HSI2/VAL1 and HSL1/VAL2 function redundantly to regulate seed dormancy by controlling DOG1 expression in Arabidopsis

DELAY OF GERMINATION1 (DOG1) represents a major quantitative locus for the genetic regulation of seed dormancy in Arabidopsis. Accumulation of DOG1 in seeds leads to deep dormancy and delayed germination. Here, we report that the conserved B3 DNA binding domains of the transcriptional repressors HIGH-LEVEL EXPRESSION OF SUGAR INDICIBLE GENE2/ VIVIPAROUS-1/ABSCISIC ACID INSENSITIVE 3-LIKE1 (HSI2/VAL1) and HSI2-LIKE1/ VIVIPAROUS-1/ABSCISIC ACID INSENSITIVE 3-LIKE2 (HSL1/VAL2), which play critical roles in the developmental transition from seed maturation to seedling growth, interact with RY elements in the DOG1 proximal promoter leading to repression of DOG1 transcription during germination and seedling establishment. DOG1 expression is partially de-repressed in hsi2/val1 (hsi2-2) but not in hsl1/val2 (hsl1-1) knockout mutants and is strongly upregulated in a hsi2/val1 hsl1/val2 double mutant, indicating that HSI2/VAL1 and HSL1/VAL2 act redundantly to repress DOG1 expression. HSI2/VAL1 and HSL1/VAL2 form homo- and hetero-dimers in vivo, and dimerization is dependent on the HSI2/VAL1 PHD-like domain. Complementation of hsi2-2 with HSI2/VAL1 harboring a disrupted plant homeodomain (PHD)-like domain results in stronger de-repression of DOG1 expression than the hsi2-2 knockout, indicating that the PHD-like domain plays a critical role in mediating functional interactions between HSI2/VAL1 and HSL1/VAL2. Both HSI2/VAL1 and HSL1/VAL2 interact with components of polycomb repressive complex 2 (PRC2), including CURLY LEAF and MULTICOPY SUPPRESSOR OF IRA1 (MSI1), along with LIKE HETERCHROMATIN PROTEIN 1 (LHP1), which are involved in the deposition and expansion of histone H3 lysine 27 trimethylation (H3K27me3) marks in repressive chromatin. Thus, HSI2/VAL1 HSL1/VAL2-dependent recruitment of PRC2 leads to silencing of DOG1 through the deposition of H3K27me3.


57
Seed dormancy is an adaptive mechanism that allows for the dispersal and survival of seeds over 58 distance and time, and ensures that germination occurs under favorable conditions (Finch-Savage 59 et al., 2006). Seed dormancy is a complex trait that is regulated by both phytohormones and 60 genetic factors. Abscisic acid (ABA) plays an important role in initiating and enhancing seed 61 dormancy, while the dormant state is reversed by gibberellins, which promote germination under 62 favorable conditions (Koornneef et al., 2002;Liu et al., 2010). Recently, DELAY OF 63 GERMINATION1 (DOG1) was reported to be a major quantitative trait locus for the genetic 64 regulation of seed dormancy in Arabidopsis (Alonso-Blanco et al., 2003;Bentsink et al., 2006, C (FLC) and AGAMOUS-LIKE15 (AGL15) (Qüesta et al., 2016;Yuan et al., 2016;Chen et al., 95 2018). The HSI2 PHD domain is also required for HSI2 accumulation at the AGL15 locus and 96 repression of AGL15 expression (Chen et al., 2018). HSI2 and HSL1 can form homodimers and 97 heterodimer in vivo and HSL1 shows partial redundancy to HSI2 in the regulation of FLC 98 (Chhun et al., 2016;Yuan et al., 2016). HSI2 is reported to recruit polycomb repressive complex 99 2 (PRC2) to the AGL15 and FLC loci (Qüesta et al., 2016;Yuan et al., 2016;Chen et al., 2018).

