RETINOBLASTOMA-RELATED Genes Specifically Control Inner Floral Organ Morphogenesis and Pollen Development in Rice

Members of a cell-cycle-related gene family maintain floral meristem function to ensure floral organ origination and regulate tapetum degradation to promote pollen formation in rice.

Abstract

RETINOBLASTOMA-RELATED (RBR) is an essential gene in plants, but its molecular function outside of its role in cell cycle entry remains poorly understood. We characterized the functions of OsRBR1 and OsRBR2 in plant growth and development in rice using both forward- and reverse-genetics methods. The two genes were coexpressed and performed redundant roles in vegetative organs but exhibited separate functions in flowers. OsRBR1 was highly expressed in the floral meristem and regulated the expression of floral homeotic genes to ensure floral organ formation. Mutation of OsRBR1 caused loss of floral meristem identity, resulting in the replacement of lodicules, stamens, and the pistil with either a panicle-like structure or whorls of lemma-like organs. OsRBR2 was preferentially expressed in stamens and promoted pollen formation. Mutation of OsRBR2 led to deformed anthers without pollen. Similar to the protein interaction between AtRBR and AtMSI1 that is essential for floral development in Arabidopsis, OsMSI1 was identified as an interaction partner of OsRBR1 and OsRBR2. OsMSI1 was ubiquitously expressed and appears to be essential for development in rice (Oryza sativa), as the mutation of OsMSI1 was lethal. These results suggest that OsRBR1 and OsRBR2 function with OsMSI1 in reproductive development in rice. This work characterizes further functions of RBRs and improves current understanding of specific regulatory pathways of floral specification and pollen formation in rice.

RB (Retinoblastoma) was originally identified as a tumor suppressor gene in animals (Friend et al., 1986). A basic and core function of its protein is to control cell proliferation via regulating cell cycle entry. RB protein is also involved in regulating cell differentiation and organ specification (Nead et al., 1998; Liu et al., 1999; Thomas et al., 2003; Toppari et al., 2003), but most of the function and mechanism of RB still remains unknown. RBR (RB-Related) genes are widely distributed in plants (Gutzat et al., 2012). The mechanism of controlling the cell cycle is also highly conserved in RBR genes (Miskolczi et al., 2007; Gutzat et al., 2012). However, plants possess a set of special cellular structures and developmental patterns that are distinguished from those of animals, such as the cell wall, cell totipotency, postembryonic organ formation, and the development of stem cells and meristem from a specific narrow area. These characteristics of plants may reflect differences between plant RBR and animal RB in regards to spatiotemporal expression pattern and functional mechanism (Ach et al., 1997; Sabelli et al., 2005; Miskolczi et al., 2007) as well as specific molecular interactions and responses (Mironov et al., 1999; Wildwater et al., 2005). In addition, differences also likely exist between monocot and dicot plants in RBR function and regulatory mechanisms because of varying RBR gene number (Lendvai et al., 2007; Miskolczi et al., 2007).

RBR genes are essential for plant growth and development and exert a broad range of cellular functions apart from cell cycle control; however, RBR gene number is small in plant genomes, with only one RBR gene in dicots and two to four in monocots (Lendvai et al., 2007). The study of RBR gene function is limited, and much of its molecular function outside the cell cycle entry remains unknown, probably due to either the gametophytic lethality of loss-of-function mutation of single-copy RBR (such as in Arabidopsis [Arabidopsis thaliana]) or the redundant roles among different RBR family members (such as in maize [Zea mays]). Nevertheless, by inhibiting or enhancing the expression of RBR genes, significant and important progress has still been achieved in interpreting the functions and mechanisms of RBR genes in plants. Studies have shown that in higher plants RBR genes are essential for various biological processes such as vegetative growth (Borghi et al., 2010; Perilli et al., 2013), reproductive development (Ebel et al., 2004; Johnston et al., 2008; Sabelli et al., 2013), and axillary bud dormancy (Shimizu-Sato et al., 2008). Among these, RBR function in gamete decision in Arabidopsis (Ebel et al., 2004; Johnston et al., 2008) and in endosperm formation in maize (Sabelli et al., 2013) have been characterized in the most detail. Loss of AtRBR function causes female gametophyte lethality due to ovule abortion and male gametophyte defects causing a proportion of pollen grains to be malformed and nonviable (Ebel et al., 2004; Johnston et al., 2008). Maize has four RBR genes with at least two functionally distinct types (Sabelli et al., 2005, 2009, 2013). Of these, RBR1 plays a major role during maize endosperm development (Sabelli et al., 2013).

The multifunctionality of an RBR gene is thought to be facilitated through interactions of its protein with various binding partners (Park et al., 2005; Desvoyes et al., 2006; Borghi et al., 2010; Gutzat et al., 2012; Perilli et al., 2013; Sabelli et al., 2013; Desvoyes et al., 2014). To date, the precise mechanism of RBR function in organ specification remains poorly understood. Only a small number of cooperative interactions involving RB protein (pRb) in animals or RBR in plants have been characterized. One of the best-characterized molecular mechanisms is the interaction of RBR with the Polycomb repressive complex2 (PRC2) during plant-specific developmental events, which contains the important component MSI1 (Multicopy Suppressors of the Iral; Köhler et al., 2003; Schönrock et al., 2006; De Lucia et al., 2008; Derkacheva et al., 2013; Kuwabara and Gruissem, 2014; Steinbach and Hennig., 2014). Studies have found that MSI1 and RBR cooperate to regulate developmental transitions and cell fate determination (Ebel et al., 2004; Johnston et al., 2008; Jullien et al., 2008; Dumbliauskas et al., 2011). In Arabidopsis, both AtRBR and AtMSI1 are required for floral meristem (FM) activation and floral organ primordia initiation (Hennig et al., 2003; Borghi et al., 2010). Reduction of AtMSI1 expression dramatically affects reproductive development, preventing the primary shoot apical meristem from developing into organs after the transition to flowering and causing flowers developing on floral shoots from axillary meristems to display a progressive loss of floral morphology, including a reduction in the size of petals and stamens and the development of carpel-like sepals (Hennig et al., 2003). Partial loss of AtRBR function and partial loss of AtMSI1 function result in similar abnormal flower phenotypes (Hennig et al., 2003; Borghi et al., 2010), suggesting that the two genes probably play similar roles in regulation of floral organ formation.

Rice (Oryza sativa) possesses two RBR genes, namely OsRBR1 and OsRBR2. These genes are similar in sequence and possess the same conserved domains, although they belong to different clades in the phylogenetic tree of RBR genes (Lendvai et al., 2007). To date, the functions and regulatory mechanisms of both OsRBR1 and OsRBR2 remain poorly understood. In this study, we characterized the functions of these two RBR genes in reproductive development in rice through both forward and reverse genetics methods. We found that the two genes had overlapping roles in the vegetative developmental phase but functioned differently in the reproductive developmental phase, with OsRBR1 being a specific and critical determiner for the establishment and maintenance of FM function, whereas OsRBR2 is an essential factor for stamen and pollen development. In addition, OsRBR1 and OsRBR2 physically associated with OsMSI1, suggesting that their functions might be dependent on interaction with OsMSI1 at the protein level. Our results provided further insight into the distinctive function of RBR in plant development and the specific regulatory pathways of floret and pollen formation in rice.

