A 65-kb deletion survey identifies a distal cis-regulatory region for red-light induction of Ghd7, a key rice floral repressor

Significance

The Ghd7 (Grain number, plant height, and heading date 7) gene, a rice key floral repressor under long-day conditions is acutely transcribed by red light signals via phytochrome photoreceptors. Ghd7 and its orthologs play important roles in crop diversification among major cereals. Using CRISPR/Cas9 genome editing, we generated a series of targeted deletion mutants in the upstream regions of Ghd7, covering a 65-kb genomic region. This approach led to the identification of a distal cis -regulatory region for Ghd7 red light induction within a 228-bp segment approximately 28 kb upstream of the transcription start site, supported by robust experimental evidence. Our findings enhance the current understanding of Ghd7 regulation and provide a model for dissecting phytochrome signaling pathways in monocots.

Abstract

The Ghd7 (Grain number, plant height, and heading date 7) gene integrates red light signals and circadian rhythms to control floral repression under long-day conditions in rice. CRISPR/Cas9 systems were employed to create a series of deletion mutant lines in the upstream regions of Ghd7, covering a 65-kb genomic region from its transcription start site (TSS). These deletions ranged from 2 to 25 kb in size. Three deletion lines, those from 0 to −3 kb (0/−3 K), −20 to −40 kb (−20/−40 K), and −26 to −30 kb (−26/−30 K) from the TSS, resulted in early flowering, similar to Ghd7 knockout lines. The −20/−40 and −26/−30 K lines exhibited a loss of acute Ghd7 morning induction. Night-break experiments consistently supported these findings, suggesting that the key cis-regulatory region for red light responses was located within the 3.7-kb region in −26/−30 K. In seedlings of the 0/−3 K deletion line, which retains the −29 to −86 bp region, Ghd7 showed a diurnal pattern similar to wild type. This suggests that the deleted region in 0/−3 K is dispensable for both circadian rhythms and red-light responses. Further analyses of two deletion lines within the −26/−30 K region allowed us to narrow down the core cis-regulatory elements, responsive to morning-light signals, within a 228-bp segment located at 28-kb upstream of the TSS in Ghd7.

Flowering time is a key determinant of harvest timing, making it a vital trait across all crop species, with profound effects on both yield and quality. The Grain number, plant height, and heading date 7 (Ghd7) gene serves as a principal floral repressor, governing both photoperiodic and ambient-temperature responses in rice (1–3). Ghd7 encodes a CO, CO-like, and TOC1 (CCT) domain protein (1) and forms a floral repressor complex with the Heading date 1 (Hd1) gene, an ortholog of the Arabidopsis CONSTANS gene, and the Grain number, plant height, and heading date 8 (Ghd8)/Heading date 5 (Hd5) gene (4–8). This repressor complex binds to the promoter region of the Early heading date 1 (Ehd1) gene, a floral promoter gene, subsequently repressing the transcription of the florigen genes, Heading date 3a (Hd3a) and Rice Flowering-locus T 1 (RFT1), under long-day (LD) and low-temperature conditions (3, 4, 9).

Among the natural variations of Ghd7, a loss-of-function allele is utilized in almost all rice cultivars grown in Hokkaido, the northern island of Japan. These cultivars show early flowering phenotypes, despite the region’s longer photoperiods and lower ambient temperatures (1, 3, 10, 11). In wheat, Vernalization 2 (VRN2) is a major genetic player that controls vernalization responses (12). VRN2 also encodes a CCT domain floral repressor and is recognized as an ortholog of Ghd7. Recently, SbGhd7, an ortholog of Ghd7 in sorghum, has been identified as a primary genetic contributor to hybrid vigor (13). In maize, loss-of-function alleles of ZmCCT genes were likely selected during the primary domestication processes (14). While several CCT domain-containing floral repressors may exist in dicotyledonous plants, those evolutionarily close to Ghd7 appear to be absent in dicot species. This suggests that the transcriptional repression system mediated by Ghd7-type proteins is unique to monocotyledonous (or Poaceae) plants. Additionally, previous studies have shown that Ghd7 has some extent of pleiotropic effects, increasing plant height and grain number, alongside its central role in flowering-time regulation (1, 15, 16).

Hd3a transcription is triggered under photoperiods shorter than the critical day length of 13.5 h (2). Ghd7 plays a central role in the accurate day-length recognition in rice, as defective mutant alleles of Ghd7 lead to the derepression of Hd3a and RFT1 genes. Ghd7 is induced in response to red light, with sensitivity to red light varying throughout the day (2, 17). Furthermore, its red-light sensitivity peaks at different times under LD and short-day (SD) conditions: at dawn for LD and at midnight for SD (2). Extensive genetic analysis using single and double mutants of rice phytochrome genes revealed that the Ghd7 transcription in response to red light is likely regulated by the phyA homodimer or the phyB/phyC heterodimer (17). Thus, Ghd7 expression is primarily induced in the morning through two integrated regulatory mechanisms: phytochrome-mediated red-light induction and circadian-clock regulated changes in red light sensitivity (gating effects). The cis-regulatory elements (CREs) that control Ghd7 transcription likely include components for both red-light signaling by phytochromes and circadian clock regulation. Recently, OsELF3-1, a rice ortholog of Arabidopsis EARLY FLOWERING 3, was found to be involved in the repression of Ghd7, also indicating its interaction with phytochromes (18, 19). Consistently, a ChIP-seq analysis suggested that OsELF3-1 binds to the upstream region of Ghd7 (20). In addition, some repressive elements mediated by Ehd3 and Trithorax1 may also exist in the Ghd7 promoter (21, 22). Therefore, understanding the transcriptional regulation of Ghd7 is valuable not only for the basic science using rice as a monocotyledonous model plant system but also for cereal breeding, including wheat, sorghum, maize, and rice. For example, Zhou et al. demonstrated that multiplex promoter-editing lines of key rice flowering genes, Ghd7, DTH8, and Hd1, hold remarkable potential for agricultural applications (23). However, no key transcription factors have been reported to bind to the Ghd7 promoter in vivo, and no functional CREs in the Ghd7 promoter have been identified. In Arabidopsis thaliana, phytochrome-mediated transcription is generally thought to be primarily regulated by phytochrome interacting factors (PIFs) including PIF4 (24). On the other hand, the responsible CREs, which are considered direct targets of phytochrome signaling, have not yet been genetically identified (24). Analysis of field transcriptome data in rice (25, 26) identified Ghd7 as one of the rare genes that acutely respond to morning light under field conditions. Combined with the phytochrome mutant analysis, Ghd7 emerges as a potential model gene for elucidating transcriptional regulation by phytochrome signals in plants.

