Seed dormancy is an important adaptive trait that enables germination at the proper time, thereby ensuring plant survival after germination. In Arabidopsis, considerable variation exists in the degree of seed dormancy among wild-type accessions (ecotypes). In this paper, we identify a plant-specific HD2 histone deacetylase gene, HD2B (At5g22650), as a genetic factor associated with seed dormancy. First, genome-wide association mapping of 113 accessions was used to identify single nucleotide polymorphisms that possibly explain natural variation for seed dormancy. Integration of genome-wide association mapping and transcriptome analysis during cold-induced dormancy cycling identified HD2B as the most plausible candidate gene, and quantitative RT-PCR analysis demonstrated that HD2B expression was up-regulated by cold and after-ripening (dry storage of mature seed), treatments that are known to break seed dormancy. Interestingly, quantitative RT-PCR analysis in 106 accessions revealed that the expression of HD2B in imbibed seeds was significantly suppressed in most of the dormant accessions compared with less-dormant accessions, suggesting that suppression of HD2B expression may be important to maintain seed dormancy in dormant accessions. In addition, transgenic seeds of a dormant Cvi-0 accession that carried a 2.5 kb genomic DNA fragment of HD2B cloned from a less-dormant Col-0 accession (ColHD2B/Cvi-0) exhibited reduced seed dormancy accompanied by enhanced expression of HD2B when after-ripened or cold-imbibed. Endogenous levels of gibberellin were found to be increased in the imbibed seeds of after-ripened ColHD2B/Cvi-0 compared with wild-type Cvi-0. These results suggest that HD2B plays a role in seed dormancy and/or germinability in Arabidopsis thaliana.
Despite their sessile nature, plants have the ability to choose their habitats through precise germination responses to environmental factors (Donohue, 2005). Germination at the right time is the first requirement for successful growth and survival, because it influences subsequent life-history traits . In this sense, seed dormancy, which arrests seed growth and development even under conditions favorable to germination, is thought to be an important adaptive trait of plants (Bewley, 1997; Finch-Savage and Leubner-Metzger, 2006; Holdsworth et al., 2008). After-ripening (AR) refers to a dry storage period, which enhances dormancy release. Additional factors, such as low temperatures (Penfield et al., 2005; Kendall et al., 2011) and nitrate (Alboresi et al., 2005; Matakiadis et al., 2009), may affect seed dormancy during seed maturation, dry storage and after imbibition. Because seed dormancy is one of the most important plant traits that affect crop yields and their industrial use , an understanding of the complex molecular mechanisms of seed dormancy is highly desirable (Gubler et al., 2005).
Increasing lines of evidence have demonstrated that the phytohormones abscisic acid (ABA) and gibberellin (GA) play essential roles in environmental regulation of seed dormancy and germination (Nambara and Marion-Poll, 2005; Finkelstein et al., 2008; Yamaguchi, 2008). ABA is essential to induce and maintain dormancy, while GA is required for germination. Dormancy-affecting factors, such as AR, low temperatures and nitrate, are known to affect the expression of genes involved in ABA catabolism and GA biosynthesis, such as CYP707A2 and gibberellin 3β-oxidase (GA3ox) (Cadman et al., 2006; Finch-Savage et al., 2007; Carrera et al., 2008). Expression of GA3ox genes is also changed in Arabidopsis mutants with altered primary seed dormancy (Penfield et al., 2005) or AR responses (Yano et al., 2009). On the other hand, genetic studies on reduced dormancy (rdo) mutants in Arabidopsis have indicated an ABA-independent mechanism of seed dormancy (Peeters et al., 2002). RDO4/HISTONE MONOUBIQUITINATION 1 (RDO4/HUB1) encodes a C3HC4 RING finger protein that appears to be involved in histone H2B monoubiquitination (Liu et al., 2007). More recently, RDO2 was shown to encode a transcription elongation factor, TFIIS (Liu et al., 2011). RDO4/HUB1 and RDO2 are thought to interact with the RNA polymerase II-associated factor 1 complex (PAF1C), suggesting the important role of transcription in seed dormancy.
