RLB (RICE LATERAL BRANCH) recruits PRC2-mediated H3K27 tri-methylation on OsCKX4 to regulate lateral branching

Abstract

Lateral branches such as shoot and panicle are determining factors and target traits for rice (Oryza sativa L.) yield improvement. Cytokinin promotes rice lateral branching; however, the mechanism underlying the fine-tuning of cytokinin homeostasis in rice branching remains largely unknown. Here, we report the map-based cloning of RICE LATERAL BRANCH (RLB) encoding a nuclear-localized, KNOX-type homeobox protein from a rice cytokinin-deficient mutant showing more tillers, sparser panicles, defected floret morphology as well as attenuated shoot regeneration from callus. RLB directly binds to the promoter and represses the transcription of OsCKX4, a cytokinin oxidase gene with high abundance in panicle branch meristem. OsCKX4 over-expression lines phenocopied rlb, which showed upregulated OsCKX4 levels. Meanwhile, RLB physically binds to Polycomb repressive complex 2 (PRC2) components OsEMF2b and co-localized with H3K27me3, a suppressing histone modification mediated by PRC2, in the OsCKX4 promoter. We proposed that RLB recruits PRC2 to the OsCKX4 promoter to epigenetically repress its transcription, which suppresses the catabolism of cytokinin, thereby promoting rice lateral branching. Moreover, antisense inhibition of OsCKX4 under the LOG promoter successfully increased panicle size and spikelet number per plant without affecting other major agronomic traits. This study provides insight into cytokinin homeostasis, lateral branching in plants, and also promising target genes for rice genetic improvement.

The nuclear-localized, KNOX-type homeobox protein RICE LATERAL BRANCH fine-tunes cytokinin levels to regulate lateral branching.

Introduction

Plants evolved with branches to compete for more resources such as nutrients, sunlight, and living space. Rice (Oryza sativa L.) is one of the most important crops in China and serves as a staple food for over 60% of the population (Fitzgerald et al., 2009). In rice production, tillers and panicle branches enable plants to bear more grains, which essentially contribute to the rice yield (Wang and Li, 2011).

The branches of rice are composed of tillers initiated from axillary meristems (AMs) and inflorescence branches from branch meristems (BMs; Du et al., 2017). Upon entering the reproductive growth stage, panicle differentiation is initiated immediately under appropriate environmental conditions. During this process, shoot apical meristem (SAM) firstly transforms into inflorescence meristem (IM). Then, primary branch meristem (PBM) is formed at the base of IM, while secondary branches meristem (SBM) is differentiated from the base of the primary branches, and finally spikelet meristem (SPM) forms at the top of the primary and secondary branches (Ikeda et al., 2004).

In recent years, tremendous progress has been made to illustrate the molecular mechanisms underlying rice lateral branching. Transcription factors monoculm 1 (MOC1) and monoculm 3 (MOC3) are crucial regulators for tiller bud formation (Li et al., 2003; Lu et al., 2015). MOC1 interacts with MOC3 to activate FLORAL ORGAN NUMBER1 transcription, thus to positively regulate the elongation of tillering buds (Shao et al., 2019). IDEAL PLANT ARCHITECTURE1 encodes a promoter binding protein OsSPL14 under the governing of microRNA OsmiR156. OsSPL14 has been found to control tillering and panicle branching in an opposite manner. Mutation of OsSPL14 resulted in reduced tillering, which was achieved by directly suppressing OsTB1, a negative growth regulator of the of tiller side buds (Jiao et al., 2010; Miura et al., 2010). However, OsSPL14 positively regulates rice panicle length by directly activating DENSE PANICLE 1 transcription (Lu et al., 2013). Up to date, numerous key genes controlling panicle morphology of rice have been documented as well. For examples, RFL (LEAFY) gene, which is highly expressed in AM on tiller buds, panicles, and branches, has been known as a master regulator in the transformation from rice vegetative growth to panicle meristem and inflorescence formation. Knockdown of RFL significantly decreased the number of tillers and panicle branches (Rao et al., 2008). Another positive regulator ABERRANT PANICLE ORGANIZATION 1 (APO1) was found to promote branching and spikelet formation when it was over-expressed, and apo1 mutants show smaller panicles with decreased branch number and abnormal flower morphology (Ikeda et al., 2007). LAX1 (Lax panicle 1) is involved in the initiation and maintenance of panicle AM by deciding the formation of primordium in rice axillary buds (Komatsu et al., 2001). lax1-2 exhibited severely impeded initiation and maintenance of branches, lateral spikelets and terminal spikelets (Komatsu et al., 2001). A similar function was reported on LAX2 (Lax panicle 2), of which the mutant displayed sparse panicle, while lax1lax2 had more severe phenotype than each of the single mutant (Tabuchi et al., 2011). Kobayashi et al (2012) suggested that panicle phytomer2/OsMADS34, OsMADS14, OsMADS15, and OsMADS18 synergistically determine the morphology formation of IM (Kobayashi et al., 2012).

Plant branching is coordinated by various phytohormones including auxin, cytokinin, and strigolactone, among which adenine-derived phytohormone cytokinin has been implicated as a major promoter of bud outgrowth (Barbier et al., 2019). Exogenous application of cytokinin either on the axillary buds or adjacent regions boosts bud outgrowth in a dose-dependent manner (Sachs and Thimann, 1967; Dun et al., 2012; Barbier et al., 2019). In addition, it is also demonstrated that the cytokinin level corresponds to shoot branching by using various genetic mutants of cytokinin synthesis and signaling genes (Muller et al., 2015). The nucleobase forms of cytokinin, including trans-, cis-, and dihydro-zeatin (Z) as well as isopentenyl adenine (iP), are the active forms that could be recognized by cytokinin receptors (Jameson and Song, 2016). Like many other hormones, cytokinin levels are orchestrated by a balance of biosynthesis, activation, inactivation, re-activation, and degradation. The first step of cytokinin biosynthesis is catalyzed by ATP/ADP isopentenyl transferases, of which the high-order mutants in Arabidopsis have reduced cytokinin level and smaller IM size (Miyawaki et al., 2006). In rice, LONELY GUY (LOG) encoding a phosphoribohydrolase enzyme has been found to catalyze the final step of cytokinin biosynthesis by converting inactive cytokinin nucleotides into active free-base forms. Loss-of-function of LOG resulted in early termination of BMs and IMs with reduced panicle branching complexity. It was proposed that LOG specifically localizes in shoot meristem tips to fine-tune the concentrations and the spatial distribution of bioactive cytokinins, representing a significant mechanism for the maintenance of rice BM activity (Kurakawa et al., 2007). Similar phenotypes were also observed on the triple log3/4/7 and septuple log1/2/3/4/5/7/8 mutants in Arabidopsis, which suggested the key roles of cytokinin activation enzymes in plant branching (Kuroha et al., 2009; Tokunaga et al., 2012). In contrast, cytokinin oxidases (CKXs) act as key enzymes in the turn-over of cytokinin through cleaving their N6 side chains, thus to modulates rice branching by altering endogenous cytokinin levels (Galuszka et al., 2000). Elevated cytokinin catabolism usually leads to simplified plant branches. For examples, over-expression of CKXs in Arabidopsis and Nicotiana benthamiana substantially decreased the cytokinin level and flower numbers (Werner et al., 2001; Werner and Schmulling, 2009), whereas rice varieties with lower CKX2 level produced larger panicles with more branches and spikelets (Ashikari et al., 2005; Li et al., 2013). Additionally, zeatin O-glucosyl transferases controlling cytokinin reversible inactivation, β-glucosidases regulating cytokinin reactivation, UDP glycosyltransferases responsible for cytokinin irreversible N-glycosylation as well as cytokinin signaling components have also been implicated in SAM initiation and size determination (Skylar and Wu, 2011; Han et al., 2014; Azizi et al., 2015).