100
HSI2 was also reported to bind HISTONE DEACETYLASE6 (HDA6) and MEDIATOR13 101 (MED13) and HSL1 was reported to interact with HDA19, indicating that these factors may also 102 be involved in the repression of seed maturation genes during germination in Arabidopsis 103 (Chhun et al., 2016;Zhou et al., 2013). The molecular mechanisms of HSL1 function and its 104 relationship with HSI2 has not been fully characterized. 105 106 LIKE HETEROCHROMATIN PROTEIN1 (LHP1) recognizes H3K27me3 (trimethylation of 107 lysine at histone H3) repressive marks through its chromodomain (Turck et al., 2007;Zhang et 108 al., 2007;Exner et al., 2009) like complexes and may act as a bridge between PRC1 and PRC2 (Xu et al., 2008;Bratzel et al., 113 2010). In Arabidopsis, LHP1 co-localizes with H3K27me3 across the genome, and is responsible 114 for the expansion of H3K27me3 associated with the stabilization of transcriptional repression 115 (Turck et al., 2007;Zhang et al., 2007;Exner et al., 2009). Expression of many tissue specific 116 genes is upregulated in lhp1 loss-of-function mutants (Lafos et al., 2011;Libault et al., 2005), 117 5 with decreased levels of H3K27me3 seen at direct target gene loci, including FLC ( Yuan et al.,118 2016; Veluchamy et al., 2016), indicating that LHP1-dependent gene repression correlates with 119 deposition of H3K27me3. Molitor et al. (2014) reported that DOG1 is negatively regulated by 120 ALFIN1-like proteins, which contain PHD domains and can form protein complexes with LHP1 121 at the DOG1 locus to replace H3K4me3 (trimethylation of lysine 4 at histone H3) marks 122 associated with gene activation with H3K27me3, leading to its transcriptional downregulation 123 and promotion of seed germination.

125
Here we report that the proximal region of the DOG1 promoter is directly targeted by both HSI2 HSI2 and HSL1 redundantly repress DOG1 to regulate seed dormancy 134 Reverse transcription quantitative PCR (RT-qPCR) analysis showed that DOG1 expression was 135 not significantly affected, relative to wild-type (WT), in freshly harvested hsl1-1 Arabidopsis 136 seeds but its expression increased by approximately 3-fold in hsi2-2 and by 14-fold in the hsi2 137 hsl1 double mutant ( Figure 1A), indicating HSI2 and HSL1 play overlapping roles in the 138 repression of DOG1 expression. Analysis of the rates of germination of freshly-harvested seeds 139 from these Arabidopsis lines confirmed that loss-of-function mutations in both hsi2 and hsl1 are 140 necessary for significant increases in seed dormancy to be detected ( Figure 1B). The gain-of-141 function mutant dog1-5, which shows strongly delayed germination, and the loss-of-function 142 mutant dog1-3, which shows early germination, served as controls in this assay. These results

143
indicate that HSI2 and HSL1 act redundantly to repress DOG1 expression in seeds. Though their 144 functions overlap, the native HSI2 gene is able to fully complement HSL1, while HSL1 cannot 145 fully replace the function of HSI2. Furthermore, while seed dormancy is promoted by DOG1 146 expression, a relatively large increase in DOG1 expression, as seen in the hsi2 hsl1 double 147 mutant, appears to be necessary to significantly affect seed dormancy.
148 149 HSI2 directly regulates DOG1 by binding to its promoter 150 Since expression of DOG1 is upregulated and H3K27me3 enrichment at the DOG1 locus is 151 reduced in hsi2-2 knockout mutant Arabidopsis seedlings (Veerappan et al., 2012(Veerappan et al., , 2014, we 152 predicted that DOG1 could be a direct regulatory target of HSI2. To test this hypothesis, a 153 rescued hsi2-2 Arabidopsis line that expresses HA epitope-tagged HSI2 under the control of the 154 native HSI2 promoter (HSI2pro:HSI2-HA) was developed. Repression of DOG1 was restored in 155 this line with expression reduced, relative to hsi2-2, to levels similar to WT plants ( Figure 1C).

156
To detect and quantify the enrichment of HSI2 at the DOG1 locus, a number of different primer 157 sets were used to analyze promoter (P) and coding (C) regions of the DOG1 gene by chromatin 158 immunoprecipitation quantitative PCR (ChIP-qPCR) ( Figure 1D). Significant HSI2 enrichment 159 was detected at the proximal promoter regions (P1 and P2) of DOG1 with the highest level of The B3 domain of HSI2 (B3 HSI2 ) was reported to play an essential role in the regulation of 173 AGL15 and FLC by binding to RY elements in the proximal promoter or first intron of these 174 genes, repectively (Qüesta et al., 2016;Yuan et al., 2016;Chen et al., 2018). We examined the 175 function of the B3 HSI2 domain in the regulation of the DOG1 promoter by transient expression 176 assays. An effector construct that encodes HSI2 with a mutated B3 domain (HSI2mB3) was co-177 infiltrated with the DOG1pro:LUC reporter construct into N. benthamiana leaves (Figure 2A).