Years ago, our group conducted preliminary experiments involving a complementation test using a genomic fragment containing the Ghd7 gene body and approximately 4 kb of its promoter. However, the complementation was only partial, suggesting that critical cis-regulatory elements essential for Ghd7 transcriptional induction might be located much farther upstream than commonly expected. These data were not able to be published since another group published the same gene as Ghd7 without clear results of the genomic complementation test (1). Using CRISPR/Cas9 genome editing technology, we initially generated deletion lines up to approximately 10 kb from the transcription start site (TSS), but no substantial changes in heading date were observed. Therefore, we extended the deletion to 65 kb upstream of TSS. In this study, we used a series of CRISPR/Cas9-generated deletion mutants covering this 65-kb upstream region to investigate the regulatory mechanisms of Ghd7 transcription and to identify the responsible CREs. Using these deletion lines, we demonstrated that a critical cis-regulatory region involved in the red-light induction of Ghd7 is located within a 228-bp segment around −28 kb, notably distant from the TSS of Ghd7. Additionally, the region from the TSS to −3 kb is not required for the diurnal fluctuation of Ghd7 expression.

Results

Generation of Ghd7 Promoter Deletion Lines.

To identify the core genomic regions regulating Ghd7 (Os07g0261200) transcription, we first generated a series of deletion mutants in the upstream region of the Ghd7 gene using the CRISPR/Cas9 system. In genome editing, introducing guide RNAs at two separate positions can result in corresponding deletions of the region between them in rice plants (27). Therefore, we designed guide RNAs at both ends of target regions, producing a series of deletion mutant lines upstream of Ghd7 (Fig. 1A). There is a locus termed Os07g0261732, located 52 kb upstream of Ghd7, which is hardly expressed and considered to be a pseudogene, as almost no reads were detected according to the RNA-seq data in RAP-DB (https://rapdb.dna.affrc.go.jp/). Another locus, Os07g0262200, located 65 kb upstream of Ghd7, was found to be expressed, as evidenced by RNA-seq reads. Therefore, we investigated a region up to 65 kb upstream from the TSS, which may be involved in the transcriptional regulation of Ghd7.

Fig. 1.

Ghd7 promoter deletion mutants used in this study. (A) CRISPR/Cas9-mediated deletion series of the Ghd7 upstream region. TSS represents the transcription start site. Os07g0261732 at −52 kb is a putative pseudogene, and Os07g0262200 is considered the closest upstream gene to Ghd7. Guide RNAs (indicated by short red lines) were designed to flank the target deletion region. (B) Heading date of Ghd7 upstream deletion mutants. Data for the −10/−20 K line are presented separately due to cultivation at a different date from the other lines. NB and KO indicate control WT plant (“Nipponbare”) and a Ghd7 knockout (KO) line, respectively. Error bars represent SD. Asterisks indicate values significantly lower than NB (**P < 0.01, ***P < 0.001, Welch’s t test). P and n values are provided at the bottom of the figure.

As shown in Fig. 1A, we initially designed seven mutant lines with distinct deletions ranging from 2 kb to 25 kb upstream of Ghd7. In total, approximately 900 regenerated plants were obtained, with 96 to 175 regenerated plants per deletion line. Among these, 1 to 20 plants exhibited deletions in the targeted regions (SI Appendix, Table S1). The mutants were named based on their deletion ranges as follows: TSS to −3 kb (0/−3 K); −0.5 to −3 kb (−0.5/−3 K); −3 to −5 kb (−3/−5 K); −5 to −10 kb (−5/−10 K); −10 to −20 kb (−10/−20 K);−20 to −40 kb (−20/−40 K); and −40 to −65 kb (−40/−65 K) (Fig. 1A). These deletions were confirmed by PCR and Sanger sequencing of the corresponding PCR products.

Deletions were less frequent in lines 0/−3 K, −0.5/−3 K, −10/−20 K, and −20/−40 K, as 175, 144, 180, and 128 regenerated plants were examined, with only 1, 2, 1, and 2 plants with designed deletions were obtained, respectively. Deletions occurred relatively more frequently in lines −3/−5 K, −5/−10 K, and −40/−65 K, as 96 regenerated plants were examined per line, with 14, 20, and 10 plants exhibiting the target deletions, respectively. For all deletion lines, the T₂ lines with homozygous deletions were selected. We also designed a guide RNA targeting the first exon of Ghd7 and generated frameshift-KO lines in Ghd7 with short deletions, all of which flowered earlier than the control cultivar, Nipponbare (NB). In the following experiments, KO-64 was used as a KO line unless otherwise noted.

Phenotypic Analysis of the Deletion Lines.

To evaluate the impact of the deletions on flowering time, we compared the heading date and plant height of these mutants with those of NB under both LD and SD conditions (Fig. 1B and SI Appendix, Fig. S2A). Under LD, days-to-heading for 0/−3 K, −20/−40 K, and KO were significantly earlier than for NB, flowering on average at 108, 95, and 104 d after germination, respectively, compared to 125 d for NB, while the −3/−5 K line showed a slight delay (Fig. 1B). Days-to-heading remained unchanged in the other lines, including the −0.5/−3 K line, where we had originally expected a critical effect on flowering time, since some key CREs are often thought to be located near the TSS. Under SD, the days-to-heading for NB were about 65 d after sowing, and neither the upstream deletion lines nor the KO line flowered earlier than NB. When these plants were grown under four distinct photoperiod and ambient temperature conditions using growth chambers, we obtained similar flowering-time phenotypes except the 0/−3 K line (SI Appendix, Fig. S1). In the 0/−3 K line, days-to-heading were not changed except the low temperature and SD conditions, suggesting certain responses sensitive to ambient environments in 0/−3 K.

Plant height for −20/−40 and −10/−20 K was lower than for NB under both LD and SD conditions (SI Appendix, Fig. S2A). The plant height of 0/−3 K was also lower than NB under LD. The changes in plant height corresponded to the changes in flowering time.

Gene Expression Analysis of the Deletion Lines.

