Conditional effects of the epigenetic regulator JUMONJI 14 in Arabidopsis root growth

The nuclear DNA of eukaryotes is packaged as chromatin, the constitutive unit of which is the nucleosome, which is ∼146 base pairs of DNA wrapped around a core histone hetero-octamer (Kouzarides, 2007). Chromatin ‘openness’, and thus its accessibility for gene expression regulators, depends on covalent post-translational modifications at the protruding N-terminal tails of histones 3 (H3) and 4 (H4) (Kouzarides, 2007; Mosammaparast and Shi, 2010). These reversible modifications include the histone methylation state, which strongly influences gene expression levels and is controlled by specific methylases and demethylases (Kouzarides, 2007; Mosammaparast and Shi, 2010). Such methylation occurs at specific lysine (K) residues and the combinatorial methylation state determines the efficacy of transcription initiation, elongation and termination. For example, H3K9 and H3K27 methylation generally inhibit gene expression, whereas H3K4 and H3K36 methylation generally activate transcription. Moreover, there is a quantitative impact of methylation level, which ranges from mono- to trimethylation and is typically catalyzed by SET domain-containing proteins (Kouzarides, 2007; Mosammaparast and Shi, 2010). Trimethylated H3K4 (H3K4^me3) is most associated with transcribed genes, because it recruits activated RNA polymerase II to their 5′ end (Bannister and Kouzarides, 2005; Kouzarides, 2007). Lysine demethylation is catalyzed by various enzymes, which do not act on all methylation states (Bannister and Kouzarides, 2005; Kouzarides, 2007; Shi et al., 2004). Removal of methyl groups from trimethylated histone lysines is performed by proteins that contain a catalytic Jumonji C (JmjC) domain, which can also reduce other methylation states (Klose et al., 2006a,b; Tsukada et al., 2006). JmjC domain proteins can be sorted into distinct classes according to the presence of other conserved domains, which appear to determine specificity for a given histone methyl-lysine (Klose et al., 2006a).

The histone modification machinery is conserved in eukaryotes. For example, the genome of the model plant Arabidopsis thaliana encodes 21 JmjC domain-containing proteins, which can be grouped according to their overall structure and the presence of additional domains, features that likely confer specificity as well as promiscuity (Hong et al., 2009; Lu et al., 2008). For example, JMJ11 (also known as EARLY FLOWERING 6) and JMJ12 (also known as RELATIVE OF EARLY FLOWERING 6) contain additional JmjN and zinc-finger domains and target H3K27 and H3K9 (Lu et al., 2011; Yu et al., 2008), whereas the smaller JMJ30 and JMJ32 proteins contain no additional conserved domains and are specific for H3K27 (Gan et al., 2014; Lu et al., 2008). One group of proteins, comprising JMJ14, JMJ15, JMJ16 and JMJ18, is particular in that its members contain a phenylalanine- and tyrosine-rich, so-called FYR domain. The FYR domain is formed by the FYR-N and FYR-C subdomains, which constitute a single-folded functional module (García-Alai et al., 2010). Association of a JmjC domain with an FYR domain is apparently only found in plants (Klose et al., 2006a; Lu et al., 2008). Interestingly, the FYR domain is typical for TRITHORAX proteins, which are H3K4-specific methyltransferases (García-Alai et al., 2010; Klose et al., 2006a; Lu et al., 2008; Saleh et al., 2008). Matching this observation, H3K4-specificity has been demonstrated for JMJ14, JM15 and JMJ18 (Lu et al., 2010; Yang et al., 2012a,b).

In A. thaliana, histone demethylases have been mostly implicated in life-cycle transitions, notably flowering time (Gan et al., 2014; Greenberg et al., 2013; Jeong et al., 2009; Ko et al., 2010; Lu et al., 2010; Searle et al., 2010; Yang et al., 2012a,b). For example, JMJ14, JMJ15 and JMJ18 all affect flowering. However, genetic loss- and gain-of-function scenarios suggest functional diversification within this group despite the similarity of the encoded proteins. Whereas jmj14 loss-of-function mutants flower early (Greenberg et al., 2013; Jeong et al., 2009; Lu et al., 2010; Searle et al., 2010), jmj15 mutants have no flowering phenotype (Yang et al., 2012b) and jmj18 mutants flower slightly late (Yang et al., 2012a). Moreover, JMJ15 or JMJ18 ectopic over-expression triggers early flowering (Yang et al., 2012a,b), suggesting that JMJ14 and JMJ15/18 act antagonistically. JMJ14 has also been implicated in post-transcriptional gene silencing (Le Masson et al., 2012), as a downstream component in the cell-to-cell spread of a mobile RNA silencing signal (Searle et al., 2010) that is directly linked to H3K4 methylation (Greenberg et al., 2013). Here, we provide evidence for a role of JMJ14 in A. thaliana root growth.