178
The DOG1pro:LUC reporter gene was strongly repressed by co-expression with intact HSI2; 179 however, co-expression of HSI2mB3 resulted in high levels of reporter gene expression 180 indicating that the B3 domain is required for HSI2-dependent repression of the DOG1 promoter.

182
Sequence analysis indicated that RY elements are distributed in the DOG1 proximal promoter 183 between -621 and -394 bp upstream of the transcriptional start site. The region corresponding to 184 P1 in Figure 2B contains overlapping RY elements (RY1, 5' TGCATGCATG 3' ), while the P2 185 region contains a single RY element (RY2, 5' TGCATG 3' ) ( Figure 2B). The expression of 186 luciferase reporter genes under control of the intact DOG1 promoter or promoters with 187 disruptions in RY1, RY2 or both were tested in Arabidopsis protoplasts using a dual luciferase 188 assay. As show in Figure 2B, loss of both RY1 and RY2 resulted in a 2.5-fold increase in 189 expression relative to the intact promoter, while mutation of RY1 led to a 2-fold increase and 190 RY2 disruption gave a 1.5 fold increase. Thus, the DOG1 RY elements appear to act additively 191 with the RY1 element showing a somewhat stronger repressive effect than RY2. To test whether the B3 HSI2 domain can bind directly to the RY elements in the DOG1 promoter in 194 vitro, we performed electrophoretic mobility shift assay (EMSA).  DNA probes from the P1 and P2 regions of the DOG1 promoter were labeled with biotin at the 3' 196 end and incubated with His-Trx-tagged B3 HSI2 fusion protein. A strong signal indicating binding 197 of B3 HSI2 to probes containing the complex RY1 sequence element was detected, but the signal 198 for binding to probe containing the RY2 element was much weaker ( Figure 2C). Probe with a 199 mutated RY1 element failed to bind B3 HSI2 , and a B3 HSI2 polypeptide with a disrupted B3 domain  To test whether the B3 domain is required for HSI2 enrichment at the DOG1 locus, we carried  ChIP-qPCR analysis showed that, unlike intact HSI2, HSI2mB3 fails to accumulate at the 212 proximal promoter region of the DOG1 locus ( Figure 2E). Similar results were obtained for the 213 enrichment of HSI2 at the AGL15 locus, which was included as a positive control in this ChIP-214 qPCR analysis. were significantly reduced when DOG1pro:LUC was co-expressed with intact HSI2, but co-223 expression with HSI2mPHD had no effect on reporter gene expression ( Figure 3A). Transgenic  Figure 3B). Rather, expression of HSI2mPHD in hsi2-2 plants resulted in an increase in DOG1 226 expression by more than 2-fold, relative to the hsi2-2 null allele, though expression in these 227 plants was not as high as in hsi2 hsl1 double mutant seedlings ( Figure 3B). To test if the PHD 228 domain is also required for HSI2 enrichment at the DOG1 locus, we performed ChIP-qPCR 229 using WT, and HSI2pro:HSI2-and HSI2pro:HSI2mPHD-expressing Arabidopsis seedlings 230 ( Figure 3C). The results showed that, relative to intact HSI2, HSI2mPHD failed to accumulate at 231 the P1 or P2 regions of the DOG1 gene, indicating that PHD domain is essential for HSI2 232 accumulation at the DOG1 locus.  Protein sequence analysis showed that HSI2 and HSL1 share high sequence identity and contain 263 similar conserved regions (Tsukagoshi et al., 2007). Since HSI2 harbors both PHD and B3 264 domains, which are required for HSI2 accumulation at the DOG1 promoter ( Figure 2E and 3C) 265 and are essential for full repression of DOG1 expression, we predicted that HSL1 might also be 266 enriched in the chromatin of the DOG1 locus. To confirm this, a gene construct that expresses 267 GFP-tagged HSL1 under control of the native HSL1 promoter (HSL1pro:HSL1-GFP) was 268 transformed into hsl1-1 Arabidopsis protoplast for ChIP-qPCR assays. The results indicated that, 269 like HSI2, HSL1 is significantly enriched at the proximal promoter regions (P1 and P2), of the 270 DOG1 locus ( Figure 5A). To test whether the enrichment of HSL1 at the DOG1 locus depends 271 on HSI2, ChIP-qPCR was also performed in protoplasts from hsi2-2 Arabidopsis plants that 272 transiently expressed HSL1-GFP. In these assays, HSL1 was still able to accumulate at the 273 proximal promoter region of DOG1 in the absence of HSI2, indicating that HSL1 enrichment at 274 DOG1 is independent of HSI2 ( Figure 5B). Significant HSI2-independent enrichment of HSL1 275 was also detected at the AGL15 promoter ( Figure 5A, B). To confirm the hypothesis that HSL1 276 can physically bind to RY elements in the DOG1 promoter in vitro, we performed EMSA assays 277 using His-tagged HSL1-B3 fusion protein incubated with biotin-labelled probes containing RY1 278 and RY2 from the P1 and P2 regions of the DOG1 promoter, respectively ( Figure 5C).

