Jumonji domain protein JMJD5 functions in both the plant and human circadian systems

TOC1 and JMJD5 Are Coexpressed.

Despite recent advances (12), the mode of action of the central plant clock gene TOC1 is currently enigmatic. Coexpressed genes frequently share biological function (14), prompting us to identify genes that are developmentally coregulated with TOC1. We therefore obtained microarray expression data generated from 79 different Arabidopsis samples (representing a wide variety of developmental stages and diverse organs) (20) and calculated the Pearson correlation coefficients between TOC1 and each gene on the ATH1 array. The gene most highly correlated with TOC1 (r = 0.77) was EARLY FLOWERING 3 (ELF3; Table S1), a gene involved both in light input to the clock and in central clock function (21–23), thereby validating our approach. Because genes involved in the circadian clock are themselves frequently expressed with a circadian oscillation (3, 13), we examined the daily patterns of expression of our candidate genes using previously published microarray data (2). Of the 10 genes most highly correlated with TOC1, only ELF3 and a previously uncharacterized gene, JMJD5 (At3g20810), were expressed with a circadian rhythm (Fig. 1A). The expression patterns of TOC1 and JMJD5 were the third most highly correlated across all genes on the microarray (r = 0.75; Fig. 1B and Table S1). Empirical resampling of randomly permuted data shows that the chance for this level of correlation occurring between two randomly selected genes is only 1.3% (2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

JMJD5 is coexpressed with TOC1 over developmental and diurnal timescales. (A) Daily expression patterns of the 10 genes most highly developmentally correlated with TOC1. TOC1, JMJD5, and ELF3 are shown in black, with the remaining genes shown in gray. Accession numbers for each of these genes are listed in Table S1. Expression data were obtained from the work of Covington and Harmer (51). (B) Scatter plot of JMJD5 compared with TOC1 mRNA expression levels across a range of developmental time points. Each symbol represents the mean of each different biological sample; error bars represent SEM. Data were obtained from the work of Schmid et al. (20). (C) qRT-PCR analysis of JMJD5 expression under constant red light. WT Columbia (Col), jmjd5-1, jmjd5-2, and toc1-2 seedlings were compared using oligos spanning intron 1 (Fig. S1A and Table S4). Plants were entrained to 12/12-h light/dark cycles for 6 d before being moved to constant conditions with 40 μmol·m⁻²·s⁻¹ red light. JMJD5 mRNA levels were normalized to PP2a. Data are the mean of three technical replicates; SE is shown. Presented results are representative of three independent experiments.

We next investigated whether JMJD5 expression was regulated by known clock genes. In toc1-2 mutants, JMJD5 expression levels gravitated toward a median level relative to the peak of expression in WT plants (Fig. 1C). In contrast, JMJD5 mRNA was not detectable by quantitative RT-PCR (qRT-PCR) in seedlings overexpressing CCA1 (Fig. S1D). This repression was similar to but more pronounced than the reduction of TOC1 expression previously reported in these CCA1-OX plants (9) (Fig. S1E). Like TOC1, the JMJD5 promoter contains an evening element sequence immediately upstream of the transcriptional start site (TSS) in addition to two within the 5′ UTR (Fig. S1F), suggesting that CCA1 might bind to this region and directly repress JMJD5 expression. These data demonstrate that not only are JMJD5 and TOC1 coregulated across developmental and circadian time but they are both repressed by the morning-phased clock protein CCA1.

jmjd5, Like toc1, Is Conditionally Arrhythmic.

To determine whether JMJD5 is simply an output of the clock or if it is involved in the clock mechanism, we introduced the CCR2::LUC luciferase reporter construct into two lines with T-DNA insertions within the At3g20810 locus (jmjd5-1 and jmjd5-2; Fig. S1A). qRT-PCR using oligos spanning the first exon–exon junction revealed no detectable full-length mRNA in either mutant but a low level of a truncated form of the message in jmjd5-1 (Fig. 1C and Fig. S1B). Because the conserved jmjC domain is located at the C terminus of the protein (Fig. S1A), both alleles are likely loss-of-function mutations. Consistent with this possibility, circadian rhythms in luciferase activity in both mutants were ≈0.5–1.0 h shorter than in WT (Fig. 2 A and B; P < 0.05, Student's t test). This phenotype was highly reproducible and was observed when plants were grown in a variety of light conditions (constant red light, constant blue light, or constant red plus blue light; Table S2). Neither insertion allele displayed an altered rhythm under constant darkness, indicating that JMJD5 function is light-dependent (Table S2). Similar light-dependent circadian phenotypes have been reported for other clock mutations (24, 25).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