Next, we examined the diurnal gene expression in 0/−3 K, −20/−40 K, and KO, which flowered markedly earlier than NB, and in −0.5/−3 K, whose flowering time was similar to that of NB. The −3/−5 K line was not included in this analysis due to its minor flowering time difference. Plants grown for 5 wk each under LD and SD were harvested at ZT (zeitgeber time) −2, 2, 6, 10, 14, and 18. In NB, the Ghd7 expression peaked prominently at ZT2 and decreased during the dark phase under both LD and SD (Fig. 2A), consistent with previous studies (4, 28). 0/−3 K, −0.5/−3 K, and KO lines showed a critical increase at ZT2 under both LD and SD like NB, while −20/−40 K showed a strikingly reduced increase at ZT2 (Fig. 2A), suggesting that the ZT2 induction of Ghd7 transcription in the morning is absent in −20/−40 K. Furthermore, NB showed increased expression at ZT14 only under LD, whereas −20/−40 K did not.

Fig. 2.

Expression analysis of Ghd7 and its downstream genes. (A) Diurnal expression analysis of Ghd7. Rice plants were grown for 5 wk under LD and SD conditions. Leaf samples were harvested at ZT −2, 2, 6, 10, 14, and 18. X-axis values are offset for visual clarity, without altering the actual sampling times. Error bars indicate SD (n = 2 or 3). Filled circles: Normal data points; Hollow squares: Significant difference from NB (P < 0.05, Welch’s t test). (B) Expression analysis of Ghd7 and its downstream genes. Leaves of 7-wk-old plants grown under LD conditions were harvested at ZT2. Error bars indicate SD (n = 5 or 6). NB and KO denote control WT plant (Nipponbare) and a Ghd7 KO line, respectively. Asterisks indicate values significantly different from NB (*P < 0.05, **P < 0.01; Welch’s t test).

Under LD, except for ZT2 and ZT14, there were no significant differences in Ghd7 expression between NB and −20/−40 K. Under SD, Ghd7 was elevated at ZT2 even in −20/−40 K, but less so than that in NB. Additionally, except for ZT2, there was no significant difference in Ghd7 expression between −20/−40 K and NB under SD. Thus, in −20/−40 K, the increase in Ghd7 expression at ZT2 was lower than in NB under both LD and SD, although the diurnal fluctuations remained. Contrary to our expectations, 0/−3 K showed a similar expression pattern to NB in 5-wk-old seedlings despite the deletion near the TSS, while the expression was lower than NB at ZT14 under LD (Fig. 2A). Next, using plants grown under LD for 7 wk, mRNA expression was examined at ZT2 for Ghd7 and its downstream genes, i.e., Ehd1, Hd3a, and RFT1. The Ghd7 level was lower in −20/−40 K and 0/−3 K than NB, but not in −0.5/−3 K and KO (Fig. 2B). The mRNA expressions of Ehd1, Hd3a, and RFT1 were higher than NB in −20/−40 K, 0/−3 K, and KO, while it was comparable to NB in −0.5/−3 K. These results suggest that −20/−40 K and 0/−3 K resulted in earlier flowering due to increased expression of Ehd1, Hd3a, and RFT1.

Detailed Analysis of the 0/−3 K line.

As previously noted, 0/−3 K also flowered earlier and showed increased Ehd1, Hd3a, and RFT1 expression under LD in the 7-wk-old plants. To date, only one mutant line of 0/−3 K has been obtained, and this line presented a complex deletion. Two deletions were observed from +2 bp to −28 bp and from −87 bp to −2,983 bp upstream of the TSS, while the −29 bp to −86 bp region was retained in the genome (Fig. 3A). This complex deletion might be due to the introduction of three guide RNAs near the TSS. We compared 0/−3 K to two photoperiod-insensitive cultivars, ‘Aikoku’ and ‘Yukara’, both of which have an identical 1.9-kb transposon insertion at −11 bp from the Ghd7 TSS (29, 30) (Fig. 3A). We confirmed that ‘Aikoku’ and ‘Yukara’ had no mutations, deletions, or insertion near the TSS except for the 1.9-kb insertion using TASUKE+ database (https://agrigenome.dna.affrc.go.jp/tasuke/ricegenomes/). Then, the diurnal expression of Ghd7 was examined in 0/−3 K, ‘Aikoku’, and ‘Yukara’ (Fig. 3B). Although the two cultivars exhibited Ghd7 expression patterns similar to NB, retaining the acute morning induction, their basal levels were significantly lower. The 0/−3 K line showed reduced expression at ZT 19 compared to NB, yet much higher than in ‘Aikoku’ and ‘Yukara’. These results suggest that CREs regulating circadian rhythms or red-light responses are unlikely to be present between the TSS and −3 kb. Given that Ghd7 expression is higher in 0/−3 K than in the two cultivars across almost all time points, it is conceivable that the residual short segment from −29 bp to −86 bp harbors a CRE that may influence the basal expression of Ghd7, although other possibilities cannot be excluded.

Fig. 3.

Diurnal expression analysis of Ghd7 for 0/−3 K, −20/−40 K, and two related cultivars with a mutation around the Ghd7 TSS. (A) Schematic diagram of the genomic regions around the TSS of 0/−3 K, ‘Aikoku’ and ‘Yukara’. The 0/−3 K has a small deletion between +2 and −28, and a large deletion between −87 and −2,983, while retaining the region from −29 and −86. The two cultivars, ‘Aikoku’ and ‘Yukara’, carry an identical 1.9-kb insertion at −11 bp upstream of the Ghd7 TSS. (B) Diurnal expression pattern of Ghd7. Fully expanded leaves from plants grown under LD conditions for 4 wk were sampled at ZT 0, 2, 6, 10, 14, and 19. X-axis values are offset for visual clarity, without altering the actual sampling times. Error bars indicate SD (n = 2 or 3). NB indicates control WT plant (Nipponbare). Filled circles: Normal data points; Hollow squares: Significant difference from NB (P < 0.05, Welch’s t test).

Key cis-Regulatory Region for Morning Induction of Ghd7.