Despite the full awareness of the importance of cytokinin homeostasis on plant branching, the underlying mechanism by which the cytokinin biosynthesis and catabolism are fine-tuned in lateral branching remains largely unclear. In this study, a KNOX type homeobox gene RLB was map-based cloned from a mutant showing typical cytokinin deficiency phenotypes such as increased tillers and sparser panicles. We revealed that rice lateral branching is controlled by an “RLB-EMF2b-OsCKX4” module, through which RLB recruits Polycomb repressive complex 2 (PRC2) member EMF2b to OsCKX4 to epigenetically repress its transcription, which suppresses the catabolism of cytokinin, thereby promoting rice lateral branching.

Results

rlb is a cytokinin-deficient mutant with defective lateral branch development

By screening a T-DNA insertional mutant library in the background of ZH11 (O. sativa, ssp. japonica), we identified a recessive mutant rice lateral branch (rlb) with abnormal lateral branch phenotypes. During vegetative growth, rlb was dwarf and branchy with 72% more effective tillers than ZH11 (Figure 1, A and B; Table 1). In contrast to the branchy tillers, rlb displayed shorter and sparser panicles with severely reduced spikelets per panicle, largely owing to the decreased secondary branch numbers in the panicle (Figure 1, C and D; Table 1). We manually dissected the young panicles for the observation under microscopy, and found that rlb had much less differentiated spikelet primordium than ZH11 (Figure 1, E and F). Paraffin section and scanning electronic microscopy (SEM) analysis also suggested that rlb panicle differentiation was retarded and had less spikelet primordium when compared with ZH11 (Figure 1, G–J). In addition to the defected branches in tiller and panicle, abnormal phenotypes were also observed in florets and grains. It was observed that 84.5% ± 0.04% of the rlb florets had only one sterile lemma, instead of two sterile lemmas as observed in ZH11, suggesting the key roles of RLB in lateral organ development in florets (Figure 1, K and L; Supplemental Figure S1C). In comparison to the normal seed setting (86.7% ± 0.02%) of the ZH11, only 37.7% ± 0.04% of the rlb spikelets had fully matured seed set (Table 1), though both lines showed similar pollen viability as revealed by the I₂-KI staining assay (Supplemental Figure S1, A and B). By dissecting off the grain hulls, we found that the majority of the rlb unset seeds displayed expanded maternal integuments, suggesting that the seed abortion is likely caused by poor grain filling (Supplemental Figure S1D). Indeed, this speculation is confirmed with the observation of very less nutrient substances accumulated in the rlb embryo sacs during grain filling stages (Supplemental Figure S1E). In terms of the fully matured seeds, the width was substantially reduced in rlb when compared with the ZH11, implying that RLB may be involved in grain fillings (Supplemental Figure S1G).

Figure 1.

Phenotypes of ZH11 and rlb. A, Plants morphology of ZH11 and rlb after heading, bar = 10 cm. B, Tillers from three plants of ZH11 and rlb at booting stage, bar = 5 cm. C, Main panicle of ZH11 and rlb, red arrows indicate the neck-panicle node, bar = 5 cm. D, Primary branch of ZH11 and rlb, red arrows indicate the degenerate floret points from the secondary branch of rlb, bar = 2.5 cm. E and F, A single primary branch of ZH11 and rlb, bar = 100 µm. G and H, Paraffin section of ZH11 and rlb at panicle initiation stage, bar = 500 µm. I and J, SEM analysis of ZH11 and rlb panicle at early differentiation stage, the red triangles indicate floret meristems, bar = 500 µm. K and L, Morphology of florets and seeds of ZH11 and rlb, red arrows indicate the sterile lemma, bar = 1 cm. M–O, Callus regeneration of ZH11 and rlb. M, basic medium was used. N, enlargement of the red frame from M. O, basic medium with the addition of 2.0 mg L⁻¹ zeatin was used. P, Regeneration ratios of ZH11 and rlb on the basic medium with or without the addition of zeatin, 50 independent calluses were used for each treatment. Different characters represent statistical difference at P < 0.05 by Student’s t test. Q and R, Endogenous hormone levels of ZH11 and rlb. TZR, trans-zeatin-riboside. IPA, isopentenyl adenosine. IP, N6-(2-Isopentenyl) adenosine. Data are shown as means ± sd (n = 3). **P ≤ 0.01 by the Student’s t test.

Table 1.

Agronomic traits of the genetic materials

Material	Plant height (cm)	Panicle length (cm)	Effective tillers	Primary branches/panicle	Secondary branches/panicle	Spikelets/panicle	Filled grains/panicle	Seed-setting (%)
ZH11a	99.6 ± 1.8	25.3 ± 0.2	7.8 ± 0.4	12.2 ± 0.4	32.4 ± 3.4	209.2 ± 16.0	181.2 ± 9.4	86.7 ± 0.02
rlb	58.2 ± 1.5**	18.3 ± 1.0**	13.4 ± 1.1**	11.8 ± 0.8	5.6 ± 2.7**	77.0 ± 5.4**	29.2 ± 5.1**	37.7 ± 0.04**
NIP	96.7 ± 1.5	21.0 ± 1.3	13.7 ± 1.5	10.7 ± 1.2	18.7 ± 3.2	117.0 ± 6.0	104.0 ± 9.2	88.8 ± 3.4
oxRLB-6	76.0 ± 3.6**	19.4 ± 1.5	17 ± 1.0*	8.3 ± 0.6	10.3 ± 3.5*	76.0 ± 13.5*	62.7 ± 5.5**	83.4 ± 9.0
oxRLB-9	79.0 ± 5.3*	18.1 ± 1.1*	18.3 ± 2.1*	11.0 ± 1.0	9.0 ± 1.0*	96.0 ± 7.0*	81.6 ± 6.8*	85.1 ± 4.7
ZH11b	114.3 ± 4.7	23.0 ± 2	9.3 ± 0.6	12.3 ± 0.6	26.7 ± 4.2	151.3 ± 15.0	131.0 ± 6.6	86.9 ± 6.0
proLOG::ackx4-2	110.3 ± 2.1	26.5 ± 1.3	9.7 ± 1.2	14.3 ± 0.6*	47.3 ± 8.1*	234.7 ± 37.0	194.6 ± 28.0	83.1 ± 3.5
proLOG::ackx4-3	102.0 ± 8.7	25.6 ± 2.2	10.0 ± 2.6	13.7 ± 1.2	40.3 ± 4.2*	211.3 ± 17.2*	185.3 ± 28.1	87.4 ± 6.8
oxckx4-15	78.3 ± 4.7**	14.1 ± 0.7*	13.3 ± 4.0	8.0 ± 1.0**	1.7 ± 1.5**	57.3 ± 16.7**	33.0 ± 7.8**	58.2 ± 4.7**
oxckx4-21	84.0 ± 7.9**	17.3 ± 1.6*	12.0 ± 1.0*	9.3 ± 1.5	10.0 ± 3.6**	71.7 ± 21.0**	47.0 ± 7.9**	69.7 ± 22.5

^*Significant difference at P ≤ 0.05;

^**Significant difference at P ≤ 0.01, by Student’s t test, respectively (n = 5).

^a Represent control special for rlb.

^b Represent control for the following four genetic materials.

During the tissue culture of rlb for genetic transformation, we noticed that the mutant callus was stagnated in the status with multiple bud structures, which is a typical phenotype of cytokinin-deficiency as previously reported (Figure 1, M and N;Howell et al., 2003), but could barely regenerate into shoots as observed in ZH11. Therefore, we applied exogenous zeatin in various concentrations in the regeneration medium, and found that 0.5 mg L⁻¹ or 2.0 mg L⁻¹ zeatin drastically restored the regeneration ability of rlb (Figure 1, O and P). Phytohormones quantification assay revealed that cytokinin components, including TZR, zeatin, and IPA, were significantly reduced in rlb, while IP was slightly increased in the mutant (Figure 1Q). Additionally, the indole-3-acetic acid (IAA) and giberelic acid (GA3) levels remain unchanged (Figure 1R). The above results indicated that rlb is a cytokinin-deficient mutant with defective lateral branch development.

