels of both indolic and benzylic glucosinolates in all tissues (Figure 2 and Table 2). The F1 progeny showed that leaf indolic glucosinolate accumulation was not under the control of a single dominant or recessive locus (Figure 2). Further, the RILs display a striking amount of transgressive segregation for all three traits. This ranges from a 34-fold difference between high and low lines for seed benzylic glucosinolate accumulation to a 13-fold difference between high and low RILs for leaf indolic glucosinolate concentration (Table 2). Thus, there is substantial genetic variation for aromatic glucosinolate accumulation.
The heritability for most traits is high, with leaf total aliphatic glucosinolates being the highest at 0.81 (Table 2). This suggests that genetic factors segregating between the lines control ~80% of the variation seen in the experiment. In contrast, seed total aliphatic glucosinolates had a heritability of only 0.40 (Table 2). This indicates that genetics contributes ~40% of the variation while environmental and experimental factors contribute 60% of the variation. Even compounds in low abundance like 8-methylthiooctyl in the leaves (see Figure 2) had a heritability of 0.63. Thereby, the detection of even low-concentration compounds is highly repeatable and their accumulation is under genetic control.
Coregulation of glucosinolate accumulation between leaves and seeds: Previous work showed a significant correlation between the levels of leaf and seed aliphatic glucosinolates in 35 different Arabidopsis ecotypes (Kliebensteinet al. 2001b). We further tested this by examining the correlation of leaf and seed aliphatic glucosinolate concentration in the leaves and seeds of the 162 Ler × Cvi RILs. This analysis showed a positive correlation between aliphatic glucosinolates in the two different tissues (Figure 4, r = 0.19, P = 0.015, n = 162). Further, concentration of aliphatic glucosinolates was always higher in seeds than in leaves. This suggests that the glucosinolate biosynthesis in the leaf may determine the lower limit for seed aliphatic glucosinolate accumulation. Support for this theory comes from the observation that 35 Arabidopsis ecotypes displayed a similar relationship between leaf and seed glucosinolates (Kliebensteinet al. 2001b). Additionally, there was a positive correlation between indolic glucosinolate accumulation in the two different tissues (Figure 4, r = 0.40, P < 0.001, n = 162).
Mapping QTL controlling total aliphatic glucosinolate accumulation: To investigate the genetic control of quantitative variation in Arabidopsis glucosinolates, we utilized the mean total leaf aliphatic glucosinolate concentration for each of the 162 RILs to map QTL. This analysis identified three QTL that controlled 68% of the variation in leaf aliphatic glucosinolate levels (Figure 5). Two of these QTL are linked with the previously known glucosinolate biosynthetic loci, GS-Elong and GS-AOP. This was expected, as both loci were previously shown to control the concentration of aliphatic glucosinolates (Mithenet al. 1995; Kliebensteinet al. 2001a). Fine-scale mapping showed that the GS-Elong and GS-AOP QTL have LOD maxima at markers that are physically located within the GS-Elong and GS-AOP genes (D. J. Kliebenstein and T. Mitchell-Olds, unpublished data; de Quiroset al. 2000; Kliebensteinet al. 2001a). The second strongest QTL controlling leaf aliphatic glucosinolates is linked with the AFLP marker EC198L and is a previously unidentified Arabidopsis glucosinolate locus. In contrast to GS-Elong and GS-AOP, which control both the type and amount of aliphatic glucosinolate produced, there is no correlation between EC198L and the presence or absence of any specific glucosinolate (D. J. Kliebenstein and T. Mitchell-Olds, unpublished data). This suggests that the EC198L QTL is a single locus or several closely linked loci that controls the production of leaf aliphatic glucosinolates irrespective of type.
Genetic control of glucosinolate profiles in Ler × Cvi RILs. Shown is a flow chart of how the three variable glucosinolate biosynthetic loci (GS-AOP, GS-Elong, and GS-OH) alter accumulation of the predominant aliphatic glucosinolates in leaves of Ler × Cvi RILs. Also listed is the number of RILs displaying each pattern of glucosinolate accumulation. Only the 149 lines with sufficent aliphatic glucosinolate levels to accurately score the genotype are shown. Seed glucosinolates are similar except that hydroxy glucosinolates are converted to benzoyloxy glucosinolates and 4-methylsulfinylbutyl glucosinolate is converted to 4-hydroxybutyl and 4-benzoyloxybutyl glucosinolate.