There is natural variation in the degree of seed dormancy and germinability among wild-type accessions (ecotypes) or cultivars of Arabidopsis and crop species (Alonso-Blanco et al., 2003; Fujino et al., 2008; Bentsink et al., 2010; Sugimoto et al., 2010). The most popular approach to identify genes for natural variation is quantitative trait locus (QTL) studies. In Arabidopsis, QTL studies using recombinant inbred lines (RILs) have identified DELAY OF GERMINATION 1 (DOG1) as a gene contributing to natural variation in seed dormancy in Arabidopsis (Bentsink et al., 2006). DOG1 functions in a genetic pathway that is independent of ABA (Nakabayashi et al., 2012), and its expression in imbibed seeds is down-regulated by AR (Bentsink et al., 2006). In rice, qLTG301, a major QTL responsible for low-temperature germinability, has also been cloned through a QTL approach using the Hayamasari and Italica Livorno strains (Fujino et al., 2008). In addition, Sdr4, a major QTL responsible for seed dormancy and domestication, has been cloned through a QTL approach using the japonica and indica cultivars (Sugimoto et al., 2010). In Arabidopsis, FLOWERING LOCUS C (FLC), a master regulator of reproductive phase transition, has also been reported to be involved in natural variation for low-temperature germinability (Chiang et al., 2009). FLC is considered to determine seasonal habitat choice in nature by controlling both seed germination and flowering time. Although QTL studies are promising, genome-wide association (GWA) mapping has recently emerged as a more comprehensive and robust approach to study natural variation (Nordborg and Weigel, 2008; Ingvarsson and Street, 2011). In Arabidopsis, GWA mapping has identified a number of single nucleotide polymorphisms (SNPs) that may explain the variation in 107 phenotypes, including seed dormancy and germination (Atwell et al., 2010). GWA mapping has been shown to be useful to identify environmentally sensitive genetic loci underlying local adaptation in flowering time (Li et al., 2010). However, because GWA mapping uses a huge number of SNPs as explanatory variables, distinguishing true-positive SNPs from false-positive ones is a challenge (Tabangin et al., 2009). None of the candidate genes identified by GWA mapping has been biologically shown to play a role in seed dormancy or germination.
In this study, we identified a HD2 histone deacetylase (HDAC) gene, HD2B (At5g22650), as a genetic factor associated with seed dormancy and germination through a combined approach of GWA mapping and transcriptome analyses. HD2-type HDAC was initially identified in maize as a plant-specific HDAC (Lusser, 1997). Some Arabidopsis homologs of HD2 HDACs have been reported to regulate seed germination, the salt stress response and stomatal closure in the Col-0 ecotype (Sridha and Wu, 2006; Colville et al., 2011; Luo et al., 2012a). However, little was known about the role of HD2B in seed dormancy. Our results indicate that expression of HD2B is affected by cold and AR, and was down-regulated in most dormant accessions compared with less-dormant accessions. In addition, we showed that enhanced HD2B expression was associated with reduced seed dormancy and increased endogenous GA accumulation in the dormant Cvi-0 background. Our results indicated that expression of HD2B is associated with seed dormancy and/or germinability in at least some accessions in Arabidopsis.
GWA mapping of seed dormancy using seeds after-ripened for two months
It is known that seeds of the Cvi-0 accession, a dormant Arabidopsis accession, require much longer AR to reduce seed dormancy compared with less-dormant accessions, such as Col-0 and Ler, when matured at 23°C (Alonso-Blanco et al., 2003; Ali-Rachedi et al., 2004; Preston et al., 2009). To extensively investigate natural variation for seed dormancy and perform GWA mapping, we analyzed germination in 117 accessions using seeds after-ripened for 2 months. Of our 117 accessions, 108 originated from European countries, while the others originated from other regions of the world (Figure 1a and Data S1). When freshly harvested seeds were after-ripened for 2 months, then imbibed on water/agarose at 23°C in the light, germination frequencies differed substantially among the 117 accessions after 3 and 7 days of imbibition (Figure 1b). Thirty-two accessions, including Col-0 and Ler-1, germinated at rates higher than 90% after 7 days imbibition, but 27 accessions, including Cvi-0, germinated at rates lower than 10%, indicating that the former accessions had almost completely lost seed dormancy during AR but that the latter retained dormancy after 2 months of AR. We designated accessions with germination >90% as less-dormant and those with germination <10% as dormant (Data S1). Then GWA mapping was performed using the previously published AtSNPtile1 SNP dataset (Atwell et al., 2010; Li et al., 2010). Of the 117 accessions, 113 were included in the SNP datasets (Data S1). Because germination frequencies of 7-day-imbibed seeds did not show a normal distribution, we also tested arcsine- or logit-transformed parameters as phenotype parameters (Figure 2a). When GWA mapping was performed using Efficient Mixed Model Association expedited (EMMAX) (Kang et al., 2010) and PLINK software (Purcell et al., 2007) with these phenotype parameters (degrees of seed dormancy), several SNPs with low P values (<10−4) were found in all analyses (Figure 2b–d). There were 224 SNPs listed in the top 100 candidates in any of the EMMAX and PLINK analyses (Data S2). Among them, the most significantly associated SNPs were 5_7534403_G_A, 5_7534842_C_T and 5_7534974_C_T, because their adjusted determination coefficients (R2) were the highest (R2= 0.215; Figure S1a,b). These SNPs were located on the HD2B gene (At5g22650), and were detected as significantly associated SNPs whichever phenotype parameters were used for GWA mapping (Figure 2b–d; arrows). We also detected strongly associated SNPs around the middle region of chromosome 3 (Figure 2b–d; inverted T symbols), where the DOG6 QTL has been previously reported (Bentsink et al., 2010). In addition, SNPs within 500 kb of DOG1 were detected as significantly associated when arcsine- or logit-transformed parameters were used for EMMAX calculation (Figure 2c,d; arrowheads).