jmjd5 mutants have circadian phenotypes. Luciferase activity of CCR2::LUC plants monitored in continuous red light. (A) WT Columbia (Col), jmjd5-1, and jmjd5-2 seedlings were entrained to 12/12-h light/dark cycles for 6 d before being moved to constant conditions with 30 μmol·m⁻²·s⁻¹ monochromatic red light. The jmjd5-2 alleles displayed reduced luciferase activity despite expressing normal levels of CCR2 mRNA (Fig. S1C). Similar phenomena have previously been described in other T-DNA lines (49). Error bars indicate SEM and are displayed every 20 h for clarity (n ≥ 20). (B) Scatter plot of data shown in A comparing circadian period with relative amplitude error, a measure of rhythmic robustness, with a perfect cosine wave having a value of 0 as calculated by Fourier fast transform-nonlinear least squares (50). Data points are presented for WT Col, jmjd5-1, and jmjd5-2. Plants were treated as described in A. (C) Bioluminescence of WT, jmjd5-1, and toc1-2 seedlings maintained in higher intensity red light. Plants were entrained to 12/12-h light/dark cycles for 6 d before being moved to constant conditions with 120 μmol·m⁻²·s⁻¹ red light. Data are representative of two independent experiments. Error bars are plotted every 20 h and indicate SEM (n ≥ 11).

Loss-of-function toc1-2 mutants have a strong short-period phenotype in most conditions but become arrhythmic in constant red light (26). We therefore examined clock regulation of luciferase activity in jmjd5 mutants maintained in different fluence rates of constant red light. When held under low levels of red light (∼30 μmol·m⁻²·s⁻¹), jmjd5 mutants had a short period but preserved robust rhythms (Fig. 2 A and B). Intriguingly, at higher fluence rates of monochromatic red light (120 μmol·m⁻²·s⁻¹), jmjd5 seedlings became arrhythmic or had greatly dampened oscillations compared with WT plants, which had robust rhythms under these conditions (Fig. 2C and Fig. S2A). The jmjd5 mutants continued to cycle on days 4 and 5 of free-run under a high fluence rate of white light (Fig. S2B), suggesting that this phenotype was specific to red light. Thus, the jmjd5 and toc1 mutant phenotypes are similar: Both have short free-running periods in constant blue or white light but become arrhythmic in constant red light.

JMJD5 Is a Positive Regulator of CCA1 and LHY Expression.

Although it is known that TOC1 promotes CCA1 and LHY expression, the underlying mechanism is not yet fully understood (9, 12). TOC1 and JMJD5 are coexpressed and have comparable mutant phenotypes; thus, we speculated that JMJD5, like TOC1, might play a role in the regulation of CCA1 and LHY expression. To investigate this possibility, we used qRT-PCR to examine the expression of core circadian genes in toc1-2 and jmjd5 mutants. Because both the toc1 and jmjd5 phenotypes were most severe in monochromatic red light (Fig. 2C and Fig. S2A), we chose to examine clock gene expression in plants grown in high-intensity constant red light, collecting samples over a full 2 d. Consistent with the luciferase data, qRT-PCR analysis revealed that expression of the evening-phased genes GIGANTEA (GI) and TOC1 was disrupted, with a decrease in rhythmic amplitude (Fig. 3 A and B).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

qRT-PCR analysis of circadian clock-regulated genes. WT Columbia (Col), jmjd5-1, and toc1-2 mutants were compared by qRT-PCR. Levels of TOC1 (A), GI (B), CCA1 (C), and LHY (D) mRNA were assessed. Plants were entrained to 12/12-h light/dark cycles for 6 d before being moved to constant conditions with 120 μmol·m⁻²·s⁻¹ red light. mRNA levels for each gene were normalized to PP2a. Data are the mean of three technical replicates; SE is shown. Results are representative of three independent replicates.

