The Arabidopsis JMJ29 Protein Controls Circadian Oscillation through Diurnal Histone Demethylation at the CCA1 and PRR9 Loci

The Arabidopsis thaliana ecotype Col-0 was used. Plants were grown under neutral day conditions (NDs; 12 h light/12 h dark cycles) with cool white fluorescence (120 μmol photons m⁻² s⁻¹) at 22–23 °C. The jmj29-1, elf3-1, pJMJ29:JMJ29-GFP/jmj29-1, and pELF3:ELF3-MYC/elf3-1 mutants were previously described [22,23]. To investigate the biological function of JMJ29, T-DNA insertional jmj29-1 (Salk_021597) mutant seeds were obtained from the Arabidopsis Biological Research Center.

Total RNAs were isolated using TRI agent (TAKARA Bio, Singa, Japan). Two μg of total RNAs were pretreated with an RNAse-free DNAse and then used to synthesize first-strand cDNA using Moloney Murine Leukemia Virus (M-MLV) reverse transcriptase (Dr. protein, Seoul, South Korea) and oligo (dT18). Synthesized cDNAs were used for PCR amplification.

Quantitative RT-PCR reactions were conducted in 96-well blocks using the Step-One Plus Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The PCR primer sequences used in this study are listed in Supplementary Table S1. EUKARYOTIC TRANSLATION INITIATION FACTOR 4A1 (eIF4A) gene (At3g13920) was used as a reference gene for relative normalization. RT-qPCR was performed with SYBR Master Mix (Enzynomics, Seoul, South Korea) using the △△C_t method. The specificity of the RT-qPCR reactions was determined by melt curve analysis of the amplified products using the standard method installed in the system.

The BD Matchmaker system (Clontech, Mountain View, CA, USA) was employed to conduct Y2H assays. The pGADT7 vector was used for the GAL4 activation domain (AD) fusion, and the pGBKT7 vector was used for GAL4 binding domain (BD) fusion. PCR products were subcloned into the pGBKT7 and pGADT7 vectors. The recombinant expression constructs were co-transformed into yeast pJG69-4A cells harboring the LacZ and His reporter genes, and transformed cells were selected by incubating on SD/-Leu/-Trp medium and SD/-Leu/-Trp/-His/-Ade.

The full-size ELF3, ELF4, and LUX coding sequences were fused in-frame to the 5′ end of a gene sequence encoding the N-terminal half of EYFP in the pSATN-nEYFP-C1 vector (E3081). The JMJ29 gene was fused to the 3′ end of a gene sequence encoding the C-terminal half of EYFP in the pSATN-cEYFP-N1 vector (E3084).

For protoplast isolation, 14-day-old seedlings grown under the ND conditions were harvested in 20 mL 0.5 M mannitol solution (90Φ plate) and incubated for 1 h at room temperature (RT). Then, the 0.5 M mannitol solution was replaced with 20 mL enzyme solution (2% Viscozyme L, 1% Celluclast 1.5 L, 1% Pectinex Ultra SP-L in MMC, adjusted to pH 5.8 by NaOH and sterilized through a 0.2 μm syringe filtering) and incubated in the dark for 12–14 h at RT. The protoplasts were collected by centrifugation at 100 g for 7 min and washed twice with the W5 solution (0.1% glucose, 0.08% KCl, 0.9% NaCl, 1.84% CaCl₂, and 2 mM MES, adjusted to pH 5.7). For BiFC assays, 10 μg nYFP- and 10 μg cYFP-fusion plasmid DNAs were used to transfect 1.2 × 10⁶ protoplasts in 300 μL MMg solution (4 mM MES containing 0.4 M D-mannitol and 15 mM MgCl_2, adjusted to pH 5.7 with 2 M NaOH). The recombinant plasmids were co-transfected with a nuclear marker (mCherry-NLS plasmid). After 14 h incubation, the protoplasts were harvested and observed using Confocal Quantitative Image Cytometer CQ1 (YOKOGAWA).