RESULTS

Morphological Phenotype of pis1 Mutant

We identified two new spikelet mutants from an indica rice cultivar MH86 generated by irradiation or tissue culture, respectively. The two mutants showed almost the same phenotype. Genetic analysis suggested that their mutations were allelic. Both mutants exhibited normal morphology except for their spikelets, which were noticeably bulgy with opened glumes (Fig. 1, A–C). A mature wild-type floret of rice comprises a lemma and a palea, two lodicules, six stamens, and a carpel (Fig. 1D). Compared with wild type, the mutants showed obvious alteration or complete degeneration in all floral organs except for lemma (Fig. 1, E–K). Most (over 80%) of the mutant paleas developed into lemma-like organs (LLOs), in which the palea membranous margin tissue was replaced with lemma-like tissues (Fig. 1, C and E–I). Only a small proportion (below 20%) of mutant spikelets maintained a wild type palea identity (Fig. 1, J–K). The most remarkable alterations inside the mutant floret were that the lodicules, stamens, and pistil all disappeared and were replaced with either several whorls of LLOs surrounding an elongated central axis (type I; Fig. 1, E–G) or a panicle-like organ (PLO) with a few or no LLOs around it (type II; Fig. 1, H–K). These phenotypes suggested that the floral organ identity and meristem determinacy of the inner three whorls were completely lost in the mutant florets. Among a total of 500 mutant spikelets investigated, approximately two-fifths (38%) and three-fifths (62%) belonged to type I and type II, respectively. Among the type II spikelets, approximately one-fourth (24%) PLOs had only one primary branch (Fig. 1, H and I), whereas the rest had two or more (Fig. 1, J and K). Considering that most of the mutant spikelets displayed a PLO in the central zone, we named the mutant phenotype panicle in spikelet1 (pis1) and the two allelic mutants pis1-1 and pis1-2.

Figure 1.

Phenotype of the pis1-1 spikelets. A and B, Parts of wild-type (WT; left) and mutant (right) panicles. C, A typical wild-type (left) or pis1-1 (right) spikelet. D, An opened wild-type spikelet. E to G, Type I pis1-1 spikelets (opened). H to K, Type II pis1-1 spikelets (opened). White arrows in C and E to I indicate LLOs transformed from paleas. White arrowheads in G, J, and K indicate LLOs transformed from lodicules. le, Lemma; pa, palea. Scale bars = 2 mm.

Developmental Process of pis1 Spikelets

To further clarify the spikelet morphogenesis of pis1, we examined the process of floral development in pis1 and wild type by scanning electron microscopy. The floral primordia of wild-type spikelets developed sequentially from the periphery to the center in an order of empty glumes, lemma/palea, two lodicules, six stamens, and a carpel (Fig. 2, A–D). In pis1, the primordia of empty glumes and lemma emerged normally as expected (Fig. 2, E and F), which then developed into the corresponding organs (Fig. 1, C–F, H, J, and K). The primordium of palea also emerged (Fig. 2, F and H), but, as mentioned above, only a small proportion then developed into palea (Fig. 1, F and H), whereas most developed into LLOs (Fig. 1, C–E, J, and K). Inside the mutant floret, the presumptive FM looked like a hemispherical hump (Fig. 2E), which differed considerably to the flatter FM in wild type (Fig. 2A). As development progressed, the visible bulged floral organ primordia seen in the wild type (Fig. 2B) did not emerge from the mutant FM (Fig. 2F). Instead, in some cases, an axis extended from the center of the mutant FM (Fig. 2G), and multiple whorls of lemma-like primordia emerged continuously (Fig. 2, G–K), which then developed into LLOs (Fig. 1, C–E, J, and K). More frequently, one or more branch-like primordia differentiated from the mutant FM (Fig. 2L), which further developed into branches and spikelet-like organs consisting of multiple whorls of LLOs (Fig. 2, M and N), finally forming PLOs (Fig. 1, F and K). Occasionally, one or two anther-shaped primordia could be also observed in the mutant spikelet (Fig. 2O). The above observations indicate that although the lemma and palea primordia could differentiate in pis1, the origination of inner floral organ primordia was entirely aborted, resulting in complete loss of FM identity of the inner three whorls.

Figure 2.

SEM observation of pis1-1 spikelets during floral primordium initiation. A to D, Wild-type (WT) spikelets. E to O, pis1-1 spikelets. E, The presumptive floral meristem (FM) appeared as a hemispherical hump. F, No inner floral organ primordia differentiated. G, A central axis extended from the presumptive FM. H to K, Whorls of lemma-like primordia differentiated around the presumptive FM. L, Lateral branch-like primordia differentiated from the central presumptive FM. M and N, A panicle-like structure developed, consisting of multiple branches and spikelet-like organs. O, An anther-shaped primordium differentiated (indicated by white arrow). Lemma and palea were removed in D, G, H, and L. e.g., Empty glume; le, lemma; pa, palea; lop, lodicule primordia; st, stamen; stp, stamen primordia; pi, pistil; fm, floral meristem. Scale bars = 50 μm.

Mutations in pis1-1 and pis1-2

Genetic analysis showed that pis1-1 and pis1-2 were allelic mutants controlled by a single gene, and their mutant phenotype was stable across different cropping seasons. To isolate the PIS1 gene, we developed an F2 population from a cross between pis1-1 heterozygote and variety DZ60. Using publicly available rice microsatellite-series simple sequence repeat (SSR) markers as well as some new SSR markers and insertion or deletion (InDel) markers developed in this study (Supplemental Table S1), we fine-mapped PIS1 to a 32-kb interval delimited by markers InDel4 and InDel5 on the terminal region of the long arm of chromosome 8. According to the annotation from the NCBI database, the interval contained five open reading frames including OsRBR1 (Fig. 3A). Sequencing analysis showed that a single base transition from G to A in the third exon of OsRBR1 resulted in a premature stop codon (TAG) in pis1-1 (Fig. 3A). No base alteration was found between wild type and pis1-1 in the other four genes. Furthermore, a 28-bp deletion was detected in the eleventh exon of OsRBR1 in pis1-2, resulting in a premature transcription termination (PTT; Fig. 3A). These results strongly suggested that OsRBR1 is a very likely candidate for PIS1.

Figure 3.

Map-based cloning and functional analysis of OsRBR1. A, Map-based cloning and functional domain analysis of PIS1. The number under each marker indicates recombination events that occurred between the marker and PIS1. A single base transversion (G to A) and 28-bp deletion were found inside the third and eleventh exons in pis1-1 and pis1-2, respectively, both of which resulted in premature termination of RBR transcription. B, Genetic complementation test of the pis1-1 mutant. The wild-type phenotype was recovered in the transgenic plants. C, Constitutive overexpression of OsRBR1. The transgenic plants showed normal panicle phenotype. COM, Complementation; OE, overexpression.

