Shoot branching is one of the critical agronomic traits, contributing to biomass and grain yield in crops (Tian and Jiao, 2015). Rice tillers are specialized shoot branches that are formed on the unelongated basal internode and grow independently from the main culm (Li et al., 2003). Rice tillering affects panicle development such as the number of primary and secondary branches, so that branching determines grain number per panicle (Wang and Li, 2011).

Tillers arise from the axillary bud apical meristem (AM), which develops in the axils of leaves during the vegetative stage in rice (Wang and Li, 2011). The development of tillers undergoes two processes: formation of the axillary bud and outgrowth of the axillary (tiller) bud (Wang and Li, 2011). At Stage 1, the AM consists of cytoplasm-dense cells which form as a bulge in the axil of the leaf primordium. The bulge develops into a cone-like structure at Stage 2. The dome and prophylls become gradually more evident at Stage 3, and the axillary bud is formed at Stage 4 (Tanaka et al., 2015).

In rice, Oryza sativa homeobox1 (OSH1), which is preferentially expressed in the AM, is required for the initiation of AM formation and the maintenance of undifferentiated cell fate (Tanaka et al., 2015). LAX PANICLE1 (LAX1) also plays an essential role in AM formation by genetically functioning together with LAX PANICLE2 (LAX2) and MONOCULM1 (MOC1) (Tabuchi et al., 2011). LAX1 is expressed in the leaf axil at Stage 4 and the transcription factor accumulates to a high level in two to three layers of cells in the boundary region between the axillary bud and the shoot AM (SAM). Although differentiation of the meristematic cells is initiated in lax1 mutants, AM formation is defective (Oikawa and Kyozuka, 2009).

TILLERS ABSENT1 (TAB1; OsWUS) encodes a transcriptional regulator containing a homeodomain WUS box and an ethylene-responsive element-binding factor (EAR) motif and is expressed in the pre-meristem zone. Mutation of TAB1 causes reduction in OSH1 expression in the pre-meristematic region and defects in AM formation (Tanaka et al., 2015). Another factor involved in AM formation is a LEAFY ortholog, RICE FLORICULA/LEAFY (RFL), also called ABERRANT PANICLE ORGANIZATION 2 (APO2) (Deshpande et al., 2015). RFL is preferentially expressed in the axils of leaves at the juvenile vegetative stage and maintains AM specification by promoting expression of LAX1 and CUP SHAPED COTYLEDON (CUC) genes (Deshpande et al., 2015).

Not all of the axillary buds develop immediately into tillers and some of them stay dormant until maturity. Final tiller number depends mainly on the capability of axillary bud outgrowth (Wai and An, 2017; Wang and Li, 2011), which is affected by various environmental factors and hormones (Deshpande et al., 2015).

Auxins and cytokinins influence the outgrowth of tiller buds (Wai and An, 2017). Auxins are synthesized at the shoot apex and transported by the polar auxin transport system. In rice, a polar auxin transporter, OsPIN1, is essential for auxin-dependent tiller bud emergence. Reduction of OsPIN1 gene expression causes an increase in tiller number due to the disturbance of auxin-dependent tiller bud inhibition (Xu et al., 2005). Similarly, OsPIN2 also controls tiller number (Chen et al., 2012). Another auxin transporter gene OsPIN3 appears to be involved in tiller development because expression of this gene is reduced in the culm of RFL knock-down plants (Deshpande et al., 2015). Expression of auxin-responsive genes such as OsIAA7 and OsIAA20 is also affected in RFL knock-down plants (Deshpande et al., 2015). Endogenous cytokinins in axillary buds are considered to act as activators of axillary bud outgrowth (Yeh et al., 2015). In rice, Cytokinin Oxidase2 (OsCKX2), which is an enzyme which degrades cytokinins, inhibits shoot branching by reducing endogenous cytokinin levels (Yeh et al., 2015).