RLB is an allele of homeobox gene Osh15

Since the mutant phenotype is not co-segregated with the T-DNA insertion, we adopted a map-based cloning strategy to identify the corresponding gene. In the genetic population derived from the cross of rlb and TN1 (O. sativa ssp indica), all the F₁ plants grew normally, and the numbers of the normal and mutant lines in F₂ followed a 3:1 ratio (χ² = 2.54 < χ²_0.05 = 3.84, n = 278), suggesting that the rlb phenotype is controlled by a single recessive Mendel locus. The gene was primarily mapped in the short arm of chromosome 7, and finally delimited to a 44.2-kb region between molecular markers Ind8 and Ind14, which harbors seven candidate genes according to the gene annotations in Gramene database (http://gramene.org/;Figure 2A). Sanger sequencing revealed the occurrence of a 24-bp deletion on the fourth exon of ORF1 (LOC_Os07g03770), which potentially lead to an eight amino acids (LKFQLLKK) deletion within the ELK domain of the resulting protein (Figure 2, B and C). To check the effects of mutation on the transcription of RLB, two sets of primers covering different regions of RLB were developed for the reverse transcription quantitative real time PCR (RT-qPCR) experiment. Primer R1 was designed on the 5′ side of the deletion, while primer R2 locates exactly in the deleted region (Figure 2B). As indicated by the primer set F1 and R2, the mutation drastically reduced the expression of the intact RLB transcript. Nevertheless, rlb displayed increased mRNA abundance of the truncated transcript of RLB, suggesting that RLB is governed by a feedback regulation mechanism (Figure 2D). Furthermore, a RLB antibody was synthesized to check the RLB protein level in the mutant and ZH11. The immunoblot showed a clear major band in the expected size of RLB, suggesting that the antibody is highly specific (Supplemental Figure S2A). In comparison to ZH11, rlb showed around 20% off in RLB protein abundance. LOC_Os07g03770 has been previously reported as Osh15/d6/Oskn3 controlling rice height and grain shattering (Sato et al., 1999; Yoon et al., 2017). Phylogenetic analysis of its homologs revealed that RLB is a KNOX type homebox protein showing 89.23% sequence identify with maize (Zea mays) Rough Sheath 1 (RS1), which has been reported as a cell fate determinator in the leaf development (Schneeberger et al., 1995). In rice, OsH1/OsKN1 (LOC_Os03g51690) shared the highest similarity with RLB (Supplemental Figure S3). Despite that those reported osh15 allelic mutants showed similar phenotypes as rlb, the potential roles of RLB in rice lateral branch development have not been investigated yet. To validate the genetic mapping results, a DNA fragment covering the native promoter and complete genic sequence of RLB was introduced into the rlb. A total of five independent transformants were obtained and all were rescued to the WT morphology in the plant, panicle, and floret architecture. Therefore, we concluded that RLB is Osh15 (LOC_Os07g03770; Figure 2, E and F).

Figure 2.

Map-based cloning and functional verification of RLB. A, RLB was preliminarily mapped between markers RM14584 and RM5711 on chromosome 7 (Chr.7) using 24 F₂ individuals. The region narrowed down to about 3.6 CM between RM3831 and RM5490 using 48 F₂ mutant individuals. Further fine-mapping analysis with a total of 882 F₂ segregants restricted the locus within a 44.2-kb region between Ind8 and Ind14. Seven putative ORFs in the mapped region were sequenced. A mutation was found in ORF1 (indicated in red). All F₂ mutant type plants were from the cross of rlb/TN1. B, Schematic presentation of RLB gene structure and the mutation site. Black rectangles represent exons, white rectangles represent untranslated region (UTR), and black arrows indicate the location of primers. C, RLB protein structure and deduced mutation in the amino acid sequence. The 237th to 244th residues were deleted in rlb. D, qRT-PCR analysis of RLB expression level in ZH11 and rlb using various primer pairs. E and F, Genetic complementation test of RLB. Plant morphology (E) and main panicles (F) morphology of ZH11, rlb and the genetic complementation lines (rlbcom-1 and rlbcom-3). Bar = 10 cm in (E) and bar = 5 cm in (F). Seed from rlbcom lines was shown with normal sterile lemma as indicated by the red arrows in (F). G, Plant morphology of Nipponbare and RLB over-expression lines (oxRLB-6 and oxRLB-9). Bar = 10 cm. H, qRT-PCR analysis of RLB expression level in OxRLB lines. I and J, Morphology of the mutated florets from RLB over-expression lines. Bar = 0.5 cm. K, Main panicles morphology of RLB over-expression lines. Bar = 5 cm. For D and H, data are shown as means ± sd (n = 3). **P ≤ 0.01 by Student’s t test.

We ectopically expressed RLB using CAMV35S promoter in Nipponbare (O. sativa ssp. Japonica) background. Two representative lines (OxRLB-6 and OxRLB-9) with over 20-folds of upregulation in both transcription and protein levels were used for phenotype characterization (Figure 2, G and H;Supplemental Figure S2A). Similar to the rlb mutants, over-expression of RLB resulted in dwarf and abnormal plant architecture, including more tillers, less secondary panicle branches, which consequently decreased the spikelets per panicle and seed set per panicle (Figure 2, G and K;Table 1). Notably, some of OxRLB florets showed more than two palea/lemma-like structures, but normal sterile lemma, which is somehow different from the observations in rlb (Figure 2, I and J).

RT-qPCR assay indicated that RLB is a constitutively expressed gene (Figure 3A). With a special interest in panicle development, we further checked the expression level of RLB in panicles at eight development stages (s1–s8 from early to mature stages), and found that the expression level in young panicle was higher than that in the mature panicle. With the progress in developmental stage, the expression level of the gene gradually decreased (Figure 3B). Transient expression of RLB-eGFP protein in rice protoplast suggested that RLB might be a nuclear-located protein, which is consistent with its annotation as a transcription factor (Figure 3C). Meanwhile, expression of the mutated version of RLB recombinant protein RLB(M)-eGFP showed the same subcellular localization as the intact RLB (Figure 3C), implying the ELK domain is not responsible for the protein localization in cellular compartments.

Figure 3.

Expression pattern of RLB. A and B, qRT-PCR analysis of the relative expression levels of RLB in different tissues (A) and panicle developmental stages (B). Flag-l-s, flag leaf sheath. Flag-l, flag leaf. B, Relative expression level of RLB in eight panicle development stage. For A and B, data are shown as means ± sd of three biological replicates. C, Subcellular localization of RLB. pro35s::RLB-eGFP, pro35s::RLB(M)-eGFP and control vector p35s::eGFP were transformed into rice protoplasts respectively, and observed using a confocal laser-scanning microscope. Green fluorescence shows eGFP, red fluorescence shows the chloroplast auto-fluorescence, Bars = 5 μm.

