.gif?ad=16248&adview=true)
Variation in the epigenetic silencing of FLC contributes to natural variation in Arabidopsis vernalization response
Abstract
Vernalization, the cold-induced acceleration of flowering, involves the epigenetic silencing of the floral repressor gene FLOWERING LOCUS C (FLC). We investigated the molecular basis for variation in vernalization in Arabidopsis natural accessions adapted to different climates. A major variable was the degree to which different periods of cold caused stable FLC silencing. In accessions requiring long vernalization, FLC expression was reactivated following nonsaturating vernalization, but this reactivation was progressively attenuated with increasing cold exposure. This response was correlated with the rate of accumulation of FLC histone H3 Lys 27 trimethylation (H3K27me3). Thus, variation in epigenetic silencing of FLC appears to have contributed to Arabidopsis adaptation.
Keywords
Arabidopsis thaliana (Arabidopsis) grows in a wide range of environments, so it provides an excellent system to dissect the molecular variation underpinning adaptive variation, a major goal in ecological and evolutionary genetics. Flowering time is a key trait in adaptation, as it is vital for reproductive success. Molecular and genetic analyses in Arabidopsis have begun to elucidate the network of pathways that regulate flowering (Simpson and Dean 2002; Henderson and Dean 2004; Sung and Amasino 2005). An important component in this network is the floral repressor FLOWERING LOCUS C (FLC), a MADS-box protein that represses the expression of genes needed to switch the meristem to a reproductive fate (Michaels and Amasino 1999; Sheldon et al. 1999). Vernalization—the acceleration of flowering by exposure to prolonged cold—represses FLC expression, thus promoting flowering, whereas FRIGIDA (FRI) activates FLC expression and represses flowering (Michaels and Amasino 1999; Sheldon et al. 1999). In the natural context, therefore, the activities of FRI and FLC actively delay flowering until winter has passed. Plants that have such a vernalization requirement are classified as winter annuals. There is considerable variability in flowering time and vernalization response between accessions, and indications of latitudinal clines in these traits support the idea that this variation is important for adaptation (Caicedo et al. 2004; Lempe et al. 2005; Shindo et al. 2005; Stinchcombe et al. 2005; Werner et al. 2005).
Our understanding of the molecular basis of vernalization in Arabidopsis has improved over recent years. Prolonged cold reduces FLC expression, and this repression is maintained by an epigenetic mechanism as plants continue to grow in spring and summer. Mutant screens identifying seedlings unable to suppress and epigenetically maintain FLC repression have identified proteins that mediate vernalization: VERNALIZATION 2 (VRN2), a homolog of the Drosophila Polycomb group gene (PcG) Su(Z)12 (Gendall et al. 2001); VERNALIZATION 1 (VRN1), a plant-specific protein with DNA-binding (B3) domains (Levy et al. 2002); and VERNALIZATION INSENSITIVE 3 (VIN3), a PLANT HOMEODOMAIN (PHD)-finger-containing protein (Sung and Amasino 2004). Prolonged cold induces expression of VIN3 but other factors must also be cold-induced, as constitutive expression of VIN3 does not result in cold-independent early flowering (Sung and Amasino 2004). The activities of VIN3, VRN2, and VRN1 result in histone modifications characteristic of PcG-induced chromatin silencing—namely, reduced histone acetylation and increased histone K27 and K9 methylation (Bastow et al. 2004; Sung and Amasino 2004). These modifications probably recruit LIKE HP1 (LHP1), the Arabidopsis HETEROCHROMATIN PROTEIN (HP1) homolog that is also involved in the stabilization of FLC repression (Mylne et al. 2006; Sung et al. 2006).
This understanding of vernalization has opened up the opportunity to analyze how this process has changed during the adaptation of Arabidopsis accessions to very different growth conditions. Here, we describe the analysis of vernalization response in Arabidopsis natural accessions chosen from a range of habitats differing in winter conditions. The expression of the molecular target of vernalization, FLC, and an early component of the response, VIN3, were analyzed in order to understand the molecular basis of the observed phenotypic variation. Interestingly, no correlation was found between their expression and vernalization response. Instead, variation in the epigenetic maintenance of silencing of FLC was found to be the parameter correlated with vernalization response, and Quantitative Trait Locus (QTL) analysis suggests that at least part of this variation maps to FLC itself. Thus, variation in epigenetic silencing of FLC appears to have been a significant factor in the adaptation of Arabidopsis to different winter conditions.