Since we found that the upstream region from −20 to −40 kb is important for the ZT2 expression (or morning induction) of Ghd7, we further investigated to narrow it down. Ghd7 is regulated by circadian clock and red light via the phytochromes in rice (2, 17). In A. thaliana, it is well known that PIFs, basic helix–loop–helix (bHLH) type transcription factors that interact with phytochromes, as well as other circadian clock genes like LUX and PRRs, bind to DNA motifs containing G-box (core sequence: “CACGTG”) (31–40). Therefore, we searched for DNA sequences containing a G-box between the −20 and −40 kb from the TSS of Ghd7. There are six G-boxes at positions −21,707, −23,048, −23,525, −24,277, −26,457, and −39,245 bp from the TSS between −20 and −40 kb. By genome editing, we then obtained rice plants with mutations in the third, fourth, and fifth G-boxes at −23,525, −24,277, and −26,457 bp from the TSS, respectively, and named as G-box −23,525, G-box −24,277, and G-box −26,457, respectively (Fig. 4A). The first G-box at −21,707 bp was included in the repeat region (RAP-DB), so we did not generate genome-edited rice plants because we thought it would be difficult to introduce mutations specifically in this G-box. Also, the rice plants with the targeted mutation were not obtained for the second G-box at −23,048 bp and the sixth G-box at −39,245 bp from the TSS. As for heading date, there was no difference among G-box −23,525, G-box −24,277, and NB when grown under LD (Fig. 4B). Upon using sgRNA for the fifth G-box at −26,457 bp from the TSS, interestingly, several lines with distinct mutations were obtained by genome editing as indicated Fig. 4B. Two of the obtained lines for the G-box −26,457, “701-10,” and “781-18,” had the almost same heading dates as NB when grown under LD, but the other line obtained using the G-box −26,457 as the target, termed as “811,” had an earlier heading date. Genome sequencing of the line 811 revealed its biallelic deletion mutations around the G-box at −26,457 bp: a deletion of approximately 3.7 kb from −26,448 to −30,138 bp, including the G-box at −26,457 bp (−26/−30 K), and a 28 bp deletion, also including the G-box at −26,457 bp (28-del). After self-crossing, the heterozygous lines and the homozygous lines of −26/−30 K and 28-del were isolated. Under LD, the homozygous −26/−30 K line, the heterozygous line, and the homozygous 28-del line flowered in this order, with the homozygous 28-del line showing no significant difference from NB in heading date (Fig. 4B). The expression of Ghd7 in −26/−30 K showed a similar expression pattern under both LD and SD as in −20/−40 K, with reduced induction at ZT2 compared to NB (Fig. 4C). Furthermore, the expressions of Ehd1, Hd3a, and RFT1 in −20/−40 K and −26/−30 K were higher than that of NB (Fig. 4D). These results indicate that the 3.7-kb region in the −26/−30 K line contains critical cis-regulatory regions which are responsible for the ZT2 induction of Ghd7. In addition, the third, fourth, and fifth G-boxes at −23,525 bp, −24,277 bp, and −26,457 bp, respectively, are not likely to be involved in heading date.

Fig. 4.

Genome-editing mutations targeting G-boxes between −20 kb and −40 kb. (A) CRISPR/Cas9-mediated mutations targeting the G-boxes between −20 kb and −40 kb upstream of Ghd7. “G” indicates an intact G-box, “X” represents a mutated G-box. The guide RNA targeting the G-box at −26,457 yielded two types of transformants: one with specific G-box mutations (designated 28b-del) and another with a large deletion (−26,448 to −30,138, designated −26/−30 K). (B) Heading date of genome-edited rice strains under LD conditions. The lines examined include G-box −23,525 (317-5), G-box −24,277 (44-1, 542-2), G-box −26,457 mutants (701-10, 781-18), and one heterozygous line (811). Line 811 carries a 28-base deletion including the G-box at −26,457 on one chromosome and a 3.7-kb deletion (−26,448 to −30,138) on the other. Homozygous lines for each mutation and their heterozygous combination were also examined (−26/−30 K, 28b-del, and biallelic). NB denotes control WT plant Nipponbare. Error bars represent SD (n = 5 to 38). Asterisks (**) indicate values significantly different from NB (P < 0.01, Welch’s t test). (C) Diurnal expression analysis of Ghd7 in −26/−30 K under LD and SD conditions. Fully expanded leaves from 5-wk-old plants were sampled at ZT −2, 2, 6, 10, 14, and 18. X-axis values are offset for visual clarity, without altering the actual sampling times. Error bars indicate SD (n = 2 or 3). Filled circles: Normal data points; Hollow squares: Significant difference from NB (P < 0.05, Welch’s t test). (D) Expression analysis of Ghd7 and the downstream genes. Leaf samples were collected from 7-wk-old plants grown under LD conditions at ZT 2. Error bars represent SD (n = 5 or 6). Asterisks indicate values significantly different from NB (*P < 0.05, **P < 0.01; Welch’s t test).

Night Break Experiments on the Early-Flowering Mutants.

Ghd7 expression is induced by red light through phytochrome signaling, with sensitivity varying according to day length and circadian phase. This sensitivity peaks at ZT0 under LD and at midnight during the dark phase under SD (2, 17). To explore whether the reduced ZT2 induction of Ghd7 in the −20/−40 K and −26/−30 K lines was due to a decreased red-light response or a shift in peak sensitivity timing, we monitored Ghd7 induction across various circadian phases using −20/−40 K, −26/−30 K, and another early-flowering mutant 0/−3 K. In the NB control, red light responses were strong at ZT−3 and ZT0 under LD and at ZT−8 and ZT−4 under SD, aligning with previous studies (Fig. 5B). However, the low red-light responses in the mutant lines made it difficult to confirm any shifts in peak sensitivity timing. In the −20/−40 K and −26/−30 K lines, Ghd7 expression remained low at all time points, with minor induction between ZT−7 and ZT0 under LD and between ZT−11 and ZT−4 under SD (Fig. 5B). Unlike these mutants, the 0/−3 K line showed a red-light response at ZT0 closer to that of NB, suggesting partial retention of wild-type responsiveness at this phase (Fig. 5B). The low expression after red light pulses in the 0/−3 K line may suggest reduced basal transcription of Ghd7 rather than an altered red-light response. These findings indicate that ZT2 induction of Ghd7 is linked to a time-sensitive red-light response mediated by phytochromes and that a 3.7-kb upstream region plays a key role in this regulation.

Fig. 5.

Temporal expression analysis of Ghd7 with red-light irradiation. (A) Schematic diagram of light treatment and sample collection for expression analysis. After growing for 18 d under LD conditions or for 22 d under SD conditions, plants were transferred to continuous dark conditions at dusk. Plants were then exposed to a 10-min red light pulse, and fully expanded leaves were collected 2 h after irradiation for RNA preparation. Black, gray, and red bars represent dark conditions, subjective light conditions, and red-light irradiation, respectively. Arrows show when samples were collected. (B) qRT-PCR analysis of Ghd7 under LD and SD conditions. The x-axis corresponds to the treatment and sampling patterns described in (A). The red bar (Red light) and gray bar (Dark) indicate treatments with and without red light pulses, respectively. Error bars indicate SD (n = 2 to 3). NB indicates control WT plant, Nipponbare.