OsRBR1 contains 18 exons encoding a protein of 1011 amino acids. The PTT in pis1-1 resulted in a truncated OsRBR1 consisting of only 81 amino acids of the N terminus, which had lost all the functional domains (Fig. 3A), whereas the OsRBR1 coding sequence in pis1-2 encodes 674 amino acids of the original N terminus plus a new 30-amino acid sequence, which lacked the B- and C-terminal domain (Fig. 3A). This sequence analysis suggests that both pis1-1 and pis1-2 likely contain null alleles. To validate the candidate gene, we constructed the plasmid pBWA(v)H-OsRBR1 carrying an 8574-bp wild type genomic DNA sequence, which included the entire OsRBR1 gene and its upstream promoter region. By introducing the plasmid into pis1-1, the mutant phenotype was rescued (Fig. 3B), confirming that OsRBR1 is PIS1. To further characterize the function of the OsRBR1 gene, we introduced the plasmid pCXUN carrying a 3122-bp cDNA sequence of OsRBR1 driven by the Ubi promoter into variety ZH11 and obtained seven transgenic lines with strong overexpression of OsRBR1 (Supplemental Fig. S1). None of the overexpression lines exhibited a visibly altered phenotype during either vegetative growth (Supplemental Fig. S1) or reproductive development (Fig. 3C), suggesting that increased OsRBR1 expression may not obviously affect the development of rice.

Expression of Genes Related to Flower Morphogenesis in pis1

The pis1 mutant phenotype and expression pattern of OsRBR1 suggested that the gene is involved in the control of FM identity and floral organ specification. To investigate the molecular mechanism behind this, we detected the transcript abundance of 12 important floral homeotic genes by reverse transcription quantitative PCR (RT-qPCR) in young panicles of 2 to 30 mm in length from pis1-1 and wild-type plants. All of the 12 genes examined exhibited statistically significant differential expression between pis1-1 and wild type. Most genes (including class B genes OsMADS2, OsMADS4, and OsMADS16; class C genes OsMADS3 and OsMADS58; class E genes OsMADS7, OsMADS8, and OsMADS19; and Zinc-finger genes OsJAG and DL) were downregulated, and only two class E genes (OsMADS1 and OsMADS5) were upregulated (Fig. 4). These results suggest that OsRBR1 could regulate the expression of floral-organ-specification-related genes during floral organ differentiation.

Figure 4.

Expression of main floral organ specification genes in pis1-1 young panicles. The error bars indicate sd. The difference between wild type (the expression levels of wild type set to 1) and the mutant were statistically significant (P < 0.05, Student’s t test) in all the genes analyzed.

Phenotype Resulting from OsRBR2 Mutation

To characterize the function of OsRBR2, we used the CRISPR/Cas9 system to create OsRBR2 loss-of-function mutants in MH86. In total, 26 transgenic lines with effective sequence editing in the open reading frame (third exon) of OsRBR2 were obtained (Table 1), which could be classified into seven types. These included 12 lines of one-base insertion between the 209th and 210th bases (types 1–4), 10 lines of four-base deletion from the 205th to the 208th base in the third exon (type 5), two lines of three-base deletion from the 205th to the 208th base (type 6), and two lines of 15-base deletion from the 205th to the 219th base (type 7). Both the one-base insertion and the four-base deletion could cause PTT, leading to a truncated polypeptide of either 80 or 68 amino acids, preventing formation of the “pocket structure” and therefore creating nonfunctional proteins (Fig. 5; Supplemental Fig. S2B). Hence, OsRBR2 mutated by the one-base insertion and the four-base deletion created null alleles, which we denoted Osrbr2-1 and Osrbr2-2, respectively. The three-base deletion and the 15-base deletion mutations only resulted in the loss of one or a few amino acids, which did not affect any functional domain of the RBR protein. Therefore, these mutant alleles were probably still functional. Indeed, these two types of mutants did not display phenotypic changes in comparison with wild type.

Table 1. Results of OsRBR2 Sequence Editing.

Del, deletion; Ins, insertion; ORF, Open reading frame; PTT, premature transcription termination

Type	Line	DNA Sequence	ORF	Protein	Phenotype
0	MH86 (CK)	caagctgaaggagacca	2937 base	978 amino acids	Normal
1	3/22/24	caagcttgaaggagacca	Ins 209th, PTT	80 amino acids	Male sterile
2	8/9/23/25	caagcatgaaggagacca	Ins 209th, PTT	80 amino acids	Male sterile
3	5/12/13/20	caagcctgaaggagacca	Ins 209th, PTT	80 amino acids	Male sterile
4	21	caagcgtgaaggagacca	Ins 209th, PTT	80 amino acids	Male sterile
5	1/2/4/6/10/11/15/17/18/26	c—-tgaaggagaccca	Del 205th-208th, PTT	68 amino acids	Male sterile
6	7/14	c—ctgaaggagaccca	Del 205th-207th	977 amino acids (Del 69th)	Normal
7	16/19	c—————-a	Del 205th-219th	973 amino acids (Del 69th-73th)	Normal

Figure 5.

Phenotypes of OsRBR2 knockout and overexpression transformants. A, Wild-type (left) and Osrbr2 (right) whole plants after heading. B and C, Panicles of wild type (left) and Osrbr2 (right) at heading stage (B) and maturation stage (C). D, Spikelets of wild type (left) and Osrbr2 (right). E and F, Spikelets of wild type (E) and Osrbr2 (F) with lemma and palea removed. G to J, I₂-KI stained anthers and pollen grains of wild type (G and I) and Osrbr2 (H and J). K to L, Mature embryo sacs of wild type (K) and Osrbr2 (L). M to O, Plants (M) and panicles (N and O) of wild type (left) and a transformant overexpressing OsRBR2 (OE-6, right). OE, Overexpression; lo, lodicule; st, stamen; pi, pistil; AC, antipodal cell; CC, central cell; EC, egg cell; SC, synergid cell; PN, polar nucleus. Scale bars = 10 cm (A and M), 1 cm (B, C, N, and O), 1 mm (D–F), and 100 μm (G–L).

Osrbr2-1 and Osrbr2-2 showed similar phenotypes. Almost all important agronomic traits appeared normal, including vegetative growth, plant structure and shape, heading date, plant height, tiller number, and so on (Fig. 5A). However, Osrbr2-1 and Osrbr2-2 plants were completely sterile and were unable to produce seeds under different experimental conditions (Fig. 5, B and C). The mutant spikelet and inner floral organs all appeared normal except for the anthers, which were much smaller in size and lighter in color compared with those of wild type (Fig. 5, D–F) and contained no pollen grains (Fig. 5, G–J). Observation by whole-mount stain-clearing laser-scanning confocal microscopy indicated that there was no apparent difference between Osrbr2 and wild type in female gametophytic development, and the mature embryo sacs of Osrbr2 also looked normal (Fig. 5, K and L). In addition, when pollinated with wild-type pollen grains, Osrbr2 spikelets could produce normal seeds. These results suggest that the pistils in Osrbr2 are normal and the sterility of Osrbr2 is the result of pollen abortion. Interestingly, Osrbr2 did not show obvious panicle enclosure (Fig. 5A) as usually seen in rice male sterile lines.

To further assess the function of the OsRBR2 gene, we introduced the plasmid pTCK303 carrying a 2989-bp cDNA sequence of OsRBR2 driven by the Ubi promoter into ZH11. Ten transgenic lines with obvious overexpression of OsRBR2 were obtained (Supplemental Fig. S3, A–C). Similar to that of OsRBR1, the OsRBR2 overexpression lines exhibited no visible phenotypic alterations (Fig. 5, N and O), suggesting that increased OsRBR2 expression may not obviously affect the development of rice.