RLB directly represses the transcription of CKX gene OsCKX4

RNA-sequencing (RNA-seq) experiments on the young panicle (0.5 cm in length) of rlb and ZH11 was carried out to investigate the potential downstream genes of RLB. A total of 1,616 differentially expressed genes (DEGs) were identified, including 1,190 upregulated and 426 downregulated in rlb (Supplemental Table S1). The DEGs are significantly enriched in the kyoto encyclopedia of genes and genomes (KEGG) pathways such as “plant hormone signal transduction” and “zeatin biosynthesis,” which is consistent to the cytokinin-deficient phenotypes observed on rlb (Supplemental Figure S4B). Gene Ontology analysis of the upregulated genes and downregulated genes revealed that both gene groups are enriched in “metabolic process,” “cell,” and “binding” in terms of “biological process,” “cellular component,” and “molecular function,” respectively (Supplemental Figure S4, A and B). To verify the RNA-seq results, we further conducted RT-qPCR analysis on 10 phytohormone-related DEGs in 0.5-cm young panicle cDNA samples. It was found that all the 10 genes showed similar inclination of the transcriptional levels as revealed by the RNA-seq, suggesting that our RNA-seq results are highly reliable (Figure 4A). Meanwhile, CKX4, SD-g, OsGH3.1, GA2ox1, and OsGH3.8 showed significant upregulation in rlb and downregulation in OxRLB when compared with ZH11, suggesting that these genes are potential downstream targets of RLB. Additionally, we conducted RT-qPCR analysis on 22 inflorescence and sterile lemma development-related genes in ZH11 and rlb. As shown in Supplemental Figure S5, majority of the tested genes were differentially expressed between ZH11 and rlb. Particularly, the transcription of OsMADS1 was significantly elevated, while these of FZP (Frizzy panicle) and multi-floret spikelet1 (MFS1) were dramatically reduced in rlb. It has been known that over-expression of OsMADS1 plants caused dwarfism, distorted panicles, and decreased numbers of branches and spikelets, which is similar to the observation on rlb (Wang et al., 2017). Meanwhile, FZP and MFS1 have been reported as positive regulators of rice sterile lemma development (Ren et al., 2013, 2018). Hence, the results implied that RLB potentially targets these genes to regulate rice flower development.

Figure 4.

RLB directly binds to the OsCKX4 promoter and suppresses its transcription. A, qRT-PCR analysis of the relative expression level of some reported development-related genes in ZH11, rlb, and OxRLB line. Data are shown as means ± sd of three biological replicates. B, ChIP-qPCR analysis of the RLB binding sites on OsCKX4. Positions of region P1–P10 are shown in (C). The enrichment values were normalized to IgG-immunoprecipitated DNA, actin primers was used as a CK. C, Schematic presentation of the OsCKX4 gene structures. Black boxes: coding region, blank box: untranslated region, line: promoter. +1: Transcription starting site, red dots: TGACA motif. Short black lines P1–P10 indicate the tested regions in ChIP-qPCR. D, EMSA assay showing RLB directly bind to the promoter of OsCKX4. The 5-, 10-, 50-, and 100-fold excess nonlabeled probes were used for competition. E, EMSA assay showing G-box mutation and RLB(M) weaken the bind between the protein and the promoter of OsCKX4. F, LUC transient transcriptional activity assay in rice protoplast. Reporter: proCKX4::19fLUC:tNOS; effectors: pro35S::RLB:tNOS and pro35S::RLB(M):tNOS. The fLUC/rLUC ratio represents the relative activity of 35S promoter. For A, B, and F data are shown as means ± sd of three biological replicates. **P < 0.01 by the Student’s t test, *P < 0.05 by the Student’s t test. For E different characters represent statistical difference at P < 0.05 by Student’s t test.

It has been well known that CKX/dehydrogenase genes are extensively involved in the catabolism of cytokinin in rice. The differential expression of OsCKX4 in rlb and OxRLB intrigued us to test whether it is direct targeted by RLB. First, chromatin immunoprecipitation quantitative real time PCR (ChIP-qPCR) was conducted on 10 different regions of OsCKX4 in ZH11 and rlb young panicles (Figure 4, B and C). RLB was significantly enriched on P3 region containing a conserved G-box cis element for the binding of homeobox transcription factors. In rlb, the intensity of RLB binding to the P3 region was apparently attenuated when compared to ZH11, possibly because the deletion of the 8 AA in RLB(M) compromised its DNA binding ability in vivo (Figure 4B). Subsequently, we purified GST-RLB and GST-RLB(M) proteins from Escherichia coli for electrophoretic mobility gel shift assay (EMSA) (Supplemental Figure S2, B and C). As a result, GST-RLB retarded the shift speed of P3 probe, while the protein–DNA binding was substantially weakened when non-labeled, competitive probes were added (Figure 4D). Mutation of the G-box significantly impaired the binding, suggesting that the TGACA motif is the core element for RLB attachment on the OsCKX4 promoter (Figure 4E). In consistence with the ChIP-qPCR result, the mutated GST-RLB(M) was not able to bind with the probe. Hence, the ELK domain of RLB may be responsible for its DNA binding ability (Figure 4E). Finally, a transient LUC transactivation assay was conducted by using proCKX4::19fLUC::tNos as a reporter. In the rice protoplast, effectors pro35S::RLB::tNos and pro35S::RLB(M)::tNos both significantly repressed the transcription of the reporter when compared with the pro35S:: tNos negative control, demonstrating that RLB is a transcription repressor on OsCKX4. In agreement to the upregulation of OsCKX4 in rlb, pro35S::RLB(M)::tNos showed weaker suppression than the intact RLB effector, possibly due to the attenuated DNA binding ability as revealed by the ChIP assay (Figure 4F).

Insitu hybridization assay indicated that OsCKX4 mRNA is abundantly accumulated in primary and secondary panicle meristems of the developing panicles, as well as the floret primordial including palea, lemma, lodicule, and stamens. Such an expression pattern is very similar to that of its homolog OsCKX11, which has been reported to regulate SAM activity and panicle branching (Zhang et al., 2021; Figure 5, A and D). As a negative control, the sense probe showed no signal (Figure 5E). RT-qPCR results showed that OsCKX4 is also constitutively transcribed among various rice tissues with the highest level detected in developing seeds (Figure 5F). To dissect the biological roles of OsCKX4, we generated over-expression lines (OxCKX4) with 100-folds of upregulation (Figure 5G). In comparison to ZH11, OxCKX4 almost phenocopied rlb with dwarf, more tillers, sparser, and smaller panicles, but did not show any defects in floret morphology (Figure 5, H and I; Table 1).

Figure 5.

Identification of OsCKX4 gene. A–E, mRNA in-situ hybridization alaysis of OsCKX4 transcript in developing panicle. Longitudinal sections of ZH11 panicles in 0.1 cm (A), 0.3 cm (B), and 0.5 cm (C) were hybridized with digoxigenin-labeled antisense OsCKX4 RNA probes. D, enlargement of the red frame from C. E, Sense probe was used as a negative contronl. FM, floret meristem; St, stamen; Pl, palea; Le, lemma; Lo, lodicule; Sl, sterile lemma. For A–C and E, bars = 100 μm, for D, bar = 25 μm. F, Expression pattern of OsCKX4 in various tissues. Data are shown as means ± sd of three biological replicates. G, The relative expression level of OsCKX4 in ZH11 and the two representative OxCKX4 lines. For F and G, data are shown as means ± sd (n = 3). **P < 0.01 by the Student’s t test. H and I, Agronomic performance of the OxCKX4 lines and ZH11 lines. For H, bar = 10 cm. For I, bar = 5 cm.