35S:ELF3-HA and 35S:JMJ29-GFP constructs were co-transfected into protoplasts isolated from 14-day-old Arabidopsis seedlings, and the transfected protoplasts were incubated in the dark for 16 h at RT. The protoplasts were collected by centrifugation at 300 g for 5 min. About 2 × 10⁶ protoplasts were homogenized in IP buffer (25 mM Tris-HCl, 150 mM NaCl, 0.5 % Triton X-100, 1 mM EDTA, and 1 % protease inhibitor, adjusted to pH 7.5), incubated for 30 min on ice, and centrifuged at 15,000 g for 10 min at 4 °C to collect supernatant. Then, 50 μL of Protein A/G Agarose Beads (SC-2003, Santa Cruz Biotechnology, Santa Cruz, CA, USA) were added, and the mixture was incubated for 3 h at 4 °C with gentle shaking (50 rpm). The beads were removed by centrifugation at 14,000 g for 5 min at 4 °C, and antibodies were then added to the supernatant. After overnight incubation at 4 °C, 50 μL of protein A-Sepharose (P9424, Sigma-Aldrich, USA) was added and incubated for an additional 6 h at 4 °C. The beads were collected by centrifugation at 100 g for 3 min at 4 °C and washed five times with IP buffer. The proteins were eluted by boiling in SDS-PAGE sample buffer for 5 min and subjected to western blot analysis. The anti-HA (ab9110, Abcam, Cambridge, UK) and anti-GFP (ab290, Abcam, Cambridge, UK) were used for immunoprecipitation and western blot analyses.

The pJMJ29:JMJ29-GFP/jmj29-1, pELF3:ELF3-MYC/elf3-1, jmj29-1, and elf3-1 plants [22,23] were used for ChIP assays. Plant materials used for the experiments were grown under NDs for 2 weeks. Harvested plant materials were fixed in 1% formaldehyde for 20 min with vacuum infiltration and ground in liquid nitrogen. Chromatin solubilized by the nuclei lysis buffer (50 mM Tris-HCl pH 8.0, 10 mM EDTA pH 8.0, 1% SDS, 1 mM PMSF, and 1× PIs) was sonicated at 4 °C to generate ~500 bp fragments using a Bioruptor Pico (Diagenode). The chromatin solutions were ultrasonicated for 10 cycles (30 s ON and 30 s OFF for each cycle on full power). The antibodies, including anti-GFP (ab290, Abcam, Cambridge, UK), anti-MYC (05-724, Millipore, Billerica, USA), anti-H3 (04-928, Millipore, Billerica, MA, USA), anti-H3K4me3 (07-473, Millipore, Billerica, USA), and anti-H3K9me2 (ab1220, Abcam, Cambridge. UK), and agarose A/G beads (SC-2003, Santa Cruz Biotechnology, Santa Cruz, CA, USA) were used for ChIP. The fragmented DNAs were purified using a DNA elution kit. Precipitated DNA level was quantified by quantitative real-time PCR using specific primer sets listed in Supplementary Table S2. ChIP-qPCR values were normalized as a percentage of the input DNA.

For transient expression assays using Arabidopsis protoplasts, reporter and effector plasmids were employed [17,24,25]. The reporter plasmid contained a minimal 35S promoter sequence as well as the GUS gene [26]. The core elements on the CCA1 and PRR9 promoters were subcloned into the reporter plasmid. To construct the p35S:JMJ29, p35S:ELF3, p35S:ELF4, and p35S:LUX effector plasmids, coding regions were subcloned into the effector vector containing the CaMV 35S promoter. Recombinant reporter and effector plasmids were co-expressed into Arabidopsis protoplasts by a polyethylene glycol (PEG) method [27]. GUS activities were measured based on the fluorometric method. A CaMV 35S promoter-luciferase construct was also co-transformed as an internal control to normalize GUS activity. The luciferase assay was performed using the D-luciferin (BT11-1000Na, BioThema, Handen, Sweden).