Strigolactone (SL), a carotenoid phytohormone, suppresses axillary bud outgrowth (Wang et al., 2013). Rice mutants defective in the biosynthesis or signaling of SLs display increased branching phenotypes (Wai and An, 2017; Zhang et al., 2010; Zhao et al., 2014; Zou et al., 2006). DWARF10 (D10), encoding carotenoid cleavage dioxygenase 8, is involved in inducing axillary bud outgrowth by reducing auxin levels and transport (Zhang et al., 2010). HIGH-TILLERING DWARF1 (HDT1), encoding carotenoid cleavage dioxygenase 7, also stimulates the outgrowth of axillary buds (Zou et al., 2006). A SL-biosynthesis gene, DWARF27 (D27), is predominantly expressed in axillary buds and inhibits outgrowth of tiller buds (Lin et al., 2009), whereas DWARF3 (D3), encoding a nuclear-localized F-box protein, inhibits the outgrowth of axillary buds by forming a Skp, Cullin, F-box containing (SCF) complex with DWARF14 (D14) and DWARF53, that are involved in SL signaling (Ishikawa et al., 2005; Jiang et al., 2013; Nakamura et al., 2013).

Several transcription factors control outgrowth of axillary buds. Rice TEOSINTE BRANCHED1 (OsTB1) is a negative regulator of the outgrowth of axillary buds, while OsMADS57 is a positive factor controlling tiller outgrowth (Guo et al., 2013; Minakuchi et al., 2010). Loss-of-function of OsTB1 causes thin culms and excessive tillering, while overexpression of the gene results in reduced tillering (Minakuchi et al., 2010; Takeda et al., 2003). OsTB1 interacts with OsMADS57 to repress expression of D14, a gene involved in SL signaling, to regulate axillary bud outgrowth; OsMADS57 affects tiller outgrowth by suppressing the expression of D14. This negative regulation by OsMADS57 is inhibited by physical interaction with OsTB1 to make balancing of D14 expression for tiller outgrowth (Guo et al., 2013).

Although many of the chromatin-remodeling factors are found to control plant development such as flowering time and panicle architecture in rice, their roles in shoot branching have not been studied. Polycomb repressive complex 2 (PRC2), which represses target gene expression by regulating the trimethylation of histone 3 lysine 27 (H3K27me3) of the chromatin of the target gene, plays essential roles as an epigenetic repressor (Mozgova and Henning, 2015). The rice VERNALIZATION INSENSITIVE 3-like (VIL3) protein OsVIL2 associates with the PRC2 complex to suppress target gene expression by mediating H3K27me3 (Yang et al., 2013; 2019). The osvil2 null mutants exhibit pleiotropic phenotypes including alteration of leaf angle, reduced tiller number, floral organ defects, and a change in flowering time (Yang et al., 2013; Yoon et al., 2018). In the current study, we demonstrate that OsVIL2 induces tillering by controlling the chromatin state of OsTB1.

Plant materials and growth condition

We previously reported the T-DNA null mutants, osvil2-1 and osvil2-2, that were isolated from a T-DNA-tagged line pool in Oryza sativa japonica cv. Dongjin (Jeon et al., 2000; Jeong et al., 2002; Yang et al., 2013). We have previously generated transgenic plants carrying the OsVIL2 promoter-GUS construct (Yang et al., 2013; 2019; Yoon et al., 2018). Plants were grown in a growth room under long-day conditions (14-h light at 28°C/10-h dark at 23°C).

RNA extraction and reverse transcription-polymerase chain reaction (RT-PCR) analysis

The cDNAs were synthesized from 2 μg of total RNA isolated from basal parts of the shoots using RNAiso Plus (Takara, Japan). Moloney murine leukemia virus reverse transcriptase (Promega, USA), RNasin Ribonuclease Inhibitor (Promega), 10 ng of oligo (dTs) and 2.5 mM deoxiribonucleotide triphosphates were included in the reaction mixture. Quantitative RT-PCR (RT-qPCR) was performed with a Rotor-Gene 6000 (Corbett Research, Australia) using SYBR Green I Prime Q-Mastermix (GeNet Bio, Korea) as described previously (Cho et al., 2016; 2018b). The internal control was OsUbi1. The ΔΔCT method was used to calculate levels of relative expression (Yoon et al., 2014). Primers used in this study are presented in Supplementary Table 1.