RLB recruits PRC2 to epigenetically repress CKX4 transcription via H3K27me3

In Eukaryotes, PRC2-mediated H3K27me3 represented a significant epigenetic repression mechanism in development (Angel et al., 2011). The repression effect of RLB on OsCKX4 provided us with a strong clue of the involvement of PRC2-mediated epigenetic regulation in this genetic pathway. Using yeast two-hybrid (Y2H) technique, we first investigated the protein–protein interaction of RLB with the five of the six rice PRC2 complex members, namely FIE2 (LOC_Os08g04270), EMF2a (LOC_Os04g08034), EMF2b (LOC_Os09g13630), CLF (LOC_Os06g16390), and EZ1 (LOC_Os03g19480). Due to the failure in FIE1 (LOC_Os08g04290) vector construction, we did not include FIE1 into this experiment. It was found that RLB was interactive to EZ1, EMF2a, and EMF2b, among which mutant of EMF2b has been reported with similar phenotypes to rlb (Figure 6A;Xie et al., 2015). Therefore, we performed in vitro pull-down and biomolecular fluorescence complementation (BiFC) assays to confirm further the RLB–EMF2b interaction. As a result, His-EMF2b was successfully pulled down with GST-RLB by using GST beads (Figure 6B), and strong YFP fluorescence signal was observed in the nucleus of N. benthamiana cells where nYFP-EMF2b and RLB-cYFP pair was transformed (Figure 6C). A previous study has shown that EMF2b is highly expressed in the rachis BMs, PBMs, and SPMs throughout the flower development (Conrad et al., 2014). In consistence to their results, our insitu hybridization assay obtained a similar pattern with high EMF2b levels detected in primary, secondary BMs as well as floret meristems (Figure 6, D–F), which is largely overlapped with that of OsCKX4 (Figure 5, A–E). Subsequently, we checked the distribution of H3K27me3, a suppressing histone modification catalyzed by PRC2 complex, on different regions of OsCKX4 by ChIP-qPCR. In most of the tested regions, especially in the P3 region where the G-box cis-element is located, H3K27me3 modification was drastically reduced in the rlb, suggesting that the targeting of H3K27me3 in OsCKX4 promoter depends on the RLB-mediated recruitment of PRC2 complex (Figure 6I).

Figure 6.

RLB recruits PRC2 to CKX4 to deposite H3K27me3. A, Yeast two-hybrid assay revealing the interaction of RLB and PRC2 members. Y2H assays were performed using the matchmaker GAL4 two-hybrid system and selected on synthetic medium lacking Leucine, Tryptophan and Histidine with 100 ng mL⁻¹ Aureobasidin A applied. B, Pull-down assay of GST-RLB and His-EMF2b. 6XHis-EMF2B was incubated with GST or GST-RLB in GST beads and was pulled down from the RLB-GST conjugated GST beads. C, BiFC analysis of RLB and EMF2B in rice protoplasts. Representative cells are used for imaging by laser-scanning confocal microscopy at excitation and emission wavelength 514 nm and 527 nm, respectively. GFP signal image gain value is 600. Positive interaction of nYFP-RLB::EMF2B-cYFP is indicated by the green fluorescence signal. Vector nYFP-RLB::-cYFP and nYFP::EMF2B-cYFP were used as control. Bars = 50 μm. D–F, mRNA in-situ hybridization alaysis of EMF2b transcript in developing panicle. Longitudinal sections of ZH11 panicle in 0.5 cm were hybridized with digoxigenin-labeled EMF2b antisense RNA probes. E, enlargement of the red frame from D. F, Sense probe was used as a negative control. PBM, primary branch meristem. FM, floret meristem; St, stamen; Pl, palea; Le, lemma; Lo, lodicule; Sl, sterile lemma. For D and F, bars = 100 μm, for E, bar = 25 μm. G and H, Plant and panicles morphology of the wild-type (Dongjin) and emf2b mutant. Bar = 10 cm for plant and 5 cm for panicle, respectively. I, ChIP-qPCR analysis of H3K27me3 deposition on OsCKX4 in ZH11 and rlb. Positions of the tested regions are indicated in Figure  4C. All values are shown as means ± sd with biological triplicates. **P < 0.01 by the Student’s t test.

We adopted a T-DNA insertional mutant of EMF2b from other lab and grew them together with the WT Dongjin (O. sativa ssp japonica). In comparison to rlb or OxCKX4 plants, emf2b displayed similar dwarf, smaller, and sparser panicle phenotypes in a more severe manner. However, it is notable that emf2b had fewer tillers than Dongjin (Figure 6, G and H).

LOG promoter-driving repression of OsCKX4 improves grain number per plant

Given the sparser panicle of OxCKX4, we hypothesized that specific knockdown of OsCKX4 in panicle would increase the grain numbers without affecting other agronomic traits. To attest this hypothesis, we constructed a proLOG::aCKX4 vector to have the antisense strand of OsCKX4 expressed under the driving of the panicle-related promoter of LOG, and transformed it into ZH11 background (Figure 7, A and B). RT-qPCR analysis indicated that the expression of OsCKX4 was substantially reduced in the panicle of proLOG::aCKX4 lines (Figure 7C). However, we also found the downregulation of OsCKX4 in the leaf tissue, possibly due to the leaking expression of OsCKX4 antisense strand in other tissues (Figure 7D). Then, a comprehensive analysis was conducted on the major agronomic traits of the proLOG::aCKX4 lines in T₂ generation along with ZH11 (Figure 7, E–K;Table 1). Although no obvious differences in plant height, tiller numbers, flowering date were found between proLOG::aCKX4 lines and ZH11, the transgenic lines had larger panicle size bearing over 20% more spikelets per plant than the WT, indicating that suppression of OsCKX4 in panicle might be an effective strategy to improve the panicle branching in rice. Nevertheless, the lower seed setting and grain weight traded off the contribution of the increased spikelets in proLOG::aCKX4 lines, which finally showed an equal yield potential to ZH11 (Figure 7K).

Figure 7.

Performance of proLOG::aCKX4 lines. A and B, Plant and main panicles morphology of ZH11 and the two representative proLOG::aCKX4 lines. For A, bar = 10 cm. For B, bar = 5 cm. C and D, The relative expression level of OsCKX4 in panicle and leaf of ZH11 and proLOG::aCKX4 lines. E–K, Agronomic traits of ZH11 and proLOG::aCKX4 lines. E, Tiller number. F, Total spikelets per plant. G, Seed setting ratio. H, 1,000-grain weight. I, Seed length. J, Seed width. K, Yield per plant. All values are means ± sd, with biological triplicates for gene expression analysis and five biological repeats for agronomic traits analysis. For seed length and width analysis 100 full seeds were used. *P ≤ 0.05, **P ≤ 0.01 by the Student’s t test.

Discussion

RLB promotes rice lateral branching by repressing cytokinin catabolism

Since the identification of the first plant homeobox gene KNOTTED1 (KN1) in maize in 1991 (Vollbrecht et al., 2000), plant homeobox genes have been widely identified and classified into six subgroups including Knotted homeobox type (KNOX; Mukherjee et al., 2009). KNOX genes are majorly expressed in SAMs of plants, where lateral organs such as tillers and panicles are derived (Hay and Tsiantis, 2009), and have been highlighted as indispensable in SAM maintenance. A few examples are KN1 in maize and SHOOT MERISTEMLESS in Arabidopsis, of which the knock-out mutants failed to form SAM and have abnormal branching (Long et al., 1996; Vollbrecht et al., 2000). As a rice KNOX type homeobox gene, OsH15 has been identified as OsH15/D6/OsKN3 for decades. It was found that OsH15 is expressed in regions between two lateral organs newly formed in the vegetative, inflorescence and floral meristem (Sato et al., 1998.). Sato et al (1999) revealed key roles of OsH15 in internode elongation, as internodes of osh15 mutants were shortened and showed abnormal-shaped epidermal and hypodermal cells (Sato et al., 1999). Meanwhile, ectopical expression of OsH15 resulted in abnormal leaf, flower morphologies, and defected shoot regeneration from calli of rice and N. benthamiana (Sato et al., 1998; Sentoku et al., 2000; Ito et al., 2001). Recently, it was also reported that the OsH15 may inhibit the monolignol biosynthesis in abscission zone to regulate rice grain shattering (Yoon et al., 2017).

Distinct from the previous studies, this study functionally characterized RLB, a new allele of OsH15, with a focus on lateral branch development. Eight amino acids deletion in the ELK domain of RLB gave rise to obvious phenotypes in the branch architecture such as more tillers, sparser panicles with less secondary branches and single sterile lemma structure (Figure 1, A–F, K–L; Table 1). Except for the differences in florets, RLB over-expression lines exhibited almost identical traits as the mutants, indicating that fine-tuning of RLB is essential for rice development (Figure 2, G and K; Table 1). In addition to the branching phenotypes, rlb showed severely defected grain filling which led to extremely low seed set, and rlb callus could barely regenerate into shoots during tissue culture, implying versatile roles of RLB in rice development (Figure 1, M and N;Supplemental Figure S1, D and E). Given the key roles of cytokinin in plant branching and the typical cytokinin-deficiency phenotypes in rlb, we attempted to interpret the phenomena from the view of cytokinin homeostasis. Indeed, rlb had significantly decreased endogenous cytokinin contents, and more importantly, application of exogenous zeatin drastically restored the callus regeneration rate (Figure 1, O–Q). Hence, the rlb phenotype is majorly, if not completely, attributed to the cytokinin deficiency.

