The mr1 mutant was identified in the ethyl methanesulfonate (EMS) mutant population of Shuangkezao (SKZ) (cv. indica), and the mutant character was inherited stably after multiple generations of self-bred and selection. The mr1 mutant was crossed with Nipponbare to construct a mapping population. The resulting F₁ grains were sown and self-crossed to construct the F₂ population, which used for gene mapping and genetic analysis. All these materials were planted under natural conditions in the paddy fields of China National Rice Research Institute.

In the flowering stage, we selected the fresh spikelets of wild-type and mr1 mutant and dissected them with tweezers, then analyzed the phenotypic characteristics of them by NIKON SMZ1500 stereoscope.

Paraffin sections were used for histological analysis, according to the method of Ren et al., 2019b. We selected the spikelets of wild-type and mr1 mutant at the heading stage and fixed them in mixed solution (50% absolute ethanol, 0.9 mol/L acetic acid, and 3.7% formaldehyde) at 4°C for more than 16 h, and dehydrated them with different concentrations of ethanol. These materials were then infiltrated with xylene and embedded in paraffin. A rotary microtome was used to cut the paraffin into 8 μm thick pieces and the slides were put at 42°C for several days. We chose some perfect slides to stain with safranine and fast green, and then dehydrated them with different concentrations of ethanol. These stained slides were infiltrated with xylene and covered with neutral resins. At last, we observed these slides under an optical microscope.

According to the method described by Mizukami and Ma, 1992, the scanning electron microscopy (SEM) analysis was performed. We selected the young spikelets of wild-type and mr1 at the stage of floral organ differentiation and fixed them in 3% glutaraldehyde solution at 4°C for about 24 h. Then the materials were dehydrated with different concentrations of ethanol, and incubated in the mixture of ethanol and isoamyl acetate for 1 h. The materials were dried and wrapped with gold powder. These coated samples were finally observed under a HITACHI S-3500 SEM with a −30°C cooled stage.

Among the progenies from the cross between the mr1 mutant and Nipponbare, we selected a total of 558 F₂ plants which were showed the mutant phenotype for mapping of the MR1 gene (Supplementary Table 1). The simple sequence repeat (SSR) markers we used were obtained from the GRAMENE website.¹ And for fine mapping, we developed more new in/del primers by comparing the sequences of the indica 93-11 and japonica Nipponbare. Primer sequences used are listed in Supplementary Table 2.

The roots, culms, leaves, internodes, inflorescences at various developmental stages, and all floral organs (pistils, stamens, lodicules, paleae, and lemmas) of the wild-type and mr1 mutant were collected, the total RNA was extracted from these tissues or organs using the RNeasy Plant Mini Kit (Axygen). The 2 μg total RNA was reverse transcribed to cDNA by using the Reverse Transcriptase Kit (Invitrogen) with genomic DNA eraser (Takara) in a 20 μl reaction volume. For the quantitative reverse transcription-PCR (qRT-PCR) analysis, the cDNA samples were used as templates after being diluted 10-fold. The qRT-PCR was carried out on a StepOne-Plus System (Applied Biosystems) using the SYBR Green PCR Mix Kit (Applied Biosystems) (Ren et al., 2019b). At least three biological replicates were carried out, and the relative expression levels were quantified using a relative quantitation method. Primer sequences used for qRT-PCR are listed in Supplementary Table 2.

Healthy young panicles from the wild-type and mr1 mutant were put in 70% mixed solution (50% absolute ethanol, 0.9 mol/L acetic acid, and 3.7% formaldehyde), then dehydrated in gradient concentrations of ethanol and xylene, and finally embedded in paraffin. Hybridization and immunological detection of these samples were performed as previous description (Ren et al., 2018). The probes were prepared and labeled by a DIG RNA Labeling Kit from Roche according to the manufacturer’s recommendations. Primer sequences used for in situ hybridization are listed in Supplementary Table 2.

The mature grains were air-dried at room temperature and then were crushed to rice powder for physicochemical properties analysis. The total starch content was examined by using a starch assay kit (Megazyme) (Cai et al., 2018). The amylose content measurement was performed according to the description by Liu et al. (2009), and the total protein content measurement was performed according to the description by Kang et al. (2005). To analyze the gelatinization temperature of the samples, we used a differential scanning calorimeter (DSC1). First, we put 5 mg rice powder into an aluminum cup and fixed it with 10 μl distilled water. Then we used a machine to seal the cup. At last, we put the cup into the DSC1, and it run under the 10°C min^–1 heating rate and in the temperature range of 35–100°C. According to the method described by Cai et al. (2018), we examined the swelling and gelatinization characteristics of endosperm starch in urea solutions. All measurements contained three biological replications. Analysis of variance (ANOVA) was used to test whether there are significant differences between these samples.