Integration of GWA mapping and transcriptome analyses identified HD2B as the most plausible candidate associated with seed dormancy
Use of the Arabidopsis eFP browser (Winter et al., 2007; Bassel et al., 2008) initially suggested that expression of HD2B may be regulated by cold and AR, treatments that are known to reduce seed dormancy. This suggested that integration of GWA mapping and transcriptome datasets is useful to restrict candidates. To integrate GWA mapping and transcriptomics, we focused on cold-induced dormancy cycling in the dormant Cvi-0 accession. It is known that short-term cold imbibition releases the primary seed dormancy of Cvi-0 seeds, whereas long-term cold imbibition induces secondary dormancy of the seeds (Cadman et al., 2006). Accordingly, more than 80% of Cvi-0 seeds that had been imbibed at 2°C for 4 or 7 days in the dark germinated after transfer to 23°C in the light for 7 days, but Cvi-0 seeds that had been imbibed at 2°C in the dark for a longer period (21 days) did not germinate after transfer to 23°C in the light (Figure 3a). In contrast, imbibition at 23°C in the dark did not affect seed dormancy. Thus, this experimental system distinguished seed dormancy from germination , because changes in seed dormancy occurred during imbibition in the dark (Figure 3a, top) but not after the seeds were subjected to a light and temperature shift (Figure 3a, bottom).
First, we determined the expression patterns of 10 652 genes during imbibition at 2 and 23°C in the dark using the Affymetrix ATH1 GeneChip. When the expression levels of the genes were regressed onto the frequency of germination after transfer to 23°C in the light (degree of seed dormancy) using a linear model, 886 genes were associated with seed dormancy (P < 0.001) (Figure 3b and Figure S2a–c). Of these, 246 genes were up-regulated more than fivefold by short-term (4 and 7 days) cold imbibition relative to 23°C-imbibed seeds, and 17 genes were down-regulated to a similar extent (Figure 3c,d). These genes, designated dormancy-associated genes, included the HD2 HDAC genes HD2B, HD2A (At3g44750) and HD2D (At2g27840) (Data S3). Quantitative RT-PCR analysis confirmed that the expression levels of these HD2 genes were more than tenfold up-regulated by short-term cold imbibition (Figure S3a), corresponding with changes in seed dormancy. However, the expression of HD2B, HD2A and HD2D did not respond to illumination and a temperature shift after cold imbibition (Figure S3b). In contrast, expression of GA3ox2, a light-inducible GA biosynthesis gene, was more than 60-fold up-regulated by light and warmth in seeds that had been cold-treated for 4 days but not in seeds that had been cold-treated for 21 days (Figure S3b), indicating that expression of GA3ox2 depends on the degree of seed dormancy before transfer to light and warmth. The dormancy-associated genes also included histone H2A/H2B chaperones (NAP1-RELATED PROTEIN1, 2) (Zhu et al., 2006) (Data S3). In addition, ribosomal protein genes were highly enriched among cold up-regulated dormancy-associated genes (10.2% of all genes). Genome-wide co-expression analysis suggested that HD2B was co-expressed with most ribosomal genes (Figure S2d,e). The transcriptome dataset was integrated with GWA mapping by simply comparing the –log10 (P value) obtained in each analysis. Of the 194 828 SNPs analyzed in GWA mapping, 79 287 non-redundant SNPs corresponded to the 10 072 genes whose expression patterns were determined by ATH1 GeneChip analysis. Interestingly, HD2B had a significantly high –log10 (P value) in each analysis, and was clearly distinguished from other genes (Figure 3e). Whichever phenotype parameters were used, HD2B was identified as a distinguished candidate (Figure S4a–c). These results indicate that HD2B is the most plausible candidate associated with seed dormancy.
Finally, to perform GWA mapping of HD2B polymorphisms in more detail, we sequenced the 2.5 kb genomic region of HD2B in 83 accessions, and identified 177 polymorphisms (Data S4). GWA mapping using the newly identified polymorphisms indicated that SNPs of HD2B were highly associated with seed dormancy (P < 10−5) (Figure 4a). In addition, most of the significantly associated HD2B SNPs were located within the promoter, first exon or first intron of the HD2B genomic region (Figure 4b). Among them, the most significantly associated polymorphism was 5_7534229_A_G. All of the accessions carrying the allele ‘A’ at this polymorphism were dormant, including Lu-1, Mr-0 and Cvi-0 (Figure 4c), but most of the less-dormant accessions including Col-0 and Ler-1 carried the allele ‘G’.