Seeds were sown in 1/2 MS-solid medium and stratified in darkness at 4 °C for 3 days, and germinated seedlings were entrained for 3 weeks under ND conditions at 22–23 °C. To examine circadian oscillations in protoplasts, mesophyll protoplasts were isolated from the lower epidermal layer of 3-week-old Arabidopsis seedlings. Whole seedlings were soaked in 10 mL of an enzyme solution (400 mM mannitol, 20 mM KCl, 20 mM MES-KOH [pH 5.7], 10 mM CaCl₂, 1% Cellulase R10, 0.5% Macerozyme R10, and 0.1% bovine serum albumin) for 16 h in the dark. The isolated protoplasts were filtered through sterile 100 mm stainless mesh and resuspended in W5 solution. The pCCA1:LUC and p35S:ELF3 plasmids were prepared by PEG purification. Then, the recombinant constructs were transiently introduced into Arabidopsis protoplasts via PEG-mediated transformation [23].

Luminescence rhythms were monitored using the Tristar2 LB 942 Multimode Microplate Reader (Berthold Technologies, Wildbad, Germany). The circadian period was estimated using the Fast Fourier Transform-Nonlinear Least Squares (FFT-NLLS) suite of programs available in the Biodare2 software.

Accumulating evidence has supported that histone demethylase activity is important for the rhythmic oscillation of histone methylation at the core clock promoters [16]. Although several JMJ proteins have been characterized as crucial regulators of circadian oscillation [20,21,28], we wanted to identify additional catalytic enzymes responsible for the H3K4me3 dynamics. Among others, we noticed the JMJ29 gene, which encodes the JHDM2/KDM3 group Jumonji-C domain-containing histone demethylase [22,29], because JMJ29 expression displayed circadian rhythm with a peak at dusk (Figure 1A) [30].

We then obtained a JMJ29-deficient jmj29-1 mutant [22] and analyzed rhythmic expression of the core clock oscillator genes, CCA1 and PRR9 (Figure 1B). Quantitative real-time RT-PCR (RT-qPCR) analysis revealed that jmj29-1 mutation led to alterations in circadian oscillation. Circadian period of clock gene expression was shortened in jmj29-1 mutants, compared with wild type (Figure 1B). Consistently, expression of a circadian output gene, COLD, CIRCADIAN RHYTHM, AND RNA BINDING (CCR2), also displayed advanced rhythmic phase, possibly due to a period shortening (Supplementary Figure S1). These results suggest that JMJ29 is involved in sustaining circadian oscillation under free-running conditions.

We then asked whether JMJ29 regulates circadian clock activity through direct binding to the core clock genes. To this end, we performed chromatin immunoprecipitation (ChIP) assays using the pJMJ29:JMJ29-GFP/jmj29-1 transgenic plants [16,20,21]. ChIP enrichment analysis in the promoter regions of core clock genes, which include key clock-related cis-elements, such as E-boxes and LUX-binding sites (LBSs) [8,9] (Figure 2A), showed that JMJ29 was associated specifically at the CCA1 and PRR9 promoters (Figure 2B and Supplementary Figure S2), and none of the other examined regions were targeted by JMJ29 (Figure 2B). We further explored whether JMJ29 rhythmically bound to the CCA1 and PRR9 promoters. The JMJ29 protein accumulation was likely to be correlated with the oscillation pattern of JMJ29 transcripts (Figure 1A), and consistently, binding of JMJ29 to the CCA1 and PRR9 loci was observed preferentially at dusk (Figure 2B), when expression of the two morning genes was low [2,3,14].

