Genome-wide Target Mapping Shows Histone Deacetylase Complex1 Regulates Cell Proliferation in Cucumber Fruit^{1  [OPEN]}

Abstract

Histone deacetylase (HDAC) proteins participate in diverse and tissue-specific developmental processes by forming various corepressor complexes with different regulatory subunits. An important HDAC machinery hub, the Histone Deacetylase Complex1 (HDC1) protein, participates in multiple protein–protein interactions and regulates organ size in plants. However, the mechanistic basis for this regulation remains unclear. Here, we identified a cucumber (Cucumis sativus) short-fruit mutant (sf2) with a phenotype that includes repressed cell proliferation. SF2 encodes an HDC1 homolog, and its expression is enriched in meristematic tissues, consistent with a role in regulating cell proliferation through the HDAC complex. A weak sf2 allele impairs HDAC targeting to chromatin, resulting in elevated levels of histone acetylation. Genome-wide mapping revealed that SF2 directly targets and promotes histone deacetylation associated with key genes involved in multiple phytohormone pathways and cell cycle regulation, by either directly repressing or activating their expression. We further show that SF2 controls fruit cell proliferation through targeting the biosynthesis and metabolism of cytokinin and polyamines. Our findings reveal a complex regulatory network of fruit cell proliferation mediated by HDC1 and elucidate patterns of HDC1-mediated regulation of gene expression.

Acetylation and deacetylation of Lys residues, mediated by histone acetyltransferase or histone deacetylase (HDAC) enzymes, usually leads to relaxed or tight chromatin structures. These are generally associated with transcriptional activation or repression, respectively. A growing number of studies have demonstrated the critical functions of histone deacetylation/acetylation in various biological processes, such as genome stability (Shahbazian and Grunstein, 2007; Zilio et al., 2014), transcriptional regulation (Rossi et al., 2007; Zhang et al., 2018), development (Ueno et al., 2007; Cigliano et al., 2013; Liu et al., 2014), and cell division control (Kouzarides, 1999; Jamaladdin et al., 2014). In animals, HDAC1 and HDAC2 have essential roles in cell proliferation and regulate stem cell self-renewal (Jamaladdin et al., 2014). Similarly, in plants, it has been shown that histone deacetylation activity controls root meristem cell division (Ikeuchi et al., 2015). Plant cell division is a continuous process in apical meristematic cells or quasi-meristematic cells, which generate new cells throughout the life of the plant (Girin et al., 2009). However, as HDAC enzymes are generally ubiquitously expressed, the regulation of histone deacetylation activity during cell division is currently not well understood. To carry out their intended functions, the correct assembly of HDAC proteins into various types of functional complexes, containing different regulatory subunits, is important for their diverse, and often cell-specific, roles (Utley et al., 1998; Gonzalez et al., 2007).

Recent studies in Arabidopsis (Arabidopsis thaliana) identified HISTONE DEACETYLASE COMPLEX1 (HDC1), which contains a yeast regulator of transcription3 domain, as an important subunit of the plant SWI-INDEPENDENT3 (SIN3)–HDAC machinery. In Arabidopsis, HDC1 functions as a hub to enable multiple protein interactions in HDAC complexes. Genetic analyses showed that HDC1 regulates abscisic acid (ABA) sensitivity and promotes increases in organ size (Perrella et al., 2013). Because overexpression of HDACs has been reported as being either ineffective, or as causing pleiotropic morphological abnormalities (Tian and Chen, 2001; Tian et al., 2005; Long et al., 2006), the regulation of organ size is thought to be regulated by HDC1 through site- or tissue-specific utilization of certain HDAC complexes. One study showed that HDC1 mRNA is ubiquitously expressed in all vegetative tissues in Arabidopsis (Perrella et al., 2013); however, it has yet to be determined whether the HDC1 protein accumulates in meristematic cells, thus enabling the function of the HDAC complex in cell division. Moreover, the genome-wide targets of HDC1 are still unknown, and so the regulatory mechanism by which HDC1 controls plant organ size is unclear.

The identities, interactions, and control of the many factors that dictate plant organ size and shape, which in turn are determined by cell number and size, are important ongoing questions. Fruit size and shape are important agronomic traits that are associated with crop yield. Classic fruit morphogenesis is divided into four physiological phases during early development: ovary growth, fruit set, rapid cell proliferation, and subsequent cell expansion (Ando and Grumet, 2010; Yang et al., 2013; Grumet and Colle, 2016). It is thought that the cell proliferation, and therefore the final size and shape of the fruit, is dependent on initial meristematic and quasi-meristematic characteristics (Girin et al., 2009). As morphologically complex fruit can take days or weeks to develop, the establishment of the quasi-meristem in the medial part of the fruit tissue allows for prolonged cell division after fruit set. Therefore, fruit organ growth is fine-tuned by maintenance of these medial tissues and their “quasi-meristematic” fate. However, although many quantitative trait loci related to fruit size have been identified (Qi et al., 2013; Wei et al., 2014; Bo et al., 2015; Weng et al., 2015; Pan et al., 2017), the genetic and molecular mechanisms that regulate rapid cell proliferation during fruit morphogenesis remain largely unclear.

Fleshy cucurbit fruits initiate from female floral meristems, and are known for their extreme diversity in shape (oblate to elongate) and size (Grumet and Colle, 2016; Colle et al., 2017). Here, we identified recessive allelic variation in a cucumber (Cucumis sativus) HDC1 homolog, Short Fruit2 (SF2). This weak mutation significantly represses fruit elongation by reducing cell proliferation by 70%. In contrast to a previous report concluding that it is ubiquitously expressed in all vegetative tissues, we found that SF2 protein specifically accumulates in meristematic tissues undergoing cell proliferation. We further show that the SF2 protein is expressed during early fruit development and accumulates in the early placenta, replum, and ovules, which are thought to maintain quasi-meristematic activity during early fruit development (Girin et al., 2009). Genome-wide analysis showed that SF2 directly targets and promotes histone deacetylation of genes involved in multiple phytohormone pathways and cell cycle regulation, and represses or activates their expression during cell proliferation. We reveal that HDC1 controls fruit cell proliferation through direct targeting of cytokinin (CK) and polyamine (PA) biosynthesis and metabolism. These findings enhance our understanding of the role of histone deacetylation in cell proliferation and fruit morphogenesis.

RESULTS

A Short-Fruit Phenotype Is Determined by a Single Nucleotide Change in Cucumber HDC1

We identified a cucumber ethyl methanesulfonate mutant bearing short fruit (named “sf2”) that was otherwise phenotypically similar to the wild type (Fig. 1, A–C). Preliminary genetic analysis showed that all F1 plants from a sf2 × wild type cross exhibited a wild-type phenotype, and the selfed F2 progeny segregated at an ∼3:1 ratio (249:78, wild type to mutant; χ² = 0.23), indicating that the short-fruit phenotype is determined by a single recessive gene (Fig. 1D; Supplemental Data Set S1).