In the wild-type, the floret includes a lemma, a palea, two lodicules, six stamens, and a pistil from outer to inner (Figures 1A–C). At the seedling and vegetative stage, the phenotype of the mr1 mutant was normal, and the flowering period was also not different from that of wild-type. However, the mr1 mutant appeared abnormal after heading stage. Compared to the wild-type, the mr1 mutant showed reduced paleae, and it was prone to splitting (Figures 1A,B,D,E), but the lemma was not altered (Figures 1A,B,D,E). Meantime, the stamens of mr1 mutant were reduced and no differences were found between the wild-type and mr1 mutant in the other floral organs (Figures 1C,F). According to our measurements and statistics, the palea, bop and mrp of mr1 mutant were narrower than that of wild-type, and the lemma had no significant changes (Figures 1G–I).

For histological analysis, we further carried out scanning electron microscopy (SEM) and paraffin sections for the florets of wild-type and mr1 mutant (Figure 2). We found that in the normal floret, there were five vascular bundles in the lemma, and there were three in the palea (Figures 2A–E,K). And the palea comprised of two mrps and a body of palea (bop), which was different from the lemma (Figures 2B–E,K). As the outer whorls of floral organs, palea and lemma had their distinctive cellular morphology (Figures 2A–E). They both comprised of silicified cells (sc), non-silicified cells (nsc), fibrous sclerenchyma (fs), and spongy parenchymatous cells (spc) (Figures 2D,E,K). In addition, the mrp differed from the bop, showing a distinctive smooth epidermis but lacking silicified thickening epicuticular cells (Figures 2B–E,K). Because of this unique mrp, the lemma and palea appeared a well-interlocked structure (Figures 2B–E,K).

In the mr1 mutant, the mrp was enlarged, and the bop was reduced compared with the wild-type (Figures 2F–J,L). However, the number of vascular bundles was not altered in the mr1 palea, and the mr1 lemma appeared normal vascular bundles (Figure 2I). And the reduced palea resulted in the poor interlocked between the mrp and lemma, so the hull was prone to splitting (Figures 2I,J,L). These results indicated that in the mr1 mutant, the lemma kept normal phenotype, while the palea was significantly reduced due to the decrease of bop and increase of mrp. In addition, no obvious histocytological defects were observed for three inner whorls of floral organs except stamen number in the wild-type and mr1 mutant (Figures 2B–D,I,J,L).

The development of early differentiation stages of wild-type and mr1 mutant florets was observed by SEM, the division of developmental stages we used refers to Ikeda et al. (2004). At the Sp4 stage, the primordia of lemma and palea of the wild-type floret began to develop, the primordia of lemma developed earlier, with a bulge on the top; and the lemma and palea were semicircular, and the marginal tissues were interconnected to form a circular structure, which held the floral meristem inside (Figure 3A). At this stage, no obvious differences were found between the wild-type and mr1 mutant (Figure 3E). During the Sp5 to Sp6 stages, the stamen primordia of the wild-type florets began to differentiate, and the visible spherical stamen primordia were formed synchronously except the stamen adjacent to the lemma (Figure 3B). The primordia of lemma and palea further differentiated, and their marginal tissues began to interlock (Figure 3B). We also observed that the mr1 palea was smaller than the wild-type in these stages (Figure 3F). At the Sp7 stage, the primordia of lemma and palea of the wild-type developed normally and presented a semi-closed state, wrapping the primordia of inner floral organs (Figure 3C). At Sp8, the primordia of lemma and palea of the wild-type was completely closed, and the inner whorls organs were completely wrapped (Figure 3D). Whereas, during the Sp7 and Sp8 stages, the mr1 florets showed significantly smaller palea than that of the wild-type (Figures 3G,H). These results suggested that the mr1 palea decreased from the early stage of floral organ development.