279
Consistent with the binding activity of the B3 HSI2 polypeptide, the B3 HSL1 polypeptide can bind 280 probe that contains the complex RY1 element more effectively than to probe containing the 281 single RY2 element ( Figure 5C). However, EMSA signals for B3 HSL1 binding to RY1 in these 282 assays was much weaker than for B3 HSI2 . These results demonstrate that the B3 domain of HSL1 283 can bind to RY elements in the proximal promoter region of DOG1, leading to HSI2-284 independent HSL1 enrichment at the DOG1 locus. However, the reduced EMSA signal for 285 binding of B3 HSL1 appears to indicate that its affinity for RY elements is reduced relative to the 286 B3 HSI2 .

288
Repression of DOG1 by HSI2 and HSL1 is associated with H3K27me3 enrichment 289 Previously, we showed that loss of HSI2 in hsi2-2 knockout and disruption in hsi2-4 PHD 290 domain Arabidopsis mutant seedlings resulted in de-repression of DOG1 expression and reduced 291 15 deposition of H3K27me3 at the DOG1 locus (Veerappan et al., 2012(Veerappan et al., , 2014. Since HSL1 also 292 16 targets DOG1, we used ChIP-qPCR to examine the effects of various hsi2 and hsl1 Arabidopsis 293 mutations on the levels of H3K27me3 at DOG1 ( Figure 6). While enrichment of H3K27me3 was 294 significantly reduced, relative to WT, across the DOG1 locus in three hsi2 mutant lines (hsi2-2, 295 HSI2mPHD and HSI2mB3), H3K27me3 levels in hsl1-1 seedlings were similar to WT. These 296 results indicate that intact HSI2 mediates the trimethylation of H3K27 at WT levels in the 297 absence of HSL1 and the PHD and B3 domains are required for this activity. Thus, as with 298 transcriptional repression, HSI2 is able to fully complement the loss of HSL1 in H3K27me3 299 deposition. However, H3K27me3 deposition at the DOG1 locus was more strongly reduced in 300 the hsi2 hsl1 double mutant, than in the three hsi2 mutants, with the greatest effect seen at the C1 301 region ( Figure 6) Yuan et al. (2016) reported that LHP1, which may serve as a bridge between PRC2 and PRC1 307 (Xu et al., 2008;Bratzel et al., 2010), is recruited by HSI2 to the FLC locus to promote 308 deposition of H3K27me3, leading to downregulation of FLC expression. Therefore, it seemed 309 likely that LHP1 could also be involved in the down-regulation of DOG1 expression. To test this 310 possibility, we evaluated the role of LHP1 in the repression of DOG1. Our results show that 311 DOG1 expression is significantly upregulated in lhp1 Arabidopsis seeds, relative to that in WT, 312 hsl1-1 and hsi2-2, but remained lower than in the hsi2 hsl1 double mutant ( Figure 7A). Protein-313 protein interactions between HSI2:LHP1 and HSL1:LHP1 in vivo were confirmed by BiFC and 314 CoIP assays ( Figure 7B, C) and BiFC analysis also showed that the B3 HSI2 and B3 HSL1 domains 315 are sufficient for interaction with LHP1 ( Figure 7D). ChIP-qPCR analysis using transgenic 316 Arabidopsis seedlings that express Myc-LHP1 showed significant LHP1 enrichment at both the 317 proximal promoter (P1) and exons (C1 and C2) of the DOG1 locus ( Figure 7E). Finally, ChIP-318 qPCR analysis in lhp1 mutant seedlings showed significant reductions in H3K27me3 chromatin 319 marks across the DOG1 locus, relative to WT ( Figure 7F). Shu et al. (2019) reported that CLF, a 320 histone methyltransferase component of PRC2, accumulated at FLC, AGL15 and DOG1 genes 321 that are directly regulated by HSI2 (Qüesta et al., 2016;Yuan et al., 2016;Chen et al., 2018). We   Our previously published data indicated that the PHD-like domain also plays a critical functional 364 role in HSI2-mediated transcriptional silencing (Veerappan et al., 2012(Veerappan et al., , 2014Chen et al., 2018). 365 As shown in Figure 2D, expression of HSI2mB3 in the hsi2-2 seedlings fails to rescue the 366 transcriptional repression of DOG1, resulting in expression levels similar to that in hsi2-2 plants. 367 On the other hand, expression of the HSI2mPHD mutant in the hsi2-2 knock out mutant 368 background results in DOG1 expression reaching levels approximately 2-fold higher than in 369 hsi2-2 seedlings, though not as high as in the hsi2 hsl1 double mutant ( Figure 3B). This dimerization. Therefore, it seems likely that the inability of HSI2mPHD subunits to dimerize 374 interferes with HSL1 activity. One possible explanation for this could be competition between 375 functionally intact HSL1 dimers and HSI2mPHD monomers, which have intact B3 domains and 376 may retain the ability to bind non-productively to RY elements in the DOG1 promoter. However, 377 this seems unlikely since HSI2mPHD is not enriched at the DOG1 locus ( Figure 3C).