To compare the factors regulating leaf and seed aliphatic glucosinolates, we mapped QTL controlling the accumulation of seed aliphatic glucosinolates. This analysis also uncovered three QTL, which were all seed specific (Figure 5). Together, these QTL explain 38% of the variation in seed aliphatic glucosinolate concentration. An analysis of the presence of specific glucosinolates showed that the QTL near Erecta on chromosome II is tightly linked with the presence of 2-hydroxy-3-butenyl glucosinolate (2.5 cM between Erecta and the 2-hydroxy-3-butenyl locus). This suggests that this locus may regulate the formation of 2-hydroxy-3-butenyl glucosinolate in addition to regulating total seed aliphatic glucosinolate levels. In contrast, the QTL associated with GD86L on chromosome I and BF168L on chromosome V did not correlate with the presence or absence of any specific glucosinolates.
Epistasis of QTL controlling aliphatic glucosinolate concentration: The observation that two of the three QTL controlling leaf aliphatic glucosinolate levels were linked with glucosinolate biosynthetic loci suggested that these loci might interact to control leaf aliphatic glucosinolate accumulation. We utilized ANOVA to test this possibility. This showed that GS-Elong and GS-AOP display a large and highly significant interaction controlling leaf aliphatic glucosinolate accumulation (Table 3 and Figure 6). This interaction is such that the presence of a Ler homozygote at GS-Elong or GS-AOP functions to decrease leaf aliphatic glucosinolate levels (Figure 6). In addition to interacting with GS-AOP, GS-Elong also interacts with the EC198L QTL (Table 3 and Figure 6). The same pattern of interaction is present with the RILs containing the Cvi alleles at GS-Elong and EC198L, having about twofold higher glucosinolate concentrations than the RILs with the other three allele combinations (Figure 6). In addition to the two-way interactions, GS-Elong, GS-AOP, and EC198L display a significant three-way interaction (Table 3). These interactions suggest that GS-Elong, GS-AOP, and EC198L all control leaf aliphatic glucosinolate concentration via the same pathway or mechanism. As GS-Elong and GS-AOP are biosynthetic loci, this suggests that EC198L is also somehow altering glucosinolate biosynthesis.
Means for total glucosinolate traits
Correlation between leaf and seed glucosinolate concentration. Scatter plots are shown comparing the accumulation of leaf and seed glucosinolates. (A) The relationship between aliphatic glucosinolate in the leaf (LTALIPH) and seed (STALIPH). (B) The relationship between indolic glucosinolate in the leaf (LTIND) and seed (STIND).
QTL controlling total indolic glucosinolate accumulation: Indolic glucosinolates are produced from tryptophan whereas all aliphatic glucosinolates in Arabidopsis are produced from methionine (Halkier and Du 1997). This fundamental difference suggests that their accumulation may be independently regulated. This theory is supported by the absence of a correlation between aliphatic and indolic glucosinolate levels in either leaves or seeds (leaf, n = 162, r = 0.01, P = 0.19; seed, n = 162, r = 0.01, P = 0.87). To further test this theory, we mapped QTL controlling the accumulation of total leaf and seed indolic glucosinolate concentration. This showed that there was no overlap between QTL controlling indolic and aliphatic glucosinolate accumulation (Figures 5 and 7). Thus, the two classes appear to be independently regulated in this cross.
QTL controlling aliphatic glucosinolate concentration in the leaves and seeds of the Ler × Cvi RILs. LOD plots of QTL for the accumulation of total aliphatic glucosinolates are shown. Total aliphatic glucosinolate accumulation was determined by summing all of the individual methionine-derived glucosinolates. Numbers are the percentages of difference in mean glucosinolate concentration between lines homozygous for the Cvi allele and lines homozygous for the Ler allele at the listed markers. The mean from the homozygous Ler allele is always utilized as the denominator. The horizontal line indicates the 0.05 significance threshold. Note the different LOD scales on the chromosomes.
The observation of a positive correlation between leaf and seed indolic glucosinolates suggested that some genes influence indolic glucosinolate levels in both tissues. Of the six identified QTL, only the DF119L QTL controls both leaf and seed indolic glucosinolate levels (Figure 7). The other five QTL appear to be tissue specific. As expected from the observation that Ler contains significantly higher indolic glucosinolate levels in both tissues, the major QTL were such that the Ler allele led to higher indolic glucosinolates. No epistasis was identified among any of the indolic QTL.