To further examine whether the identity of palea has changed, we used qRT-PCR to investigate the expression levels of several marker genes, such as the hull (palea and lemma) marker genes OsMADS1, OsMADS14, and OsMADS15, the lemma marker gene DROOPING LEAF (DL), and the mrp marker gene OsMADS6 in the palea and lemma of the wild-type and mr1 mutant. The results exhibited that the OsMADS14 expression showed no significant difference between the wild-type and mr1 mutant (Figure 4A). The expression patterns of OsMADS1 and OsMADS15 in the mr1 lemmas were consistent with that of wild-type lemmas, respectively, while the expression levels of OsMADS1 and OsMADS15 in the mr1 paleae were obviously reduced than that of wild-type paleae, respectively (Figure 4A). DL showed no expression difference in the wild-type and mr1 mutant, while an increase expression of OsMADS6 was found in the mr1 mutant (Figure 4A). And, we detailedly detected the expression of OsMADS6 in the bop and mrp. Compared with the wild-type, the OsMADS6 transcript was increased in the bop of mr1 mutant, whereas its expression was not obviously altered in the mrp of mr1 mutant (Figure 4B). The lower expressions of OsMADS1 and OsMADS15 in the mr1 paleae may be caused by the small bop, while the higher expression of OsMADS6 in the mr1 paleae was due to the enlarged mrp.

Next, we further investigate the temporal and spatial expression patterns of the DL and OsMADS6 gene in the wild-type and mr1 mutant florets by in situ hybridization.

In the wild-type, DL expression was found in the lemma and pistil at stage SP7, then DL mRNAs were found in the lemma at stage Sp8 (Figures 4C,D,G,H). In contrast, we observed no obvious differences of DL expression between the wild-type and mr1 mutant at similar stages (Figures 4C,D,G,H). At stages Sp7 and Sp8, in flowers of the wild-type, OsMADS6 showed similar expression patterns with the mr1 mutant; its signals were present in the mrp and pistil (Figures 4E,F,I,J). In general, these results indicated that the identity of mrp was not disturbed in the mr1 mutant, and the lemma had the similar identity in the wild-type and mr1 mutant.

The mr1 mutant was crossed with Nipponbare, all F₁ plants appeared the wild-type phenotype, showing that this mutation was a recessive trait. The phenotype of mr1 was separated in F₂ population. The total number of F₂ population was 2,367, including 1,809 normal plants and 558 mutant plants (Supplementary Table 1). The separation ratio was 3.242:1. By the chi-square test, the separation ratio was consistent with the theoretical ratio of 3:1 (Supplementary Table 1). Thus, we reached a conclusion that the trait of mr1 mutant was controlled by a single recessive gene.

In order to identify the MR1 gene, we performed a map-based cloning method. A total of 558 mutant plants in F₂ population of mr1/Nipponbare cross were used for gene mapping. A total of 240 pairs of SSR markers with an average distribution of 12 chromosomes were selected for polymorphism analysis, 112 of which showed polymorphisms between two parents, SKZ and Nipponbare. Using these 112 pairs of SSR markers for linkage analysis, we found that the marker B1-15 on chromosome 1 was obviously linked to the mr1 phenotype. Then 558 mutant plants in the F₂ population were investigated with this marker, and 32 recombinants were found. Other 20 SSR markers, which were located near the B1-15, were examined with the same approach. Then, the MR1 gene was initially located between the two polymorphic markers, B1-15 and B1-16, and they had 32 and 25 recombinants, respectively (Figure 5A). To further determine the location of the MR1 gene, 16 new in/del primers were developed based on the sequences downloaded from the rice genomic databases (Nipponbare for japonica and 93-11 for indica), and eight of them exhibited the polymorphisms. All 558 mutants were screened with these eight pairs of primers, and the recombinants were 17, 11, 8, 4, 1, 1, 3, and 12, respectively (Figure 5B). Finally, the MR1 gene was located in about 70 kb region between the markers ID8 and ID6 (Figure 5B). In this region, eight open frames were shown, and no known genes were reported (Figure 5C and Supplementary Table 3). ORF1 encoded a protein phosphatase 2C. ORF2 encoded a ZOS1-11-C2H2 zinc finger protein. ORF3, ORF4, and ORF8 encoded expression proteins. ORF5 and ORF7 encoded translation initiation factors. ORF6 encoded an OsFBX16-F-box domain protein. We next sequenced the nucleotide sequences of all genes. Unfortunately, we examined the DNA sequence, and failed to identify the nucleotide changes between wild-type and mr1 mutant.