Detailed Analysis of the 3.7-kb cis-Candidate Region.

We further analyzed the 3.7-kb region in detail. There was only one typical G-box motif (CACGTG) at −26,457 bp from the TSS within this 3.7-kb region, which was proved dispensable for ZT2 induction of Ghd7. Thus, we searched for G-box-like motif present in the 3.7-kb region and found seven G-box-like motifs, for which we designed corresponding sgRNAs (Fig. 6A). In the experiments, we used a binary vector harboring the Cas9-NG expression construct, an artificially modified Cas9 that recognizes NG as a PAM sequence rather than NGG (41, 42). Among the seven tested sgRNAs, several lines mutated in the three G-box-like motifs, P3, P5, and P8 at −27,402, −27,998, and –28,594 bp from the TSS of Ghd7 were obtained. The mutations in P3 and P8 did not affect heading dates under LD (SI Appendix, Fig. S3A). As for the P5 site, we obtained two distinct deletion mutations at P5: one has a 234-bp deletion between −27,781 and −28,014 bp from the TSS, and the other has a 586-bp deletion between −27,423 and −28,008 bp from the TSS. Both lines, termed 200-del and 600-del, flowered around 2 wk earlier than NB under LD, almost as early as the −26/−30 K and KO lines (Fig. 6 B and C). Ghd7 expressions in 200-del and 600-del were significantly reduced at ZT2 in plants grown under LD for 5 wk (Fig. 6D). The results clearly indicated that both 200-del and 600-del exhibited critically reduced Ghd7 mRNA expression, similar to the diurnal Ghd7 expression pattern seen in −20/−40 K and −26/−30 K. Thus, we have succeeded in narrowing down the core cis-regulatory region responsible for the ZT2 induction of Ghd7 into the 228 bp region from –27,781 to –28,008 bp from the TSS in Ghd7.

Fig. 6.

Analysis of 200-del and 600-del deletion mutants. (A) CRISPR/Cas9-mediated mutations targeting the 3.7-kb candidate cis-regulatory region of Ghd7. “G” indicates an intact G-box, while “L” represents a G-box-like motif with a single-base substitution to the typical G-box sequence. Eight guide RNAs targeting the G-box and the G-box-like motifs in the region were designed to introduce mutations and designated P1 to P8. Blue letters indicate guide RNAs that did not produce mutant lines. The P5 guide RNA yielded two lines with larger deletions: 600-del and 200-del. Dashed lines represent the deleted regions in the two lines. (B) Heading date surveys for 600-del and 200-del. The mutant line P3 to P4 was selected from the G-box-like mutant survey as a negative control line, for the line was confirmed to flower as early as NB in the pre-experiment. Error bars are drawn using SD. n = 5, except P3 to P4 (n = 3). NB and KO represent control WT plant Nipponbare and a Ghd7-KO line (KO-63), respectively. Asterisks (**) indicate values significantly different from NB (P < 0.01, Welch’s t test). (C) Plant photos of the 600-del and 200-del lines at 105 d after germination. All plants present in the picture were grown in an LD growth chamber (14.5 h Light, 9.5 h Dark). (D) Diurnal expression analysis of Ghd7 in 600-del and 200-del under long day conditions. Fully expanded leaves from 5-wk-old plants were sampled at ZT −2, 2, 6, 10, 14, and 18. X-axis values are offset for visual clarity, without altering the actual sampling times. Error bars indicate SD (n = 3 or 4). Filled circles: Normal data points; Hollow squares: Significant difference from NB (P < 0.05, Welch’s t test).

To clarify the roles of the identified cis-regulatory regions in Ghd7 transcription, we further performed RNA-seq on NB, −26/−30 K, 200-del, and the phytochrome B (phyB) mutant. In this experiment, although Ghd7 expression peaks are normally observed at ZT2, fully emerged leaves were sampled from 3-wk-old plants at ZT−8 under SD. This is mainly because the expression of Ghd7 downstream genes, such as Ehd1, Hd3a, and RFT1, peaks between ZT−9 and ZT−1 at night (4). In addition, previous studies have shown that Ghd7 repressor activity was observed with a time lag following Ghd7 transcription (2). We also considered that transcription factors which regulate Ghd7 are likely expressed at this time point, since Ghd7 is most sensitive to night-break treatment around ZT−8. Therefore, this sampling time allows for simultaneous monitoring of both upstream and downstream genes of Ghd7. Differentially expressed genes (DEGs) between NB and other samples were then identified. A volcano plot was first generated to examine whether the expression was changed for known upstream regulators of Ghd7, such as Phytochromes and ELF3; downstream genes of Ghd7, including Ehd1, Hd3a, and RFT1; and other related circadian clock and flowering-related genes, such as GI, PRR37, and Ghd8/Hd5 (SI Appendix, Fig. S3 C–E). In both −26/−30 K and 200-del, no expression change was observed in known upstream regulators. In contrast, the downstream genes such as Ehd1 and RFT1 showed increased expression in both mutants, and expression changes were also detected in MADS14, a downstream gene of florigen signaling. A similar trend was observed in the phyB mutant, with Ehd1, Hd3a, RFT1, and MADS14 identified as upregulated genes. A total of 1,428 DEGs were identified in 200-del, while 4,783 DEGs were detected in −26/−30 K, with 1,128 genes overlapping between the two (SI Appendix, Fig. S3F and Datasets S1–S3). Among the 1,128 overlapping genes, 162 were also detected in phyB mutants. Interestingly, Ghd8/Hd5 was identified as a DEG in this analysis.

Impact of Ghd7 Promoter Deletion on Plant Growth and Development Traits.

