HDA9-PWR-HOS15 Is a Core Histone Deacetylase Complex Regulating Transcription and Development^,

Arabidopsis HISTONE DEACETYLASE9 requires both PWR and HOS15 to regulate histone modifications, gene expression, and plant development.

Abstract

Histone deacetylases remove acetyl groups from histone proteins and play important roles in many genomic processes. How histone deacetylases perform specialized molecular and biological functions in plants is poorly understood. Here, we identify HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENES 15 (HOS15) as a core member of the Arabidopsis (Arabidopsis thaliana) HISTONE DEACETYLASE9-POWERDRESS (HDA9-PWR) complex. HOS15 immunoprecipitates with both HDA9 and PWR. Mutation of HOS15 induces histone hyperacetylation and methylation changes similar to hda9 and pwr mutants. HOS15, HDA9, and PWR are coexpressed in all organs, and mutant combinations display remarkable phenotypic resemblance and nonadditivity for organogenesis and developmental phase transitions. Ninety percent of HOS15-regulated genes are also controlled by HDA9 and PWR. HDA9 binds to and directly represses 92 genes, many of which are responsive to biotic and abiotic stimuli, including a family of ethylene response factor genes. Additionally, HOS15 regulates HDA9 nuclear accumulation and chromatin association. Collectively, this study establishes that HOS15 forms a core complex with HDA9 and PWR to control gene expression and plant development.

Posttranslational modification of histone protein plays critical regulatory roles for eukaryotic genomes. DNA is packaged by an octamer of histone subunits (two each of H2A, H2B, H3, and H4) to form a nucleosome (Luger et al., 1997). Histone acetylation influences DNA-templated reactions, defines chromosomal features, shapes nuclear architecture, and underlies epigenetic phenomena (Earley et al., 2006; Shahbazian and Grunstein, 2007; Yang and Seto, 2007; Venkatesh and Workman, 2015). In the context of transcription, acetylated histone is generally thought to promote transcription initiation by reducing histone-DNA affinity and recruiting transactivators, whereas deacetylation facilitates compaction and silencing (Struhl, 1998). Acetylation is catalyzed by histone acetyltransferases and removed by histone deacetylases (HDACs). Genome sequencing of the flowering plant Arabidopsis (Arabidopsis thaliana) revealed eighteen putative HDACs falling into three families: 12 are members of the Reduced Potassium Dependence3 (RPD3) superfamily, two are in the NAD-dependent Silent Information Regulator (SIR; SIR1/2) family, and four are in a plant-specific group (HD2-type; HD2A/B/C/D; Pandey et al., 2002). Given the size and relatedness of members within the HDAC family, an important question is how each HDAC is imparted with specific molecular and biological functions.

Purification of the first mammalian HDAC revealed homology to the yeast transcriptional corepressor Rpd3 (Taunton et al., 1996). Biochemical studies of RPD3 HDACs have shown that the molecular activities of deacetylases are specified by large multiprotein complexes in which they reside (Yang and Seto, 2008; Seto and Yoshida, 2014). Identification of specific complex compositions and bona fide interactors is not trivial. Quantitative mass spectrometry showed that HDACs in a single cell type (human T cells) can exist in several complexes with varied stability and participate in transient protein-protein interactions (Joshi et al., 2013). For example, mammalian HDAC1, HDAC2, HDAC3, HDAC4, HDAC5, HDAC7, and HDAC9 form distinct complexes, yet under certain cellular contexts all interact with the same paralogous corepressors Nuclear receptor corepressor (NCoR) and Silencing mediator for retinoid or thyroid-hormone receptors (SMRT; Alland et al., 1997; Guenther et al., 2000; Kao et al., 2000; Jones et al., 2001; Fischle et al., 2002). However, when purified from HeLa cell extracts, SMRT/NCoR are almost always found in stable and stoichiometric complex with HDAC3, Transducin β-like1 (TBL1), and G protein pathway suppressor2 (GPS2; HDAC3-SMRT/NCoR-TBL1-GPS2; Zhang et al., 2002; Yoon et al., 2003; Karagianni and Wong, 2007). SMRT/NCoR serves an essential scaffolding role for HDAC3-TBL1-GSP2, stimulating HDAC3 deacetylase activity and transcriptional repression (Yu et al., 2003; Watson et al., 2012). Some members also function independently of the HDACs with which they partner. For example, TBL1 (a tryptophan-aspartic acid [WD]40-repeat containing protein linked to several pathologies; Bassi et al., 1999; Heinen et al., 2016a, 2016b) helps assemble the HDAC3-SMRT/NCoR-TBL1-GPS2 complex (Oberoi et al., 2011). TBL1 also promotes gene activation, possibly by serving as an exchange factor between corepressors and coactivators (Perissi et al., 2004).

There are several lines of evidence that plant RPD3 HDACs, like their animal counterparts, form multiple complexes in planta. Mutation of Arabidopsis class I RPD3 (HDA6, HDA7, HDA9, HDA19) results in pleiotropic molecular and biological phenotypes (Liu et al., 2014), suggesting the existence of unique complexes for specific developmental pathways. For example, the Groucho/Tup1 corepressor TOPLESS modulates gene expression with many transcription factors (Causier et al., 2012). TOPLESS functions with HDA19 to control embryonic differentiation and also interacts with HDA6 to regulate circadian transcription (Long et al., 2006; Wang et al., 2013). HDA19 appears to form several other distinct complexes, partnering with BES1-TBL1 for control of seedling development (Ryu et al., 2014) and with MSI-Histone Deacetylase Complex1-SIN3-like proteins to regulate ABA signaling (Mehdi et al., 2016). Purification of one HDAC often leads to the identification of additional HDACs. Immunoprecipitation (IP) of HDA6 co-purified HD2B, HD2A, and HDA7 along with several putative HDAC complex members (Yu et al., 2017). HD2A, HD2B, and HD2D were copurified with the plant-specific HD2C (Chen et al., 2018). Bimolecular fluorescence screening of Histone Deacetylase Complex1 suggests direct interactions with both HDA6 and HDA19 (Perrella et al., 2016). Progress has been slowed as in vitro validation and mapping of intramolecular interactions routinely produce false negatives (Causier et al., 2012), likely due to requirements of endogenous factors for complex assembly. Thus, establishing the composition of plant HDAC complexes and their biological role remains challenging.

HIGH EXPRESSION OF OSMOTICALLY RESPONSIVE GENES 15 (HOS15) is a WD40-repeat domain-containing protein that represses the expression of genes involved in osmotic responses and has been implicated in histone deacetylation (Zhu et al., 2008). Here, we report a high confidence interaction between HOS15 and HISTONE DEACETYLASE9 (HDA9) using IP tandem mass spectrometry (IP-MS) experiments. HOS15 also associates with the SANT (Swi3, Ada2, N-Cor, TFIIIB) domain-containing protein POWERDRESS (PWR), previously reported to interact with HDA9 during the onset of leaf senescence and flowering (Chen et al., 2016; Kim et al., 2016). Mutations of HOS15, HDA9, and PWR induce changes in global histone modifications, produce similar pleiotropic developmental phenotypes, and share altered genome-wide differential gene expression. Our data support the existence of a conserved and biologically relevant core HDA9-PWR-HOS15 complex.

