|Home | About | Journals | Submit | Contact Us | Français|
Our understanding of the evolution of organismal diversity is restricted by the current resolution of the genotype-phenotype map. In particular, the genetic basis of environmentally relevant phenotypic variation among natural populations remains poorly understood. Trichomes are single-cell outgrowths on the surface of plant leaves and other aboveground organs. Consistent with trichomes' suggested function to protect plants from predators and abiotic stressors [1-3], trichome density is strikingly variable among natural populations (e.g., [2, 4]). Despite substantial progress in the genetic dissection of trichome development , how trichome number is modulated in natural populations remains enigmatic. Here, we show that the ENHANCER OF TRY AND CPC 2 (ETC2) from the single-repeat R3 MYB family is the major locus determining trichome patterning in natural Arabidopsis populations. Our study identifies a single amino acid substitution in ETC2 (K19E) as the causal quantitative trait nucleotide (QTN). We suggest that this amino acid replacement might affect the stability of the ETC2 repressor, which results in a reduced trichome number. This is consistent with the view that morphology can evolve by coding changes that can subtly modulate protein activity as well as cis-regulatory changes that alter expression patterns.
In a survey of 75 Arabidopsis accessions, we confirmed the presence of substantial variation in trichome density (Figure 1; e.g., ). Gr-1 and Can-0 were among the accessions with the lowest and highest trichome numbers, respectively. As a first step toward characterizing the genetic basis of this trait difference, we performed a quantitative trait locus (QTL) mapping experiment with these two accessions. In total, three QTL were detected (Figure 2; see Table S1 available online), but the single QTL with a large effect on chromosome 2 explained 33% of the variation in trichome number. Using a denser set of markers and additional F2 individuals with extreme trichome numbers, we fine mapped this QTL to a genomic region of 288 kb (between markers nga3023 and nga3098), which contains 87 genes (Figure S1). Interestingly, the mapping interval unambiguously excluded TTG2, which has previously been identified as a candidate gene for reduced trichome number [6, 7].
We further characterized the genetic basis of trichome number variation by introgressing the high and low alleles into a mutant background of TRIPTYCHON (TRY)or CAPRICE (CPC), two genes that were previously shown to affect trichome patterning on Arabidopsis leaves [8, 9]. In the cpc background, we observed an additive effect of the naturally occurring alleles (the interaction effect was not significant, p = 0.205), suggesting that natural variation in trichome number is not affected by cpc. For try, however, a significant interaction effect (p = 0.025) was detected, indicating that our QTL falls into the same pathway as TRY. Hence, we searched for potential candidates among the 87 genes in the mapping interval. We detected three members of the single-repeat R3 MYB family. Two of them (ETC2 and TRICHOMELESS 1; TCL1) have previously been shown to be involved in trichome patterning [10, 11], whereas TCL2 has not yet been assigned to epidermal patterning but shares 80% identity with TCL1 and 67% identity with ETC2.
Given that all three genes encode a potential repressor of trichomes, we used transgenesis to introduce additional copies of these genes to test their effects on trichome number. We analyzed the Can-0 alleles of ETC2, TCL1, and TCL2 in the background of Gr-1 and Can-0. TCL2 transformants showed no effect on trichome number, suggesting that this gene does not regulate trichome number. Approximately 50% of TCL1 transformants develop glabrous leaves in both genetic backgrounds. Although this had already been shown previously with the strong viral 35S promoter , our results demonstrate that even the endogenous promoter is sufficient to cause this effect. In the case of ETC2 transformants, we also observed glabrous plants, but the effect was substantially less pronounced than for TCL1 (Table S2). Hence, both TCL1 and ETC2 remained as potential candidates for the modulation of trichome number in natural Arabidopsis accessions.
Next we sequenced all three genes from accessions with high, medium, and low trichome number. For ETC2, we observed that a large number of single-nucleotide polymorphisms (SNPs) distinguished low trichome accessions from high and intermediate ones (Figures S2 and S3), suggesting that trichome number is affected by different ETC2 alleles.
To validate whether natural variation in trichome number is governed by ETC2, we used a complementation test by crossing Gr-1 and Can-0 to etc2-2Col. In addition, we also crossed another high-trichome-number accession (20-13) and low-trichome-number accession (Ler) to etc2-2Col. Consistent with ETC2 influencing trichome number, we found a higher trichome number for etc2-2Col/ETC2Can-0 and etc2-2Col/ETC220-13 than for etc2-2Col/ETC2Gr-1 and etc2-2Col/ETC2Ler individuals (Figure 3).
Given that SNPs covering the entire ETC2 gene distinguish high- and medium- from low-trichome-number accessions, we increased the sample size by adding extreme trichome phenotypes to identify recombinants between these two haplotypic classes. Recombination breakpoints resolved the potential QTNs to two candidates: one in the 5′UTR (−53 bp) and another (+55 bp) nonsynonymous difference in the first exon, changing lysine (K) to glutamate (E) (Figure S3).