In addition to influencing heading dates, Ghd7 has been shown to affect grain number, grain size, stem growth, and plant development (1, 15, 16, 43, 44). To examine such traits in −20/−40 K and −26/−30 K, we grew rice plants with increased spacing than in the heading date surveys. Plants were grown in a greenhouse under LD conditions, with natural sunlight and supplemental lighting (light periods exceeding 13.5 h for most of the growing season). Plant heights and leaf emergence rates were measured weekly. The KO64 KO line showed a slightly faster leaf emergence rate than NB up to flowering, while other lines exhibited similar rates to NB, indicating no differences in developmental stages (SI Appendix, Fig. S2C). The plant heights in KO and −20/−40 K lines tended to be lower than NB before heading and finally 10 to 20 cm lower than NB (SI Appendix, Fig. S2 A and C). Shoot weights in KO, −20/−40 K, and −26/−30 K lines were significantly lower than NB. Although the number of spikelet tended to be low in smaller plants (−20/−40 K), the variations were large, and no significant difference was observed in the other lines compared to NB (SI Appendix, Fig. S2D). When these plants were grown with similar spacing in a LD growth chamber (SI Appendix, Fig. S2E), they showed similar flowering times and growth traits to those grown in natural light, with significantly earlier heading in KO, −20/−40 K, −26/−30 K, and 0/−3 K and significantly lower shoot weights than NB. Spikelet numbers of the first ear ranged from 40 to 60 in the growth chamber, which was significantly fewer than the ~120 spikelets per panicle typically observed in field-grown plants (45). The present results suggest that there may be cis-regulatory regions, which regulate the expression of Ghd7 at times and tissues that affect plant size, from −20 kb to −40 kb. Shoot weight of −26/−30 K was significantly lower than that of NB, but plant height was not different from NB (SI Appendix, Fig. S2 D and E). There may be a cis-regulatory region, which regulates Ghd7 expression for plant height, from −20 kb to −26 kb, and −30 kb to −40 kb, although further validation will be needed since only one line of −26/−30 K was obtained.

Discussion

Large Deletions By Genome Editing in the Rice Ghd7 Promoter Region.

By using genome editing to introduce large genomic deletions ranging from 0.2 kb to as large as 25 kb, we were able to examine the cis-regulatory landscape of Ghd7 in rice across the entire 65-kb upstream region of its promoter. This approach enabled a comprehensive scan of this extended region, reaching up to the neighboring gene, to identify critical CREs. As a result, we narrowed down the core cis-regulatory region responsible for morning induction of Ghd7 to a 228-bp segment located 28 kb upstream of its TSS.

Numerous studies have reported deletions induced by genome editing in plants, with typical sizes ranging from tens to hundreds of base pairs and efficiencies varying between 2 and 85% (46). In the present experiment, even the lines with high deletion frequency exhibited deletion efficiencies of only 10 to 20% (SI Appendix, Table S1), which were lower than those observed for short deletions. Contrary to expectations, larger deletions did not necessarily result in lower deletion efficiencies. For example, the −40/−65 K line, which had the largest deletion of 25 kb, yielded 10 plants with deletions out of 96 regenerated plants, representing relatively high deletion efficiency. Mikami et al. indicated that using a combination of high-performance guide RNAs is important for achieving efficient deletions between two target sites (46). While our results do not demonstrate a clear relationship between guide RNA activity and deletion efficiency, they suggest that factors other than deletion length, such as variation in guide RNA performance or differences in chromatin structure at the target site, may be involved.

Identification of a Genomic Region Involved in Ghd7 Induction in the Morning.

First, we found that genome-edited rice lines with a deletion from −20 to −40 kb (−20/−40 K) flowered earlier than NB under LD conditions (Fig. 1B). The line with a 3.7-kb deletion from −26,448 to −30,138 bp (−26/−30 K) also exhibited early flowering (Fig. 4B). Ghd7 expression at ZT2 was significantly reduced in both −20/−40 and −26/−30 K under LD and SD conditions compared to NB (Figs. 2A and 4C). This reduced morning expression of Ghd7 in the mutants was followed by higher expression of Ehd1, Hd3a, and RFT1, leading to early flowering (Figs. 2B and 4D). Within the deleted 3.7-kb segment in −26/−30 K, we finally identified a 228-bp genomic region, located as far as −28 kb from the Ghd7 TSS, as a core cis-regulatory region responsible for ZT2 induction. This study provides direct genetic evidence for the existence of distal cis-regulatory regions in plants through in vivo promoter deletion mutants (SI Appendix, Fig. S4). Our findings suggest that specific CREs are crucial for ZT2 induction, while the distance of these CREs from the TSS appears less important, as several deletion lines (−0.5/−3, −3/−5, −5/−10, and −10/−20 K) showed no substantial effect on flowering time.

The 228-bp Region May Play Roles in Both the Induction and Suppression of Ghd7.

We used PlantCARE (47) to predict which types of TFs could interact with the 228 bp and identify possible cis-binding TFs (SI Appendix, Table S5). The “CAAT box” is a well-known binding site for NF-Y TFs, and in Arabidopsis, the CO protein has been reported to form a complex with some NF-YBs and NF-YCs (48, 49). A rice NF-YB gene may control Ghd7 transcription (50), and the floral repressor Ghd8/Hd5 gene, which reportedly forms protein complex with Ghd7 and Hd1 to control Ehd1 transcription, encodes another NF-YB protein (6). Therefore, these are potential targets as the cis-regulatory site for the ZT2 induction of Ghd7. In addition, genome-wide profiling of histone modifications in rice has been recently performed using Chromatin Immunoprecipitation followed by Sequencing (ChIP-seq). With these data, we analyzed the upstream genomic region of Ghd7 (SI Appendix, Fig. S5). Deng et al. revealed a diurnal pattern of histone modifications in rice by ChIP-seq (51). The H3K27ac and H3K9ac peaks, which are associated with enhancer regions or transcriptional activation, were observed within the 3.7-kb region from −26 to −30 kb between 8:00 and 12:00 (SI Appendix, Fig. S5A). In contrast, the H3K27me3 peak, linked to transcriptional repression, was observed in the 3.7-kb region across all time points. Zhao et al. reported histone modification patterns in various rice tissues (52), finding that H3K27ac and H3K27me3 peaks were present in the 3.7-kb region both in leaves, where Ghd7 expression is high, and in roots, where Ghd7 expression is low (SI Appendix, Fig. S5B). The presence of histone H3 acetylation in the 3.7-kb region supports its functional role on the diurnal expression of Ghd7, while the function of relatively constant and weak histone H3 methylation remains unclear. Recently, OsELF3-1 and OsELF3-2 proteins were found to interact with the phyB protein and repress Ghd7 expression (18, 20). According to these reports, suppression or KO of OsELF3-1/OsELF3-2 led to derepression of Ghd7, resulting in delayed flowering. ChIP-seq data showed that an OsELF3-1 binding peak was present in the 3.7-kb region and overlapped with the 228-bp region by 161 bp (20). Deletion of the 3.7-kb region and the 228-bp region, where the repressor OsELF3-1 binds, would be expected to increase Ghd7 expression and delay flowering. However, deletion of this region instead resulted in a reduced increase in Ghd7 expression in the morning and led to early flowering (Fig. 4 B and C). Based on our results, particularly the early flowering observed in the 200-del line, it is plausible that the 228-bp region contains two distinct CREs: one primarily responsible for the induction of Ghd7 transcription via phytochrome signaling and another involved in repression through the interaction with OsELF3s or the rice Evening Complex. Alternatively, these repressive CREs may reside in regions flanking the 228-bp segment.