We next examined expression of CCA1 and LHY, factors known to repress GI and TOC1 expression (9, 27). In jmjd5 mutants, peak levels of CCA1 and LHY mRNA were reduced relative to WT (Fig. 3 C and D and Fig. S3), similar to the reduced expression seen in toc1-2 plants (Fig. 3 C and D). These data indicate that like TOC1, JMJD5 promotes expression of the morning-phased clock genes CCA1 and LHY.

JMJD5 and TOC1 Interact Genetically.

Given that TOC1 and JMJD5 are coregulated and have related mutant phenotypes, we speculated they might act at a similar step in the clock mechanism and would show genetic interactions. We therefore generated plants mutant for both genes and assessed circadian function in multiple light conditions. When monitored in constant red plus blue light, jmjd5-1 toc1-2 double mutants had a free-running period ≈1.5 h shorter than toc1-2 single mutants; in comparison, jmjd5-1 single mutants had a period only 0.5 h shorter than WT controls (Fig. 4A). Therefore the combination of jmjd5-1 and toc1-2 mutations had a synergistic effect on period length. When maintained under constant red light of a moderate fluence rate (30 μmol·m⁻²·s⁻¹), jmjd5-1 toc1-2 mutants also showed a synergistic effect on rhythmicity. Under this condition, jmjd5-1 single mutants were robustly rhythmic, whereas ≈70% of toc1-2 mutants had a detectable rhythm in luciferase activity (Fig. 4C). In the double mutants, only 30% of plants had a significant circadian rhythm. Thus, TOC1 and JMJD5 showed synergistic genetic interactions for multiple clock phenotypes. In contrast, only additive effects on period and rhythmicity were observed in jmjd5-1 lhy-100 double mutants, with the double mutants having a free-running period 0.5 h shorter than the lhy-100 single mutants (Fig. 4 B and D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

JMJD5 and TOC1 interact genetically. (A) Period estimates of WT Columbia (Col), jmjd5-1, toc1-2, and jmjd5-1 toc1-2 genotypes. Plants were entrained to 12/12-h light/dark cycles for 6 d before being held in continuous 35 μmol·m⁻²·s⁻¹ red and blue light. Each genotype had a significantly different period length (a–d; P < 0.001, Tukey range test). Error bars represent SE (n ≥ 11). Presented data are representative of three independent experiments. (B) Period estimates of WT Col, jmjd5-1, lhy-100, and jmjd5-1 lhy-100 genotypes. Seedlings were treated and analyzed as described in A. Error bars represent SE (n ≥ 30). Presented data are representative of three independent experiments. (C) Rhythmicity of jmjd5-1 toc1-2 seedlings held under 30 μmol·m⁻²·s⁻¹ red light. Percent rhythmicity was defined as the fraction of seedlings that returned a period estimate with a relative amplitude error <1 as determined by Fourier fast transform-nonlinear least squares (50). Seedlings and entrainment conditions were as described in A. Populations of toc1-2 and jmjd5-1 toc1-2 double-mutant seedlings had significantly worse rhythms than WT or toc1-2 populations, respectively (*P < 0.05, Student's t test). Error bars indicate SE from three independent experiments. (D) Rhythmicity of jmjd5-1 lhy-100 seedlings held under 30 μmol·m⁻²·s⁻¹ red light. Seedlings and entrainment conditions were as described in C. Error bars indicate SE from three independent experiments.

Human Cells Deficient for JMJD5 also Have a Short-Period Phenotype.

Although the core circadian transcription factors are not shared across higher taxa, some proteins (such as casein kinase and cryptochrome) and protein modifications (such as histone acetylation and poly-ADP ribosylation) do play a role in the circadian clocks of diverse eukaryotes (11, 28, 29 and reviewed in 30, 31). JMJD5 orthologs can be found in many eukaryotic genomes (Fig. S4A and Table S3). For example, the human JMJD5 protein (also known as KDM8) (16) is highly related to the Arabidopsis protein, with 39% identity and 56% similarity at the amino acid level and similar predicted secondary and tertiary protein structures (Fig. S4 B and C). Amino acid similarity rises to 68% within the conserved jmjC domain. Given the similarities between the human and Arabidopsis JMJD5 genes, we investigated whether the human ortholog might also be involved in clock function. We designed an siRNA construct against the human JMJD5 gene and transfected this into U2OS cells expressing luciferase under the control of the BMAL1 promoter (U2OS-B6) (32). Transformation with the siRNA construct specifically knocked down endogenous levels of JMJD5 to 20% of WT (Fig. S5 A and B). Transformed cells had a significantly shorter period than U2OS-B6 cells transfected with a scrambled siRNA control (Fig. 5 A and B; P = 0.022, Student's t test) but continued to show robust rhythmicity (Fig. 5A). This period shortening was strikingly similar to the circadian phenotype observed for Arabidopsis jmjd5 mutants in most light conditions (Table S2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Arabidopsis and human JMJD5 have conserved functions. (A) Circadian rhythmicity of U2OS-B6 cells stably expressing a Bmal1 promoter-driven luciferase reporter construct. Average luciferase activity of cells transfected with either scrambled siRNA or JMJD5 siRNA was assayed following synchronization with a serum shock as described by Vollmers et al. (32) (n = 5). Data are representative of three independent experiments. (B) Circadian period of human U2OS-B6 cells stably transformed with the plant JMJD5 ortholog. U2OS-B6 cells expressing the plant JMJD5 ortholog (U2OS + AtJMJD5) had a significantly longer period than WT U2OS-B6 cells when transfected with a human JMJD5 siRNA (**P = 0.0026, Student's t test). Error bars indicate SEM (n = 5). Cells were treated as described in A. Data are representative of two independent experiments. (C) Circadian period of jmjd5-1 plants transformed with either the plant or human JMJD5 gene. Plants were entrained to 12/12-h light/dark cycles for 6 d before being moved to continuous 40 μmol·m⁻²·s⁻¹ red light. T1 plants transformed with either the plant or human JMJD5 gene had a significantly longer period than jmjd5-1 mutants (*P < 0.05, mixed-effect linear model). Data are pooled from three independent experiments. Error bars show SEM (n ≥ 20).

JMJD5 Homologs Act in Both Plant and Human Circadian Systems.

Because loss of either ortholog caused similar effects on clock function in plants and human cells, we speculated that the proteins might share a conserved cellular function. To test this possibility, we created a cell line (U2OS + AtJMJD5) that stably expressed the Arabidopsis gene in a U2OS-B6 background and monitored its circadian periodicity following knockdown of the endogenous human gene (Fig. 5B). Remarkably, U2OS + AtJMJD5 cells transfected with a human JMJD5 siRNA had a significantly longer period than U2OS-B6 cells (Fig. 5B; P = 0.0026, Student's t test), indicating that the plant protein is able to function within the human circadian system. This rescue is highly significant but not complete, because a modest decrease in period length was observed in U2OS + AtJMJD5 cells when transfected with JMJD5 siRNA compared with a scrambled siRNA control (Fig. 5B).

We also investigated whether the human ortholog was functional in planta. To test this hypothesis we introduced either the Arabidopsis or the human JMJD5 gene, expressed under the control of the plant JMJD5 promoter (Fig. S4D), into the Arabidopsis jmjd5-1 mutant background and tested for rescue of the period defect. The circadian periods of T1 seedlings transformed with either the human or plant gene were then compared with those of WT and jmjd5-1 plants (Fig. 5C and Fig. S4E). Intriguingly, seedlings transformed with either JMJD5 ortholog displayed a significantly lengthened circadian period compared with the untransformed parental control (P = 0.0013 and 0.0210 for the Arabidopsis and human genes, respectively). Use of T1 plants permitted characterization of at least 20 independent insertion events for each transgenic population but necessitated the use of hygromycin selection. Such treatment is unlikely to alter circadian period, however, because hygromycin-resistant homozygous T3 transgenics demonstrated no difference in period length in the presence or absence of hygromycin (Fig. S4F). In combination, these cross-kingdom rescue experiments demonstrate that JMJD5 functions in both the Arabidopsis and human circadian clocks via a conserved molecular mechanism.