The phloem of higher plants is an essential tissue that delivers phloem sap from source to sink organs via sieve tubes (Lucas et al., 2013). Sieve tubes are formed by interconnected, enucleated sieve elements that are metabolically supported by their neighboring companion cells. The growth regions of plants, the apical meristems, connect to the mature phloem network via protophloem. For example, in A. thaliana root tips, two protophloem strands transfer phloem sap into the meristem to support cellular growth and proliferation driven by the stem cell niche. Impaired protophloem differentiation results in discontinuous sieve tubes, and thus suboptimal phloem sap delivery. Consequently, root meristem size and root growth are strongly reduced, and this is accompanied by other secondary systemic defects (Anne and Hardtke, 2017). This phenotype is for instance observed in brevis radix (brx) or octopus (ops) mutants, in which protophloem sieve elements frequently fail to differentiate and appear as morphological ‘gaps’ (Rodriguez-Villalon et al., 2014). Gap frequency can be used as a proxy for phenotypic severity.

A forward genetic screen identified second site mutations that suppress the strongly reduced root growth of brx mutants (Depuydt et al., 2013). Besides loss-of-function mutations in BARELY ANY MERISTEM 3 (BAM3) [a receptor kinase that senses the secreted CLAVATA3/EMBRYO SURROUNDING REGION-RELATED 45 (CLE45) peptide], which fully suppress the brx phenotype (Depuydt et al., 2013; Hazak et al., 2017), a number of partial suppressors were isolated. For instance, loss of function in BIG BROTHER (BB) (an E3 ubiquitin ligase) partially rescues brx and also ops mutant phenotypes (Cattaneo and Hardtke, 2017). Whole-genome sequencing of bulked segregants (Kang and Hardtke, 2016) from another partial suppressor, line 2.91.1 (Depuydt et al., 2013) (Fig. 1A), suggested a point mutation in JMJ14 (At4g20400) as the causative locus. This second site mutation occurred at the splice junction between intron 10 and exon 11 (Fig. 1B). Sanger sequencing of cDNA fragments that encompassed exons 9 to 12 confirmed the annotated transcript in wild type, whereas intron 10 was retained in jmj14 mutant cDNA (Fig. 1C). This leads to the addition of two new amino acids after amino acid 780 and a subsequent premature stop codon at the end of the FYR-N subdomain, thus resulting in truncated JMJ14 protein that lacks the FYR-C subdomain. We named this allele jmj14-4, to distinguish it from previously published jmj14-1 and jmj14-2 T-DNA insertion alleles, and the jmj14-3 point mutant (Searle et al., 2010).

To test whether jmj14-4 represents a loss-of-function mutation, we crossed the jmj14-1 mutant, in which JMJ14 is disrupted by a T-DNA insertion in exon 6 (Searle et al., 2010), to brx. Similar to brx jmj14-4 double mutants, root growth was partially restored in brx jmj14-1 double mutants (Fig. 1A,D; Fig. S1A). Moreover, introduction of a transgene in which the JMJ14 genomic sequence was expressed under control of the JMJ14 promoter (JMJ14::gJMJ14) into brx jmj14-4 or brx jmj14-1 double mutants reverted their phenotype (Fig. 1E). These results prove that JMJ14 loss-of function partially suppresses brx phenotypes, and that the FYR-C subdomain is essential for JMJ14 function, likely by guiding JMJ14 to the correct H3K4 targets (Klose et al., 2006a).