Process of Anther and Pollen Development in Osrbr2

The process of anther development can be divided into 14 stages in rice (Zhang and Wilson, 2009). To investigate the anther and pollen development process in Osrbr2, we compared transverse sections of Osrbr2 and wild-type anthers at various developmental stages. No visible difference was observed between the anthers of Osrbr2 and wild type before stage 9. At stage 9, whereas the wild-type tapetum began to display degradation and its margin showed serration (Fig. 6A), the Osrbr2 tapetum remained intact (Fig. 6B). At stage 10, the wild-type anther was filled with round-shape vacuolated microspores and its tapetum was largely degraded (Fig. 6C), whereas the Osrbr2 anther was filled with irregular vacuolated microspores and its tapetum remained largely intact, with no degradation as is genetically programmed (Fig. 6D). At stages 11 and 12, the tapetum was almost completely degraded and the microspores had developed into falcate or round pollen grains in wild type (Fig. 6E), whereas the Osrbr2 tapetum was only partially degraded and the microspores had started to degenerate rather than develop into pollen grains (Fig. 6F). At stages 13 and 14, the wild-type anther was filled with mature pollens, and its wall layers (including tapetum, middle layer, and endothecium) had all completely disappeared (Fig. 6, G and I). In the Osrbr2 anther, however, although the wall layers had also disappeared by this stage, the microspores were all completely disintegrated, resulting in an empty anther with only debris inside (Fig. 6, H and J). These observations indicated that microspore mother cells in Osrbr2 could undergo meiosis to produce microspores; however, Osrbr2 microspores could not further develop into pollen grains and degenerated, probably due to the delay in tapetum degradation.

Figure 6.

Anther and pollen development and the expression patterns of three tapetum degradation-related genes in Osrbr2. A to J, Transverse section of anthers at different development stages in wild type and Osrbr2. Scale bars, 50 μm. K to M, Expression of TDR, EAT1, and OsC6 in young panicles and anthers (at stages 9–11) in wild type and Osrbr2. The expression levels of young in wild type set to 1; the error bars indicate sd. The difference between wild type and the mutant were statistically significant (P < 0.05, Student’s t test) in all the genes analyzed. T, Tapetum.

A number of genes are known to control tapetum degradation in rice. To validate the above cellular observations, we examined the expression of three tapetum degradation genes, including Tapetum Degeneration Retardation (TDR; Li et al., 2006; Ji et al., 2013), Eternal Tapetum1 (EAT1; Niu et al., 2013), and OsC6 (Zhang et al., 2010). Each of these three genes was highly expressed in anthers (at stages 9–11) as expected, and their expression levels were all dramatically decreased in the mutant compared with that in wild type (Fig. 6, K–M). This reduced expression of tapetum degradation genes is consistent with the observation that tapetum degradation is retarded in Osrbr2.

Effect of Double Mutation of OsRBR1 and OsRBR2

During the vegetative growth phase, no visible alterations were observed in either Osrbr1 or Osrbr2 mutants, suggesting that mutation of OsRBR1 and OsRBR2 may not affect vegetative growth. The possible reason for this could be either that the two genes have no role in vegetative growth or that they are functionally redundant in vegetative growth. To investigate these hypotheses, we attempted to create an Osrbr1 Osrbr2 double mutant by crossing pis1-1/+ plants with Osrbr2-1/+ plants. The F1 plants all showed normal phenotypes as expected. Theoretically, one-fourth of the F1 plants would have the double heterozygous genotype (pis1-1/+ Osrbr2-1/+), from which the F2 plants would be segregated at these two RBR loci simultaneously, and one-sixteenth of the F2 plants would be the double mutant. It is known from the above results that Osrbr1 should be epistatic to Osrbr2 for reproductive development because no anthers were generated in Osrbr1. So, we expected that the double mutant should have a similar floral phenotype to that of Osrbr1. In order to identify double mutants among the F2, we identified Osrbr1 plants according to their morphology first and then examined whether they were homozygous for Osrbr2 or not by sequencing. We obtained more than 3000 F2 plants from the double-heterozygous F1 plants and observed many plants that showed pis1-1 or Osrbr2-1 phenotypes among them. However, we did not find any double mutants, although we examined at least 100 pis1-1 plants for the Osrbr2-1 allele, of which one-fourth should theoretically have the homozygous genotype of Osrbr2-1. This result indicated that there were no double mutants among the F2 plants, suggesting that double mutation of the two RBR genes is lethal.

Expression Patterns of OsRBR1 and OsRBR2

We employed RT-qPCR to analyze the expression patterns of OsRBR1 and OsRBR2 (Supplemental Fig. S4). The two genes were both strongly expressed in vegetative organs (including roots, stems, and leaves) and showed similar expression levels, with the highest levels being in young leaves. The expression of OsRBR2 in leaves was maintained at a high level during leaf development, whereas OsRBR1 expression decreased rapidly with leaf growth and exhibited a very low expression level in mature leaves. In young panicles, OsRBR1 also had a high expression level, but the expression level of OsRBR2 was much lower.

To confirm the results of RT-qPCR analysis, we fused the promoters of OsRBR1 and OsRBR2 independently to the GUS reporter gene and transferred the resulting constructs into rice via Agrobacterium tumefaciens. The results of GUS activity examination agreed with the results of RT-qPCR analysis. In the vegetative growth phase, OsRBR1 displayed strong expression in actively developing organs, including young roots (Fig. 7A), young root hairs (Fig. 7, B and C), elongating stems (Fig. 7D), and developing leaves (Fig. 7F), but displayed much weaker expression in all mature vegetative organs, such as mature roots (Fig. 7C), stems (Fig. 7E), and leaves (Fig. 7G). The expression of OsRBR2 in vegetative tissues was similar to that of OsRBR1 (Fig. 7, H–N) except for in mature leaves, where its expression remained much higher than that of OsRBR1 (Fig. 7, G and N).

Figure 7.

GUS and in situ hybridization detection of OsRBR1 and OsRBR2 expression in different tissues or organs. A to G, GUS detection of OsRBR1 transcript accumulation during vegetative growth in roots (A–C), stems (– E), and leaves (F and G). H to N, GUS detection of OsRBR2 transcript accumulation during vegetative growth in roots (H–J), stems (K and L), and leaves (M and N). O to R, GUS detection of OsRBR1 transcript accumulation during reproductive development in young panicle (O), developing spikelet (P), and mature spikelet (Q and R). S to V, GUS detection of OsRBR2 transcript accumulation during reproductive development in young panicle (S), developing spikelet (T), and mature spikelet (U and V). Lemma and palea were removed in R and V. W1, X1, and Y1, RNA in situ hybridization detection of OsRBR1 expression during the differentiation and development of floral organ primordia. W2, X2, and Y2, RNA in situ hybridization with the sense probe (as negative control) at the similar stage corresponding to that shown in W1, X1, and Y1, respectively. fm, Floral meristem; le, lemma; pa, palea; lo, lodicule; st, stamen; pi, pistil. Scale bars = 50 µm.