Interactions of QTL controlling leaf aliphatic glucosinolate concentration
QTL for benzylic glucosinolates: The third glucosinolate class in Arabidopsis is the benzyl glucosinolates derived from phenylalanine, which are predominantly seed specific (Table 1). In comparison to the other two classes, very little is known about the quantitative or qualitative variation present in this class of glucosinolates. To investigate the genetic control of quantitative variation for the benzyl glucosinolates, we mapped QTL controlling the accumulation of benzyl glucosinolates in the seed. Strikingly, the major QTL controlling the accumulation of seed benzyl glucosinolate is linked to the GS-Elong locus that controls total leaf aliphatic glucosinolates (Figure 8). Our analysis of the RILs identified two different benzyl glucosinolates, which appear to differ in their carbon chain length (Kliebensteinet al. 2001b). Thus, it is possible that GS-Elong or a nearby locus controls the chain elongation of phenylalanine and this chain elongation leads to differences in accumulation. Further work is required to test if GS-Elong-controlled chain elongation is involved in benzyl glucosinolate biosynthesis. The other two QTL, near GB120C and AD182C, did not overlap with QTL from the other glucosinolate classes (Figure 8). Analysis of QTL interactions showed that there is a significant two-way interaction between AD182C and GS-Elong, suggesting that these loci control benzyl glucosinolate accumulation via the same mechanism (Table 4).
QTL for specific indolic glucosinolates: The variable used in mapping QTL for total leaf indolic glucosinolate concentration is the sum of the indolyl-3-methyl (I3M), 1-methoxy-indolyl-3-methyl (1MO-I3M), and 4-methoxy-indolyl-3-methyl (4MO-I3M) glucosinolates. However, seeds do not contain 4MO-I3M (Table 1). Mapping QTL controlling total indolic GS might obscure genetic segregation controlling the accumulation of specific glucosinolates. To test for this possibility we analyzed the individual indolic glucosinolate data for QTL and compared these QTL to those obtained using the composite trait.
This analysis showed that the use of summation variables concealed QTL controlling specific glucosinolates and obscured the effects of known QTL on specific glucosinolates (Figure 9). An example is the FD167L QTL that controls total leaf indolic glucosinolate concentration. Analyzing the specific glucosinolate concentrations showed that, while the FD167L Ler allele increases total indolic glucosinolate levels, this allele only increases the level of I3M (Figure 9). In contrast the FD167L Ler allele leads to decreased 4MO-I3M levels and has no effect on 1MO-I3M. This suggests that FD167L controls the ratio of the precursor I3M to the product 4MO-I3M and this affects the total level of indolic glucosinolates in the leaves.
In addition to hiding the effect of QTL influencing specific glucosinolates, the utilization of the summation variable for QTL mapping also hid glucosinolate-specific QTL. For example, a region at the top of chromosome I controls the accumulation of 1MO-I3M in the seed but has no effect on the total concentration of indolic glucosinolates in the seed. Similar QTL were found for 4MO-I3M in the leaf, while no I3M-specific QTL were identified. Thus, while the summation variable is a good tool to analyze total quantitative variation, it can be misleading when interpreting the quantitative variation of specific individual glucosinolates.
DISCUSSION
Genetics of glucosinolate quantitative variation in A. thaliana: Previous research into the genetics of quantitative variation in glucosinolates had investigated the accumulation of only one class of glucosinolates in one specific tissue type (Toroseret al. 1995; Uzunovaet al. 1995). To better understand the control of glucosinolate quantitative variation we examined genetic control of the accumulation of three different glucosinolate classes (aliphatic, indolic, and benzylic) in two different tissues (leaves and seeds) of A. thaliana. This analysis showed that there were a number of QTL controlling the accumulation of each of the three different classes (Figures 5, 7, and 8). Further, most of the QTL were specific for the accumulation of a specific glucosinolate class in a specific tissue type (Figures 5, 7, and 8). This large number of QTL suggests that it will be possible to utilize natural variation in A. thaliana to develop an in-depth understanding of glucosinolate biosynthesis.
Allele-specific interactions of QTL controlling leaf aliphatic glucosinolates. Bar diagrams show the average leaf aliphatic glucosinolate accumulation. Allelic states of the loci are listed below each bar. (A) Interaction of GS-Elong and GS-AOP. (B) Interaction of GS-Elong and EC198L.
Tissue specificity of QTL: The strong correlation between leaf and seed aliphatic glucosinolates suggests overlapping regulation between the two tissues. However, there was no overlap in QTL controlling aliphatic glucosinolate accumulation in these tissues (Figure 5). This suggests that there are no segregating global regulators of aliphatic glucosinolate concentration in this cross. It also suggests that the quantitative effects of GS-AOP and GS-Elong are tissue specific, whereas their qualitative effects are not tissue specific (Figure 5 and Table 1). The tissue specificity of GS-AOP could be explained by a difference in the aliphatic glucosinolates that accumulate in the two tissues. The predominant leaf aliphatic glucosinolates are the C3 and C4 alkenyl, hydroxyalkyl, or methylsulfinylalkyl glucosinolates, which are determined by the AOP and Elong enzymes (de Quiroset al. 2000; Kliebensteinet al. 2001a). In contrast, the predominant seed aliphatic glucosinolates are a mixture of C7 and C8 methylthioalkyl glucosinolates and C3 and C4 benzoyloxyalkyl glucosinolates. Thus, the impact of the individual glucosinolates derived from the AOP and Elong reactions would be reduced in the seeds in comparison to the leaves. This could explain the tissue specificity of these QTL.
