All Arabidopsis thaliana seed stocks used in this study were in the FRI^sf–2 (Lee et al., 1993) background except pEstro:JMJ30 (Yamaguchi et al., 2021), which generated a Columbia (Col-0) background. jmj30 jmj32 (Gan et al., 2014), flc-3 (Michaels and Amasino, 1999), and the reporter line FLC-GUS (Noh and Amasino, 2003; Michaels et al., 2005) were reported previously. To generate multiple mutants and mutants harboring the reporters, we performed crossings and genotyping. Arabidopsis seeds were grown on 0.5% gellan gum or 1% agar with Murashige and Skoog (MS). The plates were cultivated under constant light conditions. To examine the flowering phenotypes, the plants were cultivated in pots containing vermiculite and Metro-Mix (Sun Gro Horticulture).

We vernalized seeds 1 month after water absorption and sowed them on a plate. After vernalization, we transferred the plates to an incubator at 30° in the dark, cultivated them for 1 week, and then transferred them to a plate at 22°. For high reproducibility, incubation at 30° should be performed in the dark under our cultivation conditions.

Samples were frozen in liquid nitrogen immediately. The RNeasy Plant Mini Kit (Qiagen, Hilden, Germany) was used to extract total RNA. The RNase-Free DNase Set (Qiagen, Germany) was used to eliminate the contamination of genomic DNA from the RNA samples. Reverse-transcription PCR was performed using PrimeScript RT Master Mix (Takara, Shiga, Japan). Quantitative RT–PCR was applied as described previously (Wang et al., 2020). Arabidopsis PP2A (for FLC) and EIF4A (for JMJ30 and JMJ32) were used as the internal references. Each experiment was repeated at least three times. The relative expression level of each gene was calculated using the 2^–ΔΔCT method (Livak and Schmittgen, 2001). The primers are listed in Supplementary Table 1.

Seedlings were fixed in 90% acetone for 30 min at room temperature and subsequently stained with β-glucuronidase (GUS) staining solution. The staining method was as described previously (Gan et al., 2014; Shirakawa et al., 2014). After GUS staining, samples were transparentized as described previously (Shirakawa et al., 2009). Representative images were photographed under an AXIO Zoom V16 (ZEISS) microscope.

To test the timing of flowering, including the number of rosette or cauline leaves produced, we vernalized plants and then transferred them into soil cultivation conditions. We cultivated plants until the boltings of the primary stems and then counted the number of leaves, as described previously (Shirakawa et al., 2021).

Chromatin immunoprecipitation experiments were performed as previously described with minor modifications (Gan et al., 2014; Yamaguchi et al., 2014; Shirakawa et al., 2021). Briefly, total chromatin was extracted from the seedlings and immunoprecipitated using anti-H3K27me3 (Abcam, Cat. No. ab6002). The DNA fragments were recovered by QIAquick PCR Purification Kit (QIAGEN, Cat. No. 28106). qPCR with gene-specific primers (Supplementary Table 1) was performed on a LightCycler 480 System II (Roche) using a FastStart Essential DNA Green Master (Roche, Cat. No. 06924204001). Values of percent input of target loci were calibrated by values of percent input of AGAMOUS loci. The experiments were repeated three and six times for NV and V2W, respectively. The statistical significance was evaluated by two-tailed Student’s t-test.

In this study, one-way ANOVA followed by the Tukey–Kramer test or two-tailed Student’s t-test was performed to detect the differences as required.

First, we examined whether the expression levels of JMJ30 and JMJ32 were changed during vernalization. After water absorption by the seeds, we incubated them under various periods of cold treatment (from 0 h to 4 weeks) in the dark. Then, we germinated them on gellan gum plates and compared the expression levels of two genes, JMJ30, and JMJ32, in the seedlings at 3 days after germination (Figure 1). Interestingly, the expression levels of JMJ30 started to decrease after 6 h of cold treatment, and they reached their minimum level after 1 week of cold treatment and were maintained at the minimum level for 4 weeks (Figure 1A; labeled “f” in one-way ANOVA followed by the Tukey–Kramer test). Unlike JMJ30, the expression levels of JMJ32 were not changed by cold treatment (Figure 1B). These results suggested that a reduction of the expression levels of JMJ30 occurred quickly after cold treatment; however, the activities of JMJ30 and 32 remained after cold treatments.