Results and Discussion
Arabidopsis accessions show a wide rangeof vernalization responses
A previous study of the acceleration of flowering by 8 wk vernalization in 192 Arabidopsis accessions had shown a very wide range of responses (Shindo et al. 2005). In order to further analyze this variation we focused our effort on those collected in Sweden, as they represented a wide range of latitudes and their collection sites are very well-documented (Supplementary Fig. 1). Two accessions, Bil-7 and Lov-1, were chosen from northern Sweden and two, Var-2-6 and Ull-2-5, from southern Sweden. In addition, we analyzed Edi-0, a late-flowering accession from Edinburgh, and a control genotype, H51 (Napp-Zinn 1957). The plants were given up to 20 wk in the cold and then grown in a controlled environment room, and the flowering time was assayed by counting total leaf number (LN) (Supplemental Material). The accessions showed different vernalization characteristics (Fig. 1A). The flowering of Bil-7, Edi-0, and H51, but not Var-2-6, Lov-1, and Ull-2-5, was accelerated by 2 wk vernalization. Flowering time variation between accessions was maximal after 4 wk vernalization; the acceleration of flowering of H51 was almost complete (LN = 20), whereas the flowering of Lov-1 was still unchanged. All Ull-2-5 plants did flower after 4 wk vernalization but considerably later than H51. After 6 wk vernalization, Lov-1 flowered earlier than Ull-2-5, suggesting that Lov-1 might be insensitive to short periods of cold but then respond quickly after that. Despite responding to 4 wk of cold, Var-2-6 and Ull-2-5 still flowered late after >6 wk of cold; i.e., the vernalization response was relatively slow. This was especially true for Ull-2-5, which flowered extremely late after 8 wk vernalization and required at least 14 wk of cold to fully saturate the vernalization requirement.
Variation in vernalization response and FLC expression of the selected accessions. (A) Flowering time was assayed as final LN. Data is mean of at least three experiments with standard error. Nonvernalized plants (NV) of four Scandinavian accessions did not flower when the experiments were terminated in 5 mo since germination, whereas Edi-0 and H51 completed flowering but were still very late. (B) FLC expression after different lengths of cold. (W) Weeks of cold period. (C) VIN3 expression in the accessions assayed after different periods of cold. Semiquantitative PCR analysis was used to analyze VIN3 RNA levels (Supplemental Material). Twenty-day-old nonvernalized (NV) whole seedlings were compared with 7-d-old seedlings exposed to cold for 2, 4, 6, and 8 wk and up to 14 wk for Lov-1, Ull-2-5, and H51 and then harvested in the cold.
Our data suggest that the microclimate of each location, rather than general climatic factors varying with latitude, is likely to be a significant factor in selecting the most productive reproductive strategy. The extreme conditions found at the Bil-7 collection site—long, cold winters on a rocky mountain side—may lead to very limited growth through most of the winter months, necessitating a rapid vernalization response once winter has passed. In climates found in southern Sweden, where winter temperatures are somewhat variable, a “slow” vernalization response may be favored to ensure flowering is not initiated during an occasional warm period.
Regulators of vernalization show similar expression profiles during the prolonged cold
FLC and VIN3 are central regulators of the vernalization response in Arabidopsis (Sung and Amasino 2005). In order to determine if changes in their expression contributed to adaptive changes in vernalization response, we assayed FLC and VIN3 RNA levels in nonvernalized seedlings or seedlings that had been vernalized for different lengths of time (4, 6, 8, 10, or 12 wk), with the tissue harvested immediately at the end of the cold period. Although a wide variation in FLC levels was found in the different accessions, there was no correlation between the FLC level in nonvernalized tissue (Supplementary Fig. 2) and the observed vernalization response, in agreement with previous results (Shindo et al. 2005; Werner et al. 2005). FLC levels did not change significantly in nonvernalized seedlings as they developed (Supplementary Fig. 2), but decreased as the plants were given increasing weeks of cold, with no apparent plateau (Fig. 1B). Despite the varying initial levels of FLC in the accessions the relative decrease of FLC expression differed only slightly between them and there was no correlation between FLC decrease (Fig. 1B) and flowering time. This is most clearly shown in the comparison of Bil-7 and Lov-1—accessions that responded to cold rapidly and slowly, respectively, yet both showing very similar rates of FLC decrease over the time period studied (Fig. 1B). Interestingly, while FLC decreased in Lov-1 plants over a 4-wk cold period, flowering time was not accelerated (Fig. 1A,B). VIN3 expression that was undetectable in the nonvernalized tissue of all of the accessions analyzed, was first detected after 2 wk of cold (Fig. 1C) and then increased, with the maximum expression being reached after ∼4 wk of cold. Where we extended this analysis to 14 wk (Lov-1, Ull-2-5, H51) VIN3 expression did not appear to increase further (Fig. 1C).
