The ‘Oat-Like Rice’ Is Named after the Unique Phenotypes in Grain Shape
The ‘Oat-like rice’ is a recessive mutant, which was originally discovered in the paddy field in 2001 for its recognizable phenotypes of off-white grains in panicles at the mature stage (Fig. 1a). Oat-like rice was characterized by a remarkably elongated leafy lemma and palea displaying a characteristic open hull (Fig. 1b), and occasional the formation of conjugated twin brown rice in the grains (1.15% ± 0.31%, Fig. 1c and g), which highly resembled oat grains (Fig. 1b and c). Therefore, it was known as ‘Oat-like rice’ for its unique phenotypes in grain shape by local farmers and rice breeders in the Sichuan province of China.
Characterization of the Typical Morphological Characteristics of Oat-Like Rice
Although Oat-like rice exhibited some morphological differences from Nipponbare (NIP) including a distinct plant architecture, reduced plant height and total grain number per panicle, grain shape was the typical morphological characteristic used to distinguish Oat-like rice from NIP and other normal rice varieties (Additional file 1: Figures S1, S2 and Table S1, Fig. 1a-c, Additional file 2: Table S8). And the difference in plant architecture and plant height between Oat-like rice and NIP could be likely due to the difference of genetic backgrounds between the original wild type variety of Oat-like rice and NIP. As shown in Fig. 1c and g, some conjugated twin brown rice was occasionally observed in Oat-like rice, and Oat-like rice also had an extremely elongated lemma and palea compared with NIP. The length of the lemma and palea in Oat-like rice grains was 18.73 ± 2.83 mm and 16.98 ± 2.25 mm (Fig. 1d), respectively, which was approximately 1.7- and 1.5-fold longer than the lemma and palea, respectively, in NIP grains. To investigate the cause of the extremely elongated lemma and palea of Oat-like, we examined the lemma and palea of NIP and Oat-like grains by light microscopy. Longitudinal sections in the middle of lemma/palea showed that the parenchymal cells were much longer in Oat-like rice than in NIP (Fig. 2a and b). Further data analysis showed that the average cell length in the middle of the lemma and palea of Oat-like rice was 55.42 ± 1.28 mm and 54.37 ± 3.92 mm respectively, demonstrating an increased by approximately 108% and 72%, respectively, compared with NIP (Fig. 2c). Similar to the lemma and palea, the brown rice length of Oat-like rice was also significantly increased compared with that of NIP (Fig. 1d). Although the grains of Oat-like rice were extremely significantly wider than those of NIP due to the open hull formed by aberrant palea and lemma, the values of the grain thickness, brown rice width and brown rice thickness were extremely significantly lower than that of NIP (Fig. 1b and d). To summarize, Oat-like rice had a slender grain and brown rice, as indicated by the higher length-to-width ratio of the grain and brown rice (Fig. 1e and f). In addition, the 1000-brown rice weight of Oat-like rice was approximately half that of NIP (Fig. 1h), which maybe the result of difference in grain filling between them during the grain development after fertilization.
To further test this assumption, we performed time course analysis of grain development between Oat-like rice and NIP by measuring the values including dry weight, length, width and thickness of brown rice every 3 days from 6 to 30 DAF (Days After Fertilization). As shown in Fig. 3a and b, the trends in dry weight accumulation of both grains and brown rice between Oat-like rice and NIP were relatively similar in general. However, overall, Oat-like rice displayed an obviously slower grain-filling rate and lower dry weight accumulation than that of NIP, which eventually resulted in the significantly lower 1000-brown rice weight of Oat-like rice. In addition, during the grain filling process from 6 to 30 DAF, the dry weight of both grains and brown rice of Oat-like rice was markedly decreased at 21 DAF, which was contrary to that of NIP.
Moreover, changes in brown rice length, width and thickness between Oat-like rice and NIP were generally consist with their trends of dry weight accumulation, partly indicating by continuous increase of brown rice width and thickness from 6 to 27 DAF and subsequently slight decrease from 27 to 30 DAF (Fig. 3c). Besides, in Oat-like rice, the change curve of brown rice length was similar to that of dry weight accumulation, including a decrease at 21 DAF. In conclusion, the slower grain-filling rate and consequently lower 1000-brown rice weight of Oat-like rice suggested that its aberrant shape of brown rice may be also related to the affected grain filling during grain development. It should be noted that although some of the above phenotypic differences of grain shape and corresponding statistical significance tests between Oat-like rice and NIP may be affected by their difference of genetic backgrounds, the lemmas, paleae and grain length of Oat-like rice is extremely longer than other normal rice varieties which suggests that this phenotype should be caused by the spontaneous mutation in Oat-like rice rather than the difference of genetic backgrounds. We found that the plants showing Oat-like rice phenotype in F2 population only carried the homozygous mutation of OsMADS1Olr, while the normal plants carried the heterozygous OsMADS1Olr or the homozygous OsMADS1. It suggested that the differences of grain phenotype between Oat-like rice and NIP might be not related to backgrounds. To further test this assumption, the grain length of 141 representative rice varieties or accessions including Indica, Japonica, Aus, Aus/boro, Basmati/sadri, and Intermediate type originated or collected from 57 countries is analyzed. As shown in Additional file 1: Figure S2 and Additional file 2: Table S8, the grain length of Oat-like rice is obvious longer than all these 141 rice varieties or accessions, which is consistent with the above assumption.
