Methylthioalkylmalate synthases: genetics, ecology and evolution
- 749 Downloads
Glucosinolates display an enormous amount of structural variation, both within and between species. This diversity is thought to have evolved in response to challenges imposed on plants by their biotic environment. During the past decade, glucosinolates and myrosinase-catalyzed glucosinolate hydrolysis have become excellent examples for understanding functional diversification in plant secondary metabolism and plant defence. Methylthioalkylmalate (MAM) synthase genes and enzymes are central to the diversification of aliphatic glucosinolate structures in Arabidopsis thaliana and related plants. This review summarizes efforts to elucidate how MAM-mediated diversity in aliphatic glucosinolate structures is generated and maintained. It also attempts to put variability in methionine carbon chain elongation during glucosinolate biosynthesis into an ecological and evolutionary context.
KeywordsComplex traits Evolutionary dynamics Glucosinolate metabolism Natural variation Plant–insect interactions
Quantitative trait locus/loci
Recombinant inbred line
Glucosinolates and myrosinases are almost exclusively found in plants from the order Capparales. This order consists of more than a dozen families, including the Brassicaceae and the Capparaceae (Rodman 1991a, b; Rodman et al. 1996). More than 130 different glucosinolates have been characterized; however glucosinolate composition varies remarkably between and within species (Daxenbichler et al. 1991; Fahey et al. 2001; Kliebenstein et al. 2001a; Windsor et al. 2005; Heidel et al. 2006). In Arabidopsis thaliana, nearly 40 different glucosinolates have been identified (Kliebenstein et al. 2001a; Reichelt et al. 2002). These glucosinolates are generated from methionine, tryptophan or phenylalanine, with methionine-derived (aliphatic) glucosinolates being the predominant glucosinolate class represented in A. thaliana.
MAM substrate specificities in glucosinolate biosynthesis
A. thaliana MAM1
2C, 3C, (4C)
1, 2, (3)
A. thaliana MAM2
Benderoth et al. (2006)
A. thaliana MAM3
A. lyrata MAMa
Benderoth et al. (2006)
B. stricta MAMa
Cloning of the MAM QTL
The widely used A. thaliana accession Landsberg erecta (Ler) accumulates homomethionine-derived (3C) glucosinolates. In contrast, aliphatic glucosinolates in Columbia (Col-0), recognized for having provided the blueprint of the Arabidopsis genome (Arabidopsis Genome Initiative 2000), originate mainly from dihomomethionine (4C). This biochemical difference between Col-0 and Ler enabled initial mapping of the responsible genetic locus to a region of approximately 140 kb on chromosome 5 (Magrath et al. 1994; Campos de Quiros et al. 2000). Within this region, two tandemly arranged genes were identified as candidates, based on their sequence similarity with isopropylmalate synthase (IPMS) genes. IPMS catalyzes the condensation of 2-oxoisovalerate with acetyl-CoA to form isopropylmalate in leucine biosynthesis, a reaction similar to the MAM-catalyzed condensation of ω-methylthio-2-oxoalkanoic acids with acetyl-CoA in glucosinolate biosynthesis. The two candidate genes were termed MAM1 and MAM-L (meanwhile often referred to as MAM3), respectively (Kroymann et al. 2001).