Further characterization of fruit morphogenesis during early fruit development indicated that the wild-type fruit exhibited more rapid growth than sf2 fruit after anthesis, while there was no significant difference in fruit diameter (Fig. 1, E and F; Supplemental Fig. S1). The cell proliferation rate and duration of this process were both notably reduced in sf2 fruit compared with wild type in the longitudinal direction, resulting in ∼70% fewer cells in the sf2 mutant at 8 days after anthesis (DAA; Fig. 1G). Despite a potential compensatory mechanism resulting in an increase in cell size to make up for the reduction in cell numbers, sf2 fruit length was reduced by 50% (Fig. 1, H–J), demonstrating that the sf2 phenotype primarily involves reduced cell proliferation, which results in shorter fruit length compared with wild type.

To clone the SF2 gene, we used a combination of the “MutMap” strategy (Abe et al., 2012) and traditional linkage analysis in an F2 population. We identified only one single-nucleotide polymorphism (SNP; 2G15231244) that cosegregated with the sf2 locus as the causative SNP (Fig. 1, K and L; Supplemental Data Set S2). The nonsynonymous SNP 2G15231244 caused a G-to-A substitution within Csa2G337260, resulting in an amino acid change from Gly (G) to Glu (E) at residue 515 (G515E). Csa2G337260 encodes an HDC1 homolog (Fig. 1, M and N; Perrella et al., 2013), a subunit of the HDAC complex (Perrella et al., 2016).

To demonstrate the function of Csa2G337260, we first knocked out SF2 expression by clustered regularly interspaced shot palindromic repeat/CRISPR associated protein 9 (CRISPR/Cas9; Fig. 2A) to produce CR-sf2 mutants containing small deletions that produce a frameshift in the SF2 coding sequence resulting in knock-out of SF2 protein. We observed a serious growth inhibition of the entire CR-sf2 shoot compared with the control plants transformed with an empty vector. Seedlings did not regenerate from these shoots (Fig. 2, B–E). Consistent with the observation that the hdc1 mutant in Arabidopsis showed a 50% reduction in fresh weight compared with wild type (Perrella et al., 2013), our study suggested a general function of plant HDC1 in controlling meristematic cell proliferation.

To confirm the identity of Csa2G337260 as SF2, a DNA sequence containing a ∼1-kb region of the native promoter plus the full-length coding sequence of wild-type Csa2G337260 was transformed into the sf2 mutant. Both of the homozygous complemented plants displayed increased cell numbers and fruit length (Fig. 2, F–H; Supplemental Fig. S2). We therefore concluded that Csa2G337260 is responsible for the observed short-fruit phenotype of sf2.

The SF2 Protein Is Specifically Expressed in Meristematic Tissues

Although a previous study in Arabidopsis reported that HDC1 is ubiquitously expressed in vegetative tissues (Perrella et al., 2013), we hypothesized that the HDC1 protein may be specifically enriched in meristematic tissues. To test this, we generated an SF2 polyclonal antibody (Fig. 3, A–C), which we used to detect HDC1 accumulation in various tissues. We found that in contrast to the ubiquitous expression of SF2 mRNA, the SF2 protein was indeed specifically detected in meristematic tissues in which rapid cell proliferation was occurring, including: root tip at 1 d after germination (1-d root tip); stem tip; leaves measuring 5 mm in width (5-mm leaf); female flowers (−3 DAA of ovary); and placenta of fruit on the day of anthesis (0 DAA of fruit; Fig. 3D). This result was consistent with a role for SF2 in regulating cell proliferation.

To investigate a potential correlation between SF2 expression and cell proliferation, we investigated the expression pattern of SF2 mRNA and protein during fruit development. SF2 mRNA exhibited highly similar expression levels in wild type and sf2 mutants in the stages investigated (Fig. 3E), while the SF2 protein specifically accumulated in the fruit during stages of rapid cell division (−8 DAA to 3 DAA; Fig. 3F). These results also suggested that posttranscriptional regulation of SF2 results in spatially restricted SF2 accumulation in the tissues undergoing exponential cell division. We performed immunolabeling of fruit sections using the SF2 antibodies, which revealed that the SF2 protein primarily accumulates in the early placenta, replum, and ovules (Fig. 3G), all of which maintain quasi-meristematic activity and allow fruit cell proliferation.

SF2 Targets Active Genes and Promotes Histone Deacetylation

In Arabidopsis, HDC1 interacts with several components of the HDAC complex, such as HDA6, HDA19, SIN3-associated protein18 (SAP18), MULTICOPY SUPPRESSOR OF IRA (MSI), and SIN3-LIKE (Perrella et al., 2016). We confirmed these interactions for SF2 in cucumber (Supplemental Fig. S3, A and B; Supplemental Data Sets S3–S6), and found SF2 coimmunoprecipitated with HDA19A, HDA19B, SIN3-LIKE1, SIN3-LIKE3, SAP18, and MSI1 homologs by coimmunoprecipitation (Co-IP) and liquid chromatography with tandem mass spectrometry (LC-MS/MS). We further confirmed that SF2^G515E shows an impaired capacity for interaction with SIN3-LIKE1 and SIN3-LIKE3 by luciferase complementation imaging (LCI) assays and Co-IP assay (Supplemental Figs. S3C, S4, and S5).

Binding of the SIN3-HDAC complex to chromatin is accomplished by flexible interactions with DNA-binding proteins, primarily through SIN3 repressors (Grzenda et al., 2009). Therefore, we reasoned that the capacity of HDAC complex targeting to genomic regions might be decreased in the sf2 mutant, and that the relative histone acetylation levels might be increased. To test this, chromatin immunoprecipitation (ChIP)-sequencing (Seq) assays were performed using the SF2 antibody (Fig. 4A). By comparing the genome-wide binding profiles of SF2 in wild-type and sf2 mutant fruits, we found that the SF2-HDAC complex was primarily detected at the transcriptional start site (TSS) and in gene body regions (Fig. 4, A and B), which is similar to the pattern of human HDAC6 (Wang et al., 2009). Indeed, the SF2-HDAC complex showed strongly decreased binding levels in the sf2 mutant compared to the wild type (Fig. 4A). We further analyzed the genomic distribution of overlapping SF2 binding sites in two replicates (Fig. 4B; Supplemental Data Set S7), and identified 3,356 genes as putative SF2 target genes (Supplemental Fig. S6, A–C; Supplemental Data Set S8).

