Seed dormancy in Arabidopsis thaliana is controlled by alternative polyadenylation of DOG1 − Rakyat Biologi

The ability to postpone germination, to align it with permissive environmental conditions, greatly enhances the chances of plants survival (Koornneef et al., 2002; Penfield and King, 2009). Therefore, seed dormancy is extensively controlled by many factors including: humidity, temperature and light, and is also subject to parental memory (Chiang et al., 2011; Finch-Savage and Leubner-Metzger, 2006; Graeber et al., 2012; Nonogaki, 2014). DOG1 (Delay of Germination 1) has been identified as a major QTL (Quantitative Trait Locus) for seed dormancy variability among natural Arabidopsis accessions and dog1 T-DNA insertional mutants exhibit reduced seed dormancy (Bentsink et al., 2006). The expression of DOG1 increases during seed maturation and the mRNA disappears quickly after imbibition, although the DOG1 protein is more stable (Nakabayashi et al., 2012). A recent report indicated that DOG1 enforces seed dormancy by strengthening the seed coat, which occurs via modulation of the expression of gibberellin metabolism genes (Graeber et al., 2014). The expression of DOG1 is extensively regulated, being strongly induced by ABA (Abscisic acid) and low temperature during seed maturation (Chiang et al., 2011; Kendall et al., 2011). Other factors required for DOG1 expression include histone-modifying enzymes such as the histone H2B ubiquitin transferase HUB (Liu et al., 2007) and H3 lysine 9 methyl transferase KYP (Zheng et al., 2012). DOG1 expression is also highly dependent on transcription elongation factor TFIIS (Grasser et al., 2009).

The DOG1 gene is comprised of 3 exons, and the exon 2-exon 3 junction is subject to alternative splicing, generating 4 different forms of mRNA (Bentsink et al., 2006). The function of these alternatively spliced transcript isoforms is unknown, and their relative ratio remains unchanged during seed development (Bentsink et al., 2006). Our recent analysis of the regulation of alternative splicing of DOG1 in a mutant defective in PolII elongation suggested that the rate of transcript elongation regulates alternative splice site selection in accordance with the kinetic coupling model and implies that DOG1 splicing is co-transcriptional (Dolata et al., 2015).

To our knowledge complementation of the dog1 mutation in Arabidopsis has not been achieved using a DOG1 cDNA, while the seed dormancy phenotype of this mutant was complemented using a genomic DOG1 clone from Lepidium sativum (L. sativum) (Graeber et al., 2014). Notably, the L. sativum DOG1 gene lacks the exon 3, so can only encode a short two-exonic mRNA, with no alternatively spliced exon 2-exon 3 isoforms.

Many alternatively spliced genes are also subject to alternative polyadenylation (APA) (Di Giammartino et al., 2011). Alternative polyadenylation leads to the generation of transcripts with different 3' ends, through a series of steps catalyzed by components of RNA 3’ processing complexes, like: the Cleavage and Polyadenylation Specificity Factor (CPSF) and Cleavage Stimulation Factor (CstF) complexes, and poly(A) polymerases (Gruber et al., 2014; Proudfoot, 2011; Mandel et al., 2008). APA is common and widespread in animals, plants and other eukaryotic organisms (Tian et al., 2005; Pickrell et al., 2010; Sun et al., 2012; Wu et al., 2015; Shi, 2012). In animals APA is involved in a range of developmental processes including cell differentiation and has been implicated in cancer (Danckwardt et al., 2008; Mayr and Bartel, 2009; Lianoglou et al., 2013; Lin et al., 2012). Similarly, in plants APA has been shown to control key developmental processes.

In one case, the nuclear RNA-binding protein FCA interacts with FY, a component of CPSF complex, to promote the usage of a proximal polyadenylation site in its own gene, leading to the production of a non-functional RNA isoform (Simpson et al., 2003). In addition, FCA functions with CstF complex to promote proximal polyadenylation of the non-coding antisense transcript of FLC (Flowering Locus C), leading to suppression of FLC expression (Liu et al., 2010). Alternative polyadenylation has also been implicated in the control of pathogen resistance in Arabidopsis through the selection of proximal polyadenylation sites in the RPP7 (Recognition Of Peronospora Parasitica 7) gene (Tsuchiya and Eulgem, 2013).

Given the key function of DOG1 in Arabidopsis seed survival and the potential role of DOG1 homologs in controlling seed dormancy in other species, it is important to understand both the mechanisms of DOG1 locus regulation and the function of the encoded protein. Here we describe the process of DOG1 alternative polyadenylation that produce two alternatively polyadenylated isoforms of the DOG1 transcript. We characterize mutants in proteins that control this mechanism, and show that the proximally polyadenylated mRNA isoform is translated in vivo, and is functional as it complements the dog1 mutant. Demonstrating that alternative polyadenylation of DOG1 gene in Arabidopsis plays a fundamental role in regulation of seed dormancy.