QTL controlling indolic glucosinolate concentration in the leaves and seeds of the Ler × Cvi RILs. LOD plots of QTL controlling the accumulation of total indolic glucosinolates are shown. Total indolic glucosinolate accumulation was determined by summing all of the individual tryptophan-derived glucosinolates. Numbers are the percentages of difference in mean glucosinolate concentration between lines homozygous for the Cvi allele and lines homozygous for the Ler allele at the listed markers. The mean from the homozygous Ler allele is always utilized as the denominator. The horizontal line indicates the 0.05 significance threshold. Note the different LOD scales on the chromosomes.
Comparison of QTL for total glucosinolate levels and for individual glucosinolate levels: A number of glucosinolate QTL studies measure only total glucosinolate concentrations or sum together all the glucosinolates and then map only the composite trait (Mithen and Toroser 1995; Toroseret al. 1995). These summation approaches have the possibility of hiding QTL for individual glucosinolates and hiding the impact of differential effects of the identified QTL on individual glucosinolates. To test for both of these possibilities, we compared the QTL controlling total indolic glucosinolate accumulation to QTL for individual glucosinolates. This showed that analysis of summed total glucosinolates obscured some QTL, which controlled the accumulation of specific glucosinolates. Further, some of the QTL that influenced the total accumulation of indolic glucosinolates actually had differential effects on the accumulation of individual glucosinolates. Thus, to be thorough, it is best to map QTL for all of the individual variables as well as total glucosinolate concentration.
QTL controlling seed benzyl glucosinolate concentration in the Ler × Cvi RILs. LOD plots of QTL controlling the accumulation of total benzylic glucosinolates are shown. Total benzylic glucosinolate accumulation was determined by summing the values for both phenylalanine-derived glucosinolates. Numbers are the percentages of difference in mean glucosinolate concentration between lines homozygous for the Cvi allele and lines homozygous for the Ler allele at the listed markers. The mean from the homozygous Ler allele is always utilized as the denominator. The horizontal line indicates the 0.05 significance threshold. Note the different LOD scales on the chromosomes.
Leaf aliphatic glucosinolates determine the base of seed aliphatic glucosinolate accumulation: Previous work has shown that glucosinolates can be transported from the leaves to the developing siliques (Brudnellet al. 1999). Further, siliques are also known to have an endogenous glucosinolate biosynthetic capacity (Du and Halkier 1998). Thus, the concentration of seed glucosinolates may depend upon a combination of transport from the leaves and biosynthesis within the silique. If there were no variation in glucosinolate transport from the leaves to the seeds, the level of glucosinolates in the leaves would represent the minimal seed glucosinolate concentration. This level could then be enhanced by de novo biosynthesis in the silique. Analysis of leaf and seed aliphatic glucosinolate accumulation in the 162 Ler × Cvi RILs showed a significant positive correlation (Figure 4). Interestingly, there appears to be a threshold effect to this correlation, with the aliphatic leaf glucosinolates representing a basal concentration of aliphatic seed glucosinolates (Figure 4). Further, a study investigating the accumulation of aliphatic leaf and seed glucosinolates in 35 Arabidopsis ecotypes showed a similar positive correlation, with aliphatic leaf glucosinolates acting as a lower boundary on seed glucosinolates (Kliebensteinet al. 2001b).
Interactions of QTL controlling seed benzylic glucosinolate concentration
Side-chain modification and regulation of glucosinolate concentration: The major QTL controlling the accumulation of leaf glucosinolates in the Ler × Cvi RILs was tightly linked to the previously cloned glucosinolate biosynthetic locus, GS-AOP (Figure 5, Table 3). This locus is the result of a tandem gene duplication where differential expression of one gene produces alkenyl glucosinolates (AOP2) or the other gene generates hydroxyalkyl glucosinolates (AOP3; Kliebensteinet al. 2001a). Ler expresses the AOP3 gene and produces hydroxypropyl glucosinolate while Cvi expresses the AOP2 gene to produce alkenyl glucosinolates (Figures 2 and 3). The enzymes encoded by these genes function to modify the side chains of aliphatic glucosinolates at the end of the glucosinolate biosynthetic pathway (Figure 1). These data suggest that alternative GS-AOP haplotypes cause large differences in leaf aliphatic glucosinolate concentration. One possible explanation is that the GS-AOP locus controls glucosinolate biosynthesis by regulating metabolic flux through the pathway and that the enzyme encoded by AOP2 has a higher enzymatic rate than that encoded by AOP3. An alternative explanation is that hydroxyl glucosinolates may be more effective feedback inhibitors than alkenyl glucosinolates. We cannot currently differentiate between these alternatives. Finally, the mapping data reported here cannot exclude the possibility that another tightly linked locus is responsible for the variation in glucosinolate accumulation.