To investigate the correlation between SF2 binding and gene expression, we divided all cucumber genes into four groups based on their expression levels (high, medium, low, and silent expression), and correlated this grouping with SF2 binding levels (Fig. 4C). This revealed that SF2 is preferentially enriched in active genes compared to silent genes (Fig. 4C). As HDAC complexes are generally considered to be transcriptional repressors, SF2 enrichment in highly expressed genes might be considered counterintuitive. Nevertheless, studies of HDACs in both mammals and plants have yielded similar results, suggesting a conserved action of the HDAC complex on active genes (Wang et al., 2009; Yang et al., 2016).

We next searched for putative DNA-binding motifs in SF2 binding peaks using the software “MEME 4.0” (Bailey et al., 2006) and identified three significantly enriched consensus motifs (Supplemental Fig. S7, A–C). Comparison of these motifs against a database of known motifs, using TOMTOM (http://meme-suite.org/tools/tomtom), suggested that motif 1 and motif 2 are similar to the binding motif of the GATA1 transcription factor (T/CCDT/CCNT/CCNT/CCN, P = 4.14e-06; CANCANCAN, P = 9.71e-03; Supplemental Fig. S7, A and B). In mouse (Mus musculus), the HDAC complex physically interacts with GATA to modulate the expression of cell cycle genes during embryonic development (Trivedi et al., 2010). The third motif was similar to the binding site of high-mobility–group box protein (CC/TDCC/TDCC/TD, P = 4.98e-07; Supplemental Fig. S7C), which has been implicated in interacting with mitotic and meiotic chromosomes (Pedersen et al., 2011).

To investigate the relationship between SF2 binding and histone acetylation, a global analysis of H3K9 and H3K14 acetylation status in wild-type and sf2 mutant fruits was conducted (Supplemental Fig. S8; Supplemental Data Set S9). The level of SF2 binding increased with increasing H3K9ac and H3K14ac (Fig. 5, A and B), indicating that SF2 preferentially targets genes with high histone acetylation levels. To better understand how the impaired binding of the HDAC complex in the sf2 mutant affects histone acetylation levels, we performed a genome-wide comparison of H3K9ac and H3K14ac levels in wild type and sf2 and observed that sf2 showed increased H3K9ac and H3K14ac levels (Supplemental Fig. S9; Supplemental Data Sets S10 and S11). We further compared the normalized H3K9ac and H3K14ac levels around the SF2 binding sites in wild type and sf2 mutant (Fig. 5, C and D). The results showed that compared with wild type, H3K9ac, and H3K14ac levels increased in most regions of promoters and gene bodies in sf2 mutant, demonstrating that SF2 promotes histone deacetylation at its binding sites. Accordingly, immunoblotting demonstrated that the levels of histone H3 acetylation increased in 0 DAA of sf2 fruit (Fig. 5E), especially at Lys residues K9 and K14 (H3K9ac and H3K14ac), demonstrating that SF2 promotes histone deacetylation.

Identification of The Genetic Network Activated or Repressed by SF2

To obtain a detailed understanding of SF2 function in target gene regulation, we used RNA-Seq analysis to identify genes that are differentially expressed in 0-DAA fruits between wild-type and sf2 plants (Supplemental Data Sets S12 and S13). Gene ontology (GO) analysis revealed that the 1,494 upregulated genes in the sf2 mutant are primarily involved in photosynthesis and phytohormone responses (Supplemental Fig. S10A); and that the 1,272 downregulated genes in sf2 are enriched in the “cell cycle,” “cytokinesis,” and “DNA replication” terms (Supplemental Fig. S10B), which is consistent with the reduced cell number phenotype of sf2. The RNA-seq data were validated by reverse-transcription quantitative PCR (RT-qPCR) analysis. All the 10 selected genes involved in phytohormone biosynthetic and signaling pathways and also cell cycle regulation showed expression patterns corresponding with the RNA-seq data (Supplemental Fig. S10, C and D).

Because histone deacetylation is generally associated with repressed gene expression, to map the gene network directly repressed by SF2, the 3,356 target genes with potential SF2 binding sites were compared with the 1,494 genes upregulated in sf2. A total of 321 genes showed an overlap between the two data sets (Fig. 6A; Supplemental Data Set S14). To determine whether SF2 promotes deacetylation of these 321 core repressed genes, the genes with hyper-H3K9ac and hyper-H3K14ac (Supplemental Data Sets S10 and S11) were tested for overlap with the core repressed genes. This analysis revealed that 108 and 140 of the core SF2-repressed genes were hyperacetylated at H3K9 and H3K14, respectively, in the sf2 mutant, in either the gene body or in the promoter (Fig. 6, B and C; Supplemental Data Sets S14–S16). Furthermore, 52 genes had both hyper-H3K9ac and hyper-H3K14ac sites (Fig. 6D; Supplemental Data Set S17), while 125 out of the 321 genes did not show significantly increased H3K9ac/K14ac in the sf2 mutant (Fig. 6D), suggesting a possible modification of other Lys residues (e.g. K18, K23, K27, and K56; Fig. 5E).

Among the 321 core repressed genes, we found a strong enrichment of genes involved in hormone biosynthesis and signal transduction (Fig. 6E; Supplemental Fig. S11; Supplemental Data Set S18), indicating a role for SF2 in coordinating the expression of genes involved in phytohormone pathways (Fig. 6E). Among these genes, several are known to encode negative regulators of auxin, gibberellic acid (GA), and CK responses or biosynthesis (Fig. 6E; Supplemental Data Set S18). For example, PINOID (Csa1G537400) and PHYTOCHROME RAPIDLY REGULATED (Csa6G423410) mediate auxin transport (Friml et al., 2004) and signaling repression (Bou-Torrent et al., 2008), respectively. GIBBERELLIN INSENSITIVE (Csa1G408720) encodes a DELLA protein that represses the GA response (Murase et al., 2008) and cytokinin oxidase/dehydrogenase (CKX; Csa4G647490) encodes a CK oxidase/dehydrogenase involved in CK degradation (Werner et al., 2001). The core repressed genes also included some positive regulators of ABA, jasmonic acid (JA), and ethylene responses (Fig. 6E; Supplemental Data Set S18), including an ABA1 homolog (Csa2G277050) encoding a zeaxanthin epoxidase involved in ABA biosynthesis (Xiong et al., 2002), a PYRABACTIN RESISTANCE-LIKE4 homolog (Csa3G730890) encoding an ABA sensor (Park et al., 2009; Perrella et al., 2013), a myelocytomatosis oncogene homolog 2 (MYC2) gene (Csa3G902270) that positively regulates the JA response, and several ETHYLENE RESPONSE FACTOR transcription factor genes that participate in ethylene responses. The repression of negative regulators from the auxin, GA, and CK pathways, and positive regulators from the ABA, JA, and ethylene pathways, suggests activation of auxin, GA, and CK pathways and repression of ABA, JA, and ethylene pathways by SF2.