Histochemical analysis and GUS assay

Basal parts of the seedlings were fixed in a formaldehyde-acetic acid-alcohol solution at 4°C. Samples were dehydrated with an ethanol series, treated with a tert-butyl alcohol and infiltrated with paraffin. After embedding, samples were cut into 10-μm thick slices with a microtome (Leica Microsystems, Germany). Slices were stained with toluidine blue and observed under a BX61 microscope (Olympus, Japan) as previously reported (Yoon et al., 2017). For GUS staining, samples were incubated in a GUS staining solution containing and 0.1% X-gluc (5-bromo-4-chloro-3-indolyl-β-D-glucuronic acid, cyclohexylammonium salt), 5% methanol, 2% dimethyl sulfoxide (DMSO), 100 mM sodium phosphate (pH 7.0), 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 0.5% Triton X-100, and 10 mM EDTA (pH 8.0) (Yoon et al., 2017). Chlorophylls were removed from the stained samples by immersion in 70% ethanol for 2 h and in 95% ethanol overnight at 65°C.

In situ RNA hybridization

To prepare the RNA probe, a gene-specific fragment was amplified by PCR and cloned in pBluescript II. Digoxigenin (DIG)-labeled sense and antisense RNA probes were prepared as previously reported (Lee and An, 2012; Lee et al., 2007; Yoon et al., 2014). RNA in situ hybridization was performed as previously reported (Lee and An, 2012; Lee et al., 2007; Yoon et al., 2014). Briefly, samples were fixed in 4% paraformaldehyde solution. After dehydration, the paraffin block was infiltrated, and sectioned. Slices from the samples were rehydrated and incubated in a solution containing the DIG-labeled RNA probe overnight at 57°C. The slices were treated with anti-DIG alkaline phosphatase (Roche Molecular Biochemicals, Germany) and stained with nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyl phosphate (NBT/BCIP).

Chromatin immunoprecipitation (ChIP) assay

Basal parts of at least 10 plants at 28 days after germination (DAG) were harvested for the assay. ChIP was performed as previously reported (Haring et al., 2007). For cross-linking, 1-g samples were incubated with 3% formaldehyde solution. After nuclei were isolated, chromatin was sheared to approximately 500- to 1,000-bp length by sonication. Anti-trimethyl-histone H3 (Lys27) monoclonal antibody was used for immunoprecipitation (07-449; Millipore, USA) (Yang et al., 2019). For normalization, we used the fold enrichment method in which the values obtained from the antibody reaction were divided by values from no-antibody controls (Haring et al., 2007). The primers for this analysis are listed in Supplementary Table 2.

Plasmid vector construction

Plasmid vector for osvil2 ostb1 double mutant was constructed by the polycistronic tRNA-gRNA (PTG)/Cas9 method (Xie et al., 2015). The plasmid pRGEB32 was used for Agrobacterium-mediated rice transformation. The pGTR plasmid, which contains a gRNA-tRNA fused fragment, was used as a template to synthesize PTGs. Primers are listed in Supplementary Table 3.

Statistical analyses

The data were analyzed by one-way ANOVA, with pairwise multiple comparison tests carried out by the Tukey honestly significant difference test to compare samples, using the R program (Cohen and Cohen, 2008).

Mutations in OsVIL2 cause a reduction in tiller development

To determine whether tiller development is dependent on flowering time in rice, we studied tiller phenotypes of various flowering mutants that we had generated in japonica varieties Hwayoung and Dongjin. In general, late-flowering mutants produced more tillers, whereas early-flowering mutants had fewer tillers. However, osvil2 mutants exhibited phenotypes which flowered approximately 4 weeks later than the wild-type (WT) (Fig. 1A) and had a reduced tiller number (Fig. 1B).