HD2B expression was up-regulated by AR and suppressed in most dormant accessions
Next, the effect of AR on HD2B expression was analyzed using less-dormant Col-0 and dormant Cvi-0 accessions. When Cvi-0 seeds were after-ripened for 2, 4, 8 or 16 weeks, then imbibed at 23°C in the light for 24 h, AR increased the expression levels of HD2B in parallel with a decrease in seed dormancy (Figure 5a–c). When seeds after-ripened for 2 or 16 weeks- were compared, AR up-regulated HD2B by 4.5-fold in Cvi-0. Similar trends were also observed in less-dormant Col-0 seeds, although the expression levels of HD2B had been already high in Col-0 without long-term AR. It is known that expression of GA3ox2, AtEXPANSIN1 (AtEXP1) and HD2D is positively affected by AR, while that of ZEAXANTHIN EPOXIDASE (ZEP), an ABA biosynthesis gene, and DOG1 is negatively affected (Bentsink et al., 2006; Cadman et al., 2006; Carrera et al., 2008). In both Cvi-0 and Col-0, the expression pattern of HD2B was similar to that of GA3ox2 and AtEXP1 but opposite to that of ZEP and DOG1 during the course of AR (Figure 5c). The expression levels of HD2A and HD2D were also increased by AR (Figure S5a). In less-dormant Col-0 seeds, the expression levels of AR up-regulated genes, in particular GA3ox2 and AtEXP1, were continuously increased by AR even after seed dormancy had been released, suggesting that AR affects dormancy-related factor(s) even after seed dormancy has been apparently released.
The effect of ABA and GA deficiency on expression of HD2B was analyzed using ABA-deficient aba2-2 and GA-deficient ga1-3 mutant seeds. When 4-week after-ripened seeds were imbibed at 23°C in the light, wild-type Col-0 and ga1-3 seeds did not show radicle emergence and only small proportion of aba2-2 seeds germinated (1.83 ± 2.13%) after 24 h imbibition. After 4 days imbibition, the germination frequencies of Col-0 and aba2-2 reached more than 95%, whereas those of ga1-3 remained at 0%. Under the conditions, the expression levels of HD2B were increased by fourfold within 3 h of imbibition in wild-type Col-0 (Figure 5d). The degree of induction was greater in HD2B (4.0-fold) than GA3ox2 (2.1-fold) or AtEXP1 (1.2-fold) on 3 h imbibition. While expression of GA3ox2 and AtEXP1 was significantly enhanced or depressed in aba2-2 or ga1-3 mutants as reported previously (Ogawa et al., 2003; Seo et al., 2006), that of HD2B was not affected in these mutants, indicating that expression of HD2B was not affected by ABA and GA deficiency.
To further investigate the relationship between HD2B and seed dormancy, multiplex quantitative RT-PCR analysis was performed using seeds of 106 accessions that had been after-ripened for 2 months. These included 24 dormant and 28 less-dormant accessions (Data S1). When gene expression was analyzed after 24 h imbibition at 23°C in the light, expression levels of HD2B were positively associated with germination frequencies and negatively with seed dormancy (Figure 6). Welch's t test indicated that HD2B expression levels were significantly lower in 24 dormant accessions compared with 28 less-dormant accessions (P = 1.20e–5) (Figure 6). Of the dormant accessions, only Van-0 had HD2B expression levels more than one-third of those of Col-0. In less-dormant accessions, 13 accessions showed more than twofold higher expression levels of HD2B than any dormant accessions except Van-0. In contrast, another 10 less-dormant accessions showed low expression levels of HD2B similar to dormant accessions despite their high germination frequencies, suggesting that induction of HD2B expression may not be essential for the high germination frequencies in these accessions. We also analyzed the expression levels of GA biosynthesis genes (GA3ox1 and GA3ox2), dormancy- and germination-related genes (DOG1 and FLC), ABA signaling genes [ABA-INSENSITIVE3 (ABI3) and ABI5], and other HD2 genes (Figure 6 and Figure S5b). GA3ox1 and GA3ox2 showed significant differences in expression levels between dormant and less-dormant accessions (P < 0.001) (Figure 6), suggesting that GA biosynthesis is associated with germination in most less-dormant accessions. In addition, HD2A and HD2D showed expression patterns similar to those of HD2B (Figure S5b). Taken together, these results indicate that expression of HD2B was positively affected by AR but suppressed in most of the dormant accessions after imbibition.