PAs, which are involved in complex crosstalk with phytohormones, also play roles in regulating cell division (Kaur-Sawhney et al., 2003). Our analysis showed that SF2 targets and represses the expression of several s-adenosyl-l-Met decarboxylase (SAMDC) genes (Csa2G036680, Csa3G271350, and Csa3G271360) that encode key enzymes in the biosynthesis of PAs (Anwar et al., 2015). ChIP-qPCR and RT-qPCR assays of 15 selected genes involved in phytohormone biosynthesis and signaling pathways further verified these results (Fig. 6, F–I; Supplemental Fig. S12; Supplemental Data Set S18).

Although histone deacetylation generally represses gene expression, research in mice found that a specific subset of mouse genes could be deregulated in the absence of HDAC1, suggesting a novel function for HDAC1 as a transcriptional coactivator (Zupkovitz et al., 2006). To assess the possibility that SF2 promotes histone deacetylation and activates a target gene network, the target genes with SF2 binding sites were compared with the 1,272 genes downregulated in sf2. A total of 237 genes showed overlap between the two data sets (Fig. 7A; Supplemental Data Set S19). We further found that 100 and 86 genes were hyperacetylated at H3K9 and H3K14, respectively, in sf2 (Fig. 7, B and C; Supplemental Data Sets S19–S21), of which 43 genes had both hyper-H3K9ac and hyper-H3K14ac sites (Fig. 7D; Supplemental Data Set S22).

GO analysis revealed that the core activated genes were primarily involved in DNA replication, chromatin assembly, and the cell cycle (Fig. 7E), including TUBULIN6 (Csa1G629200; Pastuglia et al., 2006), the cyclin gene CYCA1;1 (Csa6G382370; de Jager et al., 2005), and Cell Division Cycle 20.1 (Csa5G141120; Supplemental Data Set S19; Niu et al., 2015). We also found a microtubule motor kinesin gene, KF3 (Csa4G002000), which was previously implicated in rapid cell division by Yang et al. (2013). In addition, a small subset of genes in the core activated gene network were also involved in phytohormone biosynthesis or signaling pathways. Some are known to encode positive regulators of GA and CK, including GIBBERELLIN INSENSITIVE DWARF1 (Csa7G391240), which encodes a GA receptor (Griffiths et al., 2006), and LONELY GUY5 (LOG5; Csa6G127300), which is involved in CK biosynthesis (Tokunaga et al., 2012). Some are negative regulators of ABA signaling, such as an ABA-INSENSITIVE1 homolog (Csa1G574880; Yoshida et al., 2006). ChIP-qPCR and RT-qPCR assays of 15 selected genes involved in cell cycle regulation and phytohormone pathways were used to verify these results (Fig. 7, F–I; Supplemental Fig. S13). Consistent with the results for the core repressed genes, the activation of positive regulators involved in the GA and CK pathways, and negative regulators involved in the ABA pathway, suggested the activation of GA and CK pathways and repression of the ABA pathway by SF2.

To distinguish whether the core activated genes require the HDAC complex directly or indirectly for their activation, we analyzed several candidate genes for their responsiveness to Trichostatin A (TSA), a deacetylase inhibitor. We can conclude that genes requiring HDAC complexes for their direct activation should be negatively regulated by the deacetylase inhibitor in wild type. As shown in Supplemental Fig. S14, the four core activated genes bound by SF2 were verified by ChIP-qPCR. We showed that after treatment with TSA for 10 h, in both wild type and sf2 mutant, on most of these genes, the H3K9ac and H3K14ac levels were significantly increased, and their mRNA expression levels decreased. A recent study in mammalian cells showed that HDACs are required for limiting acetylation in gene bodies, and this function facilitates efficient transcriptional elongation (Greer et al., 2015). Our results indicated that in plants, the HDC1-HDAC complex probably can also activate the core genes activated by histone acetylation/deacetylation.

Because CK and GA signaling cascades mediate cell proliferation, while ABA generally has the opposite effect, these results suggest a function of SF2 in regulating cell proliferation by direct activation of the CK and GA signaling pathways and repression of ABA signaling.

Expression of SF2 Core Targets Occurs in Regions of Cell Proliferation

To further verify whether the SF2 core target genes are indeed involved in cell proliferation, we related their expression patterns in our transcriptome data with cell proliferation during fruit development.

Given that SF2 promotes cell division, its core repressed genes should be implicated in repression of cell proliferation, and thus, negatively correlate with cell proliferation during fruit development. Our results showed that the 321 core repressed genes could be grouped into three clusters, based on their expression patterns. The genes in both cluster 1 (126 genes; 39%) and cluster 2 (110 genes; 34%) showed reduced expression levels during exponential cell proliferation (0–3 DAA), whereas only 77 genes belonged to cluster 3 (24%), showing higher expression levels during cell proliferation (0–3 DAA; Fig. 8A; Supplemental Data Set S23). This indicated that the expression patterns of most of the core repressed genes were negatively related to cell proliferation. RT-qPCR analysis of ABA1, GIBBERELLIN INSENSITIVE, CKX7, SAMDC1, ETHYLENE RESPONSE FACTOR, and SAMDC2 confirmed that their expression was significantly repressed by SF2 during fruit development (Fig. 8B).

Additionally, the 237 core activated genes were grouped into three clusters based on their expression patterns. The genes in cluster 1 (114 genes; 49%) showed elevated expression levels during exponential cell proliferation (0–3 DAA), and most of the genes in cluster 3 (40 genes; 17%) showed gradually increased expression levels after 0 DAA (Fig. 8C). The expression patterns of these core activated genes were positively related to cell proliferation. However, the genes in cluster 2 (81 genes; 35%) showed higher expression levels at −3 DAA, suggesting specific roles before anthesis (Fig. 8C; Supplemental Data Set S24). RT-qPCR analysis of MAP65-6, CYCA 1;1, ABA-INSENSITIVE1, GIBBERELLIN INSENSITIVE DWARF1, LOG5, and TUBULIN6 confirmed that their expression was activated by SF2 (Fig. 8D).

Taken together, these results support the notion that the direct target gene networks of SF2 are involved in regulating cell proliferation during fruit development.