In the Col-0 accession, MAM1 and MAM3 are separated by ca. 11.5 kb of intervening sequence. High-resolution mapping was employed to separate functional effects of these candidate genes. Col-0 was crossed with CL5, an RIL from the Col-0 × Ler population (Lister and Dean 1993). This particular line was chosen because it shared approximately 70% of its genome with the Col-0 accession but has the Ler allele at the MAM locus, thus enabling fine-mapping with near-isogenic lines (NILs). In Col-0 × CL5 F2 progeny one line was identified that had recombined between MAM1 and MAM3. This recombinant line had the Col-0 MAM1 genotype and was heterozygous at MAM3. Its glucosinolate phenotype resembled the parental Col-0 profile closely, with 4C glucosinolates predominating. Likewise, progeny from this line produced mainly 4C glucosinolates when they had the Col-0 MAM1 and the LerMAM3 genotype. Hence, the biochemical difference in short-chain aliphatic glucosinolate composition was attributable to MAM1 and not MAM3 (Kroymann et al. 2001). Further evidence for the role of MAM1 in methionine carbon chain elongation was obtained with MAM1 mutants, deficient in dihomomethionine-derived glucosinolates (Haughn et al. 1991), and with biochemical assays of heterologously expressed MAM1, which showed that the encoded protein has the capacity of condensing ω-methylthio-2-oxoalkanoic acids with acetyl-CoA (Kroymann et al. 2001; Textor et al. 2004; Benderoth et al. 2006).
Origin of MAM genes
Several groups of enzymes catalyze condensation reactions between 2-oxo acids and acetyl-CoA. These enzymes belong to enzyme class EC 2.3.3.-. Examples are citrate synthase (EC 188.8.131.52), which condenses oxaloacetate with acetyl-CoA in the TCA cycle, malate synthase (EC 184.108.40.206), which is responsible for the condensation of glyoxylate with acetyl-CoA in the glyoxylate cycle, and 2-IPMS (EC 220.127.116.11) involved in leucine biosynthesis.
Variability in the genetic composition of the MAM locus in A. thaliana
Sequence exchange between MAM1 and MAM2 genes is another factor contributing to variability in the gene composition of the MAM locus. In some accessions, sequence transfer has occurred from MAM2 to MAM1, in others sequence portions have been shifted from MAM1 to MAM2. In the most extreme cases, represented by two accessions from Tajikistan (Hodja and Condara), the gene at the position originally occupied by MAM1 has been almost completely converted to a MAM2-like gene. However, another accession from Tajikistan, Sorbo, was not affected by gene deletion or conversion events. This accession has functional MAM2 and MAM1 genes and accumulates 4C glucosinolates, indicating that MAM1 overrides MAM2 function. Likewise, all accessions with at least one functional MAM1-like gene accumulate short chain aliphatic glucosinolates generated from dihomomethionine, while accessions without a functional MAM1 produce short-chain aliphatic glucosinolates almost exclusively from homomethionine. Taken together, these data show that the MAM cluster in A. thaliana is subject to dynamic evolutionary change.
Quantitative effects of the MAM1/MAM2 polymorphism on glucosinolate profiles
Even though MAM1 and MAM2 lack the capacity to catalyze the condensation reactions in advanced methionine carbon chain elongation cycles (Benderoth et al. 2006), the QTL peaked in all cases at an interval containing MAM1 or MAM2, but not MAM3. Hence, the MAM1/MAM2 polymorphism influenced not only aliphatic glucosinolates with short carbon chains but also affected glucosinolates with long carbon chains. The biochemical basis for this effect of MAM1 versus MAM2 on long-chain aliphatic glucosinolate accumulation is not yet understood.
A complex influence of the MAM locus on glucosinolate profiles was also seen in studies with another Arabidopsis RIL population, Bay-0 × Sha (Kliebenstein et al. 2006; Wentzell et al. 2007). eQTL mapping and network analysis (further explained in Kliebenstein 2008) suggest that this influence is, to a certain degree, caused by changes in transcript levels.