RESULTS

HOS15 Interacts with HDA9

We recently reported a physical association between HDA9 and PWR using IP-MS (Chen et al., 2016). Interestingly, we identified 22 unique peptides corresponding to HOS15, a protein previously implicated in histone deacetylation (Zhu et al., 2008). To validate this interaction, we performed two additional biological replicate IP-MS experiments using previously generated C-terminal 3xFLAG-tagged HDA9 in the hda9 mutant background (HDA9-FLAG; Chen et al., 2016). HOS15 copurified with HDA9 in all three IPs (Fig. 1A; Supplemental Data S1). HOS15 contains a series of WD40 repeats and is a putative ortholog of mammalian TBL1, a stoichiometric component of the HDAC3-N-CoR/SMRT-TBL1 complex (Supplemental Fig. S1A; Guenther et al., 2000). We next performed the reciprocal experiment by determining whether IP-MS of HOS15 copurifies HDA9 and PWR. Specifically, we introduced a C-terminal 3xFLAG-tagged HOS15 driven by its native promoter into a hos15 mutant (pHOS15::HOS15-3xFLAG/hos15-2, referred to as HOS15-FLAG). This hos15 mutant is a transfer DNA (T-DNA) line containing an insertion disrupting the ninth exon of the HOS15 gene. This line also has a second-site insertion within AT4G10300 (Trehalose Resistance14). A segregating population was screened for plants containing a T-DNA insertion in HOS15, but not Trehalose Resistance14 (Supplemental Fig. S1B). This insertion allele (hos15-2) results in the loss of full-length HOS15 transcript (Supplemental Fig. S1C). IPs from three independent homozygous HOS15-FLAG lines copurified both HDA9 and PWR (Fig. 1A; Supplemental Fig. S1D; Supplemental Data S2). We also generated plants expressing C-terminal 3xHA (Hemagglutinin)-tagged HOS15 driven by its native promoter in hos15-2 (pHOS15::HOS15-3xHA/hos15-2, referred to as HOS15-HA) and crossed them with HDA9-FLAG plants. Co-IP in F1 progeny expressing both HOS15-HA and HDA9-FLAG confirmed an interaction between HOS15 and HDA9 (Fig. 1B). Bimolecular fluorescence complementation in Nicotiana benthamiana leaves also showed an interaction between HDA9 and HOS15 in plantae (Fig. 1C). Collectively, these results demonstrate that HOS15 forms a complex with HDA9 and PWR.

Figure 1.

HOS15 interacts with HDA9. A, Partial list of proteins copurified with HOS15 and HDA9 identified by mass spectrometry analyses. Asterisked preys in gray are from Chen et al. (2016). B, Co-IP of HOS15 and HDA9 in Arabidopsis F1 hybrids coexpressing HDA9-FLAG and HOS15-HA. Plants expressing only HDA9-FLAG serve as a control. C, Bimolecular fluorescence complementation (BiFC) analysis showing HDA9-HOS15 interaction in Nicotiana benthamiana leaves. YN and YC represent N-terminal and C-terminal parts of YFP, respectively. D, Heat map of prey proteins copurified with HDA9, PWR, and HOS15. Prey proteins present in four or more out of nine purifications are listed. Prey from HD2C and wild-type (Col-0) purifications are also shown for comparison. Proteins are ranked by their peptide spectral match (PSM) ratio (sum of HDA9, PWR, or HOS15 PSM divided by the sum of HD2C and Col-0 PSMs). i, Prey protein with Log₂(PSM ratio + 1) greater than 3.9. ii, Prey protein with Log₂(PSM ratio + 1) less than 3.9. Dotted line delineates a Log₂(PSM ratio + 1) of 3.9.

Survey of the HDA9-PWR-HOS15 Interaction Network

HDACs participate in extensive stable and transient protein-protein interactions (Joshi et al., 2013). To identify additional interactors of the HDA9-PWR-HOS15 complex, we sought to determine proteins copurified by both HDA9 and HOS15. Additionally, we performed IP-MS of PWR in two independent lines expressing C-terminal 3xFLAG tagged PWR in a pwr mutant background, copurifying both HDA9 and HOS15 (pPWR::PWR-3xFLAG/pwr, referred to as PWR-FLAG; Supplemental Data S3). Nine biological replicate IPs (three each from HDA9, PWR, and HOS15) recurrently purified 47 proteins, as defined by being present in at least four of nine IPs (Fig. 1D; Supplemental Data S4). To eliminate false positive interactions, comparisons were made with three FLAG IPs performed in wild type (Col-0) plants that do not express FLAG fusion protein (Supplemental Data S5). IP of HD2C-FLAG was also included as a control (Chen et al., 2018). We next employed a semiquantitative approach to rank bait-prey relationships. Peptide spectral match (PSM) values from HDA9, PWR, and HOS15 IPs were summed (excluding bait PSM) and divided by the sum PSM values from Col-0 and HD2C IP to create a PSM ratio (see “Materials and Methods”). The PSM ratio was applied to the 47 proteins and ranked from high (left) to low (right; Fig. 1D). HDA9, PWR, and HOS15 held the highest PSM ratios, suggesting this complex is abundant in planta. IPs of HD2C and Col-0 did not detect any peptides of PWR or HOS15. Furthermore, the only HDAC purified by PWR and HOS15 was HDA9. A natural division occurred between AT3G56940 (log₂[PSM ratio + 1] = 3.95) and AT3G11830 (log₂[PSM ratio+1]=3.68; Supplemental Data S4), in which most proteins below this threshold were also identified in Col-0 or HD2C IPs (dotted line, Fig. 1D). We therefore analyzed these protein lists separately. Gene Ontology (GO) analysis of the 30 proteins with Log₂(PSM ratio + 1) < 3.9 (Fig. 1Dii) showed significant enrichment in terms related to the chloroplast (P = 5.6E-8), thylakoid (P = 2.0E-6), and ribosome (P = 5.5E-5; Supplemental Table S1). Given the abundance of these proteins in the cell and their copurification with HD2C and Col-0, these may be artifactual interactions inherent of FLAG-affinity purification of whole-cell extracts.

We therefore focused on the 15 proteins with Log₂(PSM ratio + 1) > 3.9 (Fig. 1Di). GO analyses of these proteins found terms for protein folding (P = 2.8E-9) and ATP binding (P = 3.7E-3; Supplemental Table S1). Six of these proteins belong to the T-complex protein 1 (TCP1) chaperonin, a complex of proteins responsible for assembly of protein complexes, including HDACs and cytoskeleton proteins (Spiess et al., 2004). A recent study has suggested that HOS15 can mediate degradation of HDACs during stress through ubiquitination pathways (Park et al., 2018). To test whether the 20S proteasome may regulate HDA9 protein levels, we treated plants expressing HDA9-FLAG, PWR-FLAG, and HOS15-FLAG with the 20S proteasome inhibitor MG132 (Supplemental Fig. S2A). HDA9 showed slight increased levels when treated with MG132, but PWR and HOS15 did not (Supplemental Fig. S2A). This effect was repeated with six independent biological replicates and in an independent transgenic line expressing pHDA9::HDA9-c3xHA/hda9 (HDA9-HA; Supplemental Fig. S2, B and C). These data support a regulation of HDA9 levels by proteasome degradation.