Unlike in the vegetative developmental phase, the expression patterns of the two RBR genes were quite different in the reproductive phase. OsRBR1 was strongly expressed in early young panicles (Fig. 7O). In the developing flower, OsRBR1 expression was high in lemma/palea (in particular in young tissue; Fig. 7, P and Q), lodicules, and pistil, but very low in stamens (Fig. 7R). On the contrary, OsRBR2 expression was very weak in early young panicles (Fig. 7S), developing spikelets (Fig. 7, T and U), and in all the floral organs of a mature flower, except for in anthers where it showed very strong expression (Fig. 7, U and V). To further characterize the tissue-specific expression of OsRBR1 in spikelets, we performed in situ hybridization experiments. The results were consistent with the GUS expression analyses indicating that OsRBR1 expression was initially high in the primordia of inner three-whorl organs (Fig. 7, W1 and X1) but dramatically decreased during the development of these organs (Fig. 7Y1). No hybridization signals were detected by the OsRBR1-specific sense probe (Fig. 7, W2, X2, and Y2).

Subcellular Localization of OsRBRs and Their Interactions with OsMSI1

To investigate the intracellular location of OsRBR1 and OsRBR2, we constructed an expression vector containing a fusion of GFP and OsRBR1/2 to transform onion epidermal cells using Agrobacterium. GFP fluorescence was observed specifically in the nucleus (Supplemental Fig. S5), suggesting that both OsRBR1 and OsRBR2 are localized to the nucleus.

To clarify the molecular mechanism of RBR protein involvement in reproductive development in rice, we screened the cDNA library of rice young panicle by yeast two-hybrid methods using the full-length OsRBR1 protein as bait, which did not show substantial self-activation. In total, nine proteins were identified, including OsMSI1 (Supplemental Table S1). There were five MSI-like proteins in Arabidopsis and three AtMSI1 homologs in rice. Among these, OsMSI1 showed the homology to AtMSI1 (Supplemental Fig. S6) with 85.85% sequence similarity. It has been reported that AtMSI1 is an important interaction partner of AtRBR and is essential for development in Arabidopsis (Hennig et al., 2003; Ebel et al., 2004; Johnston et al., 2008; Jullien et al., 2008; Borghi et al., 2010; Dumbliauskas et al., 2011). The high sequence homology between OsMSI1 and AtMSI1 and analogous interaction relationships of OsMSI1 and AtMSI1 with RBR proteins imply that OsMSI1 might be an important interaction partner of OsRBR and also essential for development in rice. Hence, we focused our study on OsMSI1.

To validate the interaction between OsRBR1 and OsMSI1 and to investigate whether there were interactions between OsRBR1 and the other two rice MSI proteins, we conducted bimolecular fluorescence complementation (BiFC) assays. The coding sequences of OsRBR1, Osrbr1-1, and three MSI genes were fused to the N- or C-terminal of YFP in vectors pCAMBIA1300S-YN and pCAMBIA1300S-YC, respectively. In addition, we used BiFC to investigate the interactions between OsRBR2 and the three MSI proteins. The coding sequences of OsRBR2, Osrbr2-1, and the three MSI genes were fused to the N- or C-terminal of GFP in vectors pCAMBIA1300S-GN or pCAMBIA2300S-GC, respectively. No fluorescence was observed resulting from the combined expression of YC-MSI+YN and GC-MSI+GN or YC+YN-OsRBR1 and GC+GN-OsRBR2 . YFP fluorescence was observed in the nuclei of leaf epidermis cells of Nicotiana benthamiana cotransformed with YN-OsRBR1+YC-OsMSI1 (Fig. 8, A–C), and GFP fluorescence was observed in the nuclei of rice protoplasts cotransformed with GN-OsRBR2+GC-OsMSI1 (Fig. 8, D–F). However, no fluorescence was observed resulting from the combined expression of YN-Osrbr1-1+YC-OsMSI1 (Fig. 8, G–I), GN-Osrbrr2-1+GC-OsMSI1(Fig. 8, J–L), YC-OsMSI2+YN-OsRBR1 (Fig. 8, M–O), YC-OsMSI3+YN-OsRBR1 (Fig. 8, P–R), GC-OsMSI2+GN-OsRBR2 (Fig. 8, S–U), and GC-OsMSI3+GN-OsRBR2 (Fig. 8, V–X). These results indicate that both OsRBR1 and OsRBR2 physically interact with OsMSI1 in vivo, but neither protein binds to OsMSI2 and OsMSI3.

Figure 8.

BiFC analysis of OsMSI-OsRBR interaction in vivo. Analysis of interaction between OsMSI1 and OsRBR1 (A–C), OsMSI1 and OsRBR2 (D–F), OsMSI1 and Osrbr1-1 (G–I), OsMSI1 and Osrbrr2-1 (J–L), OsMSI2 and OsRBR1 (M–O), OsMSI2 and OsRBR2 (P–R), OsMSI3 and OsRBR1 (S–U), or OsMSI3 and OsRBR2 (V–X). A, D, G, J, M, P, S, and V, Bright field. B, H, N, and T, YFP filter. E, K, Q, and W. GFP filter. C, F, I, L, O, R, U, and X, Merged images. Scale bars = 50 µm (A–I) and 10 µm (J–R).

Expression and Function of OsMSI1

We employed RT-qPCR to analyze the expression pattern of OsMSI1. The results indicated that OsMSI1 was expressed in all the tissues analyzed, including roots, stems, leaves, and panicles (Supplemental Fig. S7A). The highest expression level was found in early young panicles, but gene expression decreased with the progress of panicle development (Supplemental Fig. S7A). This expression pattern was similar to that of OsRBR1 (Supplemental Fig. S4).

To further investigate the relationship between OsMSI1 and RBR genes, we examined the expression of OsMSI1 in pis1-1 and Osrbr2-1. Compared with that in wild type, the transcription of OsMSI1 in young panicles was significantly upregulated in pis1-1 (Supplemental Fig. S7B) but was unchanged in Osrbr2 (Supplemental Fig. S7C). Instead, Osrbr2 showed significantly downregulated expression of OsMSI1 in anthers at the pollen mitosis stage (Supplemental Fig. S7C). The variation in OsMSI1 expression in the two RBR mutants mirrored the different expression patterns of the two RBR genes in the reproductive phase, in that OsRBR1 was strongly expressed in young panicles and flowers except for stamens, whereas OsRBR2 displayed opposite expression patterns (Fig. 7). These results suggest that the two RBR genes might participate in the regulation of OsMSI1 expression.

To clarify the function of OsMSI1, we attempted to use CRISPR/Cas9 system to knock out OsMSI1. In total, we performed 10 independent rounds of transformation using over 1000 calli. However, no transgenic plants were recovered. All the transformed calli became brown and died for unknown reasons (Supplemental Fig. S7D). This result implies that OsMSI1 is indispensable for rice and that loss of OsMSI1 function is lethal.

DISCUSSION

In this study, using positional cloning and CRISPR/Cas9 editing as well as morphological and microscopic observations, we demonstrated that the two rice RBR genes play critical roles in floral organ morphogenesis and pollen development. We characterized the functions of the two genes by investigating their temporal and spatial expression patterns, regulatory relationships with other genes involved in floral organ morphogenesis and pollen development, subcellular localization of proteins, and protein interactions with OsMSI1. The results of this study clarify the functions of the two RBR genes in rice and provide an insight into the possible mechanism underlying their functions.