Reduced seed dormancy is associated with enhanced HD2B expression in transgenic Cvi-0 seeds carrying ColHD2B
Because we could not obtain null T-DNA insertion mutants for HD2B, we used a transgenic approach to genetically assess whether modulation of HD2B expression affects seed dormancy or not. Col-0 is a less-dormant accession that exhibits high HD2B expression relative to other accessions, whereas Cvi-0 is a dormant accession with relatively low HD2B expression (Figure 6). Therefore, a 2.5 kb HD2B genomic DNA fragment cloned from Col-0 was transformed into dormant accession Cvi-0. The transformants, designated ColHD2B/Cvi-0, showed no changes in seed dormancy compared with the wild-type Cvi-0 during AR for up to 4 weeks (Figure 7a; −cold). However, after 8 weeks of AR, ColHD2B/Cvi-0 showed higher germination frequencies than the wild-type. In addition, when freshly harvested seeds or seeds after-ripened for 1–2 weeks were cold-stratified at 4°C for 7 days and then incubated at 23°C in the light, ColHD2B/Cvi-0 exhibited higher germination frequencies than the wild-type (Figure 7a,b). In both wild-type and ColHD2B/Cvi-0, the sensitivity of imbibed seeds to low temperature increased depending on the duration of AR, as reported previously (Finch-Savage et al., 2007). After 4 or 8 weeks of after-ripening, there was a small or no difference in seed dormancy between cold-stratified wild-type and ColHD2B/Cvi-0. In agreement with changes in seed dormancy, the expression levels of HD2B in freshly harvested seeds were more than fivefold higher in ColHD2B/Cvi-0 than in the wild-type when cold-imbibed (Figure 7c; +cold). In addition, the expression levels of HD2B in seeds after-ripened for 8 weeks were also more than threefold higher in ColHD2B/Cvi-0 than in the wild-type when imbibed at 23°C in the light (Figure 7c; − cold). In both wild-type and ColHD2B/Cvi-0, after-ripening and cold synergistically up-regulated the expression of HD2B. There was a significant correlation between germination frequencies and HD2B expression levels (Figure 7d).
Then we tested the effect of the HDAC inhibitor trichostatin A (TSA) on seed dormancy and germination. TSA inhibits HD2 and Rpd3-type HDACs (Jung et al., 1997; Ueno et al., 2007), and has been reported to delay germination in Col-0 seeds (Tanaka et al., 2008). When 4-week after-ripened dormant Cvi-0 seeds were cold-stratified with exogenous TSA and incubated at 23°C in the light, TSA treatment effectively suppressed testa rupture and cotyledon greening in a dose-dependent manner (Figure 8a), indicating that dormancy release and/or germination are suppressed by TSA in dormant Cvi-0 seeds. In contrast, when 6-month after-ripened Cvi-0 seeds were treated similarly, TSA no longer suppressed testa rupture, even at high concentrations. The frequencies of cotyledon greening were also higher in 6-month after-ripened seeds than in 4-week after-ripened seeds (Figure 8a), indicating that TSA sensitivity was reduced by AR in the Cvi-0 seeds. When the TSA sensitivity of germination was compared between wild-type Cvi-0 and ColHD2B/Cvi-0, ColHD2B/Cvi-0 seeds showed significantly lower sensitivity than the wild-type at 10 μm TSA (Figure 8b). In contrast, the ABA sensitivity of germination was not altered between wild-type Cvi-0 and ColHD2B/Cvi-0. These results suggest that TSA affects seed dormancy and/or germinability, partly through inhibition of HD2B.
Endogenous GA4 levels were increased in ColHD2B/Cvi-0
It is known that AR affects the endogenous levels of phytohormones including ABA and GA (Ali-Rachedi et al., 2004; Millar et al., 2006; Yano et al., 2009). To further investigate whether endogenous phytohormone levels were affected in ColHD2B/Cvi-0, comprehensive phytohormone analysis was performed using 8-week after-ripened seeds. When endogenous levels of six phytohormones, including ABA and gibberellin A4 (GA4), were simultaneously determined using LC-ESI-MS/MS, GA4 levels were found to be more than twofold higher in ColHD2B/Cvi-0 than in the wild-type Cvi-0 after 24 h imbibition at 23°C in the light (Figure 9a). The GA4 levels in ColHD2B/Cvi-0 seeds were comparable to those of less-dormant Col-0, despite the low germination frequencies of ColHD2B/Cvi-0. In contrast, endogenous levels of other phytohormones, including ABA, were not significantly changed in ColHD2B/Cvi-0 compared with the wild-type Cvi-0 (Figure 9a). Next, we tested the effect of TSA on phytohormone levels. TSA treatment decreased GA4 levels in all genotypes tested (Figure 9a). However, ColHD2B/Cvi-0 accumulated a higher amount of GA4 than the wild-type Cvi-0 even after TSA treatment. Then, we analyzed the expression levels of GA biosynthesis genes, GA inactivation genes, and dormancy- and germination-related genes in ColHD2B/Cvi-0 seeds (Figure 9b). In agreement with increased GA4 levels, expression levels of GA biosynthesis genes, such as GA3ox1, GA3ox2 and ent-KAURENE SYNTHASE1 (KS1), were increased in ColHD2B/Cvi-0 compared with the wild-type Cvi-0 when imbibed at 23°C in the light for 24 h. In contrast, expression levels of GA2ox2 encoding a GA inactivation enzyme were down-regulated in ColHD2B/Cvi-0 in the dark. Expression levels of ent-KAURENE OXIDASE1 (KO1), which is involved in GA biosynthesis, and ECP31, which encodes a late-embryogenesis abundant (LEA) protein, were down-regulated in ColHD2B/Cvi-0 compared with the wild-type regardless of illumination.