Mutations of HOS15 Alter Global Histone Modifications in an HDA9-Dependent Manner

RPD3 HDACs deacetylate histone substrates (Hartl et al., 2017). We asked if HOS15 is important for HDA9 deacetylation. Acid extracted histones of 3-week-old aerial tissues were immunoblotted with acetyl-Lys antibodies. Histones blotted for histone H3 Lys 27 acetyl (H3K27ac), H3K36ac, and H3K56ac antibodies showed greater band intensities in hda9 and hos15-2 mutants compared to Col-0 plants (Fig. 2A; Supplemental Fig. S3A). The same acetyl marks in a double hda9 hos15-2 mutant were similarly increased with no additivity relative to hda9 or hos15-2. Immunoblotting with an H4 pan-acetyl antibody showed no appreciable acetylation level changes (Fig. 2A; Supplemental Fig. S3A). To quantify histone modifications, we subjected independent acid-extracted histones to mass spectrometry analysis (Supplemental Data S6). The hda9, pwr, and hos15-2 single mutants showed increased levels of many histone H3 acetyl marks (Fig. 2B). The dual acetylated peptide H3.3K27acK36ac showed the greatest increase (2- to 4-fold) between Col-0 and mutants. However, the abundance of H3.3K27acK36ac (relative to all other measured tryptic KSAPTTGGVKKPHR peptides) in Col-0 is very low, 0.09% compared to ∼0.4% in mutants. Importantly, acetylated histone from an hda9 pwr hos15-2 triple mutant had increases comparable to all single mutants, suggesting HOS15 regulates histone deacetylation through an HDA9-mediated mechanism. H4 acetylated species displayed large variation between mutants (Supplemental Fig. S3B). Summing all acetylated and unmodified H4 species to generate relative abundance measurements yielded decreases of H4ac in hda9 (P < 0.05) and hda9 pwr hos15-2 (P < 0.05), increases in pwr (P < 0.01), and no change in hos15-2 (Supplemental Figure S3C). Interestingly, we also observed concomitant changes in histone methylation (Supplemental Fig. S3D). All mutants displayed decreases in the heterochromatic mark H3K9me2, along with H3K9me1/3, despite not reaching a statistically significant threshold (Kim et al., 2016). H3.1K36me2 was the sole methylation mark that increased across all mutants in a statistically significant manner. Together, this data suggests both HOS15 and PWR regulate total histone acetylation and methylation levels through a similar genetic pathway as HDA9.

Figure 2.

Loss-of-function HDA9, PWR, and HOS15 induces H3 hyperacetylation. A, Immunoblotting showing the abundance of histone modification levels using antibodies specific for acetyl-Lys residues on H3 and H4. Two biological replicates of Col-0, hda9, hos15-2, and hda9 hos15-2 mutants are shown. Coomassie blue staining shows total histone loading. B, Relative abundances of trypsin-digested acetylated histone peptides identified by mass spectrometry. Relative abundances are transformed into Log₂ values by normalizing all identified mutant histone peptides to Col-0 peptides (see Supplemental Data S6 for all measured peptides). The left column represents the abundance of each histone peptide relative to all other tryptic residues found in Col-0 plants. Acetylated histone peptide residues are listed to the right in rows. unmod, Unmodified; ac, acetylation; me1, monomethylation; me2, demethylation; me3, trimethylation. Unpaired Student’s t test, ***P < 0.001; **P < 0.01; *P < 0.05.

HDA9, PWR, and HOS15 Regulate Pleiotropic Development through the Same Genetic Pathway

HDACs partner with specific complex components, likely in a developmental and tissue-specified manner (Liu et al., 2014). We sought to determine the relevant developmental pathways controlled by the HDA9-PWR-HOS15 complex. Roles for HDA9 and PWR during organogenesis and developmental phase transitions have been reported (Kim et al., 2013, 2016; Yumul et al., 2013; van Zanten et al., 2014; Chen et al., 2016; Zheng et al., 2016). The hos15-2 mutant plants bear a striking resemblance to hda9 and pwr mutants (Fig. 3A). Transformation of untagged HOS15 driven by its native promoter rescued the hos15-2 phenotype (Supplemental Fig. S4A). Furthermore, double mutants (hda9 pwr, hda9 hos15-2, and hos15-2 pwr) and a triple mutant (hda9 pwr hos15-2) show no phenotypic additivity (Fig. 3A). HDA9, PWR, and HOS15 are coexpressed in all tissues, with slightly greater abundance in roots, cauline leaves, and inflorescences (Fig. 3B). The hda9 and pwr mutants flower early (Kim et al., 2013, 2016). All single, double, and triple mutants flowered early in our long-day greenhouse conditions (Fig. 3, C and D). Mean rosette leaf number at time of bolting for Col-0 was 20.6, while mutants ranged from 14.4 to 16.1 leaves (n = 33–36, P < 0.001; Fig. 3C). Mean days to flowering was 27.1 d for Col-0 and ranged from 21.5 to 23.4 for mutants (n = 33–36, P < 0.001; Fig. 3D). The extent of the early flowering phenotype in long-day photoperiods varies across greenhouses (Zemach et al., 2013; Kang et al., 2015). Therefore, we included additional genotypes to control for this effect (ebs1 flowers early and hda6 flowers late; Yu et al., 2011; López-González et al., 2014; Yang et al., 2018). All mutants display a dwarf stature, with corresponding mass reductions of the fifth rosette leaf (means ranged 12.2 to 19.4 mg) relative to Col-0 (mean of 33.5 mg; n = 8, P < 0.001; Fig. 3E). By visual inspection, hos15-2 mutants are smaller and display less size variation compared to hda9. Indeed, there is a statistically significant difference between hda9 and hos15-2 fifth leaf mass (P < 0.01). However, the mass reduction of hos15-2 was not statistically different from other mutants (P > 0.05) and was suppressed in hda9 hos15-2 and hos15-2 pwr double mutants (Fig. 3E). Shorter primary bolt height and delayed leaf aging were also observed (Supplemental Fig. S4B; Chen et al., 2016). PWR was so named because mutants have broad silique tip valves (Yumul et al., 2013). The hos15-2 mutant shares this characteristic valve broadening (P < 0.001; Fig. 3F). Floral abnormalities occurred frequently in all mutants, including additional organs and morphological deformities (Supplemental Fig. S4C). In addition to precocious flowering, hda9 mutants display another phase transition phenotype in which seed dormancy is reduced and germination is accelerated (van Zanten et al., 2014). To test hos15-2 and pwr phenotypes, we grew plants side-by-side, harvested fresh seed, immediately sowed on media without cold treatment, and measured germination rates every 12 h. Mutants began germinating (protruding radicles) after 24 h while Col-0 plants took 72 h, without changing germination kinetics (Fig. 3G; Supplemental Fig. S4D). Taken together, hos15-2 mutants phenocopy hda9 and pwr under all examined developmental contexts, and compound mutations are not additive. These genetic analyses indicate a common genetic pathway by which the HDA9-PWR-HOS15 complex controls organogenesis and developmental phase transitions.

Figure 3.

HDA9, PWR, and HOS15 regulate pleiotropic organogenesis and developmental phase transitions. A, Representative photograph of 3-week-old Col-0 and mutant plants (right) and genotype key (left). B, RT-PCR of HDA9, PWR, and HOS15 cDNA from various organs. ACTIN7 served as a control. C and D, Rosette leaf count (C) and day count (D) at vegetative to floral transition in long-day photoperiod (16 hr light/8 hr dark). The ebs (early flower; Yang et al., 2018) and hda6 (late flower; Yu et al., 2011) mutants grown side-by-side serve as controls for greenhouse conditions. E, Mass of fifth rosette leaf measured from 2-week-old plants. F, Width of silique tip valves. G, Germination kinetics of freshly harvested seeds on Murashige and Skoog media. N represents sample size of parents used for each pool of offspring. n denotes offspring sample size. Error bars represent sd (C–G). Unpaired Student’s t test: ***P < 0.001; **P < 0.01; *P < 0.05; ns, not significant.