SF2 Regulates Cell Proliferation through its Effect on CK and PA Homeostasis

During early fruit development, SF2 activates LOG5 but represses CKX7 (Fig. 8, B and D), suggesting a role in promoting CK synthesis. SF2 also activates the expression of SAMDC genes (Fig. 8B). SAMDC is a key enzyme in PA biosynthesis of compounds such as spermidine and spermine, which may interact with CK and play essential roles in diverse growth and developmental processes (Kaur-Sawhney et al., 2003). PAs and CKs are essential for cell division (Kaur-Sawhney et al., 2003; Anwar et al., 2015). This suggests that SF2 facilitates cell proliferation through modulation of CK and PA homeostasis by targeting metabolic and biosynthetic genes. To test this hypothesis, we measured CKX enzyme activity and CK content in 0-DAA fruits of wild type and sf2, and showed that the enzyme activity was significantly elevated in the mutant (Fig. 9A), while the content of isopentenyladenine (iP) and dihydrozeatin decreased (Fig. 9B). Exogenous treatment of the −3 DAA of female flowers with thidiazuron (n-phenyl-n′-1,2,3-thiadiazol-5-yl urea, TDZ), an inhibitor of CKX (Hare and Van Staden, 1994), partially complemented the short-fruit phenotype (Fig. 9, C and D), and the treatment had a much stronger effect on the sf2 mutant than on wild-type fruit. This is consistent with a role for SF2 in facilitating cell proliferation through modulation of CK contents via targeting of its metabolic (CKX7) and biosynthetic genes (LOG5; Fig. 9E).

We next measured PA content and found that the spermidine levels were significantly higher in sf2 than in wild type (Fig. 9F). We then sprayed the −3 DAA of female flowers with methyl-glyoxyl-bis guanylhydrazone (MGBG), a competitive inhibitor of SAMDC, and found that the treatment partially complemented the short-fruit phenotype (Fig. 9, G and H) with no significant effect on cell size (Fig. 9I). Notably, we observed that the wild-type fruit grew less after the MGBG treatment, suggesting that an appropriate level of PAs is essential for fruit cell proliferation (Fig. 9, G and H). These results provide strong support for the hypothesis that SF2 facilitates cell proliferation through modulation of PA homeostasis via targeting of SAMDC genes.

Finally, we investigated the combined effect of TDZ and MGBG by spraying female flowers with a mixture of the chemicals and observed that the treatment largely complemented the short-fruit phenotype of sf2, although the transverse diameter also increased (Fig. 9, J and K). We found that the elongation was mostly due to increased cell numbers (Fig. 9L). Although the wild-type fruit, especially the fruit neck, became much longer after the treatment (Fig. 9J), the combined effect of TDZ and MGBG on sf2 fruit, which resulted in an 80% increase in length, was greater than on wild-type fruit, which increased by 16% (Fig. 9K). These results indicated that the combined effect of TDZ and MGBG was more than additive, suggesting agonistic effects of CKs and PAs on rapid cell proliferation regulated by SF2. Moreover, we observed the fruit-length phenotypes of the treated fruit at 16 DAA and found a partial complementation of the short-fruit phenotype of sf2, which increased in length by 18%, while wild-type fruit showed no significant change in length compared to untreated plants (Fig. 9, M and N).

DISCUSSION

Evidence of a Direct Role for HDC1 in Cell Proliferation

The basis of size control of multicellular organisms is a longstanding biological question. Two main processes, cell division and cell expansion, underlie final organ size. At the cellular level, cell division is regulated by a combination of two factors: the cell division rate and the cell division duration. Both factors influence the total number of cells in tissues. For instance, in Arabidopsis, transcription factors TEOSINTE BRANCHED1, CYCLOIDEA, and PCF1, and the growth-regulating factor are involved in increasing the cell division rate (Powell and Lenhard, 2012). Conversely, the ubiquitin-binding protein DA1 (Li et al., 2008), and the E3 ubiquitin-ligases DA2 and BIG BROTHER (Xia et al., 2013), limit organ size by repressing cell division duration. In tomato (Solanum lycopersicum), several genes have been identified that regulate fruit size and shape by affecting cell division, but the molecular mechanism remains elusive (Frary et al., 2000; Liu et al., 2002; Tanksley, 2004; Wu et al., 2011; Zhang et al., 2012). In cucumber, only a few genes affecting the cell division rate or duration have been identified and cloned to date.

Here, we found a recessive allelic variation at SF2, which is homologous to AtHDC1. Both the cell division rate and duration were inhibited in fruits of sf2 mutant, resulting in 70% reduction in cell numbers in the longitudinal direction of the fruit and a short-fruit phenotype. In contrast to the ubiquitous expression pattern of SF2 mRNA in various plant tissues, our study showed that the SF2 protein was specifically expressed in meristematic tissues in which rapid cell proliferation was occurring (Fig. 3D). A knock-out of SF2 using CRISPR-Cas9 caused substantial inhibition of shoot growth (Fig. 2, A–E), consistent with HDC1 having a general function in the control of meristematic cell proliferation (Fig. 10A). We inferred that cell number represents a major determinant of fruit size, because only fruit length was significantly affected by the weak allele of sf2 (Fig. 1, A–D). In addition, because the G515E mutation in sf2 is located in the yeast regulator of transcription3 domain, which is conserved among angiosperms (Fig. 1N), the function of this mutation is expected to be conserved in plants.

HDAC proteins form various types of complexes through interactions with different regulatory subunits. Our findings indicate that HDC1 is such a regulatory subunit, and is associated with the site- or tissue-specific function of HDAC in cell proliferation regulation. We describe here an elaborate regulatory cell proliferation network, in which SF2 directly targets and represses the expression of genes in multiple phytohormone pathways, and activates the expression of genes involved in cell cycle regulation (Fig. 10B). It would be interesting to investigate the interaction pattern and regulatory targets of HDC1 in other apical meristem cells to uncover the core mechanism of regulation of plant cell division.

The HDC1 Regulatory Mechanism Involving Histone Deacetylation and Gene Regulation

Similar to the results of the genome-wide studies of HDAC proteins in humans and plants (Wang et al., 2009; Chen et al., 2016a; Yang et al., 2016), SF2, the cucumber HDC1 homolog, mainly targeted genes with high transcription levels (Fig. 4C) and high acetylation levels (Fig. 5, A and B). This suggests a conserved action of HDAC complexes in regulating active genes in eukaryotes. Consistent with the general function of HDACs proteins as transcriptional repressors, we found that the SF2-HDAC complex acts as a direct repressor of target gene networks, including those that suppress auxin, GA, and CK biosynthesis and responses, and genes that promote JA, ABA, and PA biosynthesis and responses.

Another notable outcome of this study was the identification of a target gene network that requires HDC1 directly for its transcriptional activation (Fig. 7; Supplemental Fig. S15). These genes are involved in cell cycle regulation and in phytohormone biosynthesis or signaling pathways. We showed that the SF2-activated genes have higher SF2 binding levels and histone acetylation levels in the gene body regions, and HDC1–HDAC complex is shown to be required for limiting acetylation in gene bodies (Supplemental Fig. S15). We propose that HDC1 may utilize the same mechanism as that suggested for mammalian HDACs (Zupkovitz et al., 2006; Wang et al., 2009), where HDC1 recruits an HDAC complex to active genes, to limit acetylation in gene bodies, and positively regulate transcription through elongation machinery (Greer et al., 2015). In this scenario, histone deacetylation is essential for keeping transcription at steady-state levels (Wang et al., 2002, 2009). However, there may be another regulatory layer of histone deacetylation on transcriptional activation that is presently not well understood. Our findings indicate that HDC1 employs two different mechanisms for regulating transcriptional repression and activation during fruit cell proliferation (Fig. 10B).