In the present study, we identified HD2B as a genetic factor associated with seed dormancy using an approach combining GWA mapping, transcriptome analysis, gene expression analysis and transgenic analysis. Previously, GWA studies of 107 phenotypes in Arabidopsis identified a number of SNPs that may account for natural variation in seed dormancy or germination (Atwell et al., 2010). However, none of the GWA studies had resulted in identification of novel genes, including HD2B. In GWA mapping, distinguishing true-positive from false-positive SNPs is difficult, because multiple statistical testing using a huge number of explanatory variables (SNPs) increases the risk of obtaining low P values. In this study, we overcame this difficulty by integrating GWA mapping and transcriptome datasets. We used a cold-induced dormancy cycling system in Cvi-0 to restrict candidate genes. We expected that the candidate SNPs would affect the expression of target genes and hence seed dormancy . Indeed, this approach allowed us to identify HD2B as the most plausible candidate associated with seed dormancy (Figure 3e and Figure S4a–c). The validity of this approach was further demonstrated by additional biological experiments. First we demonstrated that HD2B expression was positively affected by AR and short-term cold, consistent with a decrease in seed dormancy (Figure 5b,c). It is likely that HD2B expression is associated with seed dormancy rather than germination, because its expression was associated with changes in seed dormancy during cold-induced dormancy cycling rather than germination after transfer to light and warmth (Figure S3a,b). In addition, HD2B expression levels were significantly down-regulated in almost all of the 24 dormant accessions compared with less-dormant accessions (Figure 6). Most of the SNPs significantly associated with seed dormancy were located in the promoter, first intron or first exon of HD2B, suggesting that these SNPs affect the expression levels of HD2B and hence seed dormancy. This idea is supported by the finding that introduction of the HD2B gene cloned from Col (ColHD2B) into Cvi-0 resulted in increased HD2B expression levels with reduced seed dormancy. However, it is still not clear whether the SNPs that we identified are actually responsible for determining natural variation of seed dormancy among accessions. We could not exclude the possibility that the increased HD2B expression levels and reduced seed dormancy of ColHD2B/Cvi-0 resulted from the increase in gene dosage by the transgene rather than the allelic difference. In addition, it is likely that enhanced HD2B expression is not necessarily essential for germination in all of our less-dormant accessions, because ten of the 28 less-dormant accessions showed low expression levels of HD2B despite high germination frequencies (Figure 6). To investigate the precise relationship between SNPs, seed dormancy and HD2B expression, additional genetic studies using multiple combinations of dormant and less-dormant accessions are required. Nevertheless, our present study demonstrates that the expression of HD2B is associated with seed dormancy (determined as germinability under certain conditions) at least in some dormant accessions, including Cvi-0.
It is known that AR negatively affects ABA levels but positively affects GA levels in imbibed seeds (Ali-Rachedi et al., 2004; Millar et al., 2006; Yano et al., 2009). Our results suggestd that HD2B may affect GA4 accumulation in imbibed seeds, because endogenous GA4 levels were increased in ColHD2B/Cvi-0 seeds compared with the wild-type Cvi-0 when imbibed at 23°C in the light (Figure 9a). This notion was further supported by the observation that application of the HDAC inhibitor, TSA, decreased endogenous GA4 levels in imbibed seeds and that ColHD2B/Cvi-0 seeds, which showed enhanced expression of HD2B, accumulated higher levels of GA4 than the wild-type. Because ColHD2B/Cvi-0 showed lower TSA sensitivity of germination than the wild-type (Figure 8b), it is expected that TSA affects GA4 accumulation partly through inhibition of HD2B. However, in this context, it is considered that HD2B affects germination rather than seed dormancy, because germinating seed usually accumulates higher levels of GA4 than non-germinating seed (Toh et al., 2008; Preston et al., 2009; Yano et al., 2009). However, in most cases, germinating seeds exhibit a simultaneous decrease in ABA levels (<10 ng g−1 dry seed fresh weight) in addition to the increase in GA4 levels compared with non-germinating seeds. In contrast, ColHD2B/Cvi-0 seeds accumulated high levels of ABA, similar to the wild-type Cvi-0 (>20 ng g−1 dry seed fresh weight; Figure 9a). Although these results do not exclude the possibility that HD2B is involved in germination, ColHD2B/Cvi-0 seeds appear to differ from usual germinating seeds with regard to phytohormone accumulation. Interestingly, TSA treatment affected endogenous ABA levels in addition to GA4 levels (Figure 9a). Because TSA inhibits microtubule deacetylase (Matsuyama et al., 2002) in addition to HDACs (Yoshida et al., 1990) in mammalian cells, there is a possibility that TSA affects the endogenous ABA/GA balance in imbibed seed through inhibition of non-HDAC deacetylases. However, in Arabidopsis, the Rpd3-type HDAC genes HDA6 and HDA19 have been reported to affect seed germination through repression of embryonic phase transition genes, such as FUS3 and ABI3 (Tanaka et al., 2008). Because FUS3 negatively affects GA3ox1 expression and positively affects ABA accumulation (Gazzarrini et al., 2004), it is possible that inhibition of HDA6 and HDA19 contributes to an altered ABA/GA balance in TSA-treated seeds through de-repression of FUS3. More recently, it was reported that HD2 proteins physically interact with HDA6 and HDA19 (Luo et al., 2012a,b), suggesting that HD2 and Rpd3 HDACs act in the same protein complex. However, because expression levels of FUS3 and ABI3 were not affected in ColHD2B/Cvi-0 compared with the wild-type (Figure 9b), HD2B appears to play a distinct role from HDA6 and HDA19 in imbibed seed. To address the precise function of HD2B in seed dormancy, germination and phytohormone metabolism, it is necessary to compare global histone acetylation patterns and gene expression between wild-type Cvi-0 and Co-0HD2B/Cvi-0 seeds.