JMJ29 that belongs to the JHDM2/KDM3 group is known to primarily catalyze H3K9 demethylation [22]. However, circadian expression of CCA1 and PRR9 is known to be independent of H3K9me1/2 levels [12]. Further, JMJ29 did not significantly influence H3K9me2 accumulation at the CCA1 and PRR9 loci (Supplementary Figure S3). Thus, we suspected that JMJ29 might contribute to changes in H3K4me3 accumulation locally at the CCA1 and PRR9 promoters. ChIP assays with an anti-H3K4me3 antibody showed that, in wild-type seedlings, reduction of H3K4me3 accumulation at the CCA1 and PRR9 loci occurred around dusk (Figure 2C) when JMJ29 was actively expressed (Figure 1A). However, the reduction of H3K4me3 levels at dusk was diminished in the jmj29-1 mutant (Figure 2C). Although histone modification at the CCA1 and PRR9 loci was altered in jmj29-1, their expression was not immediately changed (Figure 1B), possibly due to extensive transcriptional feedback loops. Thus, the JMJ29 protein promotes H3K4me3 demethylation at the CCA1 and PRR9 loci and subsequently influences their expression at later stages of free running.

Since H3K4me3 is responsible for gene activation [16,31], JMJ29 most likely inhibits the expression of CCA1 and PRR9. To test this hypothesis, we carried out transient expression assays using Arabidopsis protoplasts. The CCA1 and PRR9 promoters were fused to the 35S minimal promoter in the reporter plasmid. The recombinant reporter plasmid and the effector construct expressing the JMJ29 gene were co-transfected into Arabidopsis protoplasts (Figure 2D). Co-expression of these constructs repressed the GUS activity by ~50% (Figure 2D). Although regulation of CCA1 and PRR9 expression by JMJ29 was not immediately observed in wild-type seedlings (Figure 1B), overexpression of JMJ29 clearly led to the repression of CCA1 and PRR9 (Figure 2D). These results indicate that dusk-expressed JMJ29 is recruited to the CCA1 and PRR9 promoters and represses expression by promoting H3K4me3 demethylation.

Given that JMJ29 has no selective DNA-binding domain responsible for the recognition of specific cis-elements like other JMJ proteins [22], we hypothesized that additional molecular component(s) were required for JMJ29 regulation of CCA1 and PRR9 expression. To examine this hypothesis, we conducted yeast-two-hybrid assays and tested whether JMJ29 interacts with core clock proteins. The JMJ29-GAL4 DNA BD fusion construct was co-expressed with a construct expressing a core clock protein fused to GAL4 AD in yeast cells. As a result, JMJ29 specifically interacted with ELF3 (Figure 3A), but none of the other clock proteins were associated (Figure 3A).

To support the in vivo interaction of JMJ29 with ELF3, we conducted bimolecular fluorescence complementation (BiFC) analysis in Arabidopsis protoplasts. The JMJ29 gene was fused to the sequence encoding the C-terminal half of YFP (cYFP), and the ELF3 gene was fused to the sequence encoding the N-terminal half of YFP (nYFP). The fusion constructs were then transiently co-expressed in Arabidopsis protoplasts. Co-expression of the JMJ29-ELF3 combination generated yellow fluorescence exclusively in the nucleus (Figure 3B), while expression with empty vectors did not show visible fluorescence (Figure 3B). Consistent with the fact that ELF3 is a component of the EC [32,33], JMJ29 also physically associated with other EC components, ELF4 and LUX (Supplementary Figure S4). The discrepancy of interactions between JMJ29 and EC components in yeast and plant cells might be due to the requirement of plant-specific molecular factors in the protein-protein interactions. We also confirmed in planta interactions of JMJ29 and ELF3 (Figure 3C). These results indicated that JMJ29 interacts with the EC to directly repress CCA1 and PRR9.