Ecological consequences of the MAM1/MAM2 polymorphism
Two lepidopteran insects, Spodoptera exigua and Plutella xylostella, were tested for differences in larval performance contingent on the genotype at the MAM locus in Col-0 × CL5 NILs. S. exigua is referred to as a generalist due to possession of a broad host range; it has the ability to feed on a variety of plants from different families. In contrast, P. xylostella has a narrow host range utilizing almost exclusively plants from the crucifer family. S. exigua was found to respond to the MAM1/MAM2 polymorphism. Its performance was lower on the MAM2 genotype, and larvae caused approximately 17% less damage than on the MAM1 genotype (Kroymann et al. 2003). Again, single marker analysis showed that the resistance QTL peaked in the interval containing MAM1/MAM2 but not MAM3 (Fig. 8). For P. xylostella, no significant difference was found with ANOVA. Nonetheless, larvae performed on average better on MAM2 genotypes (Kroymann et al. 2003). Furthermore, re-analysis of this data set with regression showed that Plutella larval herbivory was positively correlated with leaf aliphatic glucosinolate content (Kliebenstein et al. 2005). Hence, the MAM1/MAM2 polymorphism had contrasting effects on S. exigua versus P. xylostella. MAM2 genotypes were better defended against the generalist insect herbivore, whereas MAM1 genotypes appeared to suffer less damage from the specialist. However, although the MAM1/MAM2 polymorphism controls these differences in insect performance, it is not yet clear which biochemical parameter ultimately accounts for the observed effects, due to the complex influence of the MAM locus on aliphatic glucosinolate profiles. Differences in resistance could be attributable to leaf total aliphatic glucosinolate concentration, levels of individual glucosinolates, differences in glucosinolate composition, or any combination of these factors.
Since MAM2 genotypes produced approximately 60% more leaf and 20% more seed glucosinolates than MAM1 genotypes, this raised the question whether increased glucosinolate production in the MAM2 genotype was paid for with a reduction in plant growth rate. Such allocation costs can occur when defenses are energetically expensive, so that genotypes with strong defenses have fewer resources to invest in growth and reproduction (Purrington 2000; Tian et al. 2003). Therefore, dry weight of plant rosettes was measured at the pre-reproductive state. Quantitative analyses indicated the presence of growth rate QTL upstream and downstream of the MAM locus, but there was no trace of a significant genotype effect on biomass accumulation at the MAM locus itself (Fig. 8, Kroymann et al. 2003; Kroymann and Mitchell-Olds 2005). Thus, allocation costs appear to be of minor importance for the MAM1/MAM2 polymorphism in A. thaliana.
Generation of glucosinolate diversity
Analyses of nucleotide substitution patterns revealed that MAM1 had accumulated an excess of non-synonymous nucleotide substitutions (i.e., substitutions that alter codon meaning) after the MAMa duplication, indicating positive (Darwinian) selection. The biochemical characteristics of MAM1, MAM2 and MAMa were investigated to infer which properties of MAM1 were targeted by positive selection. MAMa from A. lyrata and from Boechera stricta (a close relative of B. divaricarpa), MAM2 from the Arabidopsis accession Ler, and MAM1 from the accession Sorbo were heterologously expressed in Escherichia coli. Enzyme assays were carried out to investigate substrate specificity, ATP-, metal ion- and pH dependence (Benderoth et al. 2006). Major differences were found only for the enzymes’ substrate specificities (Table 1). A. petraea and B. stricta MAMa, and LerMAM2 all utilized 4-methylthio-2-oxobutanoic acid (2C) for condensation with acetyl-CoA, but none accepted ω-methylthio-2-oxoalkanoic acids with more than two methylene groups as a substrate. By contrast, Sorbo MAM1 accepted 4-methylthio-2-oxobutanoic acid (2C), 5-methylthio-2-oxopentanoic acid (3C) and, with low activity, also 6-methylthio-2-oxohexanoic acid (4C) as a substrate. Thus, MAMa and MAM2 function only in the first cycle of carbon chain extension, whereas MAM1 has acquired additional capacity to carry out condensation reactions in subsequent chain elongation cycles, equivalent to a biochemical neofunctionalization. MAM1 and MAM2 substrate specificities are matched by A. thaliana glucosinolate phenotypes. Accessions with a functional MAM1 gene accumulate 4C glucosinolates, whereas accessions without a functional MAM1 (but with a functional MAM2) generate mainly 3C aliphatic glucosinolates (Kroymann et al. 2003). Thus, gene duplication, biochemical neofunctionalization and positive selection account for the generation of diversity in the carbon chain lengths of aliphatic glucosinolates in A. thaliana.