We used site-directed mutagenesis to swap the SNPs at positions −53 bp and +55 bp between Gr-1 and Can-0. As expected, the remaining sequence polymorphisms distinguishing ETC2Gr-1 and ETC2Can-0 did not have a significant effect on trichome number (p = 0.35, analysis of variance [ANOVA]). Figure 4 shows that transformants with the Gr-1 QTN in position +55 developed a lower trichome number than those with the Can-0 QTN. Hence, we conclude that the K-to-E change at amino acid 19 of ETC2 is responsible for the low-trichome-number phenotype.
The single-repeat R3 MYB family of repressors is characterized by a single DNA-binding domain and by the lack of an activation domain [8-14]. N- and C-terminal domains of CPC are responsible for its non-cell-autonomous function . Although intercellular movement has also been shown for ETC3  and is proposed for other family members, the N- and C-terminal domains are poorly conserved. Interestingly, the QTN identified in our study is not located in any of the previously described functional domains. Rather, it is located in a part of the gene that has not yet been functionally characterized. A comparison with other genes of the single-repeat R3 MYB family indicated that the position of the K19E replacement is highly conserved, with the glutamate of the low-trichome-number accessions being the derived state (Figure S4).
Like other single-repeat R3 MYB proteins (CPC, TRY, ETC1, and ETC3), ETC2 interacts with GL3 and EGL3 [10, 14, 16, 17] (Figure S5). These interactions compete with the positive regulator GL1 (e.g., ). The single-repeat R3 MYB proteins have different binding affinities to GL3 . Because K19 is conserved in all single-repeat R3 MYB proteins, it is unlikely that the K19E substitution changes the binding affinity to GL3.
Three parameters affecting trichome patterning have been suggested for the single-repeat R3 MYB family of repressors: (1) the binding strength to GL3 or to regulatory regions of the GL1 gene, (2) the movement rate to neighboring cells, and (3) stability . Although more emphasis was previously placed on movement rates, the functional divergence at ETC2 uncovered in this study suggests that the regulation of the protein stability might play a central role for trichome patterning in natural populations. The nonconservative replacement of a lysine with a glutamate suggests that the low-trichome-number allele lost a target site for posttranslational modification. Given that lysine is a potential target for ubiquitination, it is conceivable that the K19E mutation leads to loss of ubiquitination, which increases the stability of the repressor. The higher stability of the ETC2 repressor in turn could explain the lower trichome number in these accessions.
One important feature of the single-repeat R3 MYB family is that all members except ETC2 affect both trichome and root hair density [8, 10-14, 16-19]. Trichome density serves an important role in protection from predators and abiotic stressors. However, this defense function is independent of the primary function of root hairs and the uptake of minerals and water. Therefore, it seems reasonable that these two processes might be regulated independently. Because ETC2 is not expressed in roots, the K19E QTN serves as an important modulator of trichome number in natural populations while leaving the root hair patterning unaffected. Our results have direct implications for the recent debate on how adaptive morphological variation is generated. It has been argued that cis-regulatory changes drive morphological evolution due to the possibility of minimizing pleiotropic effects . Nevertheless, this view has been challenged because several examples of morphological change due to coding sequence variation have been documented . Our study also demonstrates that morphological variation could be modulated by differences in the coding sequence, but it also suggests that such changes occur in genes with a low pleiotropy. We propose that the focus on cis-regulatory changes or protein mutations is artificial and reflects our ignorance about the functional implications of natural variation. The key to the potential to shape morphological variation is probably a low pleiotropic effect, irrespective of regulatory or structural ones.
The K19E QTN occurred at a frequency of at least 20% (15 out of 75) in our collection, suggesting that this polymorphism serves an important role in shaping trichome patterning variation in natural Arabidopsis populations. The widespread occurrence of this mutation is further substantiated by several independent QTL mapping experiments, which identified one major QTL for trichome density at the position of ETC2 [7, 22, 23]. The high sequence divergence (d = 0.047) between low-trichome ETC2 alleles and the remaining alleles suggests that this haplotype is not of recent origin. Furthermore, the low-trichome-density ETC2 haplotypes harbor substantial sequence variation (Θ = 0.00333). Hence, it appears unlikely that the K19E mutation is of recent origin or that its spread has been favored by a recent positive selection event.
Our analysis of natural variation demonstrated that a single amino acid replacement at the ETC2 gene has a substantial effect on trichome patterning. At first sight, this observation contrasts with functional analyses of trichome patterning, because several studies suggested that ETC2 has only a very moderate effect on trichome density. This discrepancy can be resolved by a closer look at the ETC2 alleles that were used in the functional studies. Both Ws and Col do not contain the ETC2 allele with the strong repressor function. Hence, null mutants of these alleles led to an underestimation of the importance of ETC2 for trichome patterning. We are convinced that this is only one of many more examples to follow, which will demonstrate that a functional characterization of genes needs to account for naturally occurring alleles. Thus, we envision that the hitherto separated research areas of population genetics and functional genetics will be united in the not too distant future.