Since Ghd7 is complexly regulated by the circadian clock and light, multiple CREs are likely involved. Our comprehensive scan of the 65-kb upstream region of Ghd7 showed that regions beyond the 3.7-kb region did not substantially affect heading time, suggesting that key CREs are concentrated within this 3.7-kb region. Additional CREs may function in coordination with this region, with the four G-box-like elements, which were not examined in this study, remaining potential candidates for Ghd7 regulation.

The RNA-seq results are also consistent with our findings. In both −26/−30 K and 200-del, Ehd1 and florigen genes were found to be upregulated, and a large proportion of DEGs (1,128 out of 1,428) identified in 200-del overlapped with those in −26/−30 K. This suggests that the effect of the 200-del mutation is encompassed within the effect of the −26/−30 K deletion. On the other hand, approximately 3,700 genes exhibited expression changes specifically in −26/−30 K but not in 200-del. This could indicate that the −26/−30 K region plays additional regulatory roles and may be involved in the pleiotropic functions of Ghd7. According to Datasets S1–S3, the detection of a greater number of genes with low-expression and high fold-change values in −26/−30 K resulted in an increased number of genes as DEGs. Thus, further analysis is waited for these low-expression genes. In addition, our analysis identified Ghd8/Hd5 as an upregulated DEG. Given that the 228-bp region contains a CAAT-box, the known binding motif for Ghd8/Hd5, this raises the possibility that Ghd7 and Ghd8 are involved in a mutual feedback regulation mechanism.

The Region Around the TSS Does Not Contain cis-regulatory Elements Responsible for Light Responsiveness or Diurnal Fluctuation.

In the 0/−3 K deletion line, which lacks the promoter region up to −3 kb but retains the −29 to −86 bp short segment, Ghd7 expression remained largely comparable to that in the wild type (NB), with only moderate reductions observed at certain time points (Figs. 2A and 3B). These findings suggest that the proximal −3 kb region, including the TSS, is not essential for light or circadian regulation of Ghd7. In cultivars ‘Aikoku’ and ‘Yukara’, where a 1.9 kb transposon is inserted between the retained −29 to −86 bp region and the TSS, Ghd7 expression is significantly reduced. One possibility is that this −29 to −86 bp segment serves as a basal CRE, and that its function is impaired when spatially separated from the TSS. Alternatively, the inserted transposon may carry repressive elements that interfere with transcription. Phenotypically, the 0/−3 K line exhibited earlier heading than NB under optimal temperature conditions (Fig. 1B and SI Appendix, Fig. S2E), but this was not consistently observed under altered environmental conditions (SI Appendix, Fig. S1). Taken together, these data indicate that the regulatory influence of the 0/−3 kb region may be environmental-dependent, and its role in Ghd7 regulation and heading date control remains to be clarified.

Materials and Methods

Plant Materials and Growth Conditions.

The japonica rice (Oryza sativa ssp. japonica ‘Nipponbare’; NB) was used as a non-genome-edited rice plant. For Fig. 3, two additional japonica rice cultivars, ‘Aikoku’ and ‘Yukara’, were also used as non-genome-edited rice plants. Plants were grown under SD (10 h of light at 28 °C and 14 h of dark at 24 °C) or LD (14.5 h of light at 28 °C and 9.5 h of dark at 24 °C) conditions for Figs. 1–5 and SI Appendix, Figs. S2 A–E and S3A. For Fig. 6 and SI Appendix, Figs. S1 and S3, four distinct growth chambers were used. The photoperiod and temperature conditions in these chambers were as follows: SD/High_SD: 10 h of light and 14 h of dark, daytime temperature at 30 °C & nighttime at 26 °C; LD/High_LD: 14.5 h of light and 9.5 h of dark, daytime temperature at 30 °C and nighttime at 26 °C; Low_SD: 10 h of light & 14 h of dark, daytime temperature at 26 °C and nighttime at 17 °C; and Low_LD: 14.5 h of light and 9.5 h of dark, daytime temperature at 26 °C and nighttime at 17 °C. Red light was provided by LEDs (5 to 8 μmol m⁻² s⁻¹; SANYO). For the performance experiment in SI Appendix, Fig. S2 B–D, rice seeds were sown on May 5, 2021, and seedlings were transplanted into a greenhouse on June 4, 2021, in Tsukuba City, Ibaraki Prefecture, Japan (35°N, 140°E), under natural light with supplemental lighting from 16:00 to 19:00. The temperature was maintained at 30 °C during the day and at 24 °C during the night. Under these conditions, the day length exceeded 13.5 h until September 27, when most of the plants had headed. Three plants were planted in each 16 cm diameter pot. The photosynthetic photon flux density in the greenhouse was 300 to 900 μmol m⁻² s⁻¹ on a sunny day and 200 to 500 μmol m⁻² s⁻¹ on a cloudy day. Flowering time (heading date) was defined as the time when the first panicle emerged from the flag leaves.

Construction of CRISPR/Cas9 Lines.

The rice cultivar Nipponbare was used to create the CRISPR/Cas9-mediated deletion lines. The guide RNA sequences were selected with the assistance of CRISPR-P (http://CRISPR.hzau.edu.cn/CRISPR/) or CRISPR-Direct (https://crispr.dbcls.jp/) (SI Appendix, Table S2).