In consideration of the mr1 mutant displayed defects in palea development, we detected whether MR1 regulates the expression levels of known genes related to palea development in rice. We examined MADS-box genes comprising class A gene (DEP/OsMADS15) and class E gene (MFO1/OsMADS6), and other palea-related genes (MFS1, CFO1, CCP1, DP1, REP1, OPB, AH2, and FON4). We used qRT-PCR to analyze the expression levels of these genes in the wild-type and the mr1 mutant young panicles less than 2 cm (Figure 6). Most tested genes displayed significant changes in the mr1 mutant except REP1 and CCP1 (Figure 6). These results indicated that the reduced palea in the mr1 mutant may be associated with the changes of expression levels of these palea development-related genes, and MR1 determined the palea fate by possibly promoting MFO1 and CFO1, or suppressing DEP, MFS1, DP1, OPB, AH2, and FON4.

The length and 1,000-grain weight of grains in the mr1 mutant were significantly decreased, but the grain width was not changed than that of the wild-type, respectively (Figures 7A–G). However, the length, width, and 1,000-grain weight of brown rice in the mr1 mutant were significantly reduced than that of wild-type, respectively (Figures 7E–J). We examined several genes that determined grain size by modulating cell proliferation and expansion (Yu et al., 2020; Hao et al., 2021). Compared with the wild-type, the transcript levels of BIG1, BIG2, GW2, GS3, GL3, GIF1, and BSG1, were significantly altered in the dls1 mutant (Supplementary Figure 1), implying that the mutated MR1 disturbed the expressions of grain size-related genes.

We also investigated several physicochemical properties of starch (Figures 8A–K). In the mr1 mutant, the amylose content was increased, and the soluble sugar content was decreased that of the wild-type, respectively (Figures 8H,I). And, the total starch and total protein contents of mr1 grains were not influenced than that of the wild-type, respectively (Figures 8G,J). The starch solubility in urea solutions was also measured to examine the gelatinization properties. The powered rice samples were mixed with various concentrations (0–9 mol L^–1) of urea solutions. The results exhibited that there are obvious differences in (3–9) mol L^–1 urea between the wild-type and mr1 starch, and the mr1 starch was more difficult to gelatinize than the wild-type starch (Figure 8K). We could conclude that the physicochemical properties of starch were obviously different between the wild-type and mr1 endosperm, indicating that MR1 could affect grain quality.

As an indispensable part of developmental biology, flower development has been widely studied, and the ABCDE model, which successfully interprets the formation of four whorls in flower, has also been widely accepted. Nevertheless, grasses have highly specialized flowers and inflorescence structures, which makes it hard to explain the process of flower development with the knowledge of other model plants. In this study, the palea of mr1 mutant was reduced while all the other floral organs (pistil, lodicule, and lemma) were still normal except stamen number. Polymorphic markers were used to locate the gene related to the mutation on the long arm of chromosome 1, and there is no report about the related genes of mutation character is located near this position. Therefore, we suggested that MR1 may be a new gene related to the development of palea in rice.

The origin and identity of the palea and lemma remain a topic of heated discussion. There are some genes that affect the development of both palea and lemma. For example, DEGENERATED HULL1 (DH1) encodes a protein with a LOB domain. In the dh1 mutant, about most florets could not develop normal lemma and palea, but degenerated into membranous or fibrous organs, and with the loss of lodicules, stamens, and pistils (Li et al., 2008). TRIANGULAR HULL1 (TH1) gene encodes a DUF640 domain transcription factor, the development of palea and lemma was affected both in transverse and longitudinal direction, and they became smaller and thicker in the th1-1 mutant (Li et al., 2012). Oryza sativa JAGGED (OsJAG) encodes a C2H2 zinc finger domain transcription factor, the loss of function leads to the malformation of all floral organs, such as lemma and palea degenerated and bent inward, lodicules elongated, pistils deformed, and all stamens became pistil-like organs (Duan et al., 2010). In our study, the mr1 mutant specially affected palea development, but the lemma development is not influenced, so we speculate that although these two organs have some similarities in morphology, they have different identities and may be regulated by different genetic pathways. Firstly, the palea was usually considered to be a prophyll, and the flower occurs in the axil of it (Ren et al., 2013), and the lemma was widely considered to be a true bract surrounding the spikelet (Luo et al., 2005; Zheng et al., 2015). Secondly, they were different in structure and size. The palea had three vascular bundles, while the lemma had five, so the palea was smaller than the lemma. And the palea was composed of bop and mrp, while the lemma had only one component. In addition, there were several genes, such as OsMADS15, MFO1, DP1, and CFO1, showing that there are lots of genes which only affected the development of palea, but not that of lemma. All these facts are consistent with our idea.