Despite partial rescue of macroscopic brx phenotypes, protophloem continuity was not restored in brx jmj14 double mutants (Fig. 1F). Consistently, phloem-mediated translocation of carbofluorescin diacetate (CFDA) was not enhanced in brx jmj14 mutants compared with brx mutants (Fig. 1G). These results suggest that JMJ14 mutation suppresses the brx short root phenotype in a protophloem-independent manner, similar to what has been observed for BB mutation (Cattaneo and Hardtke, 2017). Partial rescue of brx by bb has been attributed to increased root meristem size, which is caused by enhanced meristematic cell proliferation, accompanied by mildly enhanced mature cell length. Rescue by jmj14 was largely similar, in that meristem size was restored to wild type in terms of meristematic cortex cell number (Fig. 1H) and mature cortex cell length was increased compared with brx single mutants (Fig. 1I). Yet unlike bb mutation, jmj14 mutation did not compensate for the slightly reduced formative divisions in brx (Rodriguez-Villalon et al., 2014) (Fig. 1J). Excess formative divisions that slow down root growth are thought to be responsible for the absence of enhanced root elongation in bb single mutants despite their bigger meristem (Cattaneo and Hardtke, 2017). By contrast, no significant difference in formative divisions, meristematic divisions or mature cell length were observed in jmj14 single mutants compared with Col-0 (Fig. 1H-J). Thus, jmj14 roots appeared entirely wild type, and a contribution of JMJ14 to root growth vigor was only evident in the brx background. To test whether this was a generic feature, we created jmj14-1 ops double mutants. Their phenotype was essentially similar to that of brx jmj14-1 double mutants (Fig. S1B-E), including secondary features such as lateral root density (Fig. 2A; Fig. S1F). We thus conclude that JMJ14 inhibits root meristem activity upon suboptimal root growth in tissue culture.

Matching protophloem-independent rescue by jmj14, physiological features of brx were largely retained in brx jmj14 double mutants. For instance, JMJ14 mutation did not act through dampening of the CLE45/BAM3 pathway hyperactivity in brx (Breda et al., 2019; Depuydt et al., 2013), as brx jmj14 double as well as jmj14 single mutants were fully CLE45 sensitive (Fig. 2B). Also, the reduced auxin activity in brx meristems (Gujas et al., 2012; Mouchel et al., 2006) was not markedly restored in brx jmj14 double mutants (Fig. 2C). Nevertheless, brx jmj14 root growth could not be further enhanced by brassinolide treatment, which (partially) rescues auxin response and root growth of brx single mutants (Gujas et al., 2012; Kang et al., 2017; Mouchel et al., 2006) (Fig. 2D). Rather, both jmj14 and brx jmj14 responded like wild type (Fig. 2D).

To determine whether JMJ14 could regulate known players in protophloem differentiation, we performed RNA sequencing (RNAseq) of jmj14 root tips. Applying a stringent cut-off (adjacent P-value<0.01), 1526 genes were found to be differentially expressed between jmj14 and Col-0 (Table S1). This set displayed substantial overlap (>10-fold over random expectation) with RNAseq data from bb root tips that were monitored in parallel (Fig. 2E) (Table S2). Primary brassinosteroid-related or CLE45-related genes were, however, not differentially expressed, consistent with our physiological assays. Interestingly, the overlap with genes that were described to be H3K4 hypermethylated in jmj14 seedlings (Ning et al., 2015) was rather limited (Fig. S2A). However, unlike the essentially random overlap between H3K4-hypermethylated genes in jmj14 seedlings and differentially expressed genes in bb (Fig. S2B), the corresponding overlap with genes that were upregulated in jmj14 root tips was nearly 4-fold over random expectation (Fig. S2A). These genes could represent primary JMJ14 targets in the root (Table S3). However, known dominant suppressors of brx, through enhanced and/or ectopic expression (Breda et al., 2017; Briggs et al., 2006; Rodriguez-Villalon et al., 2014), were absent from the differentially expressed genes. To monitor pertinent changes directly, we performed another RNAseq experiment with brx and brx jmj14 root tips. Again, many genes were differentially expressed between the two samples (n=1459) despite a stringent cut-off (adjacent P-value<0.01) (Table S4). Corroborating JMJ14 function, overlap between H3K4-hypermethylated genes in jmj14 seedlings (Ning et al., 2015) and the 910 differentially expressed genes that were upregulated in brx jmj14 double mutants was significant (Fig. 2F), whereas overlap with the 549 genes that were downregulated was random (Fig. 2G). Yet again, known suppressors of brx were found neither among these likely JMJ14 targets in the root (Table S5), nor among the wider set of differentially expressed genes. Finally, brx rescue by jmj14 could not be explained by ectopic expression of the functional BRX homolog BRXL1, which we monitored by reporter gene assay (Fig. S3). In summary, our results suggest that JMJ14 mutation partially rescues brx through pathways that are unrelated to those defined previously.