Mutants of RNA 3’ processing factors display weak seed dormancy phenotype

The majority of Arabidopsis genes have multiple polyadenylation sites (Wu et al., 2011), but the biological consequences of the alternative polyadenylation are in most cases unknown. FY is the Arabidopsis homologue of the yeast RNA 3' processing factor Pfs2p (Simpson et al., 2003). A previous study demonstrated that the fy-1 mutant in Landsberg erecta (Ler) background has weak seed dormancy (Jiang et al., 2012). To check if the function of FY in seed dormancy control is independent of genetic background, we assayed seed dormancy of fy-2 mutant in Col-0 background. Compared to the Col-0 wild type, fy-2 seeds showed very weak dormancy (Figure 1A) demonstrating that FY plays a significant role in controlling this process in Arabidopsis. Since FY functions with other RNA 3’ processing factors in flowering time control (Liu et al., 2010; Manzano et al., 2009), we analyzed seed dormancy of the mutants of RNA 3’ processing factors PCFS4 (PCF11P-Similar Protein 4) and ESP1 (Enhanced Gene Silence 1). Like fy mutants, the pcsf4-1 and esp1-2 mutants showed weakened seed dormancy (Figure 1B), suggesting that the observed seed dormancy defect is caused by misregulation of RNA 3’ processing of a gene or genes involved in seed dormancy control.

Next, we examined whether the weak seed dormancy phenotype of fy-2 may be due to misregulation of DOG1 transcription. RT-qPCR (reverse transcription and quantitative PCR) analysis revealed no significant change of DOG1 mRNA level in fy-2 mutant (Figure 1C). However, western blot analysis using a DOG1-specific antibody revealed a consistent reduction in the DOG1 protein level in fy-2 seeds when compared with Col-0 wild type (Wt) (Figure 1D and Supplemental Figure S1). Thus, we concluded that FY is required for proper DOG1 protein expression and hypothesized that the fy-2 mutant may be defective in DOG1 RNA processing, which leads to the suppression of translation. Therefore, we searched for potential DOG1 polyadenylation defects in fy-2. Surprisingly, 3' RACE experiments initiated from exon 3 of DOG1 showed no obvious differences in either polyadenylation site selection or the length of the poly(A) tail, between Col-0 wild type and fy-2 seeds (Figure 1E).

Alternative polyadenylation of DOG1 gene results in production of two mRNA isoforms

As we were unable to detect any defect in the use of the canonical polyadenylation site of DOG1 in the fy mutant, we searched for additional polyadenylation sites within this gene. The most common polyadenylation signal motif in plants is AAUAAA and UUGUUU positioned 19 and 7 nt respectively upstream of the cleavage site (Sherstnev et al., 2012). The identification of likely polyadenylation sites with PASPA (Ji et al., 2015), a software designed for mRNA polyadenylation site prediction in plants, revealed three potential polyadenylation clusters in the DOG1 gene. One cluster corresponds to the predicted full length polyadenylation site of DOG1, and the two others to additional internal polyadenylation sites: one in intron 1 and another in intron 2 (Figure 2A). To validate these predictions, we reanalyzed published Direct RNA Sequencing based mapping data to detect Arabidopsis polyadenylation sites in the DOG1 locus (Supplemental Figure S2) (Sherstnev et al., 2012). We identified two distinct polyadenylation clusters: a distal one used to produce the predicted full-length mRNA, which is named hereafter as the long DOG1 (lgDOG1) form, and a proximal one matching one of the predicted internal polyadenylation sites located in intron 2 that would produce a truncated mRNA, which we name short DOG1 (shDOG1). The second predicted internal polyadenylation site lies within intron 1, but analysis of RNA sequencing data revealed no indication of its usage (Supplemental Figure S2).

In summary, the lgDOG1 transcript is comprised of three exons and corresponds to the previously described Arabidopsis DOG1 mRNA (Bentsink et al., 2006), whereas the newly identified shDOG1 transcript comprises only exons 1 and 2, and therefore lacks the alternative splicing isoforms described (Figure 2B).

Next, to quantify the different DOG1 mRNA isoforms in Col-0 wild type and fy-2 mutant seeds we designed sets of primers within the DOG1 locus for use in RT-qPCR analysis. Amplification with primer set 4 showed no difference in the level of lgDOG1 mRNA between fy-2 and wild type seeds (Figure 2A and Figure 2C). In contrast, use of primer set 3 demonstrated a clear reduction of the DOG1 transcript in fy-2 seeds. The primers of set 3 span the exon 2-intron 2 border that is specific for the short alternatively polyadenylated shDOG1 mRNA isoform. Therefore, this result suggested a reduction in usage of the proximal polyadenylation site in the fy-2 mutant. To confirm this, we designed primers to amplify the full-length short and long DOG1 transcript isoforms and obtained similar results (Supplemental Figure S3).

Different mRNA isoforms generated through APA often differ in their stability (Krol et al., 2015). However, a cordycepin-dependent RNA stability assay (Golisz et al., 2013) showed that both the short and long DOG1 mRNA isoforms have a similar half-life of about 1 h (Supplemental Figure S4). This shows that usage of DOG1 proximal polyadenylation site leads to the production of a stable transcript.