Epistatic Interactions Underlie HOS15 and HDA9 Control of Gene Expression

HDACs are transcriptional repressors. However, HDAC loss-of-function usually results in differential gene expression bidirectionally. To determine if HOS15 regulates transcription of genes targeted by HDA9, we performed whole-transcriptome sequencing (RNA-seq) on 7-d-old hos15-2 and hda9 seedlings, a developmental stage in which wild type and mutants are phenotypically indistinguishable. Calling differential transcript abundance relative to Col-0 identified 791 genes up-regulated and 1490 down-regulated in hda9 and 396 up-regulated and 920 down-regulated in hos15-2 mutants (false discovery rate [FDR] < 0.01; Supplemental Data S7). Eighty-seven percent of hos15 up-regulated and 95% of down-regulated genes overlapped with hda9 mutants (Supplemental Fig. S5A). To determine contributions of each gene for the transcriptional phenotype, we also performed RNA sequencing (RNA-seq) with the hda9 hos15-2 double mutant. Differential expression called 388 up-regulated and 843 down-regulated genes relative to Col-0 (Supplemental Data S7). Down-regulated genes greatly overlapped between single and double mutants, while one-third of up-regulated genes overlapped (Supplemental Fig. S5A). GO analysis of hda9 hos15-2 up-regulated genes enriched Biological Process terms related to abiotic stress responses (including osmotic and cold responses, regulatory elements HOS15 was originally found to repress; Zhu et al., 2008). GO analysis of down-regulated genes enriched terms for diverse genomic processes including transcription, DNA replication, mitosis, repair, and chromatin modifications, along with developmental terms for flower development and photomorphogenesis (Fig. 4A). Increasing differential stringency (FDR < 0.01, Log₂ ± 1.1) greatly reduced the overlap between single and double mutant up-regulated genes, but not down-regulated genes, indicating a positive epistatic interaction (Fig. 4B). To explore this effect, we took genes differentially expressed (FDR < 0.01, Log₂ ± 1.1) in hda9 hos15-2 (137 up and 638 down) and computed standard deviations from the mean (Z-scores) for all genes in both single and double mutants and subjected them to hierarchical clustering (Fig. 4, C and D). Up-regulated genes overlapped (shared) between all mutants clustered with comparable standard deviations between each mutant (Fig. 4C, cluster i). Most genes not overlapping with single mutants showed synergistic (Fig. 4C, cluster ii) or additive (Fig. 4C, cluster iii) transcript abundance in the hda9 hos15-2 double mutant. Despite not reaching the Log₂ 1.1 cutoff, these genes were upregulated in hda9 and hos15-2 single mutants when compared to Col-0. The same analysis applied to down-regulated genes showed most shared down-regulated genes were comparably decreased across mutants, whereas only a single cluster of 40 genes had synergistic down-regulation in the double mutant (Fig. 4D; top cluster). This data shows HOS15 regulates expression of a large subset of HDA9-regulated genes. The positive epistatic effects in the hda9 hos15-2 double mutant suggest HOS15 may have repressive properties independent of HDA9.

Figure 4.

HOS15 and HDA9 coregulate the expression of a subset of genes. A, Gene Ontology analysis of genes up-regulated and down-regulated in hda9 hos15-2 (Log₂ ± 1.1, FDR < 0.01). Enriched biological process terms are ranked by P values. The numbers at the right end of the bar represent the percent of genes relative to total genes. B, Venn diagram showing the overlap of differentially expressed genes between hda9, hos15-2, and hda9 hos15-2 mutants. Genes were called differentially expressed with FDR < 0.01 and Log₂ values > 1.1 or < 1.1. C and D, Heat map of up-regulated genes (C) and down-regulated genes (D) with RPKM (reads per kilobase of transcript per million mapped reads) transformed as Z-scores among Col-0, hda9, hos15-2, and hda9 hos15-2 mutants. Genes (rows) were grouped using Euclidean hierarchical clustering. The far-left columns in both panels indicate genes shared between single and double mutants (shared, pink) or unique to the hda9 hos15-2 double mutant (not shared, blue).

Genes Bound by HDA9 and Up-regulated in hda9, pwr, and hos15 Are Core Environmental Stress-Response Genes

We next sought to find genes that are directly repressed by the HDA9-PWR-HOS15 complex. In addition to the hda9 and hos15-2 analyses, we performed an RNA-seq experiment with the pwr mutant. Differential transcript abundance called 1660 up-regulated genes and 2626 down-regulated genes (FDR < 0.01; Supplemental Data S7). Overlap of differential expression between hda9, pwr, and hos15-2 revealed 292 shared up-regulated (shared-up) and 871 shared down-regulated genes (shared-down; Fig. 5A). To test whether the 292 shared-up genes are directly repressed by HDA9-PWR-HOS15, we overlapped 3224 chromatin-bound HDA9 genes from a previous HDA9 chromatin IP (ChIP-seq) experiment (P < 1.0E-5; Supplemental Data S8; Chen et al., 2016). Ninety-two out of the 292 shared-up genes overlapped with HDA9-bound genes (representation factor [RF] = 2.7, P = 1.6E-19), suggesting that HDA9 might directly repress the transcription of these genes (Fig. 5B). We also found that 120 out of the 871 down-regulated genes showed HDA9 binding (RF = 1.2, P < 0.04; Fig. 5B). Compared to HDA9 enrichment over up-regulated genes, down-regulated genes have less HDA9 ChIP-seq reads (Fig. 5C). Interestingly, up-regulated genes had enriched HDA9 binding across the gene body, while down-regulated gene bodies were depleted for HDA9. This indicates that distinct mechanisms might be involved in HDA9-mediated transcription repression and activation. To assess contributions of each gene on transcript abundance, we computed Z-scores and hierarchical clustering of the 292 shared-up genes, revealing five clusters (Fig. 5D). Three clusters are genes with greatest up-regulation in each mutant: pwr (Fig. 5D, cluster i), hos15-2 (Fig. 5D, cluster ii), and hda9 (Fig. 5D, cluster iv). The largest cluster of genes showed comparable up-regulation in all three mutants (Fig. 5D, cluster iii). The final cluster had greatest up-regulation in hda9 and pwr mutants (Fig. 5D, cluster v). More of these genes are bound by HDA9 relative to other shared-up genes (RF = 1.6, P < 0.004; Fig. 5D, cluster v). The 92 genes bound by HDA9 and upregulated in each mutant had GO terms for responsiveness to stress stimuli, particularly biotic stresses such as chitin (P < 3.9E-17), wounding (P < 8.2E-9), and fungi (P < 2.7E-4; Supplemental Table S2). Shared-down genes had cellular compartment terms representing the nucleus (P < 2.3E-45) and a variety of terms related to the nuclear cytoskeleton matrix (Supplemental Table S3), along with many of the same genomic processes identified in the GO analysis of the hda9 hos15-2 mutant (Fig. 4A). The shared-up genes did not integrate into a single stress-response pathway. On the contrary, many genes are responsive to a variety of stress stimuli. Plant core environmental stress response (PCESR) genes are transcriptionally responsive shortly after a plant receives biotic and abiotic stimuli (Hahn et al., 2013). Over 25% of PCESR overlapped with shared-upregulated genes (RF = 29.2, P < 2.5E-18; Supplemental Fig. S5B). Quantitative real-time PCR (RT-qPCR) validated the up-regulation of a family of ethylene response factors (ERF4, ERF5, ERF6, ERF11) in each single mutant, with no additivity in the triple hda9 pwr hos15-2 mutant (Fig. 5E). Two additional PCESR stress-responsive transcription factors (Salt Tolerance Zinc finger and KINASE 2) were also confirmed to be up-regulated but with greater biological variation across replicates and mutants (Supplemental Fig. S5C). Thus, HDA9 appears to directly repress a subset of core environmental stress genes.