We crossed cpc-1Ws and try-82Ler to Gr-1 and Can-0. F2 and F3 offspring were genotyped at CPC and TRY with the following primer combinations:
Primer sequences are given in Table S5. Genetic interaction of the major QTL with CPC and TRY was tested for individuals homozygous for QTLGr-1CPCGr-1, QTLGr-1cpc-1Ws, QTLCan-0CPCCan-0, QTLCan-0cpc-1Ws, QTLGr-1TRYGr-1, QTLGr-1try-82Ler, QTLCan-0TRYCan-0, and QTLCan-0try-82Ler.
This strategy only controls for the genotypes at CPC and TRY and does not account for the genetic heterogeneity at the remainder of the genome. Hence, it would have been desirable to test a large number of independent lines for each genotype. The tight linkage of CPC with the QTL, however, complicated the analysis such that we had to restrict our analysis to two or three independent lines from each genotype. We noted that the trichome numbers for the recombinant lines carrying the wild-type alleles did not always show a good correspondence to the phenotypes of Gr-1 and Can-0, suggesting some influence of the genetic background of the mutant lines. Nevertheless, trichome density closely followed the phenotypes of Gr-1 and Can-0, so for this analysis, we used trichome density rather than trichome number.
From each line, ten individuals were phenotyped for leaves at positions 8-10. Ln-transformed trait values of the data set were close to a normal distribution (p ≥ 0.047, one-sample Kolmogorov-Smirnov test). We performed full-factorial ANOVA with the accession genotypes at the QTL and CPC or TRY as fixed effects, and leaf position as random effect. Because try-82Ler develops trichome clusters rather than evenly spaced trichomes, we analyzed trichome numbers as well as trichome initiation sites. Both results were highly similar, therefore we only report analyses for trichome initiation sites.
We crossed etc2-2Col to Can-0, Gr-1, 20-13, and Ler. The heterozygous F1 etc2-2Col/ETC2Can-0, etc2-2Col/ETC2Gr-1, etc2-2Col/ETC220-13, and etc2-2Col/ETC2Ler individuals were phenotyped. We ln-transformed the data to satisfy the normal distribution criterion (p > 0.05, one-sample Kolmogorov-Smirnov test). A factorial ANOVA design incorporating genotype as fixed factor and leaf position as random factor was carried out. A significant main effect of genotype on trichome number (p > 0.001) was further analyzed by Tukey's honestly significant difference (HSD) post hoc test (variances of groups were equal according to Levene's test; p = 0.054).
We polymerase chain reaction (PCR)-amplified the genomic region of ETC2, TCL1, and TCL2, including 2489/358 bp, 2158/351 bp, and 1616/1425 bp of flanking upstream/downstream sequences. The PCR primers carried restriction sites, which were used for subsequent cloning steps (Table S5). The EcoRI-digested amplicon of ETC2 was ligated directly into the pPZP211 vector . Amplicons of TCL1 and TCL2 were cloned into the vector pCR4-TOPO (Invitrogen) according to the manufacturer's instructions and were subsequently transferred to pPZP211 via XbaI sites. All cloned PCR amplicons were sequenced.
SNP swapping at positions −53 bp and +55 bp used the Phusion Site-Directed Mutagenesis Kit (Finnzymes) and pCR4_ETC2Gr-1 and pCR4_ETC2Can-0 as templates. ETC2Gr-1 was PCR amplified and cloned into pCR4-TOPO, and ETC2Can-0 was transferred from pPZP211_ETC2Can-0 into pCR4-TOPO. Table S5 lists the primers used. The success of site-directed mutagenesis was confirmed by sequencing. Mutagenized ETC2 alleles were transferred back via EcoRI sites into pPZP221 and were transformed into the Agrobacterium tumefaciens strain GV3101::pMP90 by electroporation. Transgenic etc2-2Col, Gr-1, and Can-0 lines were generated via the floral dip procedure . Each T1 line was phenotyped as described above (i.e., three leaves were counted for each individual). After transformation, the distribution of trait values did not deviate from normality (p = 0.198, one-sample Kolmogorov-Smirnov test). Full factorial ANOVA incorporating the −53 bp/+55 bp status and the origin of the allele (ETC2Gr-1 or ETC2Can-0) was used as fixed effect and leaf position as random effect.
We thank C. Agreiter, X. Hochwallner, U Il Im, V. Köhler, R. Lagler, R. Ranftl, A. Redweik, J. Rudolf, W. Salaberger, N. Schlager, members of the Institute of Population Genetics, and J.A. Torres-Acosta for technical assistance and support. We are grateful to M. Hülskamp for providing cpc-1 and try-82 seeds and to the Nottingham Arabidopsis Stock Centre for distributing the SALK T-DNA insertion mutants of J. Alonso and J. Ecker. K. Schmid, H. Schmuths, and H. Stenoien provided Arabidopsis accessions. A. McGregor provided helpful comments on an earlier version of the manuscript. This work was partially supported by grants from the WWTF, the Austrian Science Fund (FWF) grants P16420-B12 and L433 B17 to M.T.H., and various FWF grants awarded to C.S.
Supplemental Data include Supplemental Experimental Procedures, five tables, and six figures and can be found online at http://www.cell.com/current-biology/supplemental/S0960-9822(09)01696-0.