Previous QTL studies reported that DOG QTLs were associated with natural variation for seed dormancy (Alonso-Blanco et al., 2003; Bentsink et al., 2010). Although we detected SNPs strongly associated with seed dormancy around the DOG6 QTL region (Figure 2b–d), questions remain about the relationship between DOG QTLs and the SNPs identified in this study. Our GWA mapping failed to identify strong candidate SNPs within the DOG1 genomic region, although SNPs close to DOG1 were detected as significantly associated (Figure 2c,d). Also, no strong DOG QTL has been reported near the HD2B genomic region. We propose three explanations for these discrepancies. First, experimental conditions differed between the QTL studies and this study. Plants were grown under diurnal light/dark conditions to obtain seeds for the QTL studies, but we grew plants under continuous white-light illumination. Because maternal photoperiod affects seed dormancy in Arabidopsis (Munir et al., 2001), differences in plant growth conditions may affect dormancy behavior in each accession. Second, the AtSNPtile1 dataset may not contain the SNPs of DOG1 associated with seed dormancy. The GWA mapping reported by Atwell et al. (2010) also failed to identify strongly associated SNPs within the DOG1 locus. However, DOG1 haplotypes identified by dideoxy sequencing were recently shown to be associated with natural variation for seed dormancy in 384 accessions (Kronholm et al., 2012). Third, there were differences in experimental populations between the QTL studies and this study. While 113 accessions were used for GWA mapping in this study, the QTL studies used RIL populations derived from crosses between Ler and other accessions. Interestingly, QTL studies have demonstrated that there are accession-specific QTLs in addition to commonly detected QTLs in Arabidopsis (Bentsink et al., 2010). In addition, another QTL study has reported that variation for seed dormancy is associated with the DOG1 QTL in the population derived from the cross between Ler and Kyoto-2 but not in that of Ler and Kyoto-2 (Silady et al., 2011). Therefore, natural variation for seed dormancy may be regulated by complex genetic pathways, including both major and minor pathways. HD2B may be a minor pathway in the populations used in the QTL studies. In any case, genetic analyses should be performed using various combinations of accessions with a comprehensive genotype dataset under distinct experimental conditions. Such an approach will enable elucidation of the detailed genetic architecture underlying natural variation for seed dormancy.
Plant materials and growth conditions
The 117 Arabidopsis accessions used in this study were obtained from the RIKEN BioResource Center (http://www.brc.riken.jp/) and the Arabidopsis Biological Resource Center (Data S1). ga1-3 in the Col-0 background (Tyler et al., 2004) was a gift from Dr. Tai-ping Sun (Duke University). aba2-2 was isolated in the Col-0 background as reported previously (Nambara et al., 1998). To obtain seeds for experiments, plants were grown under constant irradiation at 23°C, as described previously (Yano et al., 2009). To ensure simultaneous germination among accessions, seeds were after-ripened by dry storage for more than 2 months, sown on half-strength Murashige and Skoog medium containing 10 μm 1-aminocyclopropane-1-carboxylic acid, then chilled at 2–5°C in the dark for 7 days. After incubation at 23°C in the light, germinated seedlings were further chilled at 2–5°C in the dark for at least 4 weeks to promote simultaneous flowering. For each accession, seeds were freshly harvested approximately 4–5 weeks after fertilization from 12 plants grown in four independent plastic pots. To prepare independent seed batches, plants were sown at different times. For after-ripening, freshly harvested seeds were dry-stored in the dark at 23°C at approximately 30% humidity. For seed imbibition, freshly harvested seeds or dry-stored (after-ripened) seeds were sown on 0.8% w/v agarose (LO-3, Takara, www.takara-bio.com/). If illuminated, imbibed seeds were incubated under continuous white light illumination (43 μmol m−2 sec−1). Dry-stored and imbibed seeds were frozen in liquid N2 and stored at −80°C until use.
Seed germination test and TSA sensitivity test
To compare seed dormancy and germination, seeds were imbibed on 0.8% w/v agarose (LO-3, Takara). In each germination test, approximately 50 seeds were used, and radicle emergence was scored as seed germination. For the TSA sensitivity test, seeds were sown on 0.8% w/v agarose containing various concentrations of trichostatin A (S1045, Selleckchem, www.selleckchem.com/), and testa rupture and cotyledon greening were scored instead of radicle emergence, because radicle emergence was difficult to determine in these seeds.