Figure 5.

HDA9 directly represses environmental stress-response genes. A, Venn diagram showing overlap of differentially expressed genes between hda9, pwr, and hos15-2 mutants. Genes were called differentially expressed with false discovery rates (FDR) < 0.01. B, Venn diagram overlap between HDA9-bound genes are from Chen et al. (2016) and “shared” differentially expressed genes (hda9 versus pwr versus hos15-2 overlap; FDR < 0.01). Hypergeometric test and representation factor (RF) were used to calculate the P value and the greatness of overlap, respectively. C, Metaplot of HDA9 ChIP-seq reads plotted over genes up- or down-regulated (Log2 ± 1.1, FDR < 0.01) in each mutant. TSS, Transcriptional start site; TTS, transcriptional termination site. Colored text of each genotype corresponds to colors of plotted lines. D, Heat map of RNA-seq reads transformed as Z-scores between Col-0, hda9, pwr, and hos15-2 mutants. Genes (rows) were grouped using Euclidean hierarchical clustering (i–v). Left column indicates genes bound by HDA9 (purple) or unbound (gray). E, RT-qPCR validation of a subset of genes up-regulated in each mutant and bound by HDA9. Relative RNA levels were first normalized to ACTIN7 and then Col-0. ERF4, AT3G15210; ERF5, AT5G47230; ERF6, AT4G17490; ERF11, AT1G28370. Four biological replicates were performed, and each replicate value is plotted as a single data point. Error bars represent sd. Unpaired Student’s t test: **P < 0.01; *P < 0.05; ns, not significant.

HOS15 Regulates HDA9 Nuclear Levels

Given that HOS15 is indispensable for HDA9-PWR functions, we sought to define a molecular role for HOS15 by studying HDA9-tagged protein in a hos15-2 mutant background. We crossed an HDA9-HA transgene into an hda9 hos15-2 double mutant and created an F2 population segregating for HDA9-HA and the hos15-2 T-DNA allele (Fig. 6A). The hos15-2 mutation had no effect on HDA9 transcript abundance or HDA9 protein levels (Fig. 6, B and C). We next hypothesized that HDA9 chromatin association may be compromised, like in pwr mutants (Chen et al., 2016). We performed HDA9 ChIP using F2 siblings with the presence of HDA9-HA transgene in HOS15 wild type (HDA9/HOS15) or mutants (HDA9-HA/hos15-2). Our previous study identified four bona fide HDA9 targets (CAT1, MBD10, MYB44, and WRKY57; Chen et al., 2016). HDA9 chromatin binding was greatly reduced in hos15-2 mutants (Fig. 6D), suggesting that HOS15 is required for HDA9 binding at chromatin at these loci. Furthermore, we performed a nuclear-cytoplasm fractionation using HDA9/HOS15 and HDA9-HA/hos15-2 and found that HDA9 nuclear accumulation is slightly decreased in the absence of HOS15 (Fig. 6E). To further test this effect, we transformed and expressed HDA9-GFP in protoplast isolated from Col-0, hos15-2, and pwr. Consistent with the western blots (Fig. 6E) and a previous report (Chen et al., 2016), HDA9-GFP signal intensity was slightly reduced in hos15-2 and pwr compared to Col-0 (Supplemental Fig. S6). Together, these data suggest that HOS15 regulates nuclear HDA9 protein level.

Figure 6.

HOS15 regulates HDA9 nuclear accumulation levels and chromatin association. A, Diagram showing the genetic cross used to generate desired genotypes. P0, parental generation; F1, first generation offspring; F2, second generation offspring; +/+, wild type; +/_, heterozygous; −/−, homozygous; +/− at least one copy of transgene. (B) RT-qPCR of HDA9 cDNA in hda9, pwr, and hos15-2 mutants. Four biological replicates were performed, and each replicate value is plotted as a single data point. Error bars represent sd. C, Immunoblotting of HDA9-HA protein in F2 offspring. Each sample is a leaf crude extract from individual sibling plants. Ponceau-S serves as a membrane transfer and loading control. D, ChIP-qPCR of HDA9-HA enrichment over HDA9 targets in Col-0, HDA9-HA/HOS15, and HDA9-HA/hos15-2 plants. Error bars represent sd of three biological replicates. TA3 served as a negative control. E, Immunoblotting of HDA9-HA protein in crude cell extract (total), cytoplasmic (cytoplasm), and nuclear (nucleus) fractions. Lanes with +/+ are plants wild type for HOS15; −/− homozygous hos15-2 mutants. Immunoblotting of histone H3 and tubulin serve as controls for nuclear and cytoplasmic fractions, respectively. Each fraction was performed with three plants, pooled in two separate biological replicates (replicates 1 and 2).

DISCUSSION

HDA9-PWR-HOS15 Is a Core HDAC Complex

HDACs play important biological roles through diverse molecular mechanisms. An important regulatory mechanism by which HDACs receive function is through protein-protein interactions in the context of large protein complexes. Despite genetic characterization, plant HDAC complexes are not fully defined, and which individual members participate in regulation remains challenging. Here, we show that HOS15 forms a complex with HDA9 and PWR. Three lines of genetic evidence support the existence of a core HDA9-PWR-HOS15 complex: mutations elicite similar patterns of changed histone acetylation and methylation (Fig. 2), produce identical developmental phenotypes (Fig. 3), and cause differential expression of a shared set of genes (Fig. 4). These data suggest that HDA9 requires both HOS15 and PWR for its molecular and biological functions. Among other functions such as binding to histone H4 (Zhu et al., 2008), we noted that HOS15 may also play a role in regulating HDA9 nuclear accumulation. While the precise mechanism is unclear, it is possible that HOS15 may be important for the nuclear import and/or regulate the stability of nuclear HDA9.

Besides acting together with HDA9, HOS15 and PWR may also have functions independent of HDA9. We note that hos15 exhibits three subtle phenotypic differences. First, hos15 mutants are smaller in stature compared to hda9 and pwr (Fig. 3A). Second, positive epistasis occurred for transcript abundance in hda9 hos15 double mutants (Fig. 4C), suggesting that HOS15 has additional repressive properties. Finally, 345 up-regulated genes are shared only between hda9 and pwr, perhaps controlled by an HDA9-PWR complex independent of HOS15. Animal HDAC1/2 form multiple complexes (Sin3, Mi-2/Nucleosome Remodeling Deacetylase, Co-repressor of Repressor element-1 Silencing Transcription factor complex), while HDAC3 appears to form a single (SMRT/NCoR) complex (Yang and Seto, 2008). Similarly, multiple plant HDA6/19 complexes have been described (Yu et al., 2003; Liu et al., 2014; Mehdi et al., 2016), while our work suggests the existence of a single HDA9-PWR-HOS15 complex. Thus, the number of complexes each HDAC participates in may reflect an ancestral feature of class I RPD3s that predates the plant-animal divergence.