The same processes are duplicated along the basal branches of the MAM gene tree (Fig. 3). MAMa, MAMb and MAMc have originated by gene duplication events, and are functionally diversified. MAMa controls short-chain aliphatic glucosinolates (Benderoth et al. 2006) and MAMb is, like its A. thaliana ortholog MAM3 (Textor et al. 2007; Knoke et al. 2008), presumably involved in the biosynthesis of long-chain aliphatic glucosinolates. However, the function of MAMc is not yet known. Analyses of nucleotide substitution patterns indicate positive selection along the deep branches of the MAM tree, connecting MAMa, MAMb and MAMc (Benderoth et al. 2006). Hence, the same events that explain diversity in short-chain glucosinolates, i.e., gene duplication and biochemical neofunctionalization, driven by positive selection, appear to account for the diversification in aliphatic glucosinolate carbon chain lengths in general.
Dihomomethionine (4C) differs by only one methylene group from homomethionine (3C). At first glance, this seems to be a minor difference. But the consequences of this small chemical difference can be profound when taken within the context of the glucosinolate-myrosinase system. After methionine carbon chain elongation and glucosinolate core structure generation, enzymes encoded at other genetic loci act to modify the carbon chain (Kliebenstein et al. 2001a, b). These loci are, like MAM, also polymorphic in A. thaliana, i.e., they harbour alleles whose gene products have diverse biochemical activities and cause different modifications of the aliphatic carbon chain. Likewise, modifying proteins like ESP (Lambrix et al. 2001) and ESM1 (Zhang et al. 2006) can alter glucosinolate breakdown identity during myrosinase-catalyzed glucosinolate hydrolysis. Taken together, different combinations of alleles at glucosinolate biosynthesis and hydrolysis loci generate quite different blends of glucosinolate hydrolysis products (Kliebenstein et al. 2005), and these may, in turn, cause different responses in attacking insect herbivores and other enemies.
Maintenance of glucosinolate diversity
While gene duplication, neofunctionalization and positive selection contribute to the generation of metabolic diversity, these processes do not satisfyingly explain how and why this diversity is maintained. Positive selection on MAM1 indicates that this gene has provided a fitness advantage to its carriers after duplication of an ancestral MAMa gene. MAM1 overrides MAM2 function, and MAM1 has retained the capacity to function in the first methionine carbon chain elongation cycle (Kroymann et al. 2003; Benderoth et al. 2006). Thus, the MAM2 is not required to sustain MAM1 activity. Why then is MAM2 still present in A. thaliana?
It could be possible that MAM2 is in the process of becoming a pseudo-gene, but that the period after MAM1 neofunctionalization was too short for a complete degeneration of MAM2. For two reasons, this is not likely. First, estimates based on the analysis of nucleotide substitution rates in the Brassicaceae (Yang et al. 1999; Koch et al. 2001) suggest that the MAMa duplication occurred more than 105 generations ago. Second, a degenerating gene is expected to accumulate nucleotide substitutions since selection no longer acts to eliminate deleterious mutations. Mutations in open reading frames are called synonymous when the amino acid sequence remains unchanged and non-synonymous when the codon usage is altered. Because of the nature of the genetic code, a coding sequence has more non-synonymous than synonymous positions. Therefore, in a degenerating gene non-synonymous changes are more likely to occur than synonymous substitutions. But after correction for the number of non-synonymous and synonymous positions, the ratio between non-synonymous and synonymous changes in a degenerating gene is expected to be close to 1. However, when MAM2 was compared to A. thaliana MAM1 or A. lyrataMAMa, MAM2 exhibited an excess of synonymous relative to non-synonymous changes, indicating purifying selection (Benderoth et al. 2006). In conclusion, MAM2 function appears to be preserved in A. thaliana.