Stability of the repression of FLC variesin the accessions
To understand why some accessions required 14 wk of cold for the full acceleration of flowering despite FLC expression being reduced significantly in all accessions by 8 wk of cold, we analyzed the stability of FLC repression. In the accessions where flowering time is rapidly accelerated by vernalization (Edi-0 and the control genotype H51), the level of FLC RNA decreased during the cold and then was relatively stable following 4 wk vernalization (Fig. 2A). The slight reduction in FLC levels immediately after transfer of plants to the 23°C controlled-environment room (0 d points in Fig. 2A)—perhaps due to either further reduction in FLC transcription or increased turnover of the mRNA as plant growth accelerated—was not correlated to variation in vernalization response. In the accessions that responded more slowly to vernalization, Lov-1 and Ull-2-5, FLC levels decreased during the cold but then increased significantly after the plants were returned to 23°C (Fig. 2A).
Variation in the maintenance of repression of FLC. FLC expression was assayed by Northern hybridization of RNA extracted from whole seedlings. (A) Nonvernalized (NV) plants grown for different days were compared with seedlings grown for 7 d, vernalized for 4 wk (4W), and then transferred to 23°C for 0, 10, 20, or 30 d. (B) Repression of FLC following different periods of cold. Thirty-day-old nonvernalized plants (at a developmental stage similar to the 30-d-post-vernalization plants) were compared with seedlings grown at 23°C in long-day photoperiods for 0, 10, 20, and 30 d following each cold treatment (4W, 6W, 8W, 10W, 12W). The 30-d value is missing for H51 as the plants had already flowered.
We extended this analysis to all six accessions, and in an independent experiment assayed FLC expression over a 30-d period as plants grew at 23°C following cold periods of different lengths (4–12 wk). After 4 wk vernalization, FLC levels had increased in Lov-1 and Ull-2-5 (by 30 d after transfer to 23°C) to the same level as nonvernalized plants, while the repression of FLC expression in H51 and Edi-0 was maintained (Fig. 2B). Bil-7 showed an intermediate effect, with FLC returning to a moderately high level after 4 wk vernalization (Fig. 2B).
After 6 wk of cold, FLC levels were lower in all of the accessions (as compared with 4 wk) and were then maintained at that level in Edi-0 and H51. However, they increased slightly in Bil-7 and considerably in Lov-1, Var-2-6, and Ull-2-5 (Fig. 2B). After 8 wk vernalization, FLC levels were stable during subsequent development in all of the accessions except for Lov-1. Indeed, even after 12 wk vernalization FLC levels increased in Lov-1 after 30 d growth post-cold (Fig. 2B). Analysis of these northerly accessions thus revealed that increasing weeks of cold quantitatively enhances the stability of FLC repression, and the variable that shows the most correlation with vernalization response is the ability to epigenetically suppress FLC expression in post-cold growth.
