Histone acetyltransferase GCN5-mediated regulation of long non-coding RNA At4 contributes to phosphate starvation response in Arabidopsis.

Phosphate availability is becoming a limiting environmental factor that inhibits plant growth and development. Here, we demonstrated that mutation of the histone acetyltransferase GCN5 impaired phosphate starvation responses (PSRs) in Arabidopsis. Transcriptome analysis revealed that 888 GCN5-regulated candidate genes were potentially involved in responding to phosphate starvation. ChIP assay indicated that four genes, including a long non-coding RNA (lncRNA) At4, are direct targets of GCN5 in PSR regulation. In addition, GCN5-mediated H3K9/14 acetylation of At4 determined dynamic At4 expression. Consistent with the function of At4 in phosphate distribution, mutation of GCN5 impaired phosphate accumulation between shoots and roots under phosphate deficiency condition, whereas constitutive expression of At4 in gcn5 mutants partially restored phosphate relocation. Further evidence proved that GCN5 regulation of At4 influenced the miRNA miR399 and its target PHO2 mRNA level. Taken together, we propose that GCN5-mediated histone acetylation plays a crucial role in PSR regulation via the At4-miR399-PHO2 pathway and provides a new epigenetic mechanism for the regulation of lncRNA in plants.

Introduction

Phosphorus is an essential macronutrient for plant metabolic processes, including energy metabolism, synthesis of nucleic acids and membranes, photosynthesis, respiration, nitrogen fixation, and enzyme regulation (Raghothama, 1999). Phosphate (Pi) deficiency is a limiting factor for plant growth and productivity in 40% of the world’s arable soils (Vance, 2001). To cope with this, plants have evolved regulatory mechanisms to maintain phosphorus homeostasis by improving Pi uptake, translocation, remobilization, and efficiency of use (Lin ; Rouached ; Chiou and Lin, 2011; Liang ; Wang ). Under Pi-limited conditions, plants usually exhibit modified RSA (root system architecture), including reduced primary root, increased number and density of root hairs, and the secretion of phosphatases and organic acids (Lambers , 2006; Fang ; Yang and Finnegan, 2010; Xu ).

Non-coding RNA is one of the key regulators involved in the phosphate starvation response (PSR) network (Rouached ; Yang and Finnegan, 2010). The expression of the miRNA miR399 is strongly induced upon Pi starvation, especially in vascular tissues of the shoot. Mature miR399 is then translocated to roots and binds to the 5'-untranslated region (UTR) of its target PHOSPHATE2 (PHO2) transcripts, leading to the degradation of PHO2 mRNA (Bari ; Lin ; Du ). PHO2, an E2 ubiquitin-conjugating enzyme that degrades PHO1 (Liu ; Secco ) and perhaps also other proteins, negatively affects shoot Pi content and Pi remobilization in plants (Bartel, 2004; Aung ; Chitwood and Timmermans, 2007; Ying ). At4 and INDUCED BY PHOSPHATE STARVATION1 (IPS1) are Pi starvation-responsive long non-coding RNAs (lncRNAs), which sequester miR399 in a ‘target mimicry’ manner (Franco-Zorrilla ). Under Pi-deficient conditions, increased At4 lncRNA inhibits the cleavage of PHO2 by miR399, and the up-regulated functionally intact PHO2 transcript results in Pi translocation from shoots to roots, which benefits the development of the RSA (Fujii ; Aung ; Bari ; Bazin and Bailey-Serres, 2015). Some transcription factors have been reported to be involved in the regulation of Pi transporters. For instance, the MYB transcription factor family member MYB2 functions as a direct transcriptional activator for miR399 in Pi starvation signaling (Baek ), and PHOSPHATE STARVATION RESPONSE1 (PHR1) initiates the up-regulation of Pi starvation-responsive genes (Rubio ). In addition, WRKY6 was identified to inhibit PHO1 expression as a transcriptional repressor (Chen ).

Studies have shown that epigenetic mechanisms play important roles in regulating plant biotic and abiotic stress responses (Ahmad ; Sahu ; Secco ). Nuclear actin-related protein ARP6 is an essential component of the SWR1 chromatin remodeling complex, which regulates transcription via deposition of the H2A.Z histone variant into chromatin. Derepression of target loci including PSR genes in the arp6 mutant is correlated with the presence of Pi starvation-related phenotypes, which provides a link between chromatin remodeling and the PSR (Smith ). In addition, recent studies demonstrate that ALFIN-LIKE 6 (AL6), a member of the Alfin1-like plant homeodomain-containing protein family (PHD finger), controls the transcription of a suite of genes critical for root hair elongation under low Pi conditions (Chandrika ). PHD fingers have been identified as effectors that specifically recognize H3K4me3 (trimethylated lysine 4 of histone H3) and H3K4me2 (Taverna ; Wysocka ; Zhang, 2006; Sims ).

Histone acetylation and deacetylation, catalyzed by histone acetyltransferases (HATs) and histone deacetylases (HDACs), are essential for gene expression regulation (Jenuwein and Allis, 2001). In Arabidopsis, mutation of the histone acetyltransferase GCN5 (general control non-repressed protein 5) causes pleiotropic defects that alter many aspects of plant development and responses to environmental conditions (Vlachonasios ; Benhamed ; Cohen ; Servet ; Hu ; Xing ; Wang , 2018; Kim ; Zheng ). Studies showed that mutation of GCN5 resulted in a long-hypocotyl phenotype and reduced light-inducible gene expression, whereas mutation of histone deacetylase HDA19, an RPD3-containing HDAC, induced opposite effects (Benhamed ). Antisense inhibition or mutation of HDA19 leads to a range of developmental abnormalities, including early senescence, serrated leaves, the formation of aerial rosettes, and delayed flowering accompanied by defects in floral organ identity (Wu ; Tian and Chen, 2001). Recently, a group reported that HDA19 modulates a subset of PSRs by regulating genes encoding SPX domain-containing proteins and genes involved in lipid remodeling (Chen ). Here, we aimed to investigate the regulation of the PSR by histone acetylation in Arabidopsis. For this purpose, we explored the effects of GCN5 on Pi-deficient responses and its functions in modulating the acetylation levels of target genes important for the PSR network.

Materials and methods

Plant materials and growth conditions

Seeds of Arabidopsis thaliana wild-type Ws, gcn5-1, and gcn5-2 mutants (Ws background) were used in this study. The gcn5-1 and gcn5-2 mutants were both T-DNA insertion mutants in the bromodomain-coding region (Bertrand ; Vlachonasios ), which were kindly provided by Professor Zhou Daoxiu. For seed germination, sterilized seeds were incubated at 4 °C for 3 d; seeds were then sown on Murashige and Skoog (MS) plates containing 1% sucrose and 0.6% agar. Seedlings were grown under 16 h light/8 h dark conditions at 22 °C in a growth room. For Pi deficiency treatment, 7-day-old seedlings were transferred to MS medium containing 0.0125 mM KH2PO4 (1% Pi content of 1× MS) for the indicated times (Baek ).

Phosphate analysis

Pi concentrations were analyzed with molybdenum–ascorbic acid as described (Fujii ) with minor modifications. Briefly, 100 mg of fresh shoots (with the old and etiolated leaves removed) and roots homogenized in liquid nitrogen were suspended in 1 ml of solution I (100 mM NaCl, 10 mM Tris, 1 mM EDTA, 0.1% β-mercaptoethanol, pH 8.0). Nine volumes of 1% glacial acetic acid were added. After incubation at 42 °C for 30 min, the solution was centrifuged at 5000 g for 10 min. A 60 μl aliquot of supernatant was mixed with 140 μl of solution II [0.35% (NH4)6Mo7O24·4H2O, 0.86 N H2SO4, 1.43% ascorbic acid] and incubated at 45 °C for 20 min. Absorbance at 820 nm was measured using a BioTek Synergy HT Multi-Mode Microplate Reader (BioTek, Vermont, USA). KH2PO4 was used for producing the standard curve.

RNA isolation, library preparation, and transcriptome sequencing

Total RNA was extracted from 13-day-old seedlings of Ws and gcn5-2 which had been subjected to Pi-sufficient and -deficient treatments, using TRIzol reagent (Invitrogen, Carlsbad, CA, USA), and three replicates for each sample were analyzed. Paired end sequencing libraries with an average insert size of 200 bp were prepared with a TruSeq RNA Sample Preparation Kit v2 (Illumina, San Diego, CA, USA) and sequenced on HiSeq2000 (Illumina) according to the manufacturer’s standard protocols. Raw data obtained from Illumina sequencing were analyzed as previously described, with minor modifications (Li ). In brief, each library was mapped independently using TopHat version 2.0.12 (http://ccb.jhu.edu/software/tophat/index.shtml) against the A. thaliana genome sequence index (TAIR 10). For differential gene analyses of each comparison, Cuffdiff version 2.2.1 was run by using the reference transcriptome along with the BAM files resulting from TopHat for each sample. The genes, with changes >2-fold and a q-value <0.05, were defined as differentially expressed genes. The groups of differentially expressed genes from RNA sequencing (RNA-seq) in this study are shown in Supplementary Table S2 at JXB online.)

Chromatin immunoprecipitations

ChIP assays were performed as described previously (Fiil ). Arabidopsis seeds were sterilized, kept for 3 d at 4 °C, and grown in vitro under long-day conditions. After 7 d growth on MS medium, the seedlings were separately treated in normal or Pi deficiency conditions, and were harvested and fixed in 1% formaldehyde for 15 min in a vacuum and then neutralized by 0.125 M glycine. After washing with sterilized water, the samples were ground in liquid nitrogen. Nuclear pellets were suspended in a buffer containing 0.25 M sucrose, 10 mM Tris–HCl, pH 8, 10 mM MgCl2, 1% Triton X-100, 5 mM β-mercaptoethanol, 0.1 mM phenylmethylsulfonyl fluoride (PMSF), and protease inhibitors (one mini tablet per ml; Roche). The suspensions were transferred to microfuge tubes and centrifuged at 12 000 g for 10 min. The pellets were suspended in 1.7 M sucrose, 10 mM Tris–HCl, pH 8, 2 mM MgCl2, 0.15% Triton X-100, 5 mM β-mercaptoethanol, 0.1 mM PMSF, and protease inhibitors, and centrifuged through a layer of the same buffer in microfuge tubes. The nuclear pellets were lysed in a buffer containing 50 mM Tris–HCl, pH 8, 10 mM EDTA, 1% SDS, and protease inhibitors. The lysed nuclei were sonicated four times for 15 s at 4 °C followed by centrifugation. The supernatants containing chromatin fragments were diluted 10-fold with 1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris–HCl, pH 8, and 167 mM NaCl. Aliquots of the dilution were used for an immunoprecipitation assay.

The GCN5 antibody was kindly provided by Professor Zhou Daoxiu, and its specificity was confirmed by protein gel blots (Benhamed , 2008). The H3K14ac (Cat no. A-4023-025) and H3K9ac (Cat no. A-4022-025) antibodies were bought from Upstate Biotechnology. CHS was amplified as an endogenous control. Immunoprecipitated DNA was analyzed by PCR using the primer sets listed in Supplementary Table S1, and amplified DNA from the chromatin fractions prior to antibody incubation were used as the controls (inputs). The fold enrichment was normalized to the chromatin inputs.

Examination of miR399

Total RNA was extracted with TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. A stem–loop reverse transcription–PCR (RT–PCR) was performed to quantify miR399 as previously described (Chen ). Quantitative real-time PCR (qRT-PCR) was conducted on a Bio-Rad IQ5 Real-Time PCR Detection System. Each reaction included 2 μl of product from the diluted reverse transcription reactions, 1.0 μl of each primer (forward and reverse), 12.5 μl of SYBR® Premix Ex Taq™ (Perfect Real Time; TaKaRa), and 8.5 μl of nuclease-free water. The reactions were incubated in a 96-well plate at 95 °C for 30 s, followed by 40 cycles of 95 °C for 5 s, 60 °C for 30 s, and 72 °C for 30 s. The specificity of the PCR-amplified miR399f product was proved by T-vector cloning combined with sequencing. All reactions were run in three replicates for each sample. The ACT8 gene served as the endogenous control.

Accession numbers

Sequence data from this study can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: At4, AT5G03545; GCN5, AT3G54610; MYB2, AT2G47190; PHO2, AT2G33770; PHO1, AT3G23430; WRKY6, AT1G62300; MYB62, AT1G68320; SBT3.5, AT1G32940; RIPK, AT2G05940; PLDP2, AT3G05630; PLIP3, AT3G62590; MGD2, AT5G20410; ACT8, AT1G49240; and CHS, AT5G13930; The RNA-seq reads used for this study are deposited at the National Center for Biotechnology Information Short Read Archive under the accession number SRP201144.

Results

gcn5 mutants display impaired Pi deficiency responses in Arabidopsis

In order to clarify the roles of histone acetylation for Pi deficiency responses, we screened Arabidopsis T-DNA insertion mutants for 12 genes, namely GCN5, SILENT INFORMATION REGULATOR2 (SRT2), AT5G61050 (MAF19.6), HDA2C, HDA5, HDA7, HDA13, HDA14, HDA15, HDA18, HDA19, and PKL, which are essential modulators of the histone acetylation level in Arabidopsis. After Pi deficiency treatment (1% Pi content of 1× MS) for 6 d, we noted that both gcn5-2 and gcn5-1 mutant alleles (the T-DNAs were inserted in the bromodomain-coding region) showed obvious yellowish-purple leaf color compared with the wild-type Ws, which is a symptom of Pi starvation and may be connected to anthocyanin accumulation and senescence in leaves (Fig. 1A). In contrast to the significant inhibition of root growth in wild-type Ws in response to Pi deficiency, the root growth of gcn5 mutants was slightly affected by Pi deficiency treatment (Fig. 1A, B; Supplementary Fig. S1). Moreover, the S:R of the fresh weight in gcn5 mutants decreased significantly under low Pi conditions relative to that of the wild-type plants (Fig. 1C), indicating the more severe response to Pi deficiency in leaves of gcn5 mutants compared with that in Ws.

Fig. 1.

GCN5 plays important roles in PSRs in Arabidopsis. (A) Seven-day-old seedlings of Ws, gcn5-1, and gcn5-2 were transferred onto MS plates with sufficient Pi (+P) or deficient Pi (–P) for an additional 6 d. (B) FW and (C) shoot to root ratio (S:R) of the FW of 13-day-old seedlings of Ws, gcn5-1, and gcn5-2 grown under Pi-sufficient or -deficient conditions. Values are the mean of three replicates for each sample. Bars with asterisks are significantly different from the wild-type Ws in each condition (*P<0.05, **P<0.01; Student’s t-test). (D) Pi concentration and (E) S:R ratio of Pi (the ratio of the total amount of Pi in the shoot and root) are shown for shoots and roots of 13-day-old seedlings of Ws and gcn5 grown under Pi-sufficient or -deficient conditions. Values are the mean of three replicates for each sample. Bars with asterisks are significantly different from the corresponding wild-type Ws in each condition (*P<0.05, **P<0.01; Student’s t-test). (F) Seven-day-old Ws seedlings were transferred onto MS plates with deficient Pi for 0 h to 5 d. Total RNA was isolated, and qRT-PCR showed the dynamic expression of GCN5. The expression of ACT8 was used to normalize mRNA levels. Error bars represent the SD values from three biological repetitions for each sample, and the experiment was repeated at least three times. Bars with asterisks are significantly different among comparisons of Pi-sufficient and -deficient treatments at different time points (**P<0.01, *P<0.05; Student’s t-test). (This figure is available in color at JXB online.)

Plants translocate a large proportion of Pi from the shoots to the roots to maximize Pi acquisition in response to a limiting Pi level (Ticconi and Abel, 2004). As shown in Fig. 1D, the gcn5 mutants showed decreased Pi concentration in shoots and roots compared with Ws in both Pi-deprived and -replete conditions, but the S:R ratio of Pi was significantly up-regulated after Pi deficiency treatment (Fig. 1E), which proved that the mutation of GCN5 may impair the uptake and allocation of Pi in plants during Pi deficiency. In addition, we examined the time course expression patterns of GCN5 during the low Pi treatment and found that GCN5 transcripts were modestly induced by Pi deficiency and peaked after 1 d of Pi deficiency treatment (Fig. 1F). Collectively, these data suggested that GCN5 contributed to the robust PSRs in Arabidopsis.

Illumina high-throughput sequencing reveals 888 genes as GCN5-regulated candidate genes for PSRs

As a histone acetyltransferase, GCN5 generally regulates the acetylation levels of hundreds of loci in Arabidopsis (Imoberdorf ; Benhamed ). Thus, genomic-scale research was needed to interpret the complicated roles of GCN5 in PSR regulation. Because the two gcn5 mutants showed similar phenotypes, the mutant we used for further analysis was gcn5-2. We applied Illumina high-throughput sequencing approaches to compare the transcription profiles of 13-day-old seedling in four conditions, designated as Ws (wild type with sufficient Pi), Ws-P (wild type with deficient Pi), gcn5 (gcn5 mutant with sufficient Pi), and gcn5-P (gcn5 mutant with deficient Pi), respectively. With three biological replicates for each sample, the reproducibility and the accuracy of transcriptome data were validated by qRT-PCR (Supplementary Fig. S2). After Pi deficiency treatment, 3036 genes showed a dramatic response to Pi deficiency in Ws, whereas only 358 genes were changed in the gcn5 mutant, designated as ‘Ws-P/Ws’ and ‘gcn5-P/gcn5’, respectively (Supplementary Table S2). Additionally, a total of 445 and 2461 genes were significantly changed in the gcn5 mutant relative to Ws in Pi-replete and Pi-deprived conditions, respectively, designated as ‘gcn5/Ws’ and ‘gcn5-P/Ws-P’ (Supplementary Table S2). Comparisons of the four pairs of transcript abundances showed that genes differentially expressed in ‘Ws-P/Ws’ were obviously less affected in ‘gcn5-P/gcn5’, and exhibited the reverse expression pattern in ‘gcn5-P/Ws-P’ (Fig. 2A). These results suggested that the response to Pi deficiency is largely impaired in the gcn5 mutant, indicating that GCN5 functions in the PSR.

Fig. 2.

Identification of the potential GCN5 targets for PSRs. (A) Heatmap of the logFC values in ‘Ws-P/Ws’, ‘gcn5-P/gcn5’, ‘gcn5/Ws’, and ‘gcn5-P/gcn5’ of genes that are differentially expressed in all four comparisons. (B) Venn diagram of the determined differentially expressed genes in each comparison, which represented potential GCN5 targets in response to Pi deficiency. (C) Summary of the over-represented GO categories (biological processes; P-value <0.05; Benjamini–Hochberg correction) for the 888 GCN5-regulated candidate genes, using AgriGO v2.0. The number of genes that are associated with that respective GO category and the corresponding P-value are represented as dots. (This figure is available in color at JXB online.)

Since GCN5 functions as a transcriptional activator (Imoberdorf ), the GCN5-regulated genes in the PSR should be incorporated in the overlap between ‘Ws-P>gcn5-P’ and ‘Ws-P>Ws’ (Fig. 2B). Among the overlapping genes, 90 genes belonged to the ‘gcn5-P>gcn5’ group which might not be regulated by GCN5. Finally, the 888 genes were selected as GCN5-regulated candidate genes involved in the PSR (Fig. 2B). Based on Gene Ontology (GO) categories, we used AgriGO v2.0 to analyze the enrichments of function-related gene groups (Tian ). The majority (≥50%) of GCN5-regulated candidate genes had a significant enrichment [false discovery rate (FDR) <0.05] of response to stimulus, cellular process, and cell part, which includes, for example, response to inorganic substances, the phosphorelay signal transduction system, and cellular response to phosphate starvation, that may be associated with PSR mechanisms. In addition to these, a large fraction of genes implicated in membrane processes, signaling, localization, immune system, cell junction, and transporter activity were also over-represented, suggesting that the cellular signal transduction-related processes were significantly influenced (Fig. 2C; Supplementary Table S3).

Four genes are identified as direct targets of GCN5 in PSR regulation

To further clarify the relationship between GCN5 and Pi deficiency response, direct targets of GCN5 after Pi deficiency treatment need to be identified. Based on the GO analysis results, we focused on 10 candidate genes from the GO category of cellular response to phosphate starvation (Table 1), including WRKY6, MYB62, and At4, which were reported to play important roles in phosphate redistribution under Pi-limited or starvation conditions (Shin ; Franco-Zorrilla ; Chen ; Devaiah ). We performed ChIP assays to examine the 10 candidate genes in 13-day-old seedlings of Ws and the gcn5 mutant under Pi deficiency by using GCN5-specific antibodies (Benhamed , 2008). Because GCN5 is mainly responsible for the acetylation of its targets at promoters (Bertrand ; Bhat ; Earley ), primers spanning core promoter regions were used for ChIP-qPCR analysis. As shown in Fig. 3A, GCN5 significantly bound to four of the candidate genes in the wild type relative to the gcn5 mutant, namely WRKY6, SBT3.5, RIPK, and At4.

Table 1.

The potential GCN5-regulated, Pi deficiency-induced genes

Gene ID	Gene name	Annotation
AT1G67600		Acid phosphatase/vanadium-dependent haloperoxidase-related protein
AT1G62300	WRKY6*	WRKY family transcription factor
AT1G19200		Cyclin-dependent kinase, putative (DUF581)
AT1G68320	MYB62*	Myb domain protein62
AT1G32940	SBT3.5	Subtilase family protein
AT2G05940	RIPK	Protein kinase superfamily protein
AT3G05630	PLDP2	Phospholipase DP2
AT3G62590	PLIP3	PLASTID LIPAS3
AT5G03545	At4*	Response to phosphate starvation, enhanced by the presence of IAA
AT5G20410	MGD2	Monogalactosyldiacylglycerol synthase2

Asterisks indicate the three well-studied genes, WRKY6, MYB62, and At4, which are involved in Pi homeostasis.

Fig. 3.

Identification of the direct targets of GCN5 and measurement of their acetylation states. Seven-day-old seedlings of Ws and gcn5 mutants were separately transferred onto MS plates with deficient Pi for 3 d. Nuclei were extracted from the cross-linked seedlings, sonicated, and immunoprecipitated with antibodies specific to GCN5, H3K9ac, and H3K14ac, respectively. Primers were designed at the core promoter region. (A) ChIP assay (anti-GCN5) to identify the direct target of GCN5. Asterisks indicated significant differences in enrichment in the gcn5 mutant compared with that in wild-type Ws. CHS, whose expression is not affected by GCN5, was used as a negative control. (B) ChIP assay (anti-H3K9ac or anti-H3K14ac) to examine the H3K9ac or H3K14ac states of the four GCN5 target genes. Asterisks indicated significant differences in histone acetylation level in the gcn5 mutant compared with that in wild-type Ws. (C) qRT-PCR to show the expression of GCN5 target genes. Error bars represent SD values from three biological repetitions for each sample, and the experiment was repeated at least three times (**P<0.01; Student’s t-test).

GCN5 is specifically responsible for H3K14ac as well as affecting the acetylation of H3K9 and H3K27 which is required for the expression of hundreds of genes (Vlachonasios ; Earley ; Benhamed ; Chen ). To further confirm the direct regulation of these four potential target genes by GCN5, we performed ChIP-qPCR assays using antibodies against H3K9ac and H3K14ac, respectively. Consistent with the known properties of GCN5 as a histone acetyltransferase, the H3K9ac and/or H3K14ac levels of the four genes decreased due to the impairment of GCN5 (Fig. 3B). As expected, the mRNA levels of these four genes also significantly decreased in gcn5 mutants compared with the wild type under Pi-deficient conditions (Fig. 3C). Taken together, we concluded that GCN5 directly associated with these four loci, modulated their H3K9ac and/or H3K14ac levels, and in turn regulated their expression.

The regulation of acetylation of GCN5 affects the induction of At4 under Pi-deficient conditions

Given the fact that H3K9/14ac levels and the expression of At4 were significantly impaired in the gcn5 mutant (our data) and that At4 participates in the control of phosphate in Arabidopsis (Burleigh and Harrison, 1999; Shin ), we considered that At4 could be an important target for GCN5 in PSRs. Consistent with the up-regulation of GCN5, At4 was significantly induced 1 d later than GCN5 and reached its highest expression level 2–3 d after phosphate withdrawal (Figs 4A, 1F), implying that At4 might be the target gene for GCN5. To investigate how the induction of GCN5 affected At4 transcription, we performed ChIP experiments using an antibody against H3K14 acetylation. In support of our hypothesis, we found that H3K14ac accumulated to a higher level at the core region of the At4 promoter under Pi-deficient conditions, and reached ~2-fold the level than that under Pi-sufficient conditions at 1–2 d time points (Fig. 4B). Collectively, these results assigned a key role for GCN5 in At4 induction under Pi deficiency conditions. By performing another ChIP assay using the GCN5 antibody, we showed that the enrichment of GCN5 protein on the At4 promoter was elevated in a time-dependent manner after Pi withdrawal and reached the highest level 2 d after treatment (Fig. 4C). To investigate further whether GCN5 binds to At4 before Pi treatment, we performed a ChIP assay to examine the H3K14ac levels of the At4 promoter and gene body regions in the wild type and gcn5 mutants. In Pi-sufficient plants, the H3K14 acetylation of At4 was decreased dramatically at the promoter region and modestly at the gene body region in the gcn5 mutant under normal conditions (Fig. 4D). Altogether, these results demonstrated that GCN5 directly interacted with the At4 promoter and promoted the robust induction of At4 expression in response to Pi deficiency by modulating its H3K14ac level.

Fig. 4.

GCN5 directly regulates At4 dynamic expression through histone acetylation under Pi deficiency conditions. (A) Seven-day-old Ws seedlings were transferred onto MS plates with sufficient Pi (+P) or deficient Pi (–P) for 0 h to 5 d. Total RNA was isolated, and qRT-PCR showed the dynamic expression of At4 transcripts. The expression of ACT8 was used to normalize mRNA levels. Error bars represent SD values from three biological repetitions for each sample, and the experiment was repeated at least three times (*P<0.05, **P<0.01; Student’s t-test). (B) We examined the H3K14 acetylation state on the At4 promoter of Ws in normal conditions and after 0, 1, 2, 3, and 4 d of Pi deficiency treatments. Error bars represent SD values from three biological repetitions for each sample, and the experiment was repeated at least three times (**P<0.01; Student’s t-test). (C) ChIP was performed to examine the enrichment of GCN5 on the At4 promoter; CHS was used as a negative control. (D) ChIP was performed to examine the H3K14ac state on the At4 promoter and gene body region before Pi-deprived treatment; P1–P2 and P3–P4 indicated the primers designed for promoter and gene body examinations, respectively. Error bars represent SD values from three biological repetitions for each sample, and the experiment was repeated at least three times (**P<0.01; Student’s t-test). (This figure is available in color at JXB online.)

The contribution of At4 to the GCN5-regulated PSR network

To confirm whether At4 is the direct target of GCN5 during the phosphate-deficient response, At4 was constitutively overexpressed in gcn5 mutants. Since it was difficult to directly obtain 35S:At4/gcn5 transgenic plants due to the low viability of gcn5 pollen, we firstly generated 35S:At4/Ws transgenic lines. Two independent 35S:At4/Ws (#19 and #23) transgenic lines which exhibited the highest At4 transcript abundance were selected for further analysis (Supplementary Fig. S3A). Notably, PHO2 mRNA levels were significantly elevated along with the overexpression of At4 in these two lines (Supplementary Fig. S3B). By crossing the gcn5 mutant with 35S:At4/Ws transgenic lines, we successfully overexpressed At4 in the gcn5 mutant.

At4 and IPS1 have been reported to function in Pi homeostasis by sequestering miR399 via ‘target mimicry’ (Shin ; Franco-Zorrilla ). There are six loci in total identified as miR399 family genes in Arabidopsis, and all of them could target PHO2, which has five sequences complementary to miR399 in its 5'UTR (Lee ; Aharoni ). Among them, miR399f was the first member which was reported to be involved in the PSR (Koornneef ). In low Pi conditions, the miR399f could be detected at a much higher level in the gcn5 mutant compared with Ws (Fig. 5A), and the abundance of miR399d and miR399e was also increased (Supplementary Fig. S3C). Moreover, miR399f abundance was decreased along with the overexpression of At4 in the gcn5 mutant background (Fig. 5A). It should be noted that the primary transcripts of miR399f were significantly down-regulated in the gcn5 mutant compared with the wild type and 35S:At4/gcn5 transgenic lines in Pi-deprived conditions (Fig. 5B). It has been reported that MYB2 encodes a transcription factor of miR399f and positively regulates its expression (Raffaele ); qRT-PCR results showed that the expression level of MYB2 was significantly decreased in gcn5 mutants in both Pi-replete and Pi-deprived conditions (Fig. 5C). These results revealed that the expression pattern of miR399f was different from that of pri-miR399f and MYB2 in gcn5 mutants. Previous studies showed that mature miR399 binds to the 5'UTR of PHO2 transcripts, leading to the degradation of PHO2 mRNA (Fujii ; Aung ). The intact PHO2 mRNA decreased in gcn5 mutants but could be rescued up to nearly the wild-type level in 35S:At4/gcn5 transgenic lines under Pi deficiency treatment (Fig. 5D). All these data indicated that the decreased expression of At4 in gcn5 mutants caused the changes in miR399 abundance and PHO2 expression pattern.

Fig. 5.

Examination of miR399 and PHO2 transcript abundance and Pi concentration in shoots and roots. Total RNA were isolated from 13-day-old seedlings of Ws, gcn5, and 35S:At4/gcn5 transgenic lines grown under Pi-sufficient or -deficient conditions. (A) The miR399f abundance was detected by stem–loop qRT–PCR. (B) Transcript abundance of pri-miR399f expression in Ws, gcn5, and 35S:At4/gcn5 transgenic lines was detected by qRT-PCR. (C) The expression level of MYB2 was detected by qRT-PCR in Ws and gcn5 under normal conditions. (D) Examination of PHO2 mRNA in Ws, gcn5, and 35S:At4/gcn5 transgenic lines grown under Pi-sufficient or -deficient conditions was performed by qRT-PCR. The expression of ACT8 was used to normalize mRNA levels. Error bars represent SD values from three biological repetitions for each sample, and the experiment was repeated at least three times (*P<0.05, **P<0.01; Student’s t-test). (E) Pi concentration and (F) S:R ratio of Pi (the ratio of the total amount of Pi in the shoot and root) are shown for shoots and roots of 13-day-old seedlings in Ws, gcn5, and 35S:At4/gcn5 plants grown under Pi-deficient conditions. Values are the mean of three replicates for each sample. Bars with asterisks are significantly different from the corresponding wild-type Ws in each comparison (*P<0.05, **P<0.01; Student’s t-test). (This figure is available in color at JXB online.)

The phosphate signaling pathway defined by At4–miR399–PHO2 contributed to Pi translocation from shoots to roots under Pi-deficient conditions, which benefits the development of RSAs (Fujii ; Aung ; Bari ; Bazin and Bailey-Serres, 2015). To investigate the phosphate translocation ability of Ws, gcn5, and 35S:At4/gcn5 under Pi-deficient condition, we measured the Pi concentration of shoots and roots, and further compared the S:R ratio of Pi. Overexpression of At4 in gcn5 could partially rescue the phosphate translocation defect between shoots and roots (Fig. 5E, F). Taken together, these data provided genetic evidence that GCN5 regulated phosphate accumulation, at least partially, by controlling the expression of At4 in Arabidopsis.

Discussion

Histone acetyltransferase GCN5 is required for PSRs in Arabidopsis

Phosphate availability is a major factor limiting the growth, development, and productivity of plants (Rouached ; Yang and Finnegan, 2010; Chiou and Lin, 2011; Gu ; Secco ). Recently, histone ‘writers’ and ‘readers’ are reported to be involved in the PSR network (Smith ; Chandrika ; Chen ). Here, we reported that the histone acetyltransferase GCN5 is required for PSRs in Arabidopsis based on the following observations: (i) compared with the wild type, gcn5 mutants exhibited severe symptoms of Pi deficiency after low Pi treatment; (ii) the GCN5 transcript was up-regulated after Pi deficiency treatment; (iii) expression of 888 genes was significantly down-regulated due to the disruption of GCN5 under Pi deficiency condition; and (iv) four genes involved in phosphate homeostasis, including the well-studied At4 and WRKY6, were identified as direct targets of GCN5, and H3K9ac and/or H3K14ac levels of these four genes were decreased in gcn5 mutants.

To control gene expression accurately, histone acetyltransferases and deacetylases often cooperate with each other in gene regulation (Jenuwein and Allis, 2001). Previous studies showed that HDA19 and GCN5 function in some common targets but lead to the opposite effects (Benhamed ; Tanaka ). Chen reported that histone deacetylase HDA19 affects leaf responses to Pi starvation, and mutants and RNAi lines of HDA19 exhibited reduced anthocyanin levels compared with the wild-type control in low Pi conditions (Chen ). The regulation of HDA19 in Pi-deprived condition is opposite to that of GCN5 which we observed in this study, which may further confirm that histone acetylation mediated by GCN5 and HDA19 functions in Pi regulation.

GCN5-mediated epigenetic regulation of lncRNA At4 plays an important role in phosphate accumulation

LncRNA refers to a functional RNA molecule >200 nt that will not be translated into proteins. Growing numbers of lncRNAs are being reported to have regulatory roles in various developmental processes (Mercer ; Bazin and Bailey-Serres, 2015). To date, at least four divergent lncRNA-mediated regulation mechanisms have been unraveled, namely target mimicry, interference with transcription, the POLYCOMB REPRESSIVE COMPLEX2-associated histone methylation, and DNA methylation (Zhang ). In Arabidopsis, the non-protein-coding genes IPS1 and At4 are induced by phosphate starvation, and they could block the repressive role of miR399 on its target gene PHO2, which then regulates the dynamic balance of phosphate in shoots (Franco-Zorrilla ). However, the underlying mechanism for the up-regulation of the lncRNA At4 after phosphate deficiency remains elusive. In this study, we found that Pi deficiency-induced GCN5 was responsible for the robust response of At4 by enhancing its association with the At4 promoter and concomitantly its H3K14 acetylation level. Moreover, the direct binding of GCN5 to At4 was reported previously (Benhamed ).

Along with the decreased At4 expression in gcn5 mutants, miR399 accumulation was >2-fold higher than that in the wild type, and this effect was decreased after overexpression of At4 in the gcn5 background in Pi deficiency condition. Further experiments with pri-miR399 and MYB2 showed opposite expression behavior compared with mature miR399, which suggested the significance of At4 in the post-transcriptional regulation of miR399f. As expected, PHO2, the target of miR399, was down-regulated in gcn5 mutants, and the S:R ratio of Pi in gcn5 mutants was significantly increased compared with the wild type. In addition, overexpression of At4 in gcn5 could partially rescue these Pi impairments. Although the phosphate translocation could be elucidated by GCN5-regulated At4 expression, the Pi concentrations of shoots and roots in gcn5 mutants were both decreased under Pi-sufficient conditions, indicating our limited understanding of the complex role of GCN5 in Pi accumulation, which requires further investigation.

We propose that GCN5 is involved in PSR by modulating At4 expression. Remarkably, the link between GCN5 and At4 provides a new epigenetic mechanism for the regulation of lncRNA in plants. However, GCN5 controls histone acetylation of hundreds of loci (Benhamed ), and in this study we identified four targets of GCN5 involved in the phosphate deficiency response. Thus, to gain a comprehensive interpretation of the role of GCN5 in Pi homeostasis, it is necessary to study other GCN5 target genes in this process. In summary, we proposed a model of GCN5-mediated regulation of lncRNA At4 contributing to PSR in Arabidopsis (Fig. 6). Briefly, Pi deficiency-induced GCN5 positively regulates the expression of lncRNA At4 by modulating its histone acetylation. This epigenetic regulation of At4 could affect mature miR399 abundance, thus the functionally intact PHO2 mRNA level, and finally the efficiency of Pi allocation and Pi accumulation in plant.

Fig. 6.

A simple model for GCN5 regulation of PSR in Arabidopsis. Pi deficiency could induce the expression of GCN5, which facilitates the expression of the GCN5 direct target At4 by up-regulating its H3K14/K9 acetylation levels. Concomitantly, followed by the inhibition of miR399, the PHO2 mRNA level was increased, resulting in the impairment of Pi accumulation in plants. The two triangles represent the external Pi supply and the histone acetyltransferase GCN5. The schematic of lncRNA At4 (light gray) and its promoter region (dark gray) shows histone H3K14/K9ac modifications. Solid black arrows represent the GCN5-involved regulation of the PSR pathway. (This figure is available in color at JXB online.)

Supplementary data

Supplementary data are available at JXB online.

Table S1. Gene-specific primer pairs used in this study.

Table S2. The groups of differentially expressed genes from RNA-seq.

Table S3. The significantly enriched Gene Ontology groups.

Fig. S1. Observation of primary root length and lateral root number of Ws and gcn5 in Pi-replete and Pi-deprived conditions.

Fig. S2. Eight genes selected to examine the accuracy of RNA-seq using qRT-PCR.

Fig. S3. Examination of At4 and PHO2 expression and mature miR399 abundance.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.