Conserved and Unique Interactors of HDA9-PWR-HOS15

The putative orthologous mammalian HDAC3-NCoR/SMRT-TBL1 complex has additional protein members (Supplemental Fig. S1A). BLAST amino acid queries against the Arabidopsis proteome found no homologs, except for the histone demethylase JMJD2A (JMJ15, AT2G34880), which was not found in any IP. Instead, a nuclear-localized protein (AT1G32810) copurified in 6/9 IPs. AT1G32810 contains a RING/FYVE/PHD-type zinc finger domain. Members of this superfamily have diverse properties and functions, including histone and DNA binding (Saurin et al., 1996). However, recombinant AT1G32810 did not bind modified histone under the tested binding conditions of a recent in vitro screen (Zhao et al., 2018). RING fingers also function as ligases in the ubiquitination pathway (Joazeiro and Weissman, 2000). As such, AT1G32810 may mediate ubiquitin-dependent HDA9 degradation (Supplemental Fig. S2). The mammalian HOS15 putative ortholog, TBL1, functions as an E3 ubiquitin ligase to regulate the components of the WNT signaling pathway and has been proposed for controlling HDAC levels at promoters (Perissi et al., 2008; Dimitrova et al., 2010). This is consistent with the observation that HOS15 copurifies proteins from the 20S proteasome (Supplemental Data S4). IPs of HDA9 also enriched components of the TCP1 Chaperonin complex. The mammalian HDAC3 complex interacts with and is reported to be assembled by ATP-dependent chaperonins (Guenther et al., 2002; Joshi et al., 2013). This may be a general requirement by RPD3 HDACs, as HDA19 also purified TCP1 chaperonins (Mehdi et al., 2016). Whether the interactions between HDA9 and HOS15/PWR depends on chaperonins remains to be tested. Our IPs likely did not detect transient, cell-type, or stimuli-induced protein interactions. In the future, labeling and quantitation may be employed to detect such weak interactions (Joshi et al., 2013).

Histone Acetylation and Methylation Cross Talk

Although HDACs can deacetylate nonhistone substrate, the RPD3 class I HDACs (HDA6, HDA9, HDA7, and HDA19) are largely responsible for deacetylation of histone Lys (Hartl et al., 2017). Specific histone substrates targeted by RPD3 HDACs have been difficult to discern in vitro (due to low activity of purified HDACs) and in vivo (due to redundancy and existence of multiple HDAC complexes; Seto and Yoshida, 2014). Under steady-state growth conditions, mutations of HOS15 and PWR induced H3 acetylation in an HDA9-dependent manner (Fig. 2). Unexpectedly, changes in specific H4 acetylated peptides were variable. It is possible that H4 deacetylation is controlled in a stimulus-dependent manner, as cold treatment induces H4 acetylation in hos15 mutants (Zhu et al., 2008). It has been proposed that chromatin modifications “cross talk” (Suganuma and Workman, 2008), i.e. their regulation and molecular functions influence one another. Increases in H3 acetylation were accompanied by decreases in heterochromatic H3K9me1/2/3 marks (Supplemental Fig. S3), consistent with immunoblotting in the literature (Kim et al., 2016). H3.1K36me2 underwent a statistically significant increase of nearly 2-fold across all mutants, but not H3.3. H3.1 is found in heterochromatin, while H3.3 is deposited in gene bodies with H3.3K36me2 having particularly enrichment in 3′ untranslated regions of actively transcribed genes (Stroud et al., 2012). As such, the change of H3.1K36me3 in mutants may not be coupled to transcription or accumulated histone acetylation. How the hda9 mutation results in changes in histone methylation is unknown. Interestingly, direct roles for plant HDACs controlling histone methylation have been reported. HDA6 interacts with the histone methyltransferases SUVH4/5/6 to silence transposons and the histone demethylase Flowering Locus D to control flowering time genes (Yu et al., 2011, 2017).

Reconciling Developmental and Transcriptional Phenotypes

HDA6 and HDA19 repress key developmental genes through multiple, distinct complexes (Liu et al., 2014). In contrast, HDA9 controls pleiotropic development within a single complex (Fig. 3). HDA9-PWR-HOS15 likely does not regulate expression of key developmental genes, as our and others’ gene expression analyses showed limited pathway-specific changes (van Zanten et al., 2014). Instead, we found genes directly bound and repressed by HDA9 are stress-responsive PCESR genes (including ERF6, Salt Tolerance Zinc finger, and KINASE 2). These PCESR genes encode transcription factors that initiate stress-response gene expression networks influencing plant growth (Mittler et al., 2006; Dubois et al., 2013; Hahn et al., 2013; Van den Broeck et al., 2017). Furthermore, HDA9 complex mutants phenocopy stressed plants in their dwarfed stature and accelerated phase transitions. It is tempting to speculate HDA9-PWR-HOS15 may repress PCESR genes to balance growth-stress tradeoffs, and the mutant developmental phenotypes are simply “symptoms” of the up-regulation of these stress-response genes. The relationship between HDA9-PWR-HOS15’s molecular function and the causal events underlying its roles for plant development are important lines of future research.

MATERIALS AND METHODS

Plant Materials

All plants were grown in incubator chambers with long-day photoperiods (16 hr light 22°C, 8 hr dark 19°C) and received 100 μmol s⁻¹ m⁻² white light. Tissues for molecular analyses were harvested in evenings (except for the MG132 treatment, see below). Floral transitions were defined as 1 cm of bolt clearance and rosette leaf counts taken daily. For leaf mass measurements, fifth rosette leaf of 15-d- old plants was clipped including petiole and weighed. For silique valve width measurements, the fifth-tenth siliques from the primary bolt were clipped, photographed, and width measured using ImageJ. The germination assay was performed as follows: Seeds of 6-week-old plants grown side-by-side were harvested in pools of four plants constituting a single (N) parental biological replicate. Unsterilized seed (n) were immediately plated on 1/2x Murashige and Skoog medium with 0.8% agar and placed into light. Germination was defined by radical protrusion and measured every 12 h.

Mutant and Transgenic Plants

Arabidopsis (Arabidopsis thaliana) Columbia-0 (Col-0) was used as a wild-type line for comparison with genetic mutants. hda9 (AT3G44680; SALK_007123) and pwr (AT3G52250, SALK_071811C) mutants were selected as previously described (Chen et al., 2016). hos15-2 (AT5G67320; SALK_064435) mutants were selected and characterized as described in Supplemental Figure S1. Additional mutant lines used include hd2c (AT5G03740; SALK_129799), hda6 (AT5G63110, axe1-5), and ebs (AT4G22140). All homozygous mutant lines were selected by genotyping (for primers, see Supplemental Table S4). Genomic DNA including 1 kb of upstream sequence from the transcription start site of HDA9, PWR, and HOS15 were cloned into pENTR/D-TOPO and then placed into pEarlyGate C-terminal epitope vectors (3xHA, 9xMYC, 3xFLAG) using Gateway technology as previously described (Chen et al., 2016; for primers, see Supplemental Table S4). Binary vectors were transformed into Agrobacterium strain AGL1 and transformed into respective mutant backgrounds. T1 lines were selected based on phenotypic rescue, expression levels, and single insertion (inferred by T2 segregation). Homozygous lines were then established (inferred by T3 segregation ratio).