Emerging evidence has shown that KNOXs may control meristem development by fine-tuning the spatial distribution and concentration of cytokinin as well as other phytohormones (Hay et al., 2004). KNOX over-expression lines usually have elevated cytokinin accumulation (Kusaba et al., 1998; Ori et al., 1999). Such an effect is mostly explained by the KNOX-mediated activation on cytokinin biosynthesis genes such as LOG, OsIPT2, and OsIPT3 in rice (Sakamoto et al., 2006; Yasui et al., 2018), LOG and IPT3 in Medicago (Azarakhsh et al., 2020), and AtIPT7 in Arabidopsis (Yanai et al., 2005). Interestingly, we revealed that the transcription of OsCKX4, a rice cytokinin catabolism gene, is directly repressed by RLB with solid pieces of evidence from the in vitro and in vivo experiments (Figure 4, B–F). In line with this result, OxCKX4 lines showed similar plant branch architectures to rlb (Figure 5, H and I; Table 1). Our results suggested dual pathways of KNOXs in promoting cytokinin accumulation by either activating genes in biosynthesis as previously reported or repressing genes in catabolism, which greatly expanded the regulatory network of KNOXs in plant development.

PRC2-mediated H3K27 tri-methylation is involved in the homeobox-governed gene regulation

H3K27me3 catalyzed by PRC2 has been widely accepted as a major histone modification for gene expression reprogramming in plant growth and development (Makarevitch et al., 2013). In rice SAM, a large number of genes are marked by H3K27me3, and massive reprogramming of H3K27me3 is implicated in leaf organogenesis and IM formation (Liu et al., 2015). However, because PRC2 protein does not have inherent DNA binding specificity, placement of the PRC2 machinery to the target loci largely relies on other transcription factors, which could bind to PRC2 and recognize the cis elements in target gene promoters (Xiao et al., 2017). In Arabidopsis, Xiao et al (2017) have profiled 233 transcription factors that can potentially recruit PRC2 to target sites, of which 16 were annotated as homeobox-containing proteins (Xiao et al., 2017).

Homeobox proteins are defined by the presence of a helix-loop-helix-turn-helix domain called homeobox, which confers the proteins transcription activation or repression functions (Burglin, 2011; Holland, 2012). In a previous study, OSH15/RLB has been characterized as a transcription repressor by using a transient transactivation system (Nagasaki et al., 2001). In consistence, this study found that RLB directly represses the transcription of OsCKX4 to suppress cytokinin catabolism. To support the hypothesis that RLB regulates target genes via recruiting PRC2-mediated H3K27me3, several lines of evidences were provided in this study: (1) RLB physically interacted with PRC2 component EMF2b in yeast, in vitro and in planta. (2) RLB and H3K27me3 co-localized in the G-box containing region in OsCKX4 promoter. (3) emf2b displayed similar but more severe phenotype as rlb and OxCKX4. From the results and discussion above, a working model of RLB regulated rice lateral branch development was proposed (Figure 8). Actually, RLB is not the sole case that homeobox proteins epigenetically regulate target genes through PRC2 or other histone modification complexes. Likely, Yang et al. (2013) found that a plant homeodomain (PHD) finger domain protein OsVIL2 directly binds to EMF2b to enrich the H3K27me3 modification in the chromatin section of a rice flowering repressor OsLFL1, thus to induce flowering (Yang et al., 2013). Another study in rice also demonstrated that KNOX protein OsWOX11 recruits the histone H3K27me3 demethylase JMJ705 to activate the transcription of target genes during the development of apical meristem. The authors suggested that the deposition of JMJ705 to specific downstream genes is facilitated by OsWOX11, even though JMJ705 has certain DNA binding ability (Cheng et al., 2018). Our story in RLB hinted that the recruitment of PRC2 or other histone modifiers might represent a general mechanism for the homeobox in gene repression and plant development.

Figure 8.

Working model for RLB in rice lateral branch development.

OsCKXs are potential targets for rice architecture improvement

Rice yield is essentially determined by three factors, namely the number of tillers per plant, the number of spikelets per panicle and 1,000-grain weight. Apart from its roles in promoting rice tillering and panicle branching, cytokinin may also positively regulate nuclear and cell divisions in the endosperm to enhance grain fillings and seed weight (Jameson and Song, 2016). It is therefore believed that knocking-out or -down of negative cytokinin pathway genes, like OsCKXs, might be an effective way to improve rice plant architecture and yield.

A genome survey in rice has identified eleven CKX members (Ashikari et al., 2005). One of the well-documented examples of CKXs is OsCKX2/Gn1a. OsCKX2 may serve as an important node in cytokinin homeostasis and panicle branching, as the regulatory pathways of numerous rice branching genes are converged in OsCKX2 (Kurakawa et al., 2007; Huang et al., 2009; Li et al., 2011; Li et al., 2013; Huang et al., 2018). Disruption of OsCKX2 led to cytokinins accumulation in IM, enhanced panicle branching and higher reproductive organ numbers, which finally improved the rice yield (Ashikari et al., 2005). With the aid of the clustered regularly interspaced short palindromic repeats-CRISPR associated proteins 9 (CRISPR/Cas9) genome editing techniques, several groups have successfully utilized the knock-out of OsCKX2/Gn1a in local landraces to breed higher yield varieties (Shen et al., 2017; Lacchini et al., 2020). For OsCKX9 gene, both mutants and over-expression lines had increased tiller number, but decreased plant height and panicle size (Duan et al., 2019). Very recently, Zhang et al. (2021) reported the roles of OsCKX11 in delaying leaf senescence, increasing grain number and coordinating the carbon sources between source and sink organs. OsCKX11 mediates the catabolism of various cytokinins with a favor for trans-zeatin and cis-zeatin. Disruption of OsCKX11 resulted in over 14% more tillers, 15% more spikelets per panicle, and 11% more yield per plant, albeit the seed setting and seed weight were moderately decreased, which showed promising potential in crop genetic improvement (Zhang et al., 2021). OsCKX4 has been reported as a cytosolic isoform of CKX. OsCKX4 dominant mutant exhibited a more robust root system with increased crown root number, and reduced plant height. OsCKX4 under the driving of a root-specific promoter RCc3 greatly enhanced rice root development, zinc content in grains and yield without affecting shoot parts (Gao et al., 2014, 2019). In recognition of the negative association of OsCKX4 transcription and panicle branching, we explored the possibility to improve rice spikelet numbers by repressing OsCKX4 under the promoter of LOG in this study. As expected, we achieved over 20% increases of spikelets per plant in proLOG::aCKX4 lines than the control variety ZH11 (Figure 7F), suggesting the feasibility of this approach. Nevertheless, as reported in the case of OsCKX9, we also observed lower seed setting and grain weight in the transgenic lines, which counteracted the yield contribution of the increased spikelets and led to an equal yield potential to ZH11 (Figure 7, G, H, and K). We proposed that this might be the side effects brought by the leaking expression of antisense OsCKX4 in stamen and developing seeds. Thus, more specific panicle promoters, like LAX1 and LAX2, will be utilized in our future study to improve rice yield.