To clarify the roles of JMJ30 and JMJ32 in vernalization, we compared the flowering time between wild-type and jmj30 jmj32 double mutants harboring the active FRIGIDA gene (hereafter, wild-type and jmj30 jmj32) (Figure 2). We did not test single mutants of jmj30 and jmj32 because they are redundantly required for the prevention of heat-induced extreme early flowering (Gan et al., 2014). In the non-vernalized conditions, jmj30 jmj32 showed a slightly early flowering phenotype [Figure 2B; the total number of leaves: 85.45 (wild type) vs. 80.45 (jmj30 jmj32)]. Under the vernalized conditions of 2 weeks, jmj30 jmj32 showed a clear early flowering phenotype [Figure 2B; the total number of leaves: 75.5 (wild type) vs. 62.35 (jmj30 jmj32)] because the difference in the total number of leaves was larger than that in the non-vernalized condition. Under the vernalized conditions of 4 weeks, jmj30 jmj32 showed an early flowering phenotype [Figure 2B; the total number of leaves: 41.1 (wild type) vs. 33.5 (jmj30 jmj32)]; however, the difference in the total number of leaves was smaller than that under the vernalized conditions of 2 weeks. Finally, under the fully vernalized conditions of 6 weeks, jmj30 jmj32 showed a similar timing of flowering as the wild type (Figure 2B). Collectively, these results suggested that JMJ30 and JMJ32 modulate the speed of vernalization.

Next, we examined whether JMJ30 and JMJ32 modulate the speed of vernalization through the expression levels of FLC. By quantitative polymerase chain reaction (qPCR) analysis, in the partially vernalized conditions (V2W and V4W), we found a significant reduction in FLC expression in jmj30 jmj32 compared with the wild type (Figure 3A). In addition, we compared the spatiotemporal expression patterns of FLC::GUS between wild type and jmj30 jmj32 (Figure 3B and Supplementary Figure 1). Under V2W and V4W conditions, the expression levels of FLC in both cotyledons and rosette leaves of jmj30 jmj32 were lower than those in wild type. These results suggested that lower expression levels of FLC triggered the early flowering phenotype of jmj30 jmj32 in the partially vernalized conditions. To clarify the genetic pathway between JMJs and FLC, we generated triple mutants, flc jmj30 jmj32. Under non-vernalized conditions, flc exhibited the extreme early flowering phenotype [Figures 3C,D; the total number of leaves: 85.45 (wild type) vs. 13.8 (flc)]. flc jmj30 jmj32 also showed the extreme early flowering phenotype [Figures 3C,D; the total number of leaves: 13.85 (flc jmj30 jmj32)]. These results suggested that FLC is genetically epistatic to JMJ30 and JMJ32 in flowering. Combined with the data in Figures 3A,B, we concluded that JMJ30 and JMJ32 act upstream of FLC. Under high-temperature conditions, JMJ30 and JMJ32 are required for the elimination of H3K27me3 from the FLC locus. We examined whether the accumulation levels of H3K27me3 on the FLC locus were changed in jmj30 jmj32 under partial vernalized conditions. We found that the accumulation levels of H3K27me3 on the nucleation region of the FLC locus were slightly but statistically significantly increased in jmj30 jmj32 under partially vernalized conditions (V2W), while no clear changes were observed under non-vernalized conditions (NV) in multiple biological replicates (Figure 3E). Taken together, these results suggested that JMJ30 and JMJ32 modulate flowering time through the regulation of FLC during vernalization.

We generated transgenic plants, pEstro:JMJ30 (Yamaguchi et al., 2021), in which JMJ30 was overexpressed when we treated them with estrogen (Figure 4). In contrast to jmj30 jmj32, pEstro:JMJ30 with estrogen treatment showed a slight late-flowering phenotype compared with the line without estrogen treatment (Figures 4A,B). Estrogen treatment induced upregulation of FLC (Figure 4C). These results suggest that overexpression of JMJ30 may be able to confer the late-flowering phenotype through the regulation of FLC. JMJ30 is one of the key factors regulating flowering time in Arabidopsis.