A change in the long/short Dog1 transcript ratio, similar to that observed in fy-2, was also detected in mutants of other factors required for RNA 3’ processing, which showed weak seed dormancy (Figure 2D and Figure 1B). This finding led us to speculate that the reduction in the shDOG1 mRNA level in these mutants, may be the underlying cause of their reduced seed dormancy.

Having established that the DOG1 gene is subjected to alternative polyadenylation, we questioned whether this process is developmentally regulated. DOG1 expression is tightly controlled during seed development and has previously been shown to peak between 9-16 days after pollination and then slowly decay (Nakabayashi et al., 2012; Zhao et al., 2015). Alternative splicing of the lgDOG1 mRNA isoform has been reported to be unaffected during seed development (Bentsink et al., 2006). Our analysis of short and long DOG1 mRNA levels showed that both forms are induced during seed development and show a similar expression profile (Figure 2E).

Short DOG1 mRNA is translated in vivo and produce a conserved protein

The previously recognized long DOG1 mRNA transcript (lgDOG1) encodes a protein of ~32 kD. The proximally polyadenylated shDOG1 transcript described in this study codes for a slightly smaller ~30 kD protein. These short and long DOG1 proteins only differ by several amino acids at their C-terminal ends, encoded by exon 3 in lgDOG1 and a sequence from intron 2 in shDOG1 (Figure 3A). Interestingly, a multiple sequence alignment with DOG1 proteins from other species showed that the sequence unique to shDOG1 is conserved at a similar level compared to that encoded by exon 2, while the amino acids unique to lgDOG1 (exon 3) are either absent due to stop codons or are weakly conserved (Figure 3A). Evolutionary conservation is often considered an indication of functionality, suggesting that the shDOG1 protein is functional.

Next, we determined which of the DOG1 mRNA isoforms is translated. DOG1 antibodies used previously and that described here do not distinguish the long and short DOG1 proteins because they were raised to peptides shared by the two isoforms (Nakabayashi et al., 2012). In addition, the small difference in protein mass does not allow distinction between the shDOG1 and lgDOG1 proteins on a western blot: only one DOG1 band is observed in samples from freshly harvested seeds (Figure 1D). Therefore, we created a GFP::DOG1 genomic fusion driven by the DOG1 promoter. Samples from transgenic plants expressing this fusion protein were used in a GFP pull-down assay (Figure 3B) with subsequent analysis by mass spectrometry. Notably, no long DOG1-specific peptides were identified, but we could clearly detect peptides unique to the shDOG1 protein (Figure 3C and Supplementary Figure S5). Although this does not exclude the possibility that the long DOG1 form is expressed but not detected, this result clearly showed that use of the proximal polyadenylation site leads to the production of a short DOG1 mRNA that is translated in vivo.

A previous study examining Arabidopsis lines constitutively overexpressing DOG1 suggested nuclear localization of this protein (Nakabayashi et al., 2012). However, this experiment did not distinguish between the short and long DOG1 forms. Therefore, we compared the localization of these DOG1 proteins. Both N- and C-terminal GFP fusions of the shDOG1 and lgDOG1 protein showed predominantly nuclear localization when transiently expressed in Arabidopsis protoplasts (Figure 3D).

We conclude that the short alternatively polyadenylated DOG1 mRNA isoform is translated in vivo, thus producing a short nuclear localized DOG1 protein. The fact that the shDOG1 is conserved whereas the lgDOG1 is less together with our inability to find long DOG1 specific peptides suggested that the shDOG1 maybe the functional DOG1 protein in respect to seed dormancy.

Short DOG1 is sufficient to inhibit germination

To directly test the ability of the short and long DOG1 proteins to control seed dormancy in Arabidopsis we created dog1-3 mutant plants with constructs carrying the shDOG1 and lgDOG1 sequences driven by the DOG1 promoter (Figure 4A). Only the shDOG1 construct was able to partially complement the weak seed dormancy phenotype of dog1-3 (Figure 4A), even though both showed a similar DOG1 expression as measured by qRT-PCR (Supplementary Figure S6). This indicated that shDOG1 protein is sufficient for seed dormancy establishment. To further support this conclusion we selected a dog1-5 T-DNA insertion mutant allele. In this mutant a T-DNA is inserted in exon 3 of the DOG1 gene (Figure 4B). RT-qPCR analysis with primers located at the exon 2-exon 3 junction, upstream of the T-DNA insertion, confirmed that lgDOG1 is no longer expressed in this mutant (Figure 4C). In contrast, levels of the shDOG1 transcript appear slightly increased in dog1-5. Accordingly, the dog1-5 mutant exhibited an increase in the DOG1 protein level, and increased rather than reduced seed dormancy (Figure 4D and Figure 4E). These findings confirmed that the short version of the DOG1 transcript generated through APA is translated and functional in controlling seed dormancy.