The complementation of the fruit length phenotype of sf2 by exogenous treatment with hormones confirmed the role of SF2 in coordinating hormone biosynthetic and signaling pathways to facilitate cell proliferation (Fig. 9). An increasing number of studies suggest a tight link between epigenetic regulation and plant hormone signaling. Several critical regulatory factors, such as the chromatin remodeling factor, PICKLE (Ogas et al., 1997), and the corepressor of HDAC complex, TOPLESS (Pauwels et al., 2010), are potential key factors in coordinating plant hormone crosstalk (Yamamuro et al., 2016). Our results point to a general role for HDC1 in coordinating phytohormone signaling through integration with the HDAC complex during cell proliferation of meristematic cells.

MATERIALS AND METHODS

Plant Materials and Mutant Identification

The cucumber (Cucumis sativus) inbred line 406 was used to make a mutant library, by treating its seeds with 1 mm ethyl methanesulfonate (cat. no. M0880; Sigma-Aldrich) diluted with 0.1 m of P buffer (pH 7.0; Chen et al., 2016b). The plants of the first mutant generation (M1) were self-pollinated and the sf2 mutant was identified in the M2 population. An F2 population was generated by crossing sf2 with wild type 406. Whole genome resequencing was carried out as reported in Xu et al. (2017). Analysis of the resequencing data revealed 1,023 SNPs between the two bulked populations. SNP-index graphs were then calculated using these SNPs. We focused on the SNP-index peak on chromosome 2, which had an average Ɗ(SNP-index) > 0.5 and had greatest density of SNPs. Linkage analysis delimited the sf2 locus to a 474-kb interval between two SNPs (2G15049360 and 2G15523545) in the candidate region, and only one SNP (2G15231244) cosegregated with the sf2 locus.

Constructs and Generation of Transgenic Plants

For complementation of sf2, the SF2 coding region together with 995 bp of promoter was amplified and cloned into pCAMBIA1300 (Rao et al., 2015; primers are listed in Supplemental Data Set S25). To obtain the SF2 CRISPR/Cas9-edited plants, the binary pBSE402 vector containing a CRISPR cassette with a functional Cas9 under a constitutive promoter (CaMV 35S) plus a 35S-GFP expression cassette was modified from pBSE401 (a gift from Qijun Chen, China Agricultural University). The single guide RNA (sgRNA) target site from the N terminus of SF2 was selected (primers are listed in Supplemental Data Set S25). The binary vector pCAMBIA-SF2 or the pBSE402-sgRNA-SF2 was then transformed into Agrobacterium tumefaciens strain EHA105 by the freeze-thaw transformation protocol (Höfgen and Willmitzer, 1988).

Agrobacterium cells harboring the pCAMBIA-SF2 or the pBSE402-sgRNA-SF2 construct were used to transform the cucumber sf2 or CU2 lines, respectively, as described previously by Hu et al. (2017). Cucumber seeds were sterilized and spread on 1× Murashige and Skoog medium (Phytotech) supplemented with 2 mg/L of 6-benzylaminopurine (Sigma-Aldrich) and 1 mg/L of ABA (Phytotech) and left for 2 d at 28°C. Shoot regeneration, elongation, and rooting were performed as described previously by Hu et al. (2017).

Genomic DNA was extracted from callus and plants using the DNeasy Plant Mini Kit (Qiagen). PCR was performed using KOD-FX (Toyobo) and the gene-specific primers listed in Supplemental Data Set S25. PCR products were cloned into pEASY-Blunt Zero (Transgen Biotech) and the SF2 alleles were identified by sequencing.

Verification of the Causative SNP using dCAPS Markers

PCR primers for derived cleaved amplified polymorphic sequence (dCAPS) markers (Supplemental Data Set S25) were designed using the software “dCAPS FINDER 2.0” (http://helix.wustl.edu/dcaps/dcaps.html; Neff et al., 1998). The PCR products were digested with restriction enzyme as described in Supplemental Data Set S25 and subsequently separated by electrophoresis in 8% polyacrylamide gels.

Measurements of Cell Area and Number

Samples from −8, −5, −3, 0, 3, 5, 8, and 16 DAA of fruits were fixed in a 70% ethanol, acetic acid, and formaldehyde (90:5:5 by volume) solution. Sections (5-mm–thick) were cut with a scalpel from different parts of the fruit (outer, middle, and inner pericarp) and embedded in paraffin (Merck), and then used to generate 8-μm–thick sections in both widthwise and longitudinal directions using a model no. CM3050S microtome (Leica), before staining with hematoxylin-eosin and imaging (Yu et al., 2001). The cell number (X), cell area (A), and average cell area in a given section was calculated using the softwares “Infinity Capture 6.0” (Lumenera Corporation) and “Image Proplus 5.1” (http://www.mediacy.com/imageproplus; Yang et al., 2013). The area of the whole-fruit cross section or longitudinal section (A′) was determined by measuring the ovary or fruit diameter and using the equation for the area of a circle (for a cross section) or an ellipse (for a longitudinal-section). The cell number in whole fruit cross sections or longitudinal-sections (X′) was calculated by using the equation X/A = X′/A′. All the measurements were made at three sites of each tissue for three sections from each fruit.

Immunoblotting

Proteins were extracted using protein extraction buffer (20 mm of Tris-HCl at pH 7.5, 150 mm of NaCl, 4 m of Urea, 10% glycerol, 5 mm of dithiothreitol [DTT], 1 mm of Phenylmethanesulfonyl fluoride [PMSF], and 1× protease inhibitor cocktail [Roche]) and 20–50-μg protein extract was fractionated on a 12% SDS-PAGE gel. The protein concentration was measured according to the Bradford method (Bradford, 1976). Immunoblotting was conducted as reported in Cai et al. (2018). The antibodies used in this study are listed in Supplemental Data Set S26. The SF2 antibody was obtained from MBL Beijing Biotech. The SF2 antibody was raised in rabbit using a synthetic peptide corresponding to amino acids 50–66 of the SF2 protein sequence, and affinity-purified. An additional Cys was added to the C terminus to improve binding.

Immunolabeling

Whole-mount in situ immunocytochemical protein localization in 0-DAA cucumber fruit was performed as described in Paciorek et al. (2006). Primary rabbit anti-SF2 antibody was used in a 1:20 dilution with Tris-buffered saline TWEEN-20 (10 mm of Tris HCl at pH 8.0, 150 mm of NaCl, 0.05% TWEEN 20, and 0.5% nonfat dry milk) and secondary Fluorescein-Conjugated Goat anti-Rabbit (ZSGB-BIO; https://www.bioz.com/result/zsgb%20bio/product/ZSGB%20Biotech) antibody in a 1:500 dilution with Tris-buffered saline TWEEN-20. Images were collected using a model no. SP8 Confocal Microscope (Leica).