What else could explain why MAM2 was retained in A. thaliana? Insect herbivory assays had shown that NILs with the LerMAM2 performed ca. 17% better against S. exigua larvae than lines with the Col-0 MAM1 (Kroymann et al. 2003). Thus, MAM2 can provide a selective advantage over MAM1 under certain conditions. In addition, secondary gene deletion events and exchange of sequence information between paralogous MAM1 and MAM2 loci have occurred frequently in the history of A. thaliana (Fig. 6). Deletion of MAM1 or conversion of MAM1 into a MAM2-like sequence both result in a switch from a 4C glucosinolate profile to a profile dominated by 3C glucosinolates. If such a switch occurred in a local population of plants that were otherwise genetically uniform, this could confer a—temporal—selective advantage to the novel genotypes, provided that the local herbivore community was accustomed to a particular blend of glucosinolates. Since A. thaliana propagates mostly by selfing and local populations were founded frequently in the history of this species, such a scenario is not unlikely, and might account for the large proportion of derived genotypes at the MAM locus.
Furthermore, the composition of local herbivore communities varies temporally, with different classes of insects—specialists, generalists, and non-feeders (i.e., herbivorous insects not utilizing glucosinolate-containing plants as a host)—occurring in different frequencies over time. This may lead to fluctuating selection on glucosinolate profiles, with periods during which phenotypes with a particular glucosinolate composition or with high glucosinolate levels increase in frequency, alternating with periods during which other types of glucosinolate profiles or low levels are advantageous. As a consequence, one would expect some form of equilibrium between different glucosinolate phenotypes. Indeed, among 51 Arabidopsis accessions whose glucosinolate profiles have been analyzed (Kliebenstein et al. 2001a; Kroymann et al. 2003; Pfalz et al. 2007), 29 produced aliphatic glucosinolates predominantly from homomethionine (3C) and 22 from dihomomethionine (4C). These data are not significantly different from the hypothesis that both phenotypes have equal frequencies in A. thaliana (N = 51; df = 1, χ² = 1.27, n.s.). Also, two independent statistical tests of molecular population genetics found evidence for balancing selection acting on the MAM2 gene (Kroymann et al. 2003). Balancing selection refers to evolutionary scenarios that maintain more genetic variation in a population than expected under neutrality (Nordborg and Innan 2002). First, a positive Tajima’s D indicated significantly more intermediate frequency nucleotide polymorphisms segregating at MAM2 than expected (Tajima 1989). Second, a McDonald and Kreitman test (McDonald and Kreitman 1991) showed that too many amino acids segregated in A. thaliana MAM2, when compared to MAMa from A. lyrata. However, this MAM2 polymorphism has no impact on glucosinolate identity (Kroymann et al. 2003; Benderoth et al. 2006), suggesting that non-neutrality at MAM2 is caused by selection on glucosinolate quantity and not quality, a hypothesis that remains to be tested.
While gene duplication, biochemical neofunctionalization and positive selection account for the generation of metabolic diversity at the MAM locus, secondary gene deletions, gene conversion and balancing selection appear to maintain biochemical diversity. Of course, deletion of MAM1 and conversion of MAM1 into a MAM2-like sequence are both one-way streets from an archetypical MAM2–MAM1–MAM3 configuration. Likewise, deletion of MAM2 or conversion of MAM2 into a MAM1-like gene both prevents future switches of chain-length phenotypes. Therefore, the Arabidopsis MAM locus appears to be in a process during which paralogous genes are being sorted among lineages, ultimately leading to plants that harbor, in combination with MAM3, either a MAM1 or a MAM2 gene.