Considering that the JMJ29 protein is involved in the temporal regulation of CCA1 and PRR9, we needed to confirm whether the EC function is also relevant in rhythmic H3K4me3 accumulation at the CCA1 and PRR9 promoters. We first employed pELF3:ELF3-MYC/elf3-1 transgenic plants [23] and performed ChIP assays using an anti-MYC antibody. In agreement with the temporal association of JMJ29 to the CCA1 and PRR9 loci (Figure 2B), ELF3 also bound to the CCA1 and PRR9 promoters in the regions where JMJ29 was recruited, preferentially at dusk (Figure 2A, 2B, and Figure 4A).

Since the EC recruits JMJ29 to repress gene expression via H3K4me3 demethylation, we analyzed H3K4me3 levels at the CCA1 and PRR9 loci in the elf3-1 mutant. ChIP-qPCR analysis for H3K4me3 accumulation revealed that the reduction of H3K4me3 levels at the CCA1 and PRR9 loci around dusk was impaired in elf3-1 (Figure 4B), similar to that in jmj29-1 (Figure 2C). We also examined expression of CCA1 and PRR9 in elf3-1 and found that their expression was arrhythmic and increased around dusk (Figure 4C), when JMJ29 and EC were active (Figure 1A) [33]. The expression amplitude of CCA1 and PRR9 in elf3-1 was a bit different from previous reports [8,34], probably because of our growth conditions.

The transcriptional repression of CCA1 and PRR9 by EC was also confirmed by transient expression assays (Supplementary Figure S5). The recombinant reporter plasmid containing the CCA1 or PRR9 promoter and the effector construct expressing ELF3, ELF4, or LUX were co-transfected into Arabidopsis protoplasts. GUS activity measurement showed that the EC components could bind to the CCA1 and PRR9 promoters and repressed expression (Supplementary Figure S5). These results indicate that the EC is responsible for the H3K4me3-dependent regulation of CCA1 and PRR9.

Our results demonstrated that the EC and JMJ29 act together in the repression of CCA1 and PRR9. We thus asked whether ELF3 repression of CCA1 and PRR9 required JMJ29. The importance of EC function for JMJ29 binding to the CCA1 and PRR9 promoters was tested using wild-type and elf3-1 mutant protoplasts transiently expressing 35S:JMJ29-GFP. ChIP analysis with the anti-GFP antibody showed that JMJ29 binding to the CCA1 and PRR9 loci was reduced in the elf3-1 mutant background compared with the wild-type background (Figure 5A). Since JMJ29 protein catalyzes H3K4me3 demethylation at the cognate regions (Figure 2C), H3K4me3 levels at the CCA1 and PRR9 loci were also measured in wild-type and jmj29-1 protoplasts transiently expressing 35S:ELF3-HA. H3K4me3 levels were reduced by overexpression of ELF3 in wild-type protoplasts, but ELF3-induced H3K4me3 demethylation was compromised in the jmj29-1 mutant protoplasts (Figure 5B).

We also performed transient gene expression assays using Arabidopsis protoplasts. Reporter GUS activity measurement indicated that ELF3 significantly repressed CCA1 and PRR9 promoter activities in wild-type protoplasts, but the ELF3 function disappeared in the jmj29-1 mutant (Figure 5C). The impaired ELF3 function was restored by complementation of JMJ29 (Figure 5C). In further support, overexpression of ELF3 led to prolonged circadian period in wild-type protoplasts, but the ELF3 function was impaired in jmj29-1 mutant protoplasts (Figure 5D). These results demonstrate that both EC and JMJ29 are simultaneously involved in repressing CCA1 and PRR9 expression in circadian cycles.

In summary, the CCA1 and PRR9 genes are regulated by rhythmic changes in H3K4me3 levels. SDG2 is possibly responsible for the raising phase of their expression [20]. After peak expression, JMJ29 is recruited via the interaction with ELF3, a component of the EC. The EC-JMJ29 induces H3K4me3 demethylation at the CCA1 and PRR9 loci to reach their basal expression during evening time (Figure 6).