OsRBR1 Specifically Controls Inner Floral Organ Formation by Regulating Floral Organ Identity Gene Expression

It was shown in this study that the mutants of OsRBR1 displayed observable defects only inside florets, whereby the lodicules, stamens, and pistil all disappeared and were replaced with either an ectopic PLO or whorls of LLOs (Fig. 1). This mutant phenotype indicated that OsRBR1 is required for the formation of inner floral organs in rice. This was confirmed by scanning electron microscopy observations (Fig. 2), which revealed the Osrbr1 (pis1) mutant FM displayed an apparently different shape from that of wild type after the emergence of lemma and palea primordia and could not generate inner floral organ primordia afterward. Instead, the pis1 FM remained at the lemma/palea differentiation stage or returned to the branch and spikelet differentiation stage, leading to the generation of whorls of LLO or a peculiar ectopic PLO in the floret center. These results suggest that the primary role of OsRBR1 in rice is to maintain the function of the FM, ensuring the activation of floral organ primordium differentiation during floral organ origination.

Floral homeotic genes are essential for floral organ formation. According to the ABCE model, class B and class C genes are required for the specification of inner floral organs, including lodicule, stamen, and pistil, and class E genes are involved in the specification of all floral organs, but different class E genes are responsible for different floral organs (Litt and Kramer, 2010; Ciaffi et al., 2011; Yoshida and Nagato, 2011; Dreni et al., 2013). In this study, we found that loss of OsRBR1 function substantially altered the expression of floral organ identity genes and compromised their expression balance (Fig. 6). The class B and class C MADS-box genes and Zinc-finger genes OsJAG and DL, which are critical for the specification of stamen and pistil, respectively (Nagasawa et al., 2003; Yamaguchi et al., 2004; Yao et al., 2008; Horigome et al., 2009; Xiao et al., 2009; Duan et al., 2010; Dreni et al., 2011; Li et al., 2011; Yun et al., 2013), were all downregulated in Osrbr1, suggesting that OsRBR1 is a positive regulator of these genes. In contrast, the two class E genes, OsMADS1 and OsMADS5, which are known to be mainly responsible for lemma specification (Jeon et al., 2000; Cui et al., 2010), were upregulated in Osrbr1, suggesting that OsRBR1 is their negative regulator. These results were consistent with the mutant phenotype, according to the ABCE model. Down-regulation of the class B and class C genes and OsJAG and DL genes would inhibit the specification of inner floral organs, whereas up-regulation of the two class E genes would enhance lemma specification. It is possible that these combined regulatory effects caused the inner floral organs to transform into LLOs or PLOs. Therefore, we hypothesize that OsRBR1 controls the formation of inner floral organs by regulating the expression of floral identity genes in rice.

A role for RBR genes in floral organ identity adds to the currently known functions of plant RBR genes. Although in Arabidopsis RBR function is necessary for floral organ initiation and repression of RBR expression can occasionally result in flowers with supernumerary organs, it does not affect floral organ identity (Borghi et al., 2010). Therefore, the function and mechanism of OsRBR1 for the formation of inner floral organs appears to be distinct and specific to rice, which might have evolved after an additional copy of RBR appeared in rice.

OsRBR2 Specifically Controls Pollen Development by Regulating the Expression of Genes for Tapetum Degradation

Defects caused by the loss of OsRBR2 function were found only in anthers, which did not contain any pollen and therefore led to male sterility (Fig. 5, H and J), indicating that OsRBR2 is required for pollen development. Observation of anther transverse sections demonstrated that the microspore cells in Osrbr2 were degraded and thus did not develop into mature pollen grains (Fig. 6). In addition, compared with that in wild type, tapetum degradation in Osrbr2 was obviously delayed, and the expression levels of genes responsible for tapetum degradation in Osrbr2 were significantly downregulated (Fig. 6). This indicated that OsRBR2 is involved in the regulation of tapetum degradation.

Tapetum is a layer of special somatic cells that is in direct contact with developing male gametophytes and provides enzymes for release of microspores from tetrads and nutrients for pollen development during tapetal programmed cell death and the consequent cellular degradation. To ensure normal pollen development, tapetal programmed cell death and cellular degradation are required to occur at a proper stage of anther development. Premature, delayed, or suppressed cellular degeneration of tapetum usually results in male sterility (Kawanabe et al., 2006; Li et al., 2006; Luo et al., 2013). Therefore, delay of tapetum degradation is likely the main, if not the only, cause of microspore abortion in Osrbr2. As heterozygote Osrbr2/+ could produce viable pollen grains and mutant progeny, the mutant allele seems to have little deleterious effect on male gametes. Hence, it is likely that OsRBR2 controls pollen development by functioning in tapetal cells, specifically positively regulating the expression of genes for tapetum degradation, rather than in pollen grains directly.

In Arabidopsis, RBR gene heterozygous (rbr/+) anthers produce a large proportion of malformed nonviable pollen grains in addition to viable pollen grains, suggesting that loss-of-function mutation of RBR also strongly affects male gametophyte development and markedly decreases pollen viability (Ebel et al., 2004; Chen et al., 2009). However, RBR mechanism in Arabidopsis is obviously quite different from that in rice. Whereas pollen development is determined by the RBR genotype of maternal sporophytic tissue (tapetum) in rice, it is determined by the RBR genotype of pollen itself in Arabidopsis.

OsRBR1 and OsRBR2 Are Redundant for Rice Ontogenesis Except Inside Florets Due to Nonoverlapping Expression

As mentioned above, the single mutations of OsRBR1 and OsRBR2 did not affect rice ontogenesis except for the development of inner floral organs and that of pollen grains, respectively. The different mutant phenotypes of OsRBR1 and OsRBR2 inside florets suggested that functions of the two genes have diversified. However, simultaneous loss of the functions of the two genes appeared to be lethal, as double mutants of OsRBR1 and OsRBR2 could not be found among a large number (>3000) of F2 plants produced by selfing the double-heterozygous genotype of the two genes (pis1-1/+ Osrbr2-1/+). This suggests that RBR function is essential for the formation and/or development of an individual rice plant, and the two RBR genes are redundant in terms of the function. Therefore, both functional redundancy and divergence exist between OsRBR1 and OsRBR2. The functional redundancy appears to apply to the whole plant, whereas the functional divergence is tissue-specific and minor in comparison with the functional redundancy. This reasonably explains why the single mutants of the two genes did not show altered phenotypes outside florets.

The functional redundancy and divergence between OsRBR1 and OsRBR2 were consistent with their expression patterns. During the vegetative phase, the two genes were both highly expressed in the young tissues of vegetative organs (Supplemental Fig. S4; Fig. 7, A–N). This enabled them to function together in vegetative development. In contrast, the expression of the two genes significantly diverged inside florets. OsRBR1 was expressed in all inner floral organs except anthers, whereas OsRBR2 was expressed only in anthers (Fig. 7, Q–Z). Obviously, these two expression patterns were nonoverlapping and complementary, which means that OsRBR1 could not affect the biological processes inside anthers and OsRBR2 could not affect the development of inner floral organs. Therefore, mutation of one gene could not be compensated by the other. This explains why Osrbr1 and Osrbr2 displayed mutant phenotypes in floral organ development and pollen development, respectively.

In Arabidopsis, loss of RBR function is also lethal. Either the rbr ovule is nonfunctional or the zygote generated from the rbr ovule cannot develop into mature embryo regardless of the genotype of the male gamete (Ebel et al., 2004). Therefore, no rbr individuals can be obtained. In this study, we did not find plants that died during vegetative growth in the F2 population, although one-sixteenth of F2 individuals were theoretically expected to be the double mutant of OsRBR1 and OsRBR2. This suggests that the double mutants should have died before developing into seedlings. Therefore, the reason behind the lethality of loss of RBR function in rice is likely similar to that observed in Arabidopsis.