Similarities and differences in chain-length variation between Arabidopsis and Brassica
In Brassica oleracea and other members of the genus, aliphatic glucosinolates can be generated from homo-, dihomo- and trihomomethionine (Velasco and Becker 2000). As in A. thaliana, there is natural variation for homo- versus dihomomethionine-derived glucosinolates among different accessions. However, in contrast to A. thaliana, accumulation of homomethionine- and accumulation of dihomomethionine-derived glucosinolates do not mutually exclude each other. Hence, B. oleracea accessions can produce aliphatic glucosinolates from homomethionine (3C), dihomomethionine (4C) or from homomethionine and dihomomethionine (3C + 4C). This biochemical polymorphism is caused by variation at two different genetic loci, BoGSL-ELONG and BoGSL-PRO (Magrath et al. 1994; Li et al. 2001). BoGSL-ELONG harbors a typical MAM gene with ten exons (Fig. 4), and, in addition, a MAM pseudogene (Li et al. 2001; Gao et al. 2005). The B. oleracea MAM gene is closely related to MAM genes from other Brassicaceae (Fig. 3). Genetic data indicate that BoGSL-ELONG is responsible for the generation of dihomomethionine-derived glucosinolates (Li and Quiros 2002). However, B. oleracea MAM does not belong to any of the MAMa, b or c subclades found in A. thaliana and close relatives (Fig. 3). The second Brassica locus, BoGSL-PRO, contains a gene with a typical 12-exon IPMS structure (Fig. 4), and phylogenetic analyses support a close phylogenetic relationship with other IPMS genes (Fig. 3). This gene is supposedly involved in the generation of homomethionine-derived glucosinolates (Gao et al. 2006). This suggests that (i) the gene duplication events leading to MAMa, b and c occurred after Arabidopsis, Boechera and Brassica diverged from a common ancestor, (ii) the ability to utilize 3C precursors for carbon chain elongation evolved independently in Brassica and Arabidopsis and (iii) the Brassica MAM function responsible for generating 3C glucosinolates evolved de novo from a MAM progenitor gene. To confirm these hypotheses it will be necessary to sample further MAM (and also IPMS) genes from close and distant Arabidopsis and Brassica relatives and to analyse the biochemical properties of the encoded proteins.
The composition of genes at the MAM locus varies between and within cruciferous species, causing substantial diversity in glucosinolate profiles. Different types of selection act on MAM gene family members, and different factors account for generating glucosinolate variability and for maintaining this diversity. Comparative analyses suggest that particular MAM substrate specificities have evolved repeatedly in different genera of glucosinolate-producing plants. The model plant A. thaliana has been invaluable for making progress in dissecting the genetic, functional and ecological basis of glucosinolate diversity. These discoveries have already facilitated the identification and cloning of genes from the glucosinolate-myrosinase system in cruciferous species for which an ab initio approach proves much more difficult, such as crop plants or wild Arabidopsis relatives. Further functional and evolutionary studies, involving additional species from the Brassicaceae but also from the Capparaceae, the second large family in the Capparales order that is capable of using methionine homologs as glucosinolate precursors, will help to better understand the complexity and evolutionary dynamics of variation in plant secondary metabolism.
We thank John D’Auria for critical reading of the manuscript and gratefully acknowledge financial support from the Max Planck Society and the DFG (grants KR2237/2-1, KR2237/2-2 and KR2237/2-3 to JK).
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
- Knoke B, Textor S, Gershenzon J, Schuster S (2008) Modelling aliphatic glucosinolate diversity in Arabidopsis thaliana. Phytochem Rev (this issue)Google Scholar
- Li G, Riaz S, Goyal S, Abel S, Quiros CF (2001) Inheritance of three major genes involved in the synthesis of aliphatic glucosinolates in Brassica oleracea. J Amer Soc Hort Sci 126:427–431Google Scholar
- Textor S, Bartram S, Kroymann J, Falk KL, Hick A, Pickett JA, Gershenzon J (2004) Biosynthesis of methionine-derived glucosinolates in Arabidopsis thaliana: recombinant expression and characterization of methylthioalkylmalate synthase, the condensing enzyme of the chain elongation cycle. Planta 218:1026–1035PubMedCrossRefGoogle Scholar
- Wang S, Basten CJ, Zeng ZB (2001–2004) Windows QTL cartographer 2.5. Department of Statistics, North Carolina State University, Raleigh, NC. www.statgen.ncsu.edu.qtlcart/WQTLCart.htm