GWA mapping and data analysis
GWA mapping was performed in 113 accessions (Data S1) using the AtSNPtile1 SNP datasets previously published (Atwell et al., 2010; Li et al., 2010). Prior to GWA mapping, the array information was updated to version TAIR10 using Perl scripts based on probe sequences and sequence datasets obtained from the Arabidopsis Information Resource (ftp://ftp.arabidopsis.org/home/tair). The TAIR10 genetic loci were searched for each intragenic and intergenic SNP using Perl scripts, and adjacent loci were searched in cases of intergenic SNPs. Computation of P values was performed for 210 253 SNPs using EMMAX (Kang et al., 2010) and PLINK (Purcell et al., 2007) software installed on a computer operating 64-bit Ubuntu® 10.04 and 11.10. For data analysis, SNPs with minor allele frequencies ≥0.05 were evaluated. Regression analysis between genotypes and germination frequencies was performed using the function ‘lm’ implemented in R 2.13.0 (R Foundation for Statistical Computing, www.r-project.org/) for each polymorphism. Manhattan plots and box plots were drawn using R 2.13.0.
Transcriptome analysis and integrated analysis with GWA mapping
Transcriptome analysis was performed using Affymetrix ATH1 GeneChip (www.affymetrix.com) as described previously (Preston et al., 2009). Affymetrix ‘present’, ‘marginal’ and ‘absent’ flags were used to indicate whether or not a gene was expressed, and 10 652 genes with ‘present’ flags in all of either the 23°C- or 2°C-treated samples were used for subsequent data analysis. Regression analysis between gene expression levels and germination frequencies was performed using the function ‘lm’ implemented in R 2.13.0 for each gene. Co-expression analysis was performed using the package ‘WGCNA’ (Langfelder and Horvath, 2008) implemented in R 2.15.0 and Cytoscape 2.8.3 (Cytoscape Consortium, www.cytoscape.org). Integration of GWA mapping and transcriptome analysis was performed by simply comparing the –log10 (P values) from ATH1 GeneChip analysis and GWA mapping. In the case of intergenic SNPs, the −log10 (P values) were compared with those of adjacent genes. Data analysis was performed using Perl scripts, R 2.13.0 and Microsoft Excel® (www.microsoft.com).
Construction of ColHD2B/Cvi-0 transgenic plants
To construct ColHD2B/Cvi-0 plants, a 2.5 kb fragment of HD2B, including the putative promoter region and 3′ untranslated region, was amplified from Col-0 genomic DNA PCR using the primers 5‘-CACCgcacttcaaggcttctttttttggtg-3‘ and 5‘-ctcaaaatagctcatccgataacaag-3‘ (the underline indicates the sequence for directional cloning into the vector) and cloned into the pENTR/D-TOPO vector (Invitrogen, www.invitrogen.com). After the sequence had been verified, the cloned genomic DNA was inserted into a pGWB1 binary vector (Nakamura et al., 2010) by the LR clonase reaction according to the manufacturer's instructions (Invitrogen). The constructed vectors were introduced into Agrobacterium tumefaciens strain GV3101 by electroporation, and transformed into the Arabidopsis Cvi-0 accession by floral dipping (Clough and Bent, 1998). The resultant T1 transformants were selected on half-strength Murashige and Skoog medium containing 40 mg ml−1 hygromycin and 25 mg ml−1 kanamycin. Four lines of ColHD2B/Cvi-0 T3 homozygotes were established and used for further analysis.
Comprehensive phytohormone analysis
Endogenous levels of gibberellin A4, abscisic acid, indole-3-acetic acid, isoleucine-conjugated jasmonic acid, trans-zeatin and salicylic acid were simultaneously quantified using an Agilent 6410 LC-ESI-MS/MS (www.agilent.com) as described previously (Yano et al., 2009). In each analysis, 30 mg of dry seeds were imbibed and then used for phytohormone extraction.
Experimental procedures for isolation of total RNA and genomic DNA, quantitative RT-PCR analysis, and dideoxy sequencing of the HD2B genomic DNA region are provided in Methods S1.
We thank Masatomo Kobayashi and Satoshi Iuchi (RIKEN BioResource Center), and Fumio Matsuda (Kobe University) for providing Arabidopsis seeds and Tsuyoshi Nakagawa (Shimane University) for the pGWB1 binary vector. We also thank Shinjiro Yamaguchi (Tohoku University) and Jutarou Fukazawa (Hiroshima University) for helpful comments and suggestions, Sachiko Oyama and Seiko Nomura for their contributions to dideoxy DNA sequencing and Affymetrix ATH1 GeneChip analysis, and Masako Tanaka for her assistance with plant growth and seed harvesting. This work was supported by the RIKEN President's Discretionary Fund to R.Y.