IP and MS

IP and MS analyses were performed as previously described (Chen et al., 2016). In brief, 3-week-old aerial tissues were mechanically ground in liquid nitrogen with pestle and mortar. Fifteen grams of powdered tissue was resuspended in IP buffer (50 mm Tris-HCl, pH 7, 150 mm NaCl, 5 mm MgCl₂, 5% [v/v] glycerol, 1 mm dithiothreitol, 1 mm phenylmethylsulfonyl fluoride [PMSF], 0.1% [v/v] NP40, Protease Inhibitor Cocktail Complete), dounce homogenized, filtered through miracloth, centrifuged, and supernatant incubated with batch FLAG M2 beads (Sigma, M8823) for 3 h at 4°C. Protein was eluted with 3xFLAG peptide (Sigma, F4799), precipitated with trichloroacetic acid, digested with trypsin, and analyzed using Orbitrap Mass Spectrometry as previously described (Chen et al., 2016). Proteins present in at least 4/9 HDA9, PWR, HOS15 IPs were listed and ranked by PSM ratio. The PSM ratio was calculated by summing the PSM values from nine IPs (three of each HDA9, PWR, and HOS15) and subtracting the bait PSM score. This value was then divided by the sum of PSM values from control IPs (HD2C and three Col-0 mocks). PSM Ratio = ∑(HDA9, PWR, HOS15 PSM)/(∑(HD2C, mock Col-0 PSM) + 1). Co-IP was performed the same as described above with the following exceptions: The starting material was 0.5 g, and the FLAG beads were directly boiled in SDS loading buffer after washes.

Bimolecular Fluorescence Complementation

Full-length HDA9 and HOS15 complementary DNA (cDNA) were in-frame cloned into plasmids containing N-terminal and C-terminal part of YFP, respectively (Yang et al., 2018). The constructs were transformed into agrobacteria. Respective bacteria cultures were adjusted to OD₆₀ = 0.1 and mixed with 1:1 ratio and then infiltrated into Nicotiana benthamiana leaves. After 2 d of growth, a slice of infiltrated leaf was imaged with a confocal microscope (Nikon, A1R HD).

Proteasome Inhibition Assay

Ten-day-old seedlings grown on 1/2 Murashige and Skoog agar plate were transferred to water supplemented with 50 µm MG132 in dimethyl sulfoxide or dimethyl sulfoxide only for 24 h. Transfer and harvest occurred at Zeitgeber Time +1 (first hour of day during a 16 hr light, 8 hr dark cycle). Seedlings were crushed with liquid nitrogen, and total protein was extracted and analyzed by immunoblotting.

Histone Extraction, Immunoblotting, and MS

Histone acid extraction was performed as previously described (Sanders et al., 2017). In brief, 3-week-old aerial tissues were mechanically ground in liquid nitrogen with pestle and mortar. Four grams of powdered tissue were suspended in nuclear isolation buffer (10 mm HEPES, pH 8.0, 1 m Suc, 5 mm KCl, 5 mm MgCl₂, 0.6% [v/v] Triton X-100, 0.4 mm PMSF, 10 µm Pepstatin, Protease Inhibitor Cocktail Complete). The homogenate was dounce homogenized and filtered with miracloth, and nuclei were isolated by centrifugation at 2600g for 10 min at 4°C. Nuclei were washed, dounce homogenized, and extracted with 0.4 N H₂SO₄. Supernatant containing acid-soluble histones was precipitated with 35% (w/v) trichloroacetic acid and washed with ice cold acetone. For immunoblotting, soluble histones were electrophoresed on 15% SDS-PAGE and normalized by Coomassie staining and histone H3 and H4 immunoblotting. Histones were transferred to nitrocellulose, blocked with 3% [w/v] bovine serum albumin, and immunoblotted using specific histone-modification antibodies (αH3, Abcam ab1791; αH3ac, Active motif 39139; αH3K9ac, Millipore 07-352; αH3K18ac, Abcam ab1191; αH3K23ac, Millipore 07-355; αH3K27ac, Active motif 39133; αH3K36ac, Active motif 39379; αH3K56ac, Millipore 04-1135; αH4, Abcam ab7311; αH4ac, Active motif 39243). Bio-Rad Clarity chemiluminescence substrate and digital imaging was used for membrane photography. Independent histone acid extraction was subjected for MS analysis as previously described (Sanders et al., 2017). In brief, detergent from extracted histone was removed by methanol-chloroform precipitation, and histones were propionylated (diluted propionic anhydride). Following an overnight trypsin digestion, histone peptides were labeled with phenylisocyanate. Histone peptides were injected onto a Dionex Ultimate3000 nanoflow HPLC coupled to a Thermo Fisher Scientific Q-Exactive mass spectrometer. Peptides were identified with Mascot and spectral libraries built with Skyline. Signal intensity of peptides were measured by integration of MS1 peak areas. Data were normalized to total peptides of respective histone peptide families and reported as percent total (relative abundance).

RNA Extraction, RT-qPCR, and RNA-Seq

RNA was extracted from 7-d-old seedlings grown on 1/2 Murashige and Skoog agar plates. Tissue was ground in liquid nitrogen and resuspended with Trizol using standard protocol (Thermo Fisher Scientific). RNA quality was assessed using spectrometer and 8 m urea acrylamide electrophoresis. Three micrograms of RNA were treated with DNaseI and one microgram reverse-transcribed using oligoDT primers and SuperScript III reverse transcriptase. Quantitative PCR was performed with SYBR Green Master Mix on a CFX96 real-time thermal cycler (Bio-Rad). The 2^−ΔCT method using ACTIN7 as an internal reference was used for relative quantification. RNA-seq libraries were constructed using Illumina TruSeq Library Preperation Kit. Libraries were sequenced using an Illumina HiSeq2000 HighOutput 1 × 50 by the UW-Madison Biotechnology Center. Reads were aligned using TopHat to TAIR10. Transcripts were configured with Cufflinks, merged, and differential calls made using CuffDiff. Normalized fragments per kilobase million, using an FDR of < 0.01 were called as statistically significant differential abundance. Two biological replicates were performed for each genotype. Statistical significance of gene list overlaps was calculated using exact hypergeometric probability. The RF was calculated by RF = x/expected # of genes, where x is genes shared between overlap and expected # of genes = (n*D)/N. n = group 1 genes, D = group 2 genes, n = total genes in Arabidopsis genome (27,655). GO analysis was performed using the Database for Annotation, Visualization, and Integrated Discovery (Huang et al., 2009).