Neither osvil2 mutants nor WT plants produced axillary tillers at 14 DAG (Figs. 1B and 1C). WT plants started to develop axillary tillers at 21 DAG, while osvil2 mutants did not produce tillers at that stage (Figs. 1B and 1D). WT continued to develop new tillers reaching a total of approximately six per plant, and no new tillers were actively produced after 63 DAG (Figs. 1B, 1E, and 1F). Because floral signals Hd3a and RFT1 start to be generated at approximately 63 DAG (Cho et al., 2018a; 2018b), these signals appear to suppress new tiller generation. In contrast to WT plants, osvil2 mutant plants produced only one axillary tiller at 42 to 56 DAG, and no further tiller development was observed subsequently (Figs. 1B, 1E, and 1F). This observation indicated that OsVIL2 promoted axillary tiller production.

OsVIL2 affects tiller outgrowth

Axillary tillers are produced by two steps. The first involves formation of axillary buds and the second involves outgrowth of the axillary buds (Wai et al., 2017). To investigate whether the phenotype of fewer tillers in osvil2 resulted from defects in bud formation or outgrowth, we examined longitudinal sections of the basal parts of the seedlings. At 14 DAG, both the WT and osvil2-1 seedlings developed AM together with several foliage leaves and prophylls at the axils (Figs. 2A and 2B). At 28 DAG, by which time the WT had actively developed tillers, the mutant had not produced a tiller, although osvil2-1 mutants generated axillary buds similar to WT (Figs. 2C and 2D). These results indicated that osvil2 mutants developed axillary buds normally, but these buds failed to grow out to form tillers.

OsVIL2 is preferentially expressed in the apical meristem

Using transgenic plants expressing the GUS reporter gene under the control of the OsVIL2 promoter, we had previously observed that OsVIL2 is strongly expressed in leaves and preferentially at the basal parts of spikelets (Yoon et al., 2018). In the present study, we found that the gene was specifically expressed at the basal part of the seedlings, especially in axillary buds and vascular tissues (Fig. 3A). To further study the expression pattern of OsVIL2 in the meristematic regions, we performed in situ RNA hybridization studies using the basal parts of seedlings at 20 DAG (Figs. 3B–3F). The experiments showed that OsVIL2 transcripts preferentially accumulated in the AM region (Figs. 3B and 3E) as well as in the axillary bud and surrounding young leaves (Figs. 3C and 3F).

OsVIL2 affects the OsTB1-dependent pathway of tiller outgrowth

To investigate the genes that were affected by osvil2 mutation, we examined expression levels of genes related to tiller development at 13 and 14 DAG when tiller outgrow did not start yet. RT-qPCR analyses of AM development pathway genes in the basal parts of culm tissues showed that the transcripts levels of CKX2, OSH1, LAX1, MOC1, CUC1, and RFL1 were not significantly changed in osvil2-1 (Supplementary Fig. S1) and osvil2-2 (Supplementary Fig. S2) null mutants. We also examined expression patterns of tiller outgrowth-related genes. Auxin-related genes IAA7, IAA20, PIN1, and PIN3 were not significantly altered in the basal parts of culm tissues of the mutant plants (Supplementary Figs. S1 and S2). However, D10 that encodes a biosynthetic enzyme for SL was significantly suppressed in osvil2 mutants (Supplementary Figs. S1L and S2L). Other SL-related genes and tiller outgrowth genes were not changed. These observations indicate that OsVIL2 does not affect most of the tiller-related genes at the stage of axillary bud initiation.

We investigated the expression patterns of genes involved in tiller development in the basal parts of culm tissues at 28 and 30 DAG, at which time point axillary tillers had developed in WT but not in the osvil2 mutant. Expression of the genes controlling axillary bud formation, namely OSH1, LAX1, MOC1, CUC1, and RFL1, were not significantly different between osvil2 null mutants and the WT (Figs. 4B–4F, Supplementary Figs. S3B–S3F). This result was expected because there was no difference in axillary bud formation between osvil2 mutants and WT.