QTL analysis of four populations suggests that allelic variation at four to five genes accounts for the range of phenotypic variability in vernalization response
To determine the genetic basis of the variation in vernalization response we performed QTL analysis on F2 populations derived from crossing four accessions—Lov-1, Var-2-6, Ull-2-5, and Edi-0—to Columbia (Supplemental Material). F2 plants were scored for flowering time after either 4 wk (for Lov-1 and Edi-0) or 8 wk (Var-2-6 and Ull-2-5) vernalization. These treatments were chosen to maximize the phenotypic differences shown by the parental accessions. Analysis of the data revealed significant QTL on chromosomes 1, 4, and 5 (Fig. 3). A common QTL in all four populations maps on chromosome 4 at FRIGIDA. This was anticipated given that Columbia carries a loss-of-function FRI allele and the four accessions carry active FRI alleles (Shindo et al. 2005). There appear to be two QTL mapping to the top of chromosome 5 in Lov-1, Var-2-6, and Ull-2-5. A major one mapping very close to FLC shows an epistatic interaction with the QTL on chromosome 4, supporting the idea that it is FLC (A. Strange and C. Dean, unpubl.). This is also consistent with the observation that in accessions requiring >12 wk of cold to saturate the vernalization requirement, FLC expression is reactivated if the plants are returned to warm temperatures prematurely. The different vernalization phenotypes of the accessions (Fig. 1A); the lateness after 8 wk vernalization of F2 individuals carrying that region from Lov-1, Var-2-6, and Ull-2-5; and the lack of a QTL on chromosome 5 in the Edi-0 × Col-0 population all suggest allelic variation at FLC in the Scandinavian accessions. The second putative QTL on chromosome 5 maps in the region of LHP1 and ART1, the latter previously implicated in determining natural variation in flowering time (Poduska et al. 2003). One or more QTL also map on chromosome 1 and, given their position, possible candidates are FT (Kardailsky et al. 1999) and FRL2 (Michaels et al. 2004) but not MAF1/FLM (Ratcliffe et al. 2001; Scortecci et al. 2001).
QTL analysis to map genes contributing to the variation in vernalization response. Each chromosome is shown separately with the markers used indicated as triangles. LOD (Logarithm of the odds) score is calculated by the MapMaker Version 3.0b with LOD 3 considered significant. The map positions of candidate genes are indicated.
A clear conclusion that emerges from this analysis is that the natural variation in vernalization response does not map to three of the cloned trans-factors regulating vernalization—VIN3, VRN1, or VRN2—that are thought to act directly on FLC and result in a stably silenced state (Bastow et al. 2004; Sung and Amasino 2004; Sung et al. 2006). A possible scenario, given the QTL detected, is that polymorphism at cis elements in FLC, required for both vernalization and up-regulation by FRI, contributes to the variation in vernalization response, together with unknown factors mapping to chromosome 1. LHP1 has recently been reported to act in FLC silencing (Mylne et al. 2006) and so remains another candidate to be analyzed.
FLC chromatin modifications accumulate differentially in the accessions
The finding that cold induced a quantitative accumulation in the epigenetic silencing of FLC suggested that vernalization-induced chromatin modifications may accumulate differently in the different accessions. To explore this further, chromatin immunoprecipitation (ChIP) experiments were undertaken on two accessions: Edi-0, a representative of an accession with a relatively rapid vernalization response, and Lov-1, a representative of a slow-responding accession. Levels of histone H3 Lys 27 trimethylation (H3K27me3), a histone modification characteristic of PcG silenced genes in Arabidopsis and other species (Cao and Zhang 2004; Schonrock et al. 2006), were analyzed in seedlings that had not been vernalized or had been vernalized for 5 wk and then grown for 15 d in a controlled-environment cabinet. Vernalization increased H3K27me3 levels in the FLC promoter region (Fig. 4A) in both accessions. However, due to a combination of a different starting level and a different rate in the acquisition of silencing marks, the accumulated levels of H3K27me3 were 2.5 times higher in Edi-0 compared with Lov-1 after 5 wk vernalization (Fig. 4B,C).
ChIP analysis (A) Genomic structure of Arabidopsis FLC and the region examined—FLC-A (as indicated). (B) PCR analysis of DNA immunoprecipitated with anti-trimethyl-histone H3 Lys 27 (H2K27me3). (C) Quantification of the H3K27me3 enrichment on FLC locus relative to the heterochromatic locus Ta3 (Johnson et al. 2002). The graph comprises the quantification of two ChIP experiments (error bars are standard deviation of six replicates). Chromatin was prepared from nonvernalized seedlings (20 d old) and seedlings that had been vernalized 5 wk following 7 d pregrowth at 23°C and then grown for 15 d after vernalization at 23°C.