The forward (F) and the reverse (R) oligonucleotides were annealed and cloned into the BbsI site of pU6gRNA vector (53). The guide RNA for Ghd7 KO was designed at the first exon of Ghd7. For the Ghd7 KO construct, the sgRNA expression construct consisting of OsU6pro:sgRNA:polyT were transferred into the binary vector pZH_OsU3gYSA_MMCas9 (53) harboring SpCas9 and hygromycin phosphotransferase gene (HPT) expression constructs using AscI and PacI sites. For the upstream deletion constructs, two to six sgRNA expression constructs were connected tandemly in pU6gRNA using AscI, EcoRV, PvuII sites. For 0/−3 K, −0.5/−3 K, −3/−5 K, −5/−10 K, and −10/−20 K, a total of six sgRNA constructs were ligated, with three constructs near each end of the deletion. For −20/−40 K, three sgRNA constructs around −20 kb and one around −40 kb were connected. For −40/−65 K, one sgRNA construct around −40 kb and one sgRNA construct around −65 kb were connected. The sgRNA expression constructs containing two to six OsU6pro:sgRNA:polyT were transferred into the binary vector pZH_OsU3gYSA_MMCas9. To obtain mutant plants with defective candidate cis-motifs within the −26/−30 K region, 8 guide RNAs targeting G-box or G-box-like motifs were designed. These guide RNAs were cloned into the pU6gRNA vector, and each sgRNA construct was introduced into the PacI and AscI-digested binary vector pPZP_ZmUBI-SpCas9-NG (41), which contains an artificially modified Cas9 that recognizes NG rather than NGG as a PAM. Rice transformation with the CRISPR/Cas9 construct was performed using Agrobacterium tumefaciens strain EHA105 as described previously (54). Transformed calli were identified by hygromycin screening, and selected calli were used to regenerate transgenic plants. Mutations and deletions of the regenerated transgenic plants were identified by PCR followed by Sanger sequencing using a pair of primers described in SI Appendix, Table S2. The deletion of −26/−30 K was confirmed by PromethION (Nanopore, https://nanoporetech.com/products/promethion). Genomic DNA was extracted using the NucleoBond HMW DNA extraction kit (Macherey-Nagel, Düren, Germany), and the library was generated by the Ligation Sequencing Kit (SQK-LSK109). The sequencing was performed by PromethION flow cell R9.4.1. and base calling was performed using Guppy 360 at High Accuracy Mode. Reads were mapped by -x map-ont of minimap2-2.18, and the deletion was confirmed by Integrated Genome Viewer (IGV). The number of regenerated plants, the number of regenerated plants whose deletions were confirmed by Sanger sequencing were described in SI Appendix, Table S1. Deleted and/or mutated regions of plants used in the present experiments were described in SI Appendix, Table S3.

RT-PCR.

For diurnal gene expression analyses in Figs. 2A and 4C, fully emerged leaves of 5-wk-old plants grown under LD and SD were harvested at ZT−2, 2, 6, 10, 14, and 18. For Figs. 2B and Fig. 4D, 7-wk-old plants grown under LD were harvested at ZT2. For Fig. 3B, 4-wk-old plants grown under LD were harvested at ZT0, 2, 6, 10, 14, and 19. For Fig. 6D, 5-wk-old plants grown under LD were harvested at ZT−2, 2, 6, 10, 14, and 18. For Fig. 5, plants were grown under LD and SD for 18 d and 22 d, respectively, and harvested 2 h after irradiation with red light. Leaves of two to four plants were mixed as one sample. Two or three samples for Figs. 2A, 3B and 4C, and 4 or 5 samples for Figs. 2B, 4D, and 6D were used for qRT-PCR as biological replicates. Total RNA was extracted from rice leaves using Sepasol-RNA I Super G (Nacalai tesque, Kyoto, Japan) or TRIzol Reagent (Invitrogen, Thermo Fisher Scientific), in accordance with the manufacturer’s instructions. Complementary DNA was then synthesized using 0.5 μg of total RNA and the ReverTra Ace qPCR RT Master Mix with gDNA Remover (TOYOBO, Osaka, Japan). RT-qPCR was performed using TaKaRa SYBR Premix Ex Taq II (Tli RNaseH Plus) (TaKaRa; https://www.takara-bio.com) for Figs. 2–5 in accordance with the manufacturer’s instructions. For Fig. 6D, RT-qPCR was performed using FAM probes and TaqMan Fast Universal PCR Master Mix (2×) No AmpErase UNG (Applied Biosystems, Thermo Fisher Scientific). Gene-specific primers and probes are listed in SI Appendix, Table S4. A rice ubiquitin gene (RUBQ2, Os02g0161900) was used for normalization.

RNA-Seq Analysis.

Total RNA was extracted from the youngest fully emerged leaves of NB, −26/−30 K, 200-del, and phyB plants grown under SD for 3 wk at ZT−8 using the ISOSPIN Plant RNA Kit (Nippon Gene, Tokyo, Japan). On-column DNase treatment was performed according to the manufacturer’s instructions.

RNA extracted from two independent plants per line underwent poly(A) purification using the NEBNext® Poly(A) mRNA Magnetic Isolation Module (New England Biolabs, MA). Using the poly(A)-purified samples, RNA-seq libraries were prepared with the VAHTS Universal V8 RNA-seq Library Prep Kit for Illumina (Vazyme, Nanjing, China) following the manufacturer’s instructions.

Sequencing was conducted by Macrogen Japan on a NovaSeq X Plus platform using a 150-bp paired-end configuration. Quality control and adapter trimming were performed using fastp (version 0.23.4) with the following parameters: --detect_adapter_for_pe --cut_front --cut_tail --cut_window_size 4 --cut_mean_quality 20 -l 36 -u 30. The trimmed reads were further quality-assessed using FastQC (version 0.12.1).

High-quality reads were aligned to the O. sativa ssp. japonica Nipponbare IRGSP-1.0 reference genome using HISAT2 (version 2.2.1). The gene annotation used for alignment was IRGSP-1.0 (version 2024-07-12), obtained from the Rice Annotation Project Database (RAP-DB, http://rapdb.dna.affrc.go.jp). SAMtools (version 1.19) was used to convert SAM files to BAM format and create index files. Mapping quality was assessed using SAMtools flagstat.

Read assignment to transcript features was performed using FeatureCounts from the Subread package (version 2.0.6). DEGs were identified using DESeq2 (version 1.40.2), with a false discovery rate (FDR) of less than 1% and |log₂(fold change)| > 1. Volcano plots for −26/−30 K, 200-del, and phyB, along with a Venn diagram, were generated in R based on the identified DEGs.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Data, Materials, and Software Availability

The RNA-seq data have been deposited in the Sequence Read Archive (SRA) database (https://www.ncbi.nlm.nih.gov/sra/) under the accession number PRJNA1273339. ChIP-seq data used in SI Appendix, Fig. S5 have been deposited in the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) by Deng et al. (51; GSE143724) and Zhao et al. (24; GSE142570). All other data are included in the manuscript and/or supporting information.