We then studied the genes controlling axillary bud outgrowth. Tiller bud outgrowth is controlled by hormones such as auxins, cytokinins, and SLs. Expression levels of auxin-related genes IAA7, IAA20, PIN1, and PIN3, that are involved in tiller outgrowth, were not significantly altered in osvil2 mutants compared to the WT (Figs. 4G–4J). Expression level of the OsCKX2 gene, encoding an enzyme catabolizing active cytokinins, was not altered (Figs. 4K). These results suggested that OsVIL2 may not be involved with auxins or cytokinins. However, expression of D10, encoding a biosynthesis enzyme for SLs, was markedly down-regulated in the osvil2 mutants (Fig. 4L, Supplementary Fig. S3L). Expression levels of additional SL-biosynthesis genes, D27 and HTD1, as well as SL-signaling genes D3 and D14 were not significantly altered in the mutant (Figs. 4M–4P).

We also analyzed expression levels of transcription factors that control axillary bud outgrowth. Expression of a MADS-box gene OsMADS57, that encodes a positive regulator of axillary bud growth, was not significantly affected by the osvil2 mutation (Fig. 4Q). However, the expression level of another transcription factor gene, OsTB1, was much higher in the mutants than in the WT (Fig. 4R, Supplementary Fig. S4R). Because OsTB1 is a major regulator that inhibits tiller bud outgrowth (Takeda et al., 2003), the reduced tiller outgrowth phenotype in the osvil2 mutants is probably due to the altered expression of OsTB1.

To study the expression patterns of OsTB1 during tiller development, we collected basal parts of culms from the WT and the osvil2 mutant at 4-day intervals starting at 12 DAG (Fig. 5). Analysis of the samples showed that the expression level of OsTB1 was gradually reduced over time in the WT, approaching almost undetectable levels at 24 DAG. However, the transcript level did not decrease in osvil2 mutants at 24 DAG but gradually increased at the later stages (Fig. 5B). This observation supports the concept that OsVIL2 is an inhibitor of OsTB1.

OsVIL2 mediates tiller outgrowth by enrichment of H3K27me3 in OsTB1 chromatin

Previous studies suggest that OsVIL2 interacts with PRC2 to repress target chromatins by regulating H3K27me3. The analysis of expression patterns of tiller outgrowth genes revealed that OsTB1 is a potential target of OsVIL2. To investigate whether OsVIL2 directly repressed OsTB1 through controlling histone modification, we measured H3K27me3 levels of OsTB1 using the ChIP assay, with the aid of H3K27me3 antibodies in osvil2-1 and osvil2-2 mutants (Fig. 6). The assay showed that the 5′ regulatory regions (P2, P3) of OsTB1 chromatin were enriched in the WT, but this enrichment was not observed in the osvil2 mutants (Figs. 6C and 6E). As a negative control, Hd3a chromatin was analyzed, which did not show any enrichment (Figs. 6D and 6F).

To confirm that OsVIL2 directly binds to OsTB1 chromatin, we used OsVIL2-Myc epitope-tagged transgenic lines. As control, we used transgenic plants expressing Myc alone. The ChIP assay showed that the 5′ regulatory regions (P1, P2, and P3) of OsTB1 chromatin were enriched by Myc antibodies in the OsVIL2-Myc epitope-tagged transgenic lines (Fig. 6G) while the Myc control plants did not show the enrichment (Fig. 6H). These results suggest that the OVIL2-PRC2 complex directly binds to the OsTB1 chromatin, thereby promoting outgrowth of tiller buds.

OsVIL2 induces tiller outgrowth by regulating OsTB1 expression

To obtain a genetic relationship between OsVIL2 and OsTB1, we generated osvil2 ostb1 double mutant by CRISPR/Cas9 multiplex editing capability with the endogenous tRNA-processing system (Figs. 7A–7D). For this construction we used polycistronic tRNA-gRNA (PTG)/Cas9 method for simultaneously targeting multiple sites (Fig. 7B). We also generated ostb1 null mutant (Supplementary Fig. S4).