Variation in the efficiency of FLC chromatin silencing accounts for natural variationin vernalization response
Our data show that quantitative differences in the epigenetic silencing of FLC, rather than the rate of FLC suppression or VIN3 induction, contribute to the natural variation in vernalization response. QTL mapping suggests FLC itself may contribute to this variation, so polymorphism in key cis elements in FLC may alter interaction with trans-acting proteins and change the efficiency of chromatin silencing. These cis elements could be in the 5′ region or intron 1 as both have been shown to be important in the maintenance of FLC repression (Sheldon et al. 2002; Sung et al. 2006). Lempe et al. (2005) have documented substantial sequence variation in FLC, in particular several large insertions in intron 1. Whatever the nature of the variation, the outcome is a differential accumulation of H3K27me3, a histone modification that increases significantly at FLC after vernalization. Combining these data with the flowering time and FLC expression analysis suggests a possible model shown schematically in Figure 5. Three phases of vernalization are proposed: an insensitive phase, a phase where silencing occurs (and histone modifications accumulate), and a phase where FLC is fully silenced. Only when sufficient histone modifications have accumulated would a fully silenced state be reached. Full silencing could be attained by 4 wk of cold in Edi-0 (and perhaps other accessions that respond rapidly to vernalization) due to a high initial level and higher accumulation rate. In contrast, the lower initial levels in Lov-1 (and perhaps other accessions that respond slowly to vernalization) necessitate a 10- to 12-wk cold period to reach the same degree of H3K27 methylation.
Model proposing that quantitative accumulation of epigenetic modifications in FLC contributes to natural variation of vernalization response. Three phases of vernalization are envisaged: an insensitive phase, a phase where silencing occurs (and histone modifications accumulate), and a phase where FLC is fully silenced. The dashed and solid lines, estimated from the ChIP analysis shown in Figure 4C, represent the accumulation of epigenetic marks in Edi-0 and Lov-1, respectively. Due to a high initial level and higher accumulation rate, full silencing would be reached after 4 wk of cold in Edi-0 and 10–12 wk of cold in Lov-1.
We are continuing the genetic analysis on these accessions to definitively identify the genes and causative variation conferring this natural variation in vernalization response. Generation of lines differing only at these loci, either through introgression or transformation, will also provide material optimal for determination of fitness consequences of that phenotypic variation. This, together with analysis of the distribution of different alleles in the Arabidopsis populations, should be very revealing with respect to the evolution of this adaptive trait.
Materials and methods
Quantification of FLC expression by Northern analysis
Vernalized plants were harvested at 0, 10, 20, and 30 d after the return to a 23°C controlled-environment room. All above-ground material was harvested and used for extraction of total RNA as described in Etheridge et al. (1999). FLC expression levels were assayed by Northern hybridization as described previously (Shindo et al. 2005). The analysis was repeated with RNA extracted from independently grown plants. Quantification of expression levels was calculated relative to TUBULIN expression and shown as values relative to that of nonvernalized H51 plants (i.e., H51 NV30 = 1.00).
ChIP and real-time quantitative PCR analysis
ChIP assays were performed as previously described (Bastow et al. 2004) with minor modifications. Chromatin preparations from Arabidopsis seedlings were immunoprecipitated with anti-trimethyl-histone H3 Ly 27 (Thomas Jenuwein, IMP Research, Vienna, Austria; rabbit 6523, 5 bleed) and magnetic Dynabeads Protein A (Dynal Biotech). ChIP DNA was analyzed by real-time quantitative PCR using an OPTICON2 instrument from MJ Research and SYBR Green JumpStart Taq ReadyMix from Sigma. Specific oligonucleotides used are described in the Supplemental Material. Thermal cycling consisted of 2 min at 94°C, followed by 40 cycles of 10 sec at 94°C, 20 sec at 60°C, and 30 sec at 72°C. After that, a melting curve was generated to check the specificity of the amplified fragment.
Acknowledgments
This work was funded by Natural and Environmental Research Council Environmental Genomics grant NER/Y/S2001/00240 and a core grant from the Biotechnology and Biological Research Council. We are very grateful to members of the Nordborg laboratory for help with QTL analysis, Norman Warthmann and Detlef Weigel for providing extensive information on SNP markers, and Thomas Jenuwein for the H3K27me3 antibody. We also thank Judith Irwin for critical reading of this manuscript.
Footnotes
-
↵3 Present address: Département de Biologie, Moléculaire Végétale, Université de Lausanne, CH-1015 Lausanne, Switzerland.
-
↵4 Corresponding author.
↵4 E-MAIL caroline.dean{at}bbsrc.ac.uk; FAX 44-0-1603-450025.
-
Supplemental material is available at http://www.genesdev.org.
-
Article is online at http://www.genesdev.org/cgi/doi/10.1101/gad.405306
-
- Received August 10, 2006.
- Accepted October 3, 2006.
- Copyright © 2006, Cold Spring Harbor Laboratory Press