RBRs Probably Regulate Reproductive Development via the Conserved RBR-MSI1 Pathway in Rice

MSI1 and MSI1-like proteins (MSIL) are essential for plant development. MSIL is ubiquitous in plants, which is contained in PRC2. It is well characterized that RBR cooperates with PRC2 via the connection of MSI1 during plant-specific developmental events. In Arabidopsis, intensive analyses of the underlying molecular network have revealed the RBR-MSI1 regulatory pathway for gametophyte development (Ebel et al., 2004; Johnston et al., 2008; Jullien et al., 2008; Dumbliauskas et al., 2011). RBR and MSI1 show similar expression patterns and roles for normal growth and development in Arabidopsis. Loss-of-function mutation or reduction of expression of either RBR or MSI1 exhibits similar effects, including lethality, inhibition of plant growth, and disruption of floral formation (Hennig, et al., 2003; Köhler et al., 2003; Ebel et al., 2004; Borghi et al., 2010). In addition, RBR can regulate the expression of MSI1 and other PRC2 genes in Arabidopsis (Johnston et al., 2008; Kuwabara and Gruissem, 2014).

In this study, we found that the two RBR genes had a close functional relationship with MSI1 in rice, similar to that found in Arabidopsis. First, OsMSI1 was expressed in young panicles and anthers during reproductive development (Supplemental Fig. S7), which temporally and spatially overlapped with the expression of the two RBR genes (Fig. 7; Supplemental Fig. S4). This would allow interaction between MSI1 and the two RBR genes either at the transcription level or at the protein level. Second, both OsRBR1 and OsRBR2 regulated OsMSI1 expression, maintaining a suitable abundance of OsMSI1 transcripts in young spikelets and anthers, respectively. Loss of OsRBR1 function resulted in the overexpression of OsMSI1, whereas loss of OsRBR2 function caused downregulated OsMSI1 expression in anthers (Supplemental Fig. S7). Third, both OsRBR1 and OsRBR2 can physically interact with OsMSI1 in vivo (Fig. 7, A–F). Although these findings do not provide direct evidence, they imply that RBR genes function with OsMSI1 to regulate floret and pollen formation in rice, similar to the case in Arabidopsis. We here propose a putative mechanism: the OsRBR1-OsMSI1 interaction controls floral organ specification by activating floral homeotic genes to arrest proliferation and promote differentiation of the central cells, so as to establish and maintain FM function; the OsRBR2-OsMSI1 interaction controls pollen development by promoting tapetal degradation at the proper time (Fig. 9).

Figure 9.

A proposed model for RBR regulation of spikelet and pollen formation in rice. OsRBR1 negatively regulates OsMSI1 expression to maintain a coordinate abundance of OsRBR1 and OsMSI1, which interact to activate the expression of floral meristem identity genes during inner floral organ primordial initiation. This gene expression activation maintains floral meristem function and promotes floral organ specification. OsRBR2 positively regulates OsMSI1 expression to maintain a coordinate abundance of OsRBR2 and OsMSI1, which interact to promote the expression of genes related to tapetum degradation during pollen development. This gene expression regulation maintains tapetal function and promotes pollen formation. le, Lemma; pa, palea; lo, lodicule; st, stamen; pi, pistil.

In conclusion, our results suggest that OsRBR1 and OsRBR2 are the master and specific regulator of FM function and pollen generation in rice, respectively. OsRBR1 and OsRBR2 probably function with OsMSI1 by regulating its expression and interacting with it at the protein level. Our results provide further insight into the conserved RBR-MSI1 system of controlling reproductive development and also extend the understanding of molecular mechanisms behind spikelet development in rice.

MATERIALS AND METHODS

Plant Materials

Rice (Oryza sativa) cultivars Minghui-86 (MH86, indica), Zhonghua-11 (ZH11, japonica), DZ60 (tropical japonica), and four mutants from MH86, namely pis1-1 (Osrbr1-1), pis1-2 (Osrbr1-2), Osrbr2-1, and Osrbr2-2, were used in the study. The two OsRBR1 mutants were obtained by ⁶⁰Co gamma-ray radiation mutagenesis and young embryo culture, respectively, whereas the two OsRBR2 mutants were generated by genome editing using the CRISPR/Cas9 system. Since the mutants could not set seeds, they were maintained through heterozygotes (pis1-1/+, pis1-2/+, Osrbr2-1/+, and Osrbr2-1/+), which were identified using their progeny lines. The homozygous normal individuals in the progeny lines of the heterozygotes were used as wild-type control. To isolate double mutants of the two genes, 15 F2 populations consisting of over 3000 plants in total were constructed through the cross between pis1-1/+ plants and Osrbr2-1/+ plants.

Scanning Electron Microscopy Observation

Young panicles (1–10 mm) were observed by scanning electron microscopy as described (Duan et al., 2010). Panicles were fixed in 2.5% (v/v) glutaric dialdehyde and washed with a sodium phosphate buffer (0.1 m, pH 7.0); further fixed in 1% (w/v) osmic acid for 1 to 2 h and again washed with the sodium phosphate buffer (0.1 m, pH 7.0); dehydrated with an ethanol series, incubated in ethanol-isoamyl acetate and then in isoamyl acetate; dried, mounted, and coated with gold; and finally observed with a XL30 ESEM scanning electronic microscope (Philips).

Positional Cloning of PIS1 (OsRBR1)

An F2 population was developed from the cross between pis1-1 heterozygote and DZ60. A total of 1148 F2 plants with the pis1 mutant phenotype were selected for gene mapping. Preliminary mapping of the target gene was conducted using the published rice microsatellite-series SSR markers (http://www.gramene.org/markers/microsat/). New SSR markers (named with prefix SSR for distinction) and InDel markers (Supplemental Table S2) were developed to narrow down the interval of the target gene. After PIS1 was fine mapped (delimited in a 32-kb region), its candidate gene (OsRBR1) was identified according to the annotation provided by the NCBI database.

Vector Construction and Plant Transformation

For genetic complementation test of OsRBR1, an 8574-bp genomic DNA sequence covering the promoter region (2000 bp), gene region (5574 bp), and downstream region (1000 bp) of OsRBR1 was amplified from MH86 and fused into the binary vector pBWA(v)H. The plasmid was introduced into pis1-1 mutant embryonic calli.

For creating mutants of OsRBR2 and OsMSI1 using the CRISPR/Cas9 system, CRISPR targets of the two genes were selected as described by Miao et al. (2013), and vector construction was performed according to the manufacturer’s instruction of regent kit (VIEWSOLID Biotech). The vectors were transferred into MH86. The genomic region surrounding the CRISPR target site of OsRBR2 was amplified and sequenced to screen for mutants.

For overexpression test of OsRBR1 and OsRBR2, a 3122-bp cDNA of OsRBR1 and a 2989-bp cDNA of OsRBR2 were amplified and fused into the downstream of maize (Zea mays) ubiquitin promoter in the binary vectors pCXUN and pTCK303, respectively. The two plasmids were introduced into ZH11, respectively.