378
Alternatively, HSI2mPHD monomers may compete with HSL1 dimers for components of the 379 HSI2/HSL1 repressive complex such as PRC2 subunits, LHP1, and histone deacetylases.  , 1992;Meinke et al., 1992;Keith et al., 1994;West et al., 1994;Lotan et al., 1998;Luerssen   This conclusion is supported by the results of ChIP experiments that showed that FUS3 binds to 403 the DOG1 locus in vivo (Wang and Perry, 2013). However, results from our transient expression 404 assays using luciferase reporter genes controlled by intact and RY element-disrupted DOG1 405 promoters showed that, rather than abolish promoter activity, loss of one or both RY elements 406 resulted in increased reporter gene expression in protoplasts ( Figure 3B). Therefore, the potential 407 regulatory relationship between HSI2/HSL1 and AFL factors at the RY elements in the DOG1 408 promoter remains to be elucidated.

506
ChIP assays were performed as previously described 27 . Chromatin was extracted from seven day 507 old seedlings grown in MS medium supplemented with 1% sucrose. The chromatin in these 508 seedlings was cross-linked with 1% formaldehyde. The resulting chromatin was sheared to 509 fragments with 500 bp (200-1000 bp) average length by sonication and used for 510 immunoprecipitation with commercially available anti-GFP (Abcam, ab290), anti-HA (Abcam, 511 ab9110) and anti-H3K27me3 (Millipore, 07-449), respectively. After reversing the cross-links, 512 immunoprecipitated DNA was analyzed by qPCR using primers for specific regions of the 513 DOG1 gene. Three independent experiments, each using 500 mg of seedlings (25-30 individual 514 plants), were performed. Three technical replications for each qPCR assay were performed, and 515 ACT2 was used as an internal control for normalization. In Arabidopsis leaf protoplasts, the ChIP 516 26 assays were performed as described previously (Zhang et al., 2014;Lee et al., 2007;Du et al., 517 2009; Xiong et al., 2013). HSL1pro:HSL1-GFP DNA was transformed into hsl1-1 and hsi2-2 518 Arabidopsis protoplasts from 14 day old leaves using the polyethylene glycol-mediated 519 transformation method. Protoplasts were incubated at room temperature for 13 h under dark 520 conditions. Protoplast chromatin was cross-linked by 1% formaldehyde in W5 medium for 20 521 min and quenched with glycine (0.2 M) for 5 min. The protoplasts were then lysed, and the DNA 522 was sheared on ice with sonication. Immunoprecipitation was performed with anti-GFP. After 523 reversing the cross-links, the purified DNA was analyzed by qPCR using primers for specific    EMSA signal seen when using the B3 HSI2 polypeptide than with the B3 HSL1 polypeptide indicates 666 that the B3 HSI2 domain may have higher affinity for the RY1 probe than the B3 HSL1 domain.