Transient Expression in Leaves of Nicotiana benthamiana

The coding region of SF2 was cloned into a binary vector (pCAMBIA1300) to fuse it with an MYC protein downstream of the 35S promoter, using the In-Fusion Cloning Kit (Clontech). The primers used are listed in Supplemental Data Set S25. Agro-infiltration for transient expression in leaves of N. benthamiana was carried out as in Ting et al. (2013). The experiments were repeated independently at least three times with similar results.

Co-IP and LC-MS/MS

Co-IP was performed using 0 DAA of wild type and sf2 fruits and anti-SF2 antibodies, as described in Wendrich et al. (2017), with some modifications. Briefly, 1 μg of the antibody bound to 0.5 mg of fruit protein was coupled to Dynabeads Protein A (Invitrogen) in PBS buffer (137 mm of NaCl, 2.7 mm of KCl, 10 mm of Na₂HPO4·12H₂O, and 2 mm of KH₂PO4). After washing with iP buffer (25 mm of Tris-HCl at pH 7.4, 150 mm of NaCl, 1% NP40, 5% glycerol, 1 mm of DTT, 1 mm of PMSF, and protease inhibitor cocktail), the beads were resuspended in 1× SDS sample buffer (50 mm of Tris at pH 6.8, 10% [v/v] glycerol, 2% [w/v] SDS, 0.1% Coomassie brilliant blue [G-250], and 2% [v/v] β-mercaptoethanol) and the coimmunoprecipitated proteins were separated using 12% SDS-PAGE and stained by the staining buffer (a liter containing 100 mL of acetic acid, 400 mL of methanol, 1 g of Coomassie brilliant blue [R-250], and 500 mL of water). Each lane of the gel was cut into three parts to avoid the IgG heavy and light chain, and subsequently used for LC-MS/MS as described in Nallamilli et al. (2013). For immunoblotting, anti-MYC antibody (MBL Beijing Biotech) or anti-FLAG antibody (Sigma-Aldrich) was used (Supplemental Data Set S26).

LCI Assay

The full-length SF2 and sf2 coding sequences were fused with the C-terminal fragment of firefly luciferase (Luc) in the pCAMBIA-Cluc vector (35S:CLuc-SF2 and 35S:CLuc-sf2); SIN3-LIKE1, SIN3-LIKE3, HDA19A, HDA19B, MSI1, and SAP18 were fused with the N-terminal fragment of Luc in the pCAMBIA-Nluc vector (35S:SIN3-LIKE1-NLuc, 35S:SIN3-LIKE3-NLuc, 35S:HDA19A-NLuc, 35S:HDA19B-NLuc, 35S:MSI1-NLuc, and 35S:SAP18-NLuc) according to Chen et al. (2008). Primers are shown in Supplemental Data Set S25. Agro-infiltration for transient expression in N. benthamiana leaves was carried out as described in Ting et al. (2013). Fifty hours after coinfiltration in N. benthamiana leaves, the leaves were sprayed with a luciferin solution (100 mM of luciferin and 0.1% Triton X-100) and images captured with a cooled charge-coupled device imaging apparatus (Chen et al., 2008). The LUC activity was measured as described in Chen et al. (2008). The assays were repeated three times.

RNA-Seq Experiment and GO Term Enrichment Analysis

Total RNA was isolated from 0 DAA of wild-type and sf2 fruits in two biological replicates using a TRIzol kit (Invitrogen), according to the instruction manual. Total RNA from wild-type fruits at different development stages (−3, 0, 1, 3, 5, and 8 DAA) was isolated in three biological replicates. Paired-end sequencing libraries with an average insert size of 250–300 bp were prepared according to the manufacturer’s instructions (Illumina), and 150-bp paired-end reads were generated using a model no. Hiseq2000 Analyzer (Illumina)

All reads were mapped to the reference genome (Huang et al., 2009) with default parameters using the software “HISAT2” (https://ccb.jhu.edu/software/hisat2/manual.shtml; Kim et al., 2015). After alignments, gene expression levels were reported as fragments per kilobase of transcript per million mapped reads, which were calculated by the software “StringTie” (https://ccb.jhu.edu/software/stringtie/; Pertea et al., 2016). Differentially expressed genes were identified through the package “DESEQ R” (http://bioconductor.org/packages/release/bioc/html/DESeq.html; Anders and Huber, 2010) with the cutoff: P value < 0.05 and fold change > 1.5. GO enrichment analyses were conducted for both the upregulated and downregulated genes using the software “TopGO” (Alexa and Rahnenfuhrer, 2010).

RT-qPCR and RT-PCR Analysis

One microgram of RNA was reverse-transcribed into complementary DNA with FastQuant RT Super Mix (Tiangen) according to the manufacturer’s instructions, followed by qPCR with SYBR Premix (Roche) using an ABI 7900 (Salk; primers are listed in Supplemental Data Set S25). Three independent biological replicates were used. Relative gene expression was calculated using the comparative 2^−△△Ct method (Livak and Schmittgen, 2001).

SF2 transcript levels in different tissues were analyzed by RT-PCR (Chiu et al., 2006) for 28 PCR cycles at an annealing temperature of 58°C. A cucumber ubiquitin gene (UBQ; Csa3G778350) was used as a reference. The primers are listed in Supplemental Data Set S25.

ChIP Assays

ChIP assays were performed with 0 DAA of wild-type and sf2 fruits essentially as described in Zhu et al. (2012). The following antibodies were used for ChIP assays: SF2 antibody (two biological replicates), anti-H3K9ac (Abcam; two biological replicates), and anti-H3K14ac (Abcam; two biological replicates; Supplemental Data Set S26). ChIP products were combined and eluted into 50 μL of Tris-EDTA buffer for ChIP-Seq (>5 ng DNA). The software “MACS 2.1.0” (model-based analysis of ChIP-Seq; http://liulab.dfci.harvard.edu/MACS/index.html) is used to identify ChIP-enriched regions for ChIP-Seq data (Zhang et al., 2008).

ChIP-qPCR was performed as described in Zhu et al. (2012). The primers are listed in Supplemental Data Set S25. The qPCR signals derived from the ChIP samples were normalized to the signals derived from the input DNA control sample. The value (percentage of input; input %) was calculated using the 2^−∆Ct method.