Compound specificity of indolic glucosinolate QTL. Diagrams show the location and strength of QTL for the traits denoted by the columns. The QTL are shown as arrows to the right of the chromosomal map. The total percentage of the variation explained by the QTL for the given trait is listed at the bottom of each column. Chromosome III is not shown as it did not contain any QTL. The arrow height represents the range of 0.05 significance for the QTL. S represents seed glucosinolates and L represents leaf glucosinolates. MO to I3M = (1-methoxy-I3M + 4-methoxy-I3M)/I3M.
Effect of methionine elongation on aliphatic glucosinolate accumulation: Arabidopsis contains aliphatic glucosinolates with differing carbon lengths, which are produced by a common elongation pathway (Hoggeet al. 1988; de Quiroset al. 2000; Kliebensteinet al. 2001b). One locus, GS-Elong, has been identified that controls C3 vs. C4 polymorphism (de Quiroset al. 2000). Previous work had shown that this locus was a major determinant of glucosinolate concentration in seeds from the Ler × Col RILs of Arabidopsis (Mithen and Campos 1996). In our analysis of the Ler × Cvi RILs, the GS-Elong locus exerted a significant influence on leaf but not on seed aliphatic glucosinolate concentrations (Figure 5). This suggests that the quantitative effect of GS-Elong is either dependent upon the genetic background in which it occurs or that the C4 alleles present in Cvi and Col are somehow different. The hypothesis that GS-Elong is dependent upon the genetic background is supported by the observation that there is an epistatic interaction between GS-Elong and both GS-AOP and EC198L (Figure 6). Further work is required to understand how the C3/C4 biochemical polymorphism controls total accumulation of aliphatic glucosinolates.
Independent regulation of glucosinolates derived from different amino acids: The accumulation of glucosinolates derived from different amino acids could be coregulated because they utilize the same enzymatic reactions in the formation of their basic glucosinolate skeleton (Figure 1; Halkier and Du 1997). However, in the Ler × Cvi RILs there is no correlation between the production of indolic, aliphatic, and benzylic glucosinolates. This agrees with previous analysis of Arabidopsis ecotypes, which also showed a lack of correlation between glucosinolates produced from different amino acids (Kliebensteinet al. 2001b). Further, only one QTL, GS-Elong, appears to coregulate the accumulation of two different glucosinolates and it is unknown if this is due to the action of a single locus or closely linked loci (Figures 5, 7, and 8). This suggests that the enzymes involved in biosynthesis are either amino acid specific or do not regulate glucosinolate accumulation. The cytochrome P450 monooxygenases are known to have specificity for certain amino acids and could thereby explain some of this independent regulation (Kahnet al. 1999; Wittstock and Halkier 2000). In contrast, sulfotransferase and glucosyltransferase have minimal side-chain specificity and might thereby have been expected to lead to glucosinolate coregulation (Halkier and Du 1997). However, this is not the case in this study.
Our study shows that glucosinolate concentration and type are regulated by a large number of naturally varying genetic loci. This information provides precise genetic tools to manipulate both glucosinolate type and amount and test their relative biological effects. Further, the observation that glucosinolate type and concentration are highly variable in Arabidopsis suggests that breakdown type and amount may be variable as well. If this is the case, then it may be possible to utilize Arabidopsis genetics to compare the relative impacts of glucosinolate type and amount and breakdown product type and amount upon plant fitness in a single species.
Acknowledgments
We acknowledge Antje Figuth for technical support in extracting all glucosinolates, Deana Pedersen for aid in plant growth and tissue collection, Michael Reichelt for assistance in compound identification, and Jonathan Gershenzon for editorial advice. This work was funded by the Max-Planck Gesellschaft and grant DEB-9527725 from the U.S. National Science Foundation.
Footnotes
-
Communicating editor: V. Sundaresan
- Received February 21, 2001.
- Accepted July 3, 2001.
- Copyright © 2001 by the Genetics Society of America