Nuclear-Cytoplasm Fractionation Assay

A half gram of 3-week-old plants was ground in liquid nitrogen with pestle and mortar and then lysed in hypotonic buffer (20 mm Tris-HCl, pH 7.5, 20 mm KCl, 2 mm EDTA, 2.5 mm MgCl₂, 25% [v/v] glycerol, 250 mm Suc, 5 mm dithiothreitol with protease inhibitor) and filtered through miracloth. Nuclei were isolated by 1500g centrifugation for 10 min. Supernatant was transferred to new tube and centrifugated at 10,000g for 10 min. The supernatant was sampled to represent the cytoplasmic fraction. The nuclei were further washed four times with wash buffer 1 (20 mm Tris-HCl, pH 7.5, 2.5 mm MgCl₂, 25% [v/v] glycerol, 0.2% [v/v] Triton X-100 with protease inhibitor) and resuspended with wash buffer 2 (20 mm Tris-HCl, pH 7.5, 10 mm MgCl₂, 250 mm Suc, 0.5% [v/v] Triton X-100, 5 mm β-mercaptoethanol with protease inhibitor). The homogenate was transferred and layered slowly on top of wash buffer 3 (20 mm Tris-HCl, pH 7.5, 10 mm MgCl₂, 1.7 m Suc, 0.5% [v/v] Triton X-100, 5 mm β-mercaptoethanol with protease inhibitor). Pellet was collected as nuclei after 45 min centrifugation at 10,000g.

Protoplast Transformation

Full-length HDA9 and DRM2 cDNA were cloned into pEG103 and pCAMBIA1300 harboring a C-terminal GFP driven by the CaMV 35S promoter, respectively. The lower epidermis of 2-week-old leaves from Col-0, hos15-2, and pwr were peeled with tape and then soaked in 10 mL enzyme solution (0.4 m mannitol, 20 mm KCl, 20 mm MES, pH5.7) containing 0.1 g cellulase R10 and 0.02 g macerozyme R10 for 2 hr with slow shake (dissolve enzymes at 55°C for 10 min, then cool to room temperature, and add CaCl₂ to 10 mm, β-mercaptoethanol to 0.5 mm, bovine serum albumin to 0.1% [w/v]). The solution was then filtered through 75 μm nylon mesh. Protoplasts were collected by 100g centrifugation for 5 min and then washed twice with W5 buffer (154 mm NaCl, 125 mm CaCl₂, 5 mm KCl, 2 mm MES, pH 5.7). After 40 min incubation on ice in 10 mL W5 buffer, protoplasts were collected and resuspended with 200 μL MMG buffer (0.4 m mannitol, 15 mm MgCl₂, 4 mm MES, pH 5.7). For each transformation, 10 μg plasmid and 220 μL Polyethylene glycol buffer (40% [w/v] PEG4000, 100 mm CaCl₂, 200 mm mannitol) were added into 200 μL protoplast and mixed well slowly. After 20 min incubation at room temperature, the reaction was stopped by adding 800 μL W5 buffer. Pellet was collected and resuspended with 2 mL W5 buffer and transferred to a 6-well plate. After slow shaking overnight, the protoplasts were imaged with a confocal microscope (Nikon, A1R HD). All images were taken with the same parameters and quantified with ImageJ.

ChIP

ChIP was performed as previously described (Chen et al., 2016) with modifications. In brief, nuclei were isolated from 2 g of 10-d-old seedlings with nuclei isolation buffer (10 mm HEPES, pH 8, 1 m Suc, 5 mm KCl, 5 mm MgCl₂, 5 mm EDTA, 0.6% [v/v] Triton X-100, 0.4 mm PMSF, and protease inhibitor) and then washed with nuclei isolation buffer II (0.25 m Suc, 10 mm Tris-HCl, pH 8, 10 mm MgCl₂, 1% Triton X-100, 1 mm EDTA, 5 mm β-mercaptoethanol, 0.4 mm PMSF, protease inhibitor). Nuclei were resuspended with 1 mL MNase digestion buffer (50 mm Tris-HCl 7.5, 50 mm NaCl, 0.1% [v/v] NP-40, 5 mm CaCl₂, 10% [v/v] glycerol, 0.4 mm PMSF, protease inhibitor) and subjected to sonication. Sheared chromatin was further digested with 1 µL MNase (New England Biolabs, M0247S) at 37°C for 10 min. Soluble nucleosomes were isolated by centrifugation at 16,000g for 10 min and transferred to new tube. Human nucleosomes isolated from HEK293 cells expressing H3.1-FLAG-HA were added as spike-in with 1:50 ratio. Supernatant was incubated with 4 µg anti-HA antibody (Invitrogen, 26183) and 1 mg protein G beads (Invitrogen, 10003D) overnight. The protein-antibody-beads complex was washed sequentially with low-salt buffer (150 mm NaCl, 0.1% [w/v] SDS, 1% [v/v] Triton X-100, 2 mm EDTA, 20 mm Tris-HCl, pH 8), high-salt buffer (500 mm NaCl, 0.1% [w/v] SDS, 1% [v/v] Triton X-100, 2 mm EDTA, 20 mm Tris-HCl, pH 8), LiCl buffer (0.25 m LiCl, 1% [v/v] NP-40, 1% [w/v] sodium deoxycholate, 1 mm EDTA, 10 mm Tris-HCl, pH 8), and TE buffer (10 mm Tris-HCl, pH 8, 1 mm EDTA). Elution products were reverse crosslinked and DNA was purified by phenol:chloroform extraction and ethanol precipitation. ChIP and input were normalized with spike-in human Glyceraldehyde 3-phosphate dehydrogenase gene before calculation, respectively.

Accession Numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers HDA9 (AT3G44680), PWR (AT3G52250), and HOS15 (AT5G67320). RNA-seq data are available at the Gene Expression Omnibus and can be found with the GEO accession number GSE121673.

Supplemental Data

The following supplemental materials are available:

• Supplemental Figure S1. Mutant and transgenic lines used for purification of HDA9, PWR, and HOS15.

• Supplemental Figure S2. HDA9 accumulates upon inhibition of the 20S proteasome.

• Supplemental Figure S3. Hos15-2 and hda9 mutations induce histone hyperacetylation and methylation change.

• Supplemental Figure S4. Pleiotropic developmental phenotypes of hda9, pwr, and hos15-2 mutants.

• Supplemental Figure S5. Characterization of differential gene expression in hda9, pwr, and hos15-2 mutants.

• Supplemental Figure S6. HOS15 and PWR regulate HDA9 nuclear levels.

• Supplemental Table S1. GO analysis of proteins copurified with HDA9, PWR, and HOS15 with different PSM ratio.

• Supplemental Table S2. GO analysis of up-regulated genes in hda9, pwr, and hos15-2 mutants.

• Supplemental Table S3. GO analysis of down-regulated genes in hda9, pwr, and hos15-2 mutants.

• Supplemental Table S4. List of primers used in this study.

• Supplemental Data S1. List of proteins copurified by immunoprecipitation of HDA9.

• Supplemental Data S2. List of proteins copurified by immunoprecipitation of HOS15.

• Supplemental Data S3. List of proteins copurified by immunoprecipitation of PWR.

• Supplemental Data S4. PSM scores of all proteins copurified by HDA9, HOS15, and PWR.

• Supplemental Data S5. List of proteins copurified by mock immunoprecipitations in wild type.

• Supplemental Data S6. Histone peptide relative abundance values as measured by mass spectrometry.

• Supplemental Data S7. List of differentially expressed genes in hda9, pwr, hos15-2, and hda9 hos15-2 mutants from RNA-seq experiments.

• Supplemental Data S8. List of genes bound by HDA9 at the promoter or gene body as determined by ChIP-seq experiments.

• Supplemental Data S9. GO analysis of down-regulated genes overlapped between had9, pwr, and hos15-2 mutants.

ACKNOWLEDGEMENTS

The authors thank James Dowell and Erik Armstrong for technical support using the mass sepctrometer. They also thank the UW-Madison Biotechnology Center DNA Sequencing Facility for high-throughput sequencing.