At heading stage in the paddy field, osvil2-1 mutant plants had 4 tillers per plant that is much smaller number compared to 33 of WT (Figs. 7E and 7F). On the contrary, tiller number of ostb1 was 48.6, more than that of WT. The double mutant osvil2 ostb1 showed the phenotypes similar to ostb1 single mutant, having 30 tiller per plant (Figs. 7E and 7F). This experiment suggests that OsVIL2 epistatic to OsTB1.

OsVIL2 is required for outgrowth of tiller buds

In this study, we provided evidence that OsVIL2 promotes tiller outgrowth by suppressing expression of OsTB1, which is an integrator of multiple pathways for tillering (Hussien et al., 2014). For instance, OsTB1 is inhibited by cytokinins and induced by auxins, whereas GAs also influence OsTB1 (Hussien et al., 2014). However, epigenetic regulation of OsTB1 has not previously been reported. We have previously shown that a chromatin-remodeling factor, OsVIL2, associates with H3K27me3 at the target genes to achieve suppression of expression of the target genes (Bemer and Grossniklaus, 2012; Yang et al., 2013). Our ChIP assay indicated that OsTB1 is a potential target of OsVIL2 for tiller outgrowth.

The expression of D10 is significantly down-regulated in osvil2 mutants

We observed that the expression level of D10 was reduced in osvil2 mutants. Because D10 is a gene involved in SL biosynthesis, suppression of D10 expression should result in reduced levels of SLs and should increase tiller number. However, tiller number was reduced in the D10 mutant. This unexpected result may be due to feedback regulation of the SL biosynthesis gene. Suppression of tiller development would inhibit production of SLs, which are major inhibitors of tiller development. D10 encodes a carotenoid cleavage dioxygenase that plays a key role in SL biosynthesis (Arite et al., 2007). Alternatively, reduction of tiller number might stimulate cytokinin signaling or inhibit auxin signaling that function upstream of D10 (Zhang et al., 2010).

OsVIL2 binds to the PRC2 complex

The PRC2 complex plays important roles in the development of plants and animals by suppressing target gene expression through histone modification (Jeong et al., 2015). OsVIL2 is highly homologous to Arabidopsis VIN3 and VIN3-like genes that contain conserved motifs of the PHD finger, FNIII and VID domains (Greb et al., 2007). The FNIII domain of OsVIL2 binds to EMF2b, which is a component of the PRC2 complex (Yang et al., 2013). Loss-of-function mutations of EMF2b result in late flowering and abnormal floral organ development, which are phenotypes similar to those observed from the osvil2 mutant (Conrad et al., 2014; Yang et al., 2013). It will be necessary to carry out further experiments to determine whether EMF2b also plays a role in tiller outgrowth and shoot branching.

Tillering ceases at the floral transition stage

We showed that new tillers were developed during the vegetative phase and that tiller development stopped at the floral transition stage. In most late-flowering mutants that have a longer vegetative period than the WT, tiller numbers were higher than in the WT, due probably to increased vegetative growth. It is also possible that the floral transition signal directly inhibits a major regulatory gene that controls tiller bud growth. Alternatively, floral signals could affect biosynthesis of or signaling by the hormones that control tiller development. Interestingly, the osvil2 mutant exhibited fewer tillers than did the WT, even though the mutant flowered later than the WT. We postulate that OsVIL2 directly promotes axillary bud growth, whereas other flowering genes influence tiller number indirectly by the action of florigens.

This work was supported in part by a grant from the Next Generation BioGreen 21 Program (Plant Molecular Breeding Center; No. PJ013210), Rural Development Administration, Republic of Korea to GA and by the Republic of Korea Basic Research Promotion Fund to JY (grant No. NRF-2018R1A6A3A11047894).