On the other hand, to investigate the cause of the open hull phenotype of Oat-like rice, we initially analyzed young spikelets at different developmental stages under a scanning electron microscope (SEM) and dissecting microscope. The lemma-palea showed an open structure at the end of Sp8 in both NIP and Oat-like rice (Fig. 4a and i) (Ikeda et al. 2004). In the subsequent developmental stages shortly after Sp8, the lemma and palea of NIP began to couple with each other and finally formed a closed spikelet (Fig. 4b-h). However, the lemma and palea of Oat-like rice remained open during all observed developmental stages of spikelets (Fig. 4j-p). As showed in Fig. 4j-p, the coordinate development of lemma and palea was disordered, as indicated by retarded palea relative to lemma and different dispositions and orientations between the lemma and palea, which caused the open hull phenotype of Oat-like rice.
In summary, Oat-like rice displayed a dramatic change in grain shape including occasionally formed conjugated twin brown rice and an open hull caused by uncoordinated development and elongation of lemmas and paleae.
Aberrant Floral Organs and Defects in the Embryo Sac Affect the Seed Setting Rate of Oat-Like Rice
Apart from the open hull phenotype formed by extremely elongated lemma and palea, occasionally formed conjugated twin brown rice and a dramatic change in grain shape, an extremely low seed setting rate was another obvious abnormal characteristic of Oat-like rice. We initially analyzed the floral organ numbers, morphologies and structures in spikelets to investigate the cause of the low seed setting rate of Oat-like rice. As shown in Additional file 1: Figure S3a-c and j, the normal spikelet of NIP consisted of a pair of empty glumes, a lemma, a palea, a pair of lodicules, six stamens and a pistil. However, the spikelet of Oat-like rice displayed various variations in floral organ numbers, morphologies and structures, including altered floral organ numbers of empty glume, lemma, palea, lodicule, stamen, and pistil in a spikelet, as well as the number of stigma in a pistil (Additional file 1: Figure S3d-i, k-l and Table S2). Furthermore, some floral organs exhibited abnormal variations of morphologies, including leafy and extremely elongated lodicules, as well as conjugated pistils, which might cause the formation of twin brown rice in one grain (Additional file 1: Figure S3d-f, h, i and k). Nevertheless, there appeared to be no obvious difference in stamen morphologies between NIP and Oat-like rice (Additional file 1: Figure S3a, b and d-f). In addition, some floral organs of Oat-like rice exhibited aberrant structural variations, including hypertrophic lodicules and misshapen carpels (Additional file 1: Figure S3k and l). In summary, aberrant floral organs of Oat-like rice might not only impede normal pollination but also affect gamete fertility and finally the seed setting rate.
To further investigate this possibility, we analyzed the pollen vitality and germination ratio, morphologies and structures of embryo sac of Oat-like rice. There were no significant differences in the pollen vitality and pollen germination ratio between Oat-like rice and NIP (Fig. 5a-e), suggesting that the low seed setting rate of Oat-like rice might not be due to male gamete fertility. These results are consistent with the observation that stamen morphologies of Oat-like rice were apparently normal in comparison to NIP (Additional file 1: Figure S3a, b and d-f). However, about 6/7 abnormal carpels were observed in Oat-like rice, in which the embryo sac did not develop and was replaced with an abnormally proliferated nucellus (Fig. 5g-i). Thus, the low seed setting rate of Oat-like rice might be partly attributed to defects in the embryo sac of abnormal carpels (Fig. 5f-j). To summarize, the seed setting rate of Oat-like rice was related to aberrant floral organs and defects in the embryo sac.
Map-Based Cloning of the OsMADS1Olr Gene
To clone the gene responsible for the unique grain shape in Oat-like rice, we constructed a mapping population derived from a cross between Oat-like rice and NIP. All the F1 plants showed a normal grain shape phenotype similar to NIP. The subsequent F2 population segregated into two phenotypes: the NIP phenotype and the Oat-like rice phenotype. However, the segregation ratio was approximately 1.50, which deviated from a typical ratio of 3:1 (χ23:1 = 432.56 > χ20.05 = 3.84, p < 0.001) (Additional file 1: Table S3), indicating that the phenotype of Oat-like rice might be controlled by a recessive quality trait locus with a segregation distortion. We initially found that the Oat-like rice locus was linked to the InDel (Insertion-Deletion) markers InDel 3–11 and InDel 3–13 on the long arm of chromosome 3 by bulked segregate analysis (BSA) (Fig. 6a). For preliminary mapping, 152 F2 individuals with the Oat-like phenotype were used, and new InDel markers were also developed based on the sequences of NIP and 93–11 (Additional file 1: Table S4). The Oat-like rice locus was first mapped to a 440-kb interval between markers FMM-17 and InDel 3–12 on chromosome 3 in rice (Fig. 6b). Then, we used 1450 F2 individuals with the Oat-like phenotype and developed additional new markers to finely map the locus and narrowed it down to an approximately 14-kb region (Fig. 6b). In the 14-kb interval, there was only one gene, OsMADS1 (LOC_Os03g11614), predicted by the genome annