Knockdown of a JmjC domain-containing gene JMJ524 confers altered gibberellin responses by transcriptional regulation of GRAS protein lacking the DELLA domain genes in tomato

Highlight Jumonji-C (JmjC) domain-containing protein JMJ524 plays important roles in tomato growth and development through conferring altered gibberellin responses by transcriptional regulation of a GRAS protein lacking the DELLA domain gene SGLD1. These results broaden our understanding of the JmjC domain-containing proteins, which not only control genic DNA methylation but are also involved in the GA response.


Introduction
Phytohormone gibberellins (GAs) are fundamentally involved in various aspects of plant growth and development, including internode elongation (Hedden and Kamiya, 1997;Xiao et al., 2006).GA exists in nature in many different isoforms, but only a few of them can actively regulate plant growth.The bioactive isoforms of GA include GA 1 , GA 3 , GA 4 , and GA 7 , in which GA 12 and GA 53 are the main precursors converted to various GA intermediates (e.g.GA 9 and GA 20 ) and bioactive GAs (e.g.GA 1 and GA 4 ) by GA 20-oxidases (GA20ox) and GA 3-oxidases (GA3ox), respectively.GA3ox catalyses the final step in the synthesis of bioactive GAs, whereas GA 2-oxidase (GA2ox) plays a major role in converting active GAs (GA 1 and GA 4 ) and their precursors (GA 9 and GA 20 ) to inactive forms (Sun, 2011;Raddatz et al., 2013).Bioactive GAs can trigger a plant GA response in which DELLA proteins play a major role.With bioactive GAs, the GA receptor GID1 can bind GA in the nucleus and/or cytoplasm (Sun, 2010).If the binding occurs in the cytoplasm, the GID1-GA complex induces traffic to the nucleus, binding the nuclearlocalized DELLA proteins.Once the DELLA protein binds to the GID1-GA complex, ubiquitin E3 ligase complex SCF recognizes and ubiquitinates the former, which is then degraded by the 26S proteasome, subsequently triggering GA responses (Achard and Genschik, 2009;Harberd et al., 2009).
DELLA proteins belong to GRAS (GAI, RGA, and SCARECROW), a subfamily of the plant-specific regulatory protein family.These proteins have important roles in exceedingly diverse processes, such as signal transduction and meristem maintenance and development (Bolle, 2004).Other than the conserved N-and C-terminal domains of all GRAS family members, DELLA proteins also contain a unique DELLA domain that is essential in its own degradation (Dill et al., 2001).Arabidopsis contains five DELLA proteins: RGA, GAI, RGA-LIKE1 (RGL1), RGL2, and RGL3.These display overlapping and distinct functions in repressing GA responses (Sun, 2011).However, only a single DELLA protein, Procera (pro) and SLENDER1 (SLR1), has been identified in tomato and rice, respectively (Ikeda et al., 2001;Bassel et al., 2004;Bassel et al., 2008).Loss-of-function mutation in pro can result in tomato displaying a GA-constitutive response phenotype, in which the mutant is taller and has non-serrated leaves, which can be resemble the wild type (WT) plants if treated with GA 3 (Carrera et al., 2012).The loss-of-function mutant of the SLR1 gene also results in a constitutive GA response and exhibits rapid extension growth in seedling and sterile phenotypes.In particular, the truncation of DELLA in the SLR1 gene produces a dwarf phenotype in rice (Ikeda et al., 2001).As a repressor of GA signalling, the rice genome contains two SLR1-LIKE (SLRL) genes encoding proteins that have insufficient DELLA domain and function, whereas the overexpression of SLRL1 in transgenic rice inhibits shoot growth with a weaker effect compared with SLR1 (Itoh et al., 2005).
Jumonji-C (JmjC) proteins play important roles in plant growth and development (Chen et al., 2011b).The JmjC domain was first defined based on the amino acid similarities of Jumonji (Jmj) proteins identified in a gene-trap screen for neural tube formation factors in mouse (Takeuchi et al., 1995).These similarities were later suggested to regulate chromatin remodelling (Clissold and Ponting, 2001).JmjC domain-containing proteins were initially predicted to be involved in demethylation of modified arginine or lysine amine groups within histones (Trewick et al., 2005;Tsukada et al., 2006).This proposition was confirmed using a technique based on biochemical purification, which identified the JmjC domain-containing protein JHDM1, an H3K36-specific demethylase (Tsukada et al., 2006).Human Jmj-type enzymes are involved in various pathological processes, including development, cancer, inflammation, and metabolic diseases (Johansson et al., 2014).For example, Jarid2/Jumonji, a member of the Jmj factor family, regulates cardiovascular development (Mysliwiec et al., 2012).Lung cancer-associated JmjC domain protein MDIG is a common feature of non-small cell lung cancer (Lu et al., 2009).Inhibition of a JmjC domaincontaining histone demethylase 1B (JHDM1B) contributes to acute myeloid leukemia cell proliferation (Nakamura et al., 2013).Moreover, the H3K27-specific JMJs (KDM6 subfamily members JMJD3 and UTX) regulate disease-relevant inflammatory responses (Kruidenier et al., 2012).JMJD5 is a participant in both human and Arabidopsis circadian systems (Jones et al., 2010).Arabidopsis JMJ30 also standardizes the circadian clock and period length (Lu et al., 2011).
Many JmjC domain-containing proteins have recently been identified in plants (Lu et al., 2008;Zhou and Ma, 2008), but only a few of them have been functionally characterized.The first plant JmjC genes characterized were Early Flowering 6 (ELF6) and Relative of Early Flowering 6 (REF6) from Arabidopsis.ELF6 acts as floral repressor, whereas REF6 functions as floral activator (Noh et al., 2004).Subsequently, more functions of different JmjC domain-containing proteins from Arabidopsis and rice were distinguished.In Arabidopsis, the improved expression of 'increase in bonsai methylation 1' (IBM1), another JmjC domain-containing protein, represses genic DNA cytosine methylation, possibly via demethylation at H3K9 (Saze et al., 2008), and it has also been confirmed to protect the transcribed genes, but not the transposons, from DNA methylation at CHG sites (Miura et al., 2009).JMJ706, a rice homologue of IBM1, encodes an H3K9 demethylase and participates in the regulation of floral organ formation (Sun and Zhou, 2008).Meanwhile, rice JmjC domain-containing protein JMJ703 is a histone lysine demethylase that is required for stem elongation by regulating the expression of cytokinin oxidase genes (Chen et al., 2013).Posttranscriptional gene silencing has been reported in Arabidopsis due to mutations in the Jmjc domain protein H3K4me2/3 demethylase JMJ14 (Le Masson et al., 2012).
Using an oligonucleotide microarray, Gong et al. (2010) previously investigated drought stress and identified a JmjC domain-containing gene, JMJ524, from tomato with different expression profiles in drought-tolerant introgression lines (ILs) and their recurrent parent M82.In this study, JMJ524 was further characterized and shown to respond to circadian rhythms and GA treatment.The knockdown of JMJ524 caused a severe GA-insensitive dwarf phenotype.The results of RNA-seq transcriptome analysis indicated that two Procera homologues, namely SlGLD1 (GRAS protein Lacking the DELLA domain) and SlGLD2, were upregulated in JMJ524-RNAi transgenic plants.Transgenic tomato plants overexpressing SlGLD1 also exhibited a GA-insensitive dwarf phenotype.These findings demonstrate that in tomato, JMJ524 might be required for stem elongation through the regulation of SlGLD1 expression.

Plant materials and growth conditions
Tomato (Solanum pennellii LA0716 and S. lycopersicum cv.M82) plants were grown in a glasshouse with natural illumination (day/ night temperature 25/18°C; relative humidity ~55%).For gene expression analysis, six-week-old seedlings were treated with 100 μM GA 3 or a mock to detect the gene response to circadian rhythm and GA.The leaves were then collected at designated time points, immediately frozen in liquid nitrogen, and stored at -80°C until use.A total of 100 µM GA 3 solution containing 0.02% Tween 20 was sprayed on the 30-d-old seedlings with an interval of 3 d for 2 weeks to determine the sensitivity of dwarf phenotypes.

Gene isolation and tomato transformation
In previous studies on drought stress in tomato ILs, differential transcript expression profiles were observed in the drought-tolerant ILs and M82 for the JMJ524 gene by using an oligonucleotide microarray (Gong et al., 2010).The tomato JMJ524 coding sequence (CDS) was amplified using the polymerase chain reaction (PCR) from S. pennellii cDNA using gene-specific primers (GSPs: JMJ524-OE) (Supplementary Table S1) based on a unigene sequence (SGN-U576053, http://solgenomics.net/).The reverse transcription PCR (RT-PCR) product was recovered, cloned into pMD 18-T vector (TaKaRa), and sequenced (BGI, China).Multiple alignment of homologous sequences of JMJ524 from other species was conducted using ClustalW (http://www.ebi.ac.uk/clustalw/).A phylogenetic tree was constructed using the neighbour-joining method with MEGA (version 5.05) software.
The RNA interference vector was constructed by amplifying a 357 bp fragment from JMJ524 CDS using GSPs with 5′-attB1 and 5'-attB2 extensions on forward and reverse primers, respectively (JMJ524-Ri) (Supplementary Table S1; 5′-attB1 and 5′-attB2 extensions are underlined).A recombination reaction was performed between the PCR product and pHellsgate 2 vector (Invitrogen, USA) using BP clonase (Invitrogen, USA) according to the manufacturer's instructions.For SlGLD1 and SlGLD2 overexpression, the binary plasmid vector pMV2 (Yang et al., 2011), which carries spectinomycin resistance and neomycin phosphotransferase II genes for bacterial and transformed plant selections, respectively, was used.Meanwhile, the binary plasmid was formed by inserting SlGLD1 or SlGLD2 cDNA in a sense orientation into the vector after the cauliflower mosaic virus 35S promoter.All plasmids were transformed into the tomato cultivar M82 by Agrobacterium tumefaciens (strain C58) mediated transformation.After regenerated shoots were screened on selection medium containing kanamycin, the transgenic plants were further verified by PCR using genomic DNA as a template and the 35S promoter forward and gene-specific reverse primers.

RNA isolation and real-time RT-PCR
Gene expression patterns were examined by isolating total RNA using the TRIzol reagent (Invitrogen, USA).DNase I-treated RNA was reverse transcribed using an M-MLV Reverse Transcription enzyme (Invitrogen, USA).The resulting cDNA was used for real-time RT-PCR.Real-time RT-PCR was performed with the SYBR Green I Master kit (Roche, Switzerland) using primers (Supplementary Table S1) specific for genes, with β-actin (SGN-U580609) transcripts as an internal control.PCR amplification consisted of an initial incubation at 95°C for 5 min, followed by 40 cycles of 95°C for 10 s, 58°C for 15 s, and 72 °C for 20 s.Data were collected during extension and melting curve acquisitions; analyses were also performed in real time.PCR products were monitored with a LightCycler 480 (Roche, Switzerland) PCR system.

Transcriptional activation analysis in yeast cells
For transactivation assays, the CDS of JMJ524 was amplified with PCR using the generalized system of preferences (GSPs) (JMJ524-Y; Supplementary Table S1), which were designed to introduce SmaI and PstI restriction sites (underlined).The PCR product was fused in frame to the yeast GAL4 DNA-binding domain of the pGBKT7 vector (Clontech, USA) after its digestion with SmaI and PstI.pGBKT7-JMJ524 and pGBKT7 (negative control) were separately transformed into the yeast strain AH109 following the manufacturer's instructions (Clontech, USA).The transformed strains were streaked on SD/-Trp or SD/-Ade/-His/-Trp medium supplemented with X-α-Gal (20 mM) to assay another yeast reporter (MEL1) gene.The transactivation activity of each gene was evaluated according to their growth status.

RNA-seq and functional assignment
Young leaves from 50-d-old WT and JMJ524-RNAi (Ri-14) plants were harvested, and total RNA was isolated using TRIzol reagent (Invitrogen, USA).At least 10 µg of total RNA per sample was enriched for mRNA by conducting purification using oligo(dT) magnetic beads.cDNA synthesis was performed by random hexamer priming after the mRNA was fragmented into 200 bp fragments.Buffer, dNTPs, RNase H, and DNA polymerase I were added to synthesize the second strand.The double-stranded cDNA was purified with a QiaQuick PCR extraction kit and washed with EB buffer prior to end repair and single nucleotide A (adenine) addition.Finally, sequencing adaptors were ligated to the fragments, which were purified via agarose gel electrophoresis and enriched by PCR amplification.The products were submitted for single-end sequencing with a read length of 49 bp using an Illumina HiSeq TM 2000.Raw sequence data were then filtered to remove low-quality tags (tags with unknown nucleotide, 'N'), empty tags (no tag sequence between the adaptors), and tags with only one copy (which might result from sequencing errors).For tag annotation, clean tags were mapped to the tomato transcriptome reference database ITAG2.3 (http://solgenomics.net/itag/release/2.3/list_files), and no more than two base mismatches were allowed in the alignment.
Differences in gene expression were compared by statistically analyzing the tag frequency in each RNA-seq tag library according to previously described methods (Audic and Claverie, 1997).False discovery rate (FDR) was used to determine the value of threshold P in multiple tests.An FDR ≤0.001 and an absolute value of the log 2 ratio ≥1 were used as thresholds to determine significant differences in gene expression.The differentially expressed genes were used in KEGG pathway enrichment analyses.Tomato transcripts were annotated by performing a BLAST search against the nonredundant database at NCBI.The values of reads per kb of exon model per million mapped reads were used to evaluate the expressed value and to quantify transcript levels.

Paraffin sectioning
All tissues from 50-d-old JMJ524-RNAi and WT seedlings were harvested and fixed in formalin/acetic acid/alcohol solution for 24 h, stained with hematoxylin, dehydrated with an ethanol gradient series (15, 30, 50, and 70%), infiltrated with paraffin using chloroform as a solvent, and gradually embedded with paraffin.The specimens were sectioned (8-10 µm) using a Leica RM2245 (Germany) semimotorized rotary microtome and subsequently mounted on microscope slides.Paraffin was removed from the specimens by immersing the slides in xylene (twice for 20 min) and embedding them with neutral resin.The slides were then placed in an oven at 42°C until dried.Photomicrographs were taken using a Nikon ECLIPSE 80i microscope.

Quantification of endogenous GAs
Tomato leaves (1 g) were frozen in liquid nitrogen, finely ground, and then extracted with 15 ml methanol containing 20% water (v/v) at 4°C for 12 h.The following labelled GAs were added as internal standards before grinding: (2.00 ng g -1 ), and [ 2 H 2 ]GA 53 (4.00 ng g -1 ).Further sample preparation and analyses were performed as previously described (Chen et al., 2011a;Li et al., 2012).

Analysis of cytosine methylation by HPLC
The cetyltrimethylammonium bromide (CTAB) plant DNA extraction method (Murray and Thompson, 1980) was modified to isolate tomato DNA.A 50 µl volume of 70% perchloric acid was added to 100 µl DNA solution (containing ~25 µg DNA) and was hydrolysed at 95°C for 50 min.The pH was adjusted to a value of 3 to 5 with 1 M KOH.The sample was subsequently spun at 12 000g for 5 min; the resulting supernatant was used for high pressure liquid chromatography (HPLC) analysis.HPLC analysis was performed using a Waters 2695 separation module equipped with a Waters 2998 DAD detector and a ZORBAX-AQ C18 column (4.6 × 250 mm, 5 μm particle size: Agilent, USA).Sample was detected at 280 nm (UV) with an injection volume of 10 μl.The mobile phase comprised 0.1% aqueous formic acid as eluent A and methanol as eluent B with the following gradient: 5-30% eluent B over 30 min with a flow rate of 1 ml min -1 .Cytosine (C) and methylcytosine (MC) contents were assessed via co-migration under the same HPLC conditions with commercial standards (Sigma).MC percentages were calculated with the following formula: %MC = (MC/(C + MC)) × 100.

Isolation, transactivation activity assay, and expression pattern of JMJ524
In previous studies on drought stress in tomato ILs, JmjC524 (SGN-U218046) was identified as a differentially expressed gene in comparison with drought-tolerant ILs and M82 (Gong et al., 2010).Using RT-PCR, the full-length (1254 bp) open reading frame of JMJ524 from the wild tomato species S. pennellii LA0716 was isolated and cloned.In silico analysis illustrated that the JMJ524 genomic DNA sequence contains five introns.According to the SGN (http://solgenomics.net) genome annotation database, four other JmjC domain-containing proteins, namely Solyc03g112600.2.1, Solyc09g065690.2.1, Solyc10g081630.1.1,and Solyc08g075510.2.1, were identified in tomato，but these proteins share low similarity with JMJ524.Using NCBI pBLAST and JMJ524 amino acid sequence, the sequences of the most homologous proteins were retrieved from other species in the database.A CLUSTAL alignment of all sequences indicated high conservation among the JmjC domain-containing proteins (Fig. 1A).The phylogenetic tree constructed based on the amino acid sequences of JMJ524 and JmjC proteins from other representative organisms (Arabidopsis thaliana, human, and rice) showed that JMJ524 was significantly correlated with its counterpart Arabidopsis AtJMJ30 protein, whereas it was most distant from human HsKDM6A protein (Fig. 1B).
JmjC domain-containing proteins have been identified in numerous eukaryotic proteins, including PHD, C2H2, ARID/ BRIGHT, and zinc fingers, which contain domains typical of transcription factors (Clissold and Ponting, 2001;Elkins et al., 2003).Transactivation activity is a defining feature for a transcription factor.To identify whether JMJ524 functions as a transcriptional activator the yeast two-hybrid analysis was used.To this end, a GAL4 DNA-binding domain JMJ524 fusion protein was expressed in yeast cells that were subsequently assayed for their ability to activate transcription from the GAL4 sequence.Yeast cells carrying either control (pGBKT7) or fusion plasmid grew well on SD/-Trp medium, indicating that the analysis system was reliable.However, on the SD/-Ade/-His/-Trp medium, the cells transformed with the control plasmid could not grow, whereas those transformed with fusion plasmid grew normally (Fig. 2).Moreover, JMJ524 promoted yeast growth in the absence of histidine and adenine and exhibited X-α-gal activity.The vector control pGBKT7 did not achieve the same result (Fig. 2).These data confirmed that JMJ524 functions as a transcriptional activator in yeast and probably acts as a transcription factor in tomato.
The expression level of JMJ524 showed obvious tissue/ organ specificity in tomato's wild relative S. pennellii, and it was higher in the leaves but lower in stem tissue (Fig. 3A).Previous studies have shown that the expression of JMJ30 (Lu et al., 2011) and JMJD5 (Jones et al., 2010) exhibit circadian regulation.In order to make clear that their counterpart JMJ524 acts in a similar way, the expression profile of JMJ524 in response to circadian rhythms was assessed using real-time RT-PCR.The results showed that the expression of JMJ524 was quickly inhibited and then began to accumulate after 4 h, continued to increase until it reached the highest level at 12 h, >90-fold higher than the initial level, but returned to the 0 h level 24 h later (Fig. 3B).Meanwhile, JMJ524 expression was also evaluated in leaves from plants that had been treated with GA 3 to determine whether it is regulated by this growth regulator.In order to eliminate the influence of the circadian rhythm, the transcription levels of JMJ524 in GA 3 -treated and untreated plants at the same time point were compared.The result of the investigation revealed that the transcript levels of JMJ524 substantially increased at 8 h after GA treatment (Fig. 3C).These expression patterns indicate that JMJ524 was regulated by GA 3 and the circadian rhythm.

Suppression of tomato JMJ524 caused a GA-insensitive dwarf phenotype
JMJ524 was functionally characterized by generating transgenic tomato lines with overexpression or RNAi silencing of JMJ524.The expression level of the JMJ524 gene in transgenic (T 0 and T 1 ) and control plants was examined at the five-leaf stage by conducting real-time PCR analyses.Two significantly downregulated JMJ524 T 1 transgenic lines, Ri-14 and Ri-19 (Fig. 4A), were selected for further analysis.The transgenic and WT plants were kept under the same conditions, and plant height was measured 50 d after germination.The JMJ524-RNAi plants showed a severe dwarf phenotype (Supplementary Figure S1A), whereas no obvious change of morphology was observed in plants overexpressing JMJ524 (data not shown).The average height of the Ri-14 line was 5.7 cm, which was 73% less than that of the control at 21.1 cm.Moreover, the leaves of the transgenic plants were smaller (Supplementary Figures S1B, C).The dwarf phenotype was more severe toward the reproductive stage, with shortened stems and wizened leaves ( Figs 4B, C).Meanwhile, results of the histological analysis of the stem cross-section cell (Fig. 4D) revealed that cell size was smaller in the JMJ524-RNAi lines, which contain more cells in their stems (Fig. 4E), implying a reduction in cell elongation in the dwarf phenotype.In plants, the dwarf phenotype is probably caused by GA deficiency.Accordingly, endogenous GA content in JMJ524-RNAi transgenic lines was investigated.As shown in the schematic representation of GA biosynthesis (Fig. 5D), the endogenous levels of three GAs (GA53, GA20, and GA4) in JMJ524-RNAi plants were significantly higher than those in the WT.High concentration of GAs could enhance plant stem elongation and induce a tall plant phenotype.These adverse results signified that the dwarf phenotype of JMJ524-RNAi transgenic plants was due to other reasons, not GA deficiency.

JMJ524 regulates the expression of GA biosynthesis and response genes
The molecular mechanism of the JMJ524-mediated dwarf plant phenotype was further examined and the molecular role of JMJ524 was studied by performing a comparative analysis of the JMJ524-RNAi and WT leaf transcriptomes (RNA-seq).Most of the RNA-seq data parameters were comparable between the two samples (Supplementary Table S2).For the WT, up to 12 503 639 clean reads, which corresponded to a total length of 612 678 311 bp, were sequenced and accounted for 89% of the tomato transcriptome reference database ITAG2.3.Similar to the WT (Supplementary Table S2), up to 11 898 385 clean reads, which corresponded to a total length of 583 020 865 bp, were obtained from the JMJ524-RNAi transgenic plants, and these accounted for 88% of the tomato transcriptome reference database ITAG2.3.These results suggest that RNA-seq transcriptome covers most of the reference genome and most of the expressed genes were detected.
The putative differentially expressed genes comparing JMJ524-RNAi and WT plants were identified by applying FDR ≤0.001 and a log 2 ratio ≥1 as thresholds to determine significance levels.Using these criteria, 3026 genes were identified as being upregulated in the JMJ524-RNAi plant compared with the WT.Meanwhile, transcripts corresponding to 293 genes decreased by more than 2-fold in the JMJ524-RNAi plant (Supplementary File S1), revealing that a far greater number of genes are upregulated when JMJ524 expression is suppressed.The result of KEGG pathway enrichment analyses using differentially expressed genes  (1).Vectors pGBKT7 and pGBKT7-53+pGADT7-T were expressed in yeast as negative (2) and positive controls (3), respectively.Yeast streaks were cultured on SD/-Trp and SD/-Ade/-Leu/-Trp media containing X-α-GAL for assaying another yeast reporter LacZ.
indicated that the absence of JMJ524 expression altered GA metabolism because the expression levels of GA biosynthesis genes, including GA20oxs and GA3oxs, were significantly greater in the JMJ524-RNAi plants than in WT, as confirmed by real-time PCR (Fig. 5B).The upregulated expression of GA biosynthesis genes in RNAi lines was consistent with the increased endogenous GA levels in JMJ524-RNAi transgenic plants.
A high concentration of GAs could not explain the dwarf phenotype of JMJ524-RNAi transgenic lines; however, besides GA biosynthesis, the top differentially expressed genes also included several GA response-related genes (Table 1).Numbers indicate endogenous GA levels in WT (upper) and transgenic plants (lower, boxed).*, P < 0.05.Data (ng g -1 fresh weight) is presented as mean ± SE from three technical replicates using two transgenic lines.This figure is available in colour at JXB online.
Considering that the dwarf phenotype associated with GA insensitivity can also be obtained by blocking the GA-GID1-DELLA signalling pathway (Sun, 2011) and that silencing of SlDELLA in tomato can produce a taller plant with decreased expression of GA20ox and GA3ox (Marti et al., 2007), we suspected that the dwarf phenotype of JMJ524-RNAi might be associated with a GA response.Thus, a GA spraying assay was used to evaluate the GA sensitivity of the JMJ524-RNAi dwarf plants.After three weeks of GA spraying, the dwarf plants could not be rescued (Fig. 5A), indicating that the GA response was blocked and that the JMJ524-RNAi dwarf plants were GA insensitive.RNA-seq and real-time PCR results showed that the differentially expressed genes associated with the GA response did not contain the SlDELLA gene in the JMJ524-RNAi dwarf plants (Fig. 5B).However, the expression of two genes encoding GRAS protein Lacking the DELLA domain proteins (SlGLD1, Solyc10g086380.1.1;and SlGLD2, Solyc10g086370.1.1)was significantly upregulated (Fig. 5B).Previous research has confirmed that overexpression of SLRL1 in transgenic rice inhibits shoot growth (Itoh et al., 2005); thus, we considered that JMJ524 might regulate stem development and a GA response by regulating SlGLD1 or SlGLD2, but not SlDELLA.
Given that the stem cell number of JMJ524-RNAi lines was significantly larger than that of the WT (Fig. 4E), we also comprehensively compared the expression of cell cyclerelated genes in the WT and JMJ524-RNAi transgenic plants using the RNA-seq transcriptome data.Four cell cycle-related genes (Solyc10g074720.1.1,Solyc12g088530.1.1,Solyc05g051410.2.1, and Solyc02g092980.2.1) were significantly and differentially expressed in the WT and JMJ524-RNAi plants with an increased expression in the JMJ524-RNAi lines (Supplementary Table S3), as confirmed through real-time PCR (data not shown).These results showed that the abnormal expression patterns of the cell cycle-related genes might be related to the abnormal stem elongation of the JMJ524-RNAi transgenic plants.The expression of these cell cycle-related genes might be indirectly regulated by a GA response.

Overexpression of SlGLD1 in tomato caused a dwarf phenotype
SlGLD1 and SlGLD2 are present in tandem in the tomato genome (Fig. 6A).The full-length ORFs of SlGLD1 and SlGLD2 were 1509 and 1533 bp, respectively.SlGLD1 shared 66% identity with SlGLD2 and 50% identity with SlDELLA, whereas SlGLD2 and SlDELLA were 45% identical at the amino acid sequence level.The C-terminal conserved domains of the SlGLDs, such as VHIID, leucine heptad repeat II, PFYRE, and SAW, showed high similarity with SlDELLA (procera) protein.Interestingly, although the N-terminal regions of the SlGLDs contained the conserved domains of DELLA proteins, such as TVHYNP and Ser/ Thr/Val-rich (polyS/T/V), the DELLA domains were deleted (Fig. 6B).Phylogenetic analysis of the SlGLDs and other Data according to previous study by RNA-seq analysis in JMJ524-RNAi (Ri14) and WT.*, genes selected for qRT-PCR.
GRAS proteins in plants using the VHIID domain showed that both SlGLD1 and SlGLD2 are categorized under the DELLA group, suggesting that they may function in a GA response similar to the Procera protein.
To investigate whether SlGLD1 and SlGLD2 function in a GA response, transgenic plants with overexpressed SlGLD1 or SlGLD2 genes were generated.A total of 35 and 26 independent transgenic lines of SlGLD1 and SlGLD2 were obtained, respectively.No visible alteration in phenotype was observed in SlGLD2-overexpressed transgenic plants, but eight of the 35 SlGLD1-overexpressed transgenic plants showed an obvious dwarfism phenotype similar to the JMJ524-RNAi transgenic lines (Figs 7C, D).Three independent transgenic lines, namely OE3, OE18, and OE33, with significantly increased SlGLD1 expression and obvious an dwarfism phenotype, were selected for further study (Fig. 7A).Gene expression analysis indicated that GA biosynthesis-related genes (GA20oxs and GA3oxs) and the GA response-related gene (GID1) were significantly upregulated in SlGLD1-overexpressed transgenic lines (Fig. 7B).Their expression patterns were similar to those in the JMJ524-RNAi transgenic plants.These results implied that JMJ524 might partially affect stem elongation and a GA response by regulating the expression of SlGLD1.

JMJ524-RNAi plants have increased DNA methylation
It has been reported that JmjC domain-containing proteins play important roles in DNA demethylation (Saze et al., 2008), so the global DNA methylation level of JMJ524-RNAi transgenic plants was evaluated by HPLC.The overall cytosine methylation levels of tomato genomic DNA prepared from the young upper leaves of JMJ524-RNAi and WT plants were measured and the 5 m C content in JMJ524-RNAi genomic DNA was shown to be significantly higher than in the WT (Fig. 8).The average cytosine methylation levels of the Ri14 and Ri19 transgenic lines were 30.7% and 31.7%,respectively, which was significantly higher than that of the control at 19.9%, confirming that JMJ524 plays a role in regulating DNA demethylation.

JmjC domain-containing proteins play diverse functions in plants
Plant JmjC domain-containing proteins have important functions in epigenetic processes, gene expression, and plant development, as well as in the interplay between histone modification and DNA methylation.More than 20 JmjC domain-encoding genes have been found in Arabidopsis and rice, and these all showed higher conservation at the amino acid level than homologues from animals (Chen et al., 2011b).Similar to other JmjC domain-containing proteins, JMJ524 is highly conserved in different species (Fig. 1A), and a phylogenetic tree revealed that JMJ524 was closely related to JMJ30 and JMJD5 from Arabidopsis and human, respectively (Fig. 1B).The deduced protein sequence of JMJ524 shares 63% similarity with AtJMJ30 but only 35% similarity with HsJMJD5(data not shown).Consistent with its counterparts AtJMJ30, AtJMJD5, and HsJMJD5 (Jones and Harmer, 2011;Lu et al., 2011), JMJ524 is also involved in a circadian clock response in tomato (Fig. 3B).However, knockdown of JMJ524 causes a severe dwarfism phenotype in tomato, which is different from that of Atjmj30 and Atjmjd5 mutants (Jones and Harmer, 2011;Lu et al., 2011), but similar to that of the Arabidopsis ibm1 mutation (Saze et al., 2008).Although no substantial similarity was found between IBM1 and JMJ524 at the amino acid sequence level, we still investigated whether JMJ524 suppression would affect cell cycle regulators, in the same way as IBM1.However, unlike the ibm1 mutation, which resulted in the epigenetic silencing of a gene encoding a homologue of a cell cycle regulator, BNS (Saze et al., 2008), the JMJ524-RNAi lines showed increased expression of several cell cycle genes (Supplementary Table S3), consistent with the observation that the stem cell number of JMJ524-RNAi lines was significantly larger than that of the WT (Fig. 4E).In summary, these data suggest that JMJ524 and IBM1 might influence plant architecture through diverse mechanisms or speciesspecific differences.

JMJ524 functions in the regulation of a GA response
Limited growth has a number of advantages for plants, such as preventing damage in cereal crops, curbing unwanted vegetative growth, improving the ratio of vegetative growth to fruit production, or reducing the size of ornamentals (Rademacher, 2000).Plant growth is driven by cell proliferation and elongation (Beemster and Baskin, 1998), and it has been confirmed that GAs regulate Arabidopsis root growth by controlling cell elongation and promoting root  Pysh et al. (1999).(C) Phylogenetic analysis of SlGLD1, SlGLD2, and other GRAS proteins from other species (species and GenBank accession numbers are shown in Supplementary List S2) generated using the VHIID region.SlGLD1 is indicated by a black circle.The phylogenetic tree was constructed using the MEGA5 program with the neighbour-joining method using 1000 bootstrap replicates.This figure is available in colour at JXB online.
cell production (Ubeda-Tomas et al., 2009).Plant growth retardants (PGRs) are used in agriculture to alter plant morphology by reducing shoot growth via a lowered rate of cell division and a reduction in cell elongation.Most PGRs that are used in agriculture or horticulture act in an antagonistic manner to GA and the phytohormone auxin (Ubeda-Tomas et al., 2009).The JMJ524-RNAi plants showed a reduction in cell elongation (Fig. 4D) that correlates with abnormal GA metabolism.
The phytohormone GA not only has a prominent role in the regulation of various developmental processes (Dill et al., 2004), but is affected by the circadian clock, which gates GA signalling through transcriptional regulation of the GA receptors (Arana et al., 2011).Here, we report on the response of JMJ524 to the circadian rhythm.This was induced by GA 3 treatment but inhibited in the pro mutant (Figs 3B, C and 5C).Moreover, the GA content increased, and GA biosynthetic genes were upregulated, in the JMJ524-RNAi transgenic lines (Fig. 5B).These data may suggest that JMJ524 is involved in circadian oscillation of GA signalling.However, the function of JMJ524 in GA biosynthesis contradicts the dwarf phenotype in the JMJ524-RNAi transgenic lines.Previous studies have shown that accumulation of DELLA protein would increase the expression levels of GA20ox and GA3ox genes, whereas downregulation of such protein following GA treatment would reduce the expression levels of these genes (Silverstone et al., 2001;Itoh et al., 2002).Based on these data, we hypothesized that the silenced JMJ524 would upregulate the DELLA gene, leading to the dwarf phenotype.Upregulation of the DELLA gene would trigger improved expression of GA biosynthesis genes, such as GA20oxs and GA3oxs, through an unknown feedback mechanism.Moreover, the expression levels of GA2oxs do not show a consistent trend (some genes are upregulated and some are downregulated) (Table 1); how GA2oxs are regulated by DELLAs or JMJ524 is still unknown.DELLAs are nuclear transcriptional regulators that repress GA signalling and restrict plant growth (Harberd et al., 2009).The DELLA motif is essential for gibberellin-induced degradation of the DELLA protein (Dill et al., 2001) and for allowing GA responses to occur (Harberd et al., 2009).Deletions of the DELLA region in DELLA proteins would confer semi-dominant dwarf phenotypes in crops (Dill et al., 2001), thereby reducing plant tolerance to cold and salt stresses (Sun, 2011).In the current study, two putative DELLA proteins lacking DELLA domains, namely SlGLD1 and SlGLD2, were identified (Fig. 6A).SlGLD1 and SlGLD2 shared high similarities with the tomato DELLA protein Procera at the amino acid level (Bassel et al., 2004) (Fig. 6B).All these DELLA homologues lack introns.SlGLD1 overexpression in tomato resulted in a dwarf phenotype and induced the expression of GA biosynthesis and response (Fig. 7), suggesting that SlGLD1 acts as a repressor of GA signalling.Given that SlGLD1 does not contain the DELLA domain, which is essential for degradation of DELLA proteins, the resulting dwarfism of the tomato plants could not be recovered by exogenous GA 3 application.However, SlGLD2 overexpression showed no dwarf phenotype, which was consistent with the studies on two DELLA-like proteins from rice (SLRL1 and SLRL2) that have high sequence homology but lack DELLA domains.SLRL1 overexpression in rice also induced a dwarf phenotype, whereas SLRL2 was not involved in GA signalling (Itoh et al., 2005).SlGLD1 also induced a dwarf phenotype that was less severe than that of the JMJ524-RNAi lines.However, both plants displayed GA insensitivity, suggesting that SlGLD1 is not the only target gene of JMJ524 regulation throughout plant development.

Knockdown of JMJ524 confers altered GA responses via transcriptional regulation of SlGLD1 by influencing DNA methylation
Recent research had shown that the JmjC-domain protein IBM1 controls genic DNA methylation in Arabidopsis (Saze et al., 2008), that the ibm1 mutation resulted in dwarf phenotypes, and that a rice histone lysine demethylase JMJ703 was also required for stem elongation (Chen et al., 2013).These reports are consistent with our observation that JMJ524 is required for stem elongation (Fig. 4), and induces the global DNA methylation level when it is inhibited (Fig. 8).
JMJ524 suppression could confer altered gibberellin responses via transcriptional regulation of SlGLD1.However, how JMJ524 regulates SlGLD1 expression remains unknown.Knockout of JMJ524 would cause a significant global increase in DNA methylation levels and inactivated genes may directly or indirectly regulate the expression of SlGLD1 and other genes.JMJ524 might also affect histone modifications, and less condensed chromatin remodelling could result in induction of the expression of SlGLD1 and some other genes (Supplementary File S2).
At least 54 genes encoded a homologous amino acid sequence to the GRAS domain in the tomato genome (Jin et al., 2014).Only a few GRAS proteins have been characterized to date.However, recent studies showed that GRAS proteins have important roles in plant development (Bolle, 2004) and environmental signals (Ma et al., 2010).Nevertheless, the mechanisms of JMJ524 and GRAS proteins in integrating environmental signals (such as the circadian rhythm) to restrict plant internode elongation remains unclear.Further studies on the relationship of JMJ524 and GA-related GRAS proteins could provide a better understanding of how genes alter the GA response by integrating environmental signals or by DNA methylation to restrict plant development.

Fig. 1 .
Fig.1.Comparison and phylogenetic analysis of amino acid sequences of plant JMJ524 genes.(A)CLUSTAL alignment of deduced amino acid sequence of JMJ524 and those of Glycine max (GmjmjC: I1LJT0), Populus trichocarpa (PtjmjC: B9GM97), Ricinus communis (RcjmjC: B9RI82), Vitis vinifera (VvjmjC: F6GU98), Arabidopsis thaliana (AtjmjC: Q9LT40), and Brassica rapa (BrjmjC: G1FE06).Black and grey boxes indicate identical and similar amino acids, respectively.(B) Phylogenetic analysis of JMJ524 (indicated by a black circle) and JmjC proteins from other species (species and GenBank accession numbers are shown in Supplementary List S1), including proteins from human, Arabidopsis thaliana, and Oryza sativa indicated by Hs, At, and Os prefixes, respectively.The phylogenetic tree was generated by ClustalW2 using standard parameters of the neighbour-joining method in MEGA (version 5.05).

Fig. 3 .
Fig. 3. Analysis of JMJ524 expression in tomato.(A)Real-time RT-PCR analysis of JMJ524 expression in different tissues (R, root; S, stem; L, leaf; FL, flower; FR, fruit) of S. pennellii LA0716.(B)The JMJ524 transcript level under circadian rhythm in S. pennellii LA0716 plants.(C) Real-time RT-PCR analysis of JMJ524 expression in response to GA treatment.Expression of GA-treated S. pennellii LA0716 plants was compared with that in untreated plants after normalization of values with reference to the tomato β-actin gene and is presented as the relative expression level.All samples were collected at the time points indicated ('h' refers to hours after treatment) from three biological replicates of each treatment.Error bars indicate SE of three replicates.

Fig. 4 .
Fig. 4. Knockdown of JMJ524 causes a severe dwarfism phenotype.(A) Expression of JMJ524 in transgenic and WT plants was examined in the fifth leaf through real-time PCR analyses.Double asterisks (**, P < 0.01) denote statistically significant differences between transgenic and WT lines.(B) Phenotype of plants (about three months old) in JMJ524-RNAi (Ri) and WT lines with magnified view in the box (C).(D) Phenotype of transverse sections of stems collected from Ri14 (left) and WT (right) plants showing epidermal cells.Scale bar, 100 μm.(E) Comparison of stem epidermal cell lengths between WT and mutant plants.Single asterisk (*, P < 0.05) denotes statistically significant differences between transgenic and WT lines.This figure is available in colour at JXB online.

Fig. 5 .
Fig. 5. Increased GA level and expression of GA biosynthetic and responsive genes in JMJ524-RNAi lines.(A) Plant height is displayed as the average of eight plants.Error bars show SEs by spraying GA 3 at 3 d intervals starting from 30 d after inoculation.(B)Expression of tomato GA biosynthetic and responsive genes in JMJ524-RNAi lines (Ri14 and Ri19) and WT.(C) Expression of JMJ524 in wild type (WT) and procera (pro) mutant plants as revealed by real-time RT-PCR analysis.(D) GA biosynthesis pathway.GA 20-oxidase and GA 3-oxidase both have significantly upregulated expression (P < 0.05).Numbers indicate endogenous GA levels in WT (upper) and transgenic plants (lower, boxed).*, P < 0.05.Data (ng g -1 fresh weight) is presented as mean ± SE from three technical replicates using two transgenic lines.This figure is available in colour at JXB online.

Fig. 6 .
Fig. 6.Schematic representation and comparison of amino acid sequences of SlDELLA, SlGLD1, and SlGLD2, as well as phylogenetic analysis of SlGLDs in the GRAS family.(A)Schematic representation of SlGLD1 and SlGLD2 genes.Boxes represent exons; triangles and diamonds represent positions of transcription start and poly(A) sites, respectively.(B)Comparison of amino acid sequences of SlDELLA, SlGLD1, and SlGLD2.Black and grey boxes indicate identical and similar amino acids, respectively.Lines above the alignment indicate locations of conserved regions in GRAS proteins as defined byPysh et al. (1999).(C) Phylogenetic analysis of SlGLD1, SlGLD2, and other GRAS proteins from other species (species and GenBank accession numbers are shown in Supplementary List S2) generated using the VHIID region.SlGLD1 is indicated by a black circle.The phylogenetic tree was constructed using the MEGA5 program with the neighbour-joining method using 1000 bootstrap replicates.This figure is available in colour at JXB online.

Fig. 8 .
Fig. 8.Total percentage of methylcytosine in JMJ524-RNAi lines (Ri14and Ri19) and the WT.Data are presented as mean ± SD from three technical replicates using two independent genomic DNA sets prepared from 50-d-old seedlings.

Table 1 .
Genes involved in GA biosynthesis and signalling response