To determine whether JMJ14 is distinct from its homologs, as reported for flowering, we first monitored their expression by reporter constructs. NLS-3xVENUS fusion protein expressed under the control of different promoters indicated similar expression patterns and levels for JMJ15, JMJ16 and JMJ18 in the root vascular cylinder (Fig. 3A), similar to JMJ14 (Fig. 3B). Moreover, available jmj16 and jmj18 T-DNA insertion loss-of-function mutants also did not display apparent root phenotypes (Fig. 3C,D). Yet, in brx jmj16 and brx jmj18 double mutants derived from crosses, brx root growth was barely alleviated (Fig. 3C,D). Because none of the homologs was differentially expressed in jmj14 root tips, JMJ14 therefore does not appear to act redundantly with JMJ16 or JMJ18 in the root. The premature stop codon in the jmj14-4 allele pointed to a possible role of the FYR domain in diversification. To test this notion, we created transgenes that expressed the JMJ14 coding region up to the FYR-N domain in fusion with a brief linker and different FYR domains under control of the JMJ14 promoter (JMJ14::JMJ14^N-14^C, JMJ14::JMJ14^N-16^C and JMJ14::JMJ14^N-18^C) (Fig. 3E). Of these three constructs, only JMJ14::JMJ14^N-14^C was able to revert brx jmj14-1 back to a brx phenotype (Fig. 3F), suggesting that JMJ14::JMJ14^N-16^C and JMJ14::JMJ14^N-18^C could not substitute for JMJ14. Thus, the FYR domain likely has a role in the functional diversification of this group of proteins and confers specificity to the action of JMJ14.

One intriguing finding from the expression analysis was that the JMJ14 promoter is only active in already differentiating and mature stele cells, but not in the root meristem itself (Fig. 3B, Fig. 4A). This contrasted with the effects of JMJ14 mutation on brx root meristems, which could not be explained by ectopic JMJ14 expression in brx (Fig. 3B, Fig. 4B). Closer inspection indicated that the JMJ14 promoter is active throughout the root vasculature, but not in the phloem, nor in Col-0 or brx (Fig. 4C,D, Movies 1 and 2). However, when the NLS-3xVENUS reporter protein was replaced by a fusion between the JMJ14 coding sequence and CITRINE (JMJ14::JMJ14-CITRINE), the fusion protein was observed throughout the root (Fig. 4E,F). This pattern was not influenced by the presence or absence of BRX (Fig. 4F,G). Transgenic expression of JMJ14 coding sequence under control of the ubiquitous UBQ10 promoter did not trigger any discernible gain-of-function phenotypes in Col-0 (Fig. S4A) and reverted the brx jmj14 root phenotype (Fig. S4B), corroborating that ubiquitous JMJ14 expression is not detrimental and indicating that JMJ14 activity is well buffered, possibly by rate-limiting interacting factors.

Because of the strong nuclear localization of JMJ14-CITRINE and its ∼135 kDa size (∼108 kDa of JMJ14 plus ∼27 kDa of CITRINE), it seemed unlikely that the wider expression pattern was due to cell-to-cell protein movement. Indeed, in bleaching experiments, JMJ14-CITRINE expression was apparently recovered cell-autonomously, without any evidence for a gradient across the bleached area (Fig. 4H). Grafting experiments to determine whether JMJ14 mRNA could be mobile indicated that this is not the case, in line with the finding that JMJ14 was not among phloem-mobile transcripts identified in high-throughput approaches (Thieme et al., 2015; Yang et al., 2019). Collectively, our data therefore indicate that the transcript region of JMJ14 contributes substantially to its expression pattern.

In summary, our study identified a hitherto unrecognized role of JMJ14 in root development, which is evident in genetically sensitized backgrounds. Our data suggest that this role is unique for JMJ14 compared with its homologs, and that this specificity is conferred by its FYR domain. Moreover, we show that cistronic sequences contribute to the JMJ14 expression pattern, which might be of interest in light of previously reported JMJ14-dependent, supposedly non-cell-autonomous phenomena.

Plant tissue culture, plant transformation, common Molecular Biology procedures [such as genomic DNA isolation, genotyping, (whole genome) sequencing], peptide treatments, phloem probe translocation, GUS staining, microscopy and physiological assays were performed according to standard procedures as previously described (Breda et al., 2017; Cattaneo and Hardtke, 2017; Depuydt et al., 2013; Kang et al., 2017; Kang and Hardtke, 2016).