Materials and methods

Experimental materials and phenotypic analysis

The mutant rlb was obtained from a T-DNA insertion mutant library of Zhonghua 11 (O. sativa ssp. japonica) background. The F₂ segregating population for RLB map-based cloning was generated through a cross between the rlb and TN1 (O. sativa ssp. indica). The mutant emf2b was obtained from a T-DNA insertion mutant library derived from Dongjin (O. sativa ssp. japonica; Xie et al., 2015). All plants used in this study were grown in the experimental field under local growing conditions of the China National Rice Research Institute located in Fuyang, Zhejiang Province, China (119°57′E, 30°05′N). The agronomic traits, including plant height, effective panicle number, panicle length, grain number per panicle, primary panicle branches number, and secondary panicle branches number, were measured with five replicates at the mature stage.

Scanning electronic microscopy

SEM was performed by following a reported protocol (Zhao et al., 2020). Briefly, the tissue samples were fixed in 2.5% (v/v) glutaraldehyde solution at 4°C for overnight and then rinsed with 0.1 M phosphate buffer at pH 7.0 for three times, 15 min each time. The samples were then fixed with 1% (v/v) osmic acid solution for 1–2 h and rinsed again using 0.1 M phosphate buffer at pH 7.0 for three times. Subsequently, the fixed samples were dehydrated by going through ethanol solution of gradient concentrations. The drying process was carried out in a Hitachi HCP-2 critical point dryer (Tokyo, Japan) and the samples were observed under a Hitachi SU-8010 scanning electron microscope (Tokyo, Japan) after coating.

Paraffin section

Paraffin sectioning was performed by following a previous report (Zhang et al., 2010). Briefly, fresh tissues were fixed in 50% (v/v) FAA (50% [v/v] ethanol, 5% [v/v] acetic acid, 3.7% [v/v] formaldehyde) for overnight, dehydrated through gradient ethanol, infiltrated by xylene, and embedded in paraffin solution. Sections were cut with a microtome in thickness of 10 µm, stained by fast green solution, and finally observed under a bright-field microscope.

Map-based cloning

To determine the genetic control of the sparse lateral branch trait, F₁ plants, F₂ populations derived from the three crosses (rlb/ZH11, rlb/TN1, and rlb/9311) were analyzed. Using 260 Simple sequence repeat (SSR) markers, which uniform distributed on the 12 chromosomes of rice and 24 F₂ plants derived from rlb/TN1, the rlb control gene was initially located on the short arm of chromosome 7, linked with markers RM4584 and RM5711. Using another 48 F₂ plants, we delimited the gene to a 3.6 CM region between RM3831 and RM5490. To fine map the gene six InDel markers Ind6, Ind7, Ind8, Ind14, Ind21, and Ind25 were developed. These six Indel markers besides RM3831 and RM5490 total eight markers were then used to genotype the 822 sparse panicle F₂ individuals derived from rlb/TN1. SSR markers were obtained from (http://www.gramene.org/), while insertion/deletion (InDel) markers were designed using Primer 5.0 after comparison of the sequences between the japonica cultivar Nipponbare and the indica cultivar 9,311 in the following public databases: RGP (http://rgp.dna.affrc.go.jp/E/toppage.html), Gramene (http://gramene.org/genome_browser/index.html) and the Gene Research Center of the Chinese Academy of Sciences (http://rice.genomics.org.cn/rice/index2.jsp). The primers were synthesized by Sunya Biotech Co. Ltd (Hangzhou, China) and listed in Supplemental Table S2.

Vector construction and plant transformation

For the complementation test, the 8,474-bp RLB genomic DNA fragment containing 1.8-kb upstream sequences and 1.1-kb downstream sequences was amplified and inserted into the pCAMBIA2300 vector (http://www.cambia.org) using EcoRI and SmaI sites. The recombinant vector was introduced into rlb callus to generate transgenic plants. For over-expression of RLB and OsCKX4 the gene coding sequence (CDS) were amplified using the leaf cDNA of ZH11 and cloned into the PU1301 vector containing 35S promoter using the KpnI and BamHI sites. The japonica rice cv Nipponbare and Zhonghua11 were used as the transformation recipient, respectively. For antisense suppression analysis of OsCKX4, pCAMBIA1300 vector was linearization with KpnI and SalI. Then, 2.0-kb fragment from LOG promoter and full CDS of OsCKX4 were cloned into the above linearized vector. All the binary vectors were transformed into Agrobacterium tumefaciens strain EHA105, then genetically transformed into Nipponbare or Zhonghua 11 callus using the Agrobacterium-mediated transformation method as described previously (Hiei et al., 1994). The primers are listed in Supplemental Table S2.

RNA isolation and gene expression analysis

The total RNA of all the tissues except developing seeds was extracted by using RNA Isolation kit (YESEN Biotech, Shanghai, China). The RNA of developing seeds was extracted according to a modified sodium dodecyl sulfate (SDS)-Trizol method (Li et al., 2016). PrimeScript RT Master Mix (Takara, Dalian, China) was used for the RNA reverse transcription. Quantitative real-time PCR was performed in a CFX 96 real-time PCR instrument (Bio-Rad, CA, USA). The PCR programs were set according to a previous report (Tong et al., 2018). Experiments were performed with biological triplicates for each sample, and UBQ (LOC_Os03g13170) was used as an internal control. The relative mRNA level of tested genes was normalized to UBQ and calculated by the 2^−ΔΔCT method (Schmittgen and Livak, 2008). All the primers used for RT-qPCR can be found in Supplemental Table S2.

Subcellular localization of RLB

For subcellular localization analysis of the RLB protein, the CDS of RLB and RLB(M) without stop codon were amplified from Zhonghua 11 and rlb, respectively. Then, the amplified fragments were cloned into pCA1301-35S-eGFP vector (linearized by XbaI site). The fusion plasmid was transformed into rice protoplasts according to a previous protocol (He et al., 2020). After 48 h of dark incubation at 28°C, the enhanced green fluorescent protein (eGFP) signals were observed under a Zeiss LSM710 confocal laser-scanning microscope (Karl Zeiss, Jena, Germany) at excitation and emission wavelength 480 nm and 505 nm, respectively. The empty pCA1301-35S-eGFP vector was used as a control. The primers of this experiment were listed in Supplemental Table S2.

Yeast two-hybrid assay

Matchmaker GAL4 two-hybrid system (Clontech, CA, USA) was used for the Y2H assays. CDS of RLB was cloned into vector pGBK as bait, and CDS of five PRC2 members (OsEZ1, OsEMF2a, OsEMF2b, OsFIE2, OsCLF) were cloned into pGAD vector as preys. The bait and prey plasmids were co-transformed into yeast strain Y2H Gold (Clontech, CA, USA) and grew on SD/–Leu–Trp–His–Ade medium containing 3 mM 3-AT by standard protocols (Clontech, CA, USA). The primers of this experiment were listed in Supplemental Table S2.

In vitro pull-down assays

Full CDS of RLB was cloned into the pGEX-4T-1 vector (GE Healthcare, Chicago, USA) to fuse with GST tag, and the CDS of EMF2B were ligated into pET28a, which linearized with BamHI and SacI sites (Thermo, Waltham, USA) to generate the recombinant protein with HIS tag. The recombinant proteins were expressed in E. coli strain Transetta (DE3; Transgen, Beijing, China), and purified using the GST-SefinoseTM Kit (Sangon Biotech, Shanghai, China) and 6×His-Tagged Protein Purification Kit (CWBIO, Beijing, China) based on the manufacturer’s instructions respectively. For the pull-down assays, 500 μg of each protein were incubated with glutathione high capacity magnetic agarose beads (Sigma-Aldrich, St Louis, USA) in pull-down buffer (50 mM Tris–HCl, pH 7.5, 5% [v/v] glycerol, 1 mM EDTA, 1 mM DDT, 1 mM phenylmethylsulfonyl fluoride (PMSF), 0.01% [v/v] Nonidet P-40, and 150 mM KCl) at 4°C for 2 h, and then washed three times with pull-down buffer. GST fusion proteins were eluted and detected by immunoblotting analysis using anti-GST (Cat: CW0085, CWBIO, Beijing, China) or anti-HIS (Cat: CW0083, CWBIO, Beijing, China). The primers of this experiment were listed in Supplemental Table S2.