CKX Activity Assay

This assay was based on bleaching of 2,6-dichlorophenolindophenol as described in Galuszka et al. (2007) and Frébort et al. (2002). Approximately 1 g of fresh fruit tissue was ground in extraction buffer (0.2 m of Tris-HCl at pH 8.0, 0.3% Triton X-100, and 1 mm of PMSF). Cell debris was removed by centrifugation at 19,500g for 10 min. The protein extract was incubated in a reaction mixture (total volume of 0.6 mL in 1.5-mL tube) at 37°C in 100 mm of McIlvaine buffer (pH 6.5; McIlvaine, 1921) with 0.5 mm of 2,6-dichlorophenolindophenol and 0.5 mm of iP (6-[γ,γ-Dimethylallylamino]) for 6 h. This reaction was stopped by the addition of 0.3 mL of 40% trichloroacetic acid and 2% 4-aminophenol was added to the supernatant, and the sample was centrifuged at 19,500g for 5 min to remove protein precipitate. The resulting concentration of Schiff base was determined using the molar absorption coefficient (e352 = 15.2 mM⁻¹ cm⁻¹; Frébort et al., 2002) and expressed as the amount of cleaved iP per total protein and reaction time. The fruit protein concentration was measured according to the Bradford method (Bradford, 1976).

Quantification of Endogenous CK Levels

Each sample contained ∼120 mg of 0-DAA fruit tissue for quantification of endogenous CK levels as previously described by Novák et al. (2003) and Riefler et al. (2006). Three independent biological replicates were analyzed.

PA Measurements

Analysis of PAs was carried out according to methods described in Flores and Galston (1982) and Ge et al. (2006). Each sample contained ∼100 mg of 0-DAA fruit tissue for quantification of endogenous PA levels. Three independent biological replicates were analyzed.

Statistical Analyses

Statistical analysis was performed using one-tailed Student’s t tests to compare the two sample groups.

Data Availability

All data that support the findings within this article are available from the corresponding author upon request.

Accession Numbers

Cucumber genomic sequence data from this article can be found in the Cucurbit Genomics Database (icugi.org) under the following accession numbers: SF2 (Csa2G337260); CKX7 (Csa4G647490); SAMDC1 (Csa2G036680); SAMDC2 (Csa3G271360); SIN3-LIKE1 (Csa5G603960); SIN3-LIKE3 (Csa5G484650); MSI1 (Csa3G127190); HDA19A (Csa6G116140); HDA19B (Csa7G029990); SAP18 (Csa3G038150).

Sequence data from this article can be found in the GenBank data libraries under accession number SRP194253.

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1 Morphological changes in wild type and sf2 during early fruit development.

• Supplemental Figure S2 Identification of two independently complemented plants.

• Supplemental Figure S3 SF2 interacts with the HDAC complex.

• Supplemental Figure S4 Quantitative detection of interactions between SF2 and HDAC components in N. benthamiana leaves.

• Supplemental Figure S5 Co-IP assays indicating the interaction between SF2 and SIN3-LIKE proteins in vivo.

• Supplemental Figure S6 Representative target genes bound by SF2.

• Supplemental Figure S7 Representative DNA motifs identified in SF2 binding sites using MEME (Bailey et al., 2006).

• Supplemental Figure S8 Genomic distribution of H3K9ac and H3K14ac binding peaks.

• Supplemental Figure S9 Genome-wide comparison of H3K9ac and H3K14ac levels in wild-type and sf2 mutant fruits.

• Supplemental Figure S10 GO enrichment analyses of genes regulated by SF2.

• Supplemental Figure S11 GO enrichment analysis of 321 core SF2-repressed genes.

• Supplemental Figure S12 Analysis of nine core SF2-repressed genes.

• Supplemental Figure S13 Analysis of nine core activated genes of SF2.

• Supplemental Figure S14 Responsiveness of core SF2-activated genes to TSA.

• Supplemental Figure S15 Analysis of SF2 binding and H3Ac levels between SF2-activated genes and SF2-repressed genes.

• Supplementary Data Set S1 Phenotypic measurements of selfed F2 plants.

• Supplementary Data Set S2 SNP-index from a MutMap analysis (Abe et al., 2012).

• Supplementary Data Set S3 LC-MS/MS identification of proteins immunoprecipitated from wild-type fruit protein extracts using IgG.

• Supplementary Data Set S4 LC-MS/MS identification of proteins immunoprecipitated from wild-type fruit protein extracts using SF2-antibody.

• Supplementary Data Set S5 LC-MS/MS identification of proteins immunoprecipitated from sf2 fruit protein extract using IgG.

• Supplementary Data Set S6 LC-MS/MS identification of proteins immunoprecipitated from sf2 fruit protein extract using SF2-antibody.

• Supplementary Data Set S7 Distribution of SF2 complex peaks.

• Supplementary Data Set S8 Peaks corresponding to SF2 binding sites.

• Supplementary Data Set S9 Distribution of H3K9ac and H3K14ac peaks.

• Supplementary Data Set S10 Hyper-H3K9ac peaks.

• Supplementary Data Set S11 Hyper-H3K14ac peaks.

• Supplementary Data Set S12 The 1,494 upregulated genes in sf2.

• Supplementary Data Set S13 The 1,272 downregulated genes in sf2.

• Supplementary Data Set S14 The 321 genes that are both SF2 targets and upregulated in sf2.

• Supplementary Data Set S15 Overlap between the 321 core repressed genes and hyper-H3K9ac-enriched genes in sf2.

• Supplementary Data Set S16 Overlap between the 321 core repressed genes and hyper-H3K14ac-enriched genes in sf2.

• Supplementary Data Set S17 Fifty-two genes with both H3K9 and H3K14 hyperacetylation among the 321 core repressed genes.

• Supplementary Data Set S18 Sixteen hormone biosynthesis and signal transduction genes from the high-confidence 321 core repressed genes that were targeted and repressed by SF2.

• Supplementary Data Set S19 The 237 genes that are both SF2 target genes and downregulated in sf2.

• Supplementary Data Set S20 Overlap between the 237 core activated genes and hyper-H3K9ac-enriched genes in sf2.

• Supplementary Data Set S21 Overlap between the 237 core activated genes and hyper-H3K14ac-enriched genes in sf2.

• Supplementary Data Set S22 Forty-three genes among the 237 core activated genes with both H3K9 and H3K14 hyperacetylation.

• Supplementary Data Set S23 Expression of 321 core repressed genes from −3 to 8 DAA during early fruit development.

• Supplementary Data Set S24 Expression of 237 core activated genes from −3 to 8 DAA during early fruit development.

• Supplementary Data Set S25 The primers used in this study.

• Supplementary Data Set S26 The antibodies used in this study.

ACKNOWLEDGMENTS

We thank Zhizhong Gong from China Agricultural University for comments on the article, PlantScribe (www.plantscribe.com) for editing this article, and Qing Li from the Chinese Academy of Agricultural Sciences for experimental assistance.

LITERATURE CITED

Author notes