For plant tissue culture, seeds were surface-sterilized, germinated and grown vertically on half strength Murashige & Skoog agar media with or without 0.3% sucrose under continuous light of ∼120 µE intensity at 22°C. All mutants were in the A. thaliana Columbia-0 (Col-0) wild-type background and, except the newly isolated jmj14-4 allele, were described previously: brx-2, jmj14-1 (SALK_135712), jmj16-3 (SALK_029530), jmj18-1 (SALK_073422) and ops-2 (SALK_139316) (Liu et al., 2019; Lu et al., 2010; Rodrigues et al., 2009; Truernit et al., 2012; Yang et al., 2012b).

All DNA fragments were amplified from Col-0 genomic DNA or cDNA using suitable oligonucleotides (see Table S6). To monitor promoter activities, ∼2 kb fragments 5′ upstream of the JMJ14 (At4g20400), JMJ15 (At2g34880), JMJ16 (At1g08620) and JMJ18 (At1g30810) ATG start codons were cloned in front of an NLS-3xVENUS reporter in the pCAMBIA1305.1 plasmid. The genomic JMJ14 fragment (including the promoter) was cloned into pCAMBIA1305.1 as well. The GUS reporter construct for the JMJ14 promoter was created in pMDC163. The hybrid protein sequences expressed under control of the JMJ14 promoter were cloned into plasmid pH7m24GW. The JMJ14 coding sequence was fused to CITRINE and put under control of the JMJ14 promoter by Gateway cloning of the three fragments into the pH7m34GW plasmid. The UBQ10 promoter and JMJ14 coding sequences were recombined into the pH7m24GW plasmid by Gateway cloning. The sequences of all DNA fragments are provided in the Table S6.

Seven-day-old roots of Col-0 and jmj14-4 grown in parallel were harvested and frozen in liquid nitrogen before total RNA extraction was performed using a Qiagen RNeasy Plant kit. cDNA synthesis was performed as described previously (Depuydt et al., 2013). A ∼1 kb fragment spanning exons 9 to 12 was PCR-amplified using oligonucleotides 5′-CGAAGAAAGTGGATGGTTGTTTAG-3′ and 5′-CACAGAAGTCCATGCATCATTAC-3′. The resulting DNA fragments were separated by gel electrophoresis, gel-extracted and Sanger sequenced directly.

For RNAseq, 7-day-old roots of the different genotypes (two independent replicates per genotype) grown in parallel on vertical plates were harvested and frozen in liquid nitrogen before total RNA extraction was performed using a Qiagen RNeasy Plant kit. cDNA synthesis, amplification, size selection, and high-throughput sequencing were performed on an Illumina HiSeq 2500 instrument (Col-0, bb and jmj14) or HiSeq 4000 instrument (brx and brx jmj14) (single read sequencing) as described previously (Depuydt et al., 2013; Rodriguez-Villalon et al., 2015). Bioinformatic analysis of the high-throughput sequencing data was performed with the HTS pipeline (David et al., 2014) (Col-0, bb and jmj14) or kallisto v.0.43.1 (Bray et al., 2016) followed by DESeq2 (Love et al., 2014) (brx and brx jmj14). The RNA sequencing raw data have been deposited in the SRA under accession numbers PRJNA559150 and PRJNA580217.

To determine JMJ14 expression levels by qPCR, 5 mm root tips were collected from 7-day-old seedlings for total RNA extraction (Qiagen), and cDNAs were produced by reverse transcriptase (Invitrogen). q PCR was performed in triplicate (Table S7) using MESA Blue qPCR MasterMix (Eurogentec). EF1a was used as the reference gene (primers 5′-AAGGCTCAAACGCCATCAAAGTTTTAAGAA-3′ and 5′-AGGTCACCAAGGCTGCAGTGAAGAA-3′). JMJ14 transcripts were quantified using primers 5′-AGTGTGCTTGACCCAACAA-3′ and 5′-AGCCCATTTATACTCTCCAAAGG-3′.

Statistical analyses were performed in GraphPad Prism software version 8.2.0. Significance of differences was determined by one-way ANOVA with subsequent post-hoc Tukey test or Fisher's exact test. The analyses are provided together with the raw data in Table S7.

We would like to thank the Lausanne Genomic Technologies Facility for whole-genome sequencing and RNAseq services.