Furthermore, in a barley mutant called leafy lemma, its lemma was completely transformed into a vegetative leaf with ligule, blade, and sheath, while the palea and other floral organs did not change at all, suggesting that the lemma in grass may not be part of the floret perianth (Pozzi et al., 2000). And in the same barley mutant, the palea and lemma exhibited different phenotypes, indicating that palea and lemma are not homologous in barley. In rice, there is a floral organ identity gene SUPERWOMAN1 (SPW1), it belongs to class B genes and is homologous to APETALA3 in Arabidopsis. The lodicules of spw1 mutant transformed into palea-like organs instead of lemma (Nagasawa et al., 2003), which indicating that the palea of rice may be equivalent to the sepal of eudicot in evolution. Here, in the mr1 florets, the palea was reduced, while the lemma maintained its identity, which seemed to confirm the view that the palea and lemma have different origins. However, some other researchers believed that the palea, and even the lemma, should be equivalent to the sepals of eudicot (Ambrose et al., 2000; Goto et al., 2001). Now, there is not corresponding mutants to support such a hypothesis, so it is still not confirmed.

There are various bop or mrp defective mutants in previous studies. MOSAIC FLORAL ORGANS1 (MFO1) and CHIMERIC FLORAL ORGANS1 (CFO1) both belong to the MADS-box genes, their mutants showed enlarged and silicified mrp, which makes the mutant palea look like bop or lemma-like organ (Ohmori et al., 2009; Sang et al., 2012). RETARDED PALEA1 (REP1) belongs to TCP gene family, in the rep1 mutant, the bop was decreased and the development of it was delayed, but in the REP1 overexpression plants, the mrp overgrew (Yuan et al., 2009). DEPRESSED PALEA1 (DP1) encodes an AT-hook DNA-binding protein, MULTI-FLORET SPIKELET1 (MFS1) encodes an AP2/ERF protein, both the bop of dp1 and mfs1 mutant was completely lost, and only two marginal regions were preserved (Jin et al., 2011; Ren et al., 2013). In recent studies, ABNORMAL HULL2 (AH2) and FLORAL ORGAN NUMBER4 (FON4) were proved to be able to identify the bop (Ren et al., 2019a,b). In the ah2 and fon4 mutants, the bop size was changed, while the mrp was not altered. In our study, detailed observations revealed that the mr1 mutant bore an enlarged mrp and a reduced bop, which suggests that the bop was partly converted to mrp, and MR1was required for bop and mrp identity. Molecular evidences also showed that the small bop of mr1 mutant indeed had the mrp identity. These findings suggested that the bop and mrp were homologous organs and a fused organ. Together, more mutants will be needed to further confirming the relationship of the lemma, bop, and mrp.

In this study, the MR1 gene was fine mapped on chromosome 1 with a physical distance of 70 kb. In this region, there were eight genes, and none of them had been cloned to be related to palea development. But according to the results of sequencing and alignment of these ORF genes between the SKZ and mr1 mutant, we failed to identify the MR1 gene. So we speculated that the mr1 mutant phenotypes might be caused by the intergenic enhancer or repressor and the alteration of promoter sequence or gene expression in the target region. Obviously, this hypothesis needed more experimental evidences to be verified. For instance, we can sequence and compare the promoter regions or 5′ and 3′ UTR sequences, and detect the expression levels of candidate genes. We believe that the further cloning and functional analysis of the MR1 gene would be helpful to enrich the molecular mechanism of rice palea development, and provide abundant evidences to reveal the relationship between its hull (palea and lemma) and the floral organs of eudicots, as well as the evolutionary relationship between the bop, mrp, and lemma.

We also observed the mr1 mutant produced smaller defective grains compared with the wild-type. Several known genes that control grain size by modulating cell proliferation and expansion have been reported. BIG1, BIG2, GW2, GS3, GL3, GIF1, and BSG1 transcripts were changed in the mr1 mutant than in the wild-type. In addition, we found that the mr1 grains had low SSC, and high AC and the mr1 starch was more difficult to gelatinize than that of the wild-type. These findings revealed that the mutation of MR1 reduced the grain size and quality, which may be caused by defective paleae. Taken together, these results suggested that MR1 had a possible effort to improve the grain yield and quality.