For determining the expression patterns of OsRBR1 and OsRBR2, a 3724-bp promoter upstream of the coding region of OsRBR1 and an 1898-bp promoter upstream of the coding region of OsRBR2 were amplified from MH86 and fused into the GUS reporter gene with the nopaline synthase terminator in pCAMBIA1391Z, respectively. The plasmid was introduced into ZH11. Histochemical assay for GUS activity in transgenic plants was performed as described (Jefferson et al., 1987).

All constructs were confirmed by sequencing, introduced into Agrobacterium tumefaciens strain EHA105, and transferred into rice with the mediation of Agrobacterium. The primers used for vector construction are listed in Supplemental Table S3.

Histological Observations

Observation of pollen development was performed by resin slicing as described (Duan et al., 2012). Rice anthers at various stages were fixed in 2.5% (v/v) glutaraldehyde solution, dehydrated using a graded ethanol series (20, 40, 60, 80, 100, 100% [v/v]) and embedded in Leica 7022 historesin with one-sixteenth volume of Hardner (Leica). Samples were sectioned to 4 μm, stained with 0.1% (w/v) Toluidine Blue-O (Sigma), and observed under an Olympus Leica DM 4000B light microscope.

Observation of embryo sac development was performed by whole-mount eosin B staining as described (Huang et al., 2017). Ovaries at various stages were collected carefully, fixed in formaldehyde, anhydrous ethanol, acetic acid overnight, washed with 50% (v/v) ethanol, and stored in 70% (v/v) ethanol at 4°C. The samples were scanned under a Leica SP8 laser scanning confocal microscope. Images of the ovaries were recorded using the software accompanying the microscope.

Subcellular Localization

OsRBR1-GFP and OsRBR2-GFP fusions were made by in-frame fusion of the 3036- and 2937-bp full-length coding sequences of OsRBR1 and OsRBR2 with GFP, respectively. The fusion genes of OsRBR1-GFP and OsRBR2-GFP were inserted into vectors pCX-DG and pGDG, respectively, and then introduced into EHA105. OsRBR1-GFP, OsRBR2-GFP, and GFP alone (as control) were transferred into rice protoplasts. Nuclear location sequence-red fluorescent protein was cotransfected for nuclear localization control. The samples were observed with a confocal laser scanning microscope (LEICA-SP8). Primer pairs for amplifying OsRBR1 and OsRBR2 cDNA are listed in Supplemental Table S3.

mRNA In Situ Hybridization

mRNA in situ hybridization was performed as described (Lai et al., 2002) with slight modification. Digoxygenin-labeled sense or antisense RNA probes were prepared following the manufacturer’s recommendation. Young panicles (2–20 mm) were fixed with formalin-acetic acid-alcohol solution and embedded in paraffin. The tissues were sliced into 8-mm sections with a rotary microtome and attached to microscope slides. Primers used for mRNA in situ hybridization were 5′-tcg​caa​acg​atg​agg​aca-3′ and 5′-tgc​ttc​gca​cgc​tgg​tac-3′.

Yeast Two-Hybrid

We constructed a cDNA library of young panicles (<5 cm) of rice. The MATCHMAKER GAL4 Two-Hybrid System (Clontech) was used to screen proteins interacting with OsRBR1. To construct the bait vector, the full-length cDNAs of OsRBR1 that did not show self-activation and toxicity were inserted into pGBKT7. The recombinant vectors were then introduced into Saccharomyces cerevisiae strain AH109 (BD Biosciences). Primers for amplifying the full-length and truncated cDNA of OsRBR1 are listed in Supplemental Table S3.

Screening of the interacting proteins of OsRBR1 was performed according to the manufacturer’s instruction of Matchmaker Gold Yeast Two-Hybrid System (Clontech, cat. no. 630489). To verify the interaction proteins of OsRBR1, the AD-Prey plasmids were rescued from yeast strains and sequenced. The full-length cDNA sequences of candidate proteins eliminating false reading proteins were amplified and inserted into vector pGADT7, which was used together with pGBKT7-OsRBR1 to cotransform yeast competent cell AH109, which was then selected on SD/-Leu/-Trp/-His and higher stringency SD/-Leu/-Trp/-His/Ade plates. The transformants containing empty plasmids pGADT7 and pGBKT7 served as a negative control. A positive interaction was judged by the growth of yeast colonies on selective media (2Trp/2Leu/2His/+3-AT).

BiFC Assay

The CDSs of OsRBR1, Osrbr1-1, and three OsMSI genes were amplified and inserted into the N terminus and the C terminus of YFP in vectors pCAMBIA1300S-YN and pCAMBIA1300S-YC to form fusion proteins YN-OsRBR1, YN-Osrbr1-1, YC-OsMSI1, YC-OsMSI2, and YC-OsMSI3, respectively. The CDSs of OsRBR2, Osrbr2-1, and three OsMSI genes were amplified and inserted into the N terminus and the C terminus of GFP in vectors pCAMBIA1300S-GN and pCAMBIA2300S-GC to form fusion proteins GN-OsRBR2, GN-Osrbr2-1, GC-OsMSI1, GC-OsMSI2, and GC-OsMSI3, respectively. Equal concentration mixture of Agrobacterium tumefaciens containing plasmids of YN-OsRBR1, YN-Osrbr1-1, and YC-OsMSI was used to cotransform Nicotiana benthamiana leaves, and that of GN-OsRBR1, GN-Osrbr2-1, and GC-OsMSI was used to cotransform rice protoplasts. The vector combinations of YN/YC-OsMSI, YC/YN-OsRBR1, GN/GC-OsMSI, and GC/GN-OsRBR1 were used as negative controls to verify the specificity of the interactions. Fluorescence in transformed leaf cells or protoplasts was examined and imaged using a confocal laser scanning microscope (LEICA-SP8). The primers used for vector construction are listed in Supplemental Table S3.

RNA Isolation and RT-qPCR Analysis

The expression of OsRBR1 and OsMSI1 in roots, leaves, stems, and young panicles and that of floral identity genes in young panicle were analyzed using RT-qPCR. Total RNA was isolated using Trizol reagent kit (Invitrogen). Reverse transcription of total RNA was performed using Primescript RT reagent kit (Takara). The cDNA samples were diluted to 8 ng/μL and 2 ng/μL. qPCR analysis was performed with three biological replicates using the SYBR Premix Ex Taq II (Takara) with a Mastercycler ep realplex sequence detection system (Eppendorf). Amplification of Actin was used as internal control to normalize all data. Primers used for RT-qPCR analysis are listed in Supplemental Table S4.

Accession Numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers XM_015793187.2 (OsRBR1), NC_029266.1 (OsRBR2), and NC_029258.1 (OsMSI1).

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Constitutive overexpression of OsRBR1.

• Supplemental Figure S2. Homology analysis of plant RBR proteins and their functional domains.

• Supplemental Figure S3. Constitutive overexpression of OsRBR2.

• Supplemental Figure S4. RT-qPCR analysis of OsRBR1 and OsRBR2 expression in different tissues or organs.

• Supplemental Figure S5. Subcellular localization of OsRBR1 and OsRBR2.

• Supplemental Figure S6. Cluster analysis of MSI proteins of rice and Arabidopsis.

• Supplemental Figure S7. Expression and functional analysis of OsMSI1.

• Supplemental Table S1. Identification the interaction protein of OsRBR1.

• Supplemental Table S2. Primers used for fine mapping of PIS1.

• Supplemental Table S3. Primers used for vector construction.

• Supplemental Table S4. Primers used for RT-qPCR.