Biomolecular fluorescence complementation assay

pDOE-BiFC vector was first linearized with BspEI and adopted for the BiFC assays (Bello et al., 2019). The CDS without stop codon of RLB was inserted into MCS3 of the vector to create a “parent” vector, which has the RLB fusion with the C terminal of YFP. The EMF2b cDNA was inserted into MCS1 of the parent vector with the BamHI site, for the N-terminal fusion. All of the above fusional expression vectors were transformed into Agrobacterium strain EHA105 by electroporation, and subsequently infiltrated into the leaf epidermal cells of 3-week-old N. benthamiana. After 48 h of incubation, the transformed leaves were observed for yellow fluorescence under a Zeiss LSM710 confocal laser-scanning microscopy (Carl Zeiss AG, Jena, Germany) at excitation and emission wavelength 514 nm and 527 nm, respectively. The primers of this experiment were listed in Supplemental Table S2.

EMSA

The probes of OsCKX4 (57 nt in length) used for Figure 4D was commercially synthesized by Sunya Biotech Co. Ltd (Hangzhou, China), and labeled with an EMSA Probe Biotin Labeling Kit (Beyotime, Shanghai, China) by following the manufacturer’s instructions. Nonlabeled probes were used as competitors. EMSA was performed as previously described (Bello et al., 2019). Reaction system in 20 μL volume was made by mixing 4 μL 5×EMSA/Gel-shift binding buffer (Beyotime, Shanghai, China), 2-μg purified protein and different concentration of nonlabeled DNA probes to 18 μL, and incubated at 25°C for 25 min. After that, 50-nmol labeled probe was added and incubated for another 30 min. The reaction products were electrophoresed using 6% (w/v) native polyacrylamide gels on ice. For the biotin-labeled probe detection, the probes were transferred to a Nylon membrane (Beyotime, Shanghai, China) and visualized by the Light Shift Chemiluminescent EMSA kit (Beyotime, Shanghai, China) under a ChemDocTM Touch Imaging system (Bio-Rad, CA, USA). The EMSA probes for Figure 4E were commercially synthesized and labeled with Cy5.5 by Tsingke Biological Technology (Beijing, China). The assay was performed as mentioned above. The fluorescence signal in the gel was visualized using an Odyssey CLx infrared fluorescence imaging system (LI-COR, Lincoln, USA) at excitation wavelength 680 nm and emission wavelength 720 nm. The sequence of the probes was listed in Supplemental Table S2.

Luciferase transient transcriptional activity assay

For luciferase transient transcriptional activity assay vector “None” and “190fLUC” were used. Vectors were linearized using BamHI/EcoRI and HindIII/BglII, respectively. Then, the CDS of RLB and RLB(M) were, respectively, cloned into vector “None” to generate the pro35S::RLB::tNOS and pro35S::RLB(M)::tNOS effectors. The 1.5-kb promoter region of OsCKX4 was cloned into 190fLUC vector as reporter. The primers used were listed in Supplemental Table S2. A total of 10 μg of plasmid (effector, reporter, internal control) was used for the transformation in rice protoplast. The method of rice protoplast extraction and transformation were the same as subcellular localization assay. All the luciferase activities were measured using the Dual-Luciferase Reporter (DLR) Assay System Kit (Promega, Wisconsin, USA). The relative luciferase activity was calculated as the ratio between rLUS1 and rLUS2. AtUbi3:rLUC was used as an internal control. The luciferase activity for each sample was detected with triple biological replicates.

ChIP-qPCR

The chromatin immunoprecipitation (ChIP) experiment was performed as previously described (Hou et al., 2019). Briefly, 3–4 g of leaf samples were crosslinked in 1% (v/v) formaldehyde under vacuum for about 30 min. Chromatins were extracted and sonicated to the size about 200–500 bp. Then, the DNA/protein complex was immune-precipitated with polyclonal RLB antibody, which was commercially synthesized by Genescript company (Nanjing, China), and H3K27me3 antibody (Cat No: Ab6002. Abcam, Cambridge, USA). After reverse cross-linking and protease K treatment, the immune-precipitated DNA was purified. The purified ChIP DNA was used as template for quantitative PCR with triple biological replicates. The quantitative PCR results were analyzed by following a method of Magna ChIP HiSens kit (Millipore, MA, USA) The data analysis was performed as described previously (Shao et al., 2019). All primers of this experiment were listed in Supplemental Table S2.

mRNA in situ hybridization

The mRNA in situ hybridization was conducted as previously described (Zhang et al., 2010). Young panicles at different development stages of the wild-type Zhonghua 11 were fixed in 50% (v/v) FAA (3.7% [v/v] formaldehyde, 5% [v/v] glacial acetic acid, and 50% [v/v] ethanol) and embedded in paraffin. The tissues were sliced into 8-mm sections using a microtome (Leica, Wetzlar, Germany). Primers used for mRNA in situ hybridization were labeled by digoxygenin with a DIG RNA Labeling Kit (Roche, Basel, Switzerland) following the manufacturer’s recommendation. Images were photographed using a Leica DM2500 microscope (Leica, Wetzlar, Germany). Primers used in this experiment are listed in Supplemental Table S2.

RNA-seq analysis

For RNA-seq analysis, 2- to 3-cm length young panicle of rlb and ZH11 were manually collected. RNA samples were extracted using TRIzol according to the manufacturer’s instructions (LSC, Hangzhou, China). Qubit 2.0 (Invitrogen, Carlsbad, CA, USA), and Agilent BioAnalyzer 2100 (Beijing, China) were used to detect the purity and integrity of the RNA samples. Then, the double-stranded cDNA library was synthesized and purified by Biomarker Technologies (Beijing, China). Illumina Hiseq platform (Illumina, San Diego, CA, USA) was used for the sequencing. Gene expression changes between the two samples were analyzed by cuffquant and cuffnorm components in cufflinks (2.2.1) software. DEGs were defined as genes with |log2Fold change| ≥1 and FDR <0.01 using EBSeq. For GO analysis, GO::TermFinder, KOBAS(2.0) was used find different expression gene enrichment, and chose P < 0.05 as the cutoff for significant GO terms.

Quantification of free plant hormone content

Extraction and determination of the IAA, GA3, and natural forms of cytokinin in each sample were performed by Nanjing webiolotech Biotechnology Co., Ltd using the method described previously (Kojima et al., 2009; Liu et al., 2010).

Accession numbers

Sequence data from this article for the cDNA and genomic DNA of RLB, OsCKX4 and EMF2b can be found in the GenBank/EMBL/Gramene data libraries under accession numbers LOC_Os07g03770, LOC_Os01g71310 and LOC_Os09g13630, respectively.

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. The rlb mutant has abnormal filling of seeds.

Supplemental Figure S2. Purity of proteins and specificity of RLB antibody.

Supplemental Figure S3. A neighbor-joining phylogenetic tree of TALE protein

Supplemental Figure S4. RNA-seq results of rlb.

Supplemental Figure S5. Relative expression level of genes related to spikelet development in ZH11 and rlb floret.

Supplemental Table S1. DEGs of rlb in young panicles.

Supplemental Table S2. Sequences of primers used in this study.

 

H.W., X.T., and J. Zhang conceived and designed the experiments; H.W., X.T., L.T., Z.L., X.L., M.Y., W.L., H.X., and J. Zhao conducted experiments; H.W., X.T., L.T., Y.W., J. Ying, W.Y., J. Yao, and J. Zhang analyzed data; Y.S. and S.W. managed field work; H.W. and J. Zhang wrote the manuscript; T.A. polished language. H.W., X.T., and L.T. contributed equally.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plphys/pages/general-instructions) is Jian Zhang (zhangjian@caas.cn).