Devernalization is a reversion of vernalized status to non-vernalized status by heat. It was reported that H3K27me3 on the FLC locus was reduced after heat treatment (Bouché et al., 2015). In addition, we previously found that heat induced the upregulation of JMJ30 and the stabilization of JMJ30 (Gan et al., 2014). Combining these results, we hypothesized that heat-activated JMJ30 might eliminate H3K27me3 from the FLC locus during devernalization. First, we established the experimental conditions for devernalization using Arabidopsis. We vernalized the seeds at 4° in the dark and then transferred them to 30° in the dark (Figure 5A). These plants showed a late flowering phenotype compared with vernalized plants [Figures 5B,C; the total number of leaves: 32.1 (V4W) vs. 59.3 (V4W + 30°C)]. Upon heat treatment, jmj30 jmj32 showed a late flowering phenotype compared with vernalized jmj30 jmj32 [Figures 5B,C; the total number of leaves: 24.3 (V4W) vs. 50.7 (V4W + 30°C)]. These results suggested that devernalization occurred even in jmj30 jmj32. Consistent with this, heat-treated jmj30 jmj32 expressed 1.7-fold higher levels of FLC than vernalized jmj30 jmj32, as the wild-type did (Figure 5D). Taken together, these results suggested that JMJ30 and JMJ32 were not key factors for devernalization in Arabidopsis, although we could not exclude the possibility that JMJ30 and JMJ32 work with other histone demethylases during devernalization.

Histone demethylases involved in the vernalization pathway have not been identified. In this study, we showed that JMJ30 and JMJ32 act as molecular brakes for vernalization through the regulation of FLC in Arabidopsis (Figures 1–4). First, the loss-of-function mutants jmj30 jmj32 exhibited an early flowering phenotype under partial vernalization conditions (Figure 2). Second, the levels of these early flowering phenotypes under different vernalized conditions were fairly consistent with the expression levels of FLC, and the genetic interaction with FLC indicated that JMJ30 and JMJ32 were upstream factors for FLC (Figure 3). Third, the deposition of H3K27me3 was enhanced in jmj30 jmj32 at the FLC locus in partial vernalized conditions. Finally, the inducible overexpression of JMJ30 caused the late-flowering phenotype via FLC regulation (Figure 4). Similar results were obtained by using a constitutive overexpression line of JMJ30 (Gan et al., 2014). We found a cold-inducible reduction in JMJ30 (Figure 1). To reduce the levels of JMJ30, plants may prepare for the start of vernalization. Future work will identify upstream factors for the cold inducibility of JMJ30.

In this study, we found that devernalization was triggered in jmj30 jmj32 by heat, resulting in the upregulation of FLC and delayed flowering (Figure 5). These results suggested that JMJ30 and JMJ32 were not essential factors for devernalization in Arabidopsis, although we do not exclude the possibility that the other three jumonji proteins, JMJ11, JMJ12, and JMJ13, redundantly function in devernalization with JMJ30 and JMJ32 (Lu F. et al., 2011; Crevillen et al., 2014; Cui et al., 2016; Yan et al., 2018). It was reported that devernalized plants exhibited lower accumulation levels of H3K27me3 on the locus than vernalized plants (Bouché et al., 2015). However, it is largely unknown whether devernalization is induced by the active demethylation of de novo deposited H3K27me3. There are two additional targets of devernalization. For the stable silencing of FLC after vernalization, the spreading of H3K27me3 to the whole FLC genomic region and the maintenance of H3K27me3, de novo deposition of H3K27me3 into newly incorporated histones during cell division/DNA replication are required. Heat may inhibit these processes. LHP1 and CLF are required for the spreading and maintenance of H3K27me3 in the FLC locus and are not required for the deposition of H3K27me3 in the nucleation region of FLC, where H3K27me3 is deposited first after cold treatment (Yang et al., 2017). Interestingly, it was reported that FLC in lhp1 and clf was slightly and gradually reactivated after vernalization. The levels of FLC reactivation are very similar to those in heat-induced (Périlleux et al., 2013) and chemical-induced reactivation (Shirakawa et al., 2021). It is interesting to question whether heat affects the activity and stability of LHP1 and CLF and whether devernalization responses occur in mutants. Future works involving time course analysis of H3K27me3 during devernalization will provide detailed insights into the molecular mechanisms of devernalization. Other epigenetic marks, such as H3K4me3 and H3K9me2 may be involved in devernalization.

In their roles in flowering, JMJ30 and JMJ32 are required for abscisic acid (ABA) and brassinosteroid (BR) responses (Wu et al., 2019a,b, 2020), acclimation to high temperature (Yamaguchi, 2021a,b; Yamaguchi et al., 2021; Yamaguchi and Ito, 2021a,b), callus formation (Lee et al., 2018) and the regulation of period length (Lu S. X. et al., 2011). In addition, it has been reported that the expression of JMJ30 and the stability of JMJ30 are regulated by heat (Gan et al., 2014). However, the upstream factors affecting JMJ30 expression and the stabilizer of JMJ30 are largely unknown. Future work will identify such factors. The functions of JMJ30 and JMJ32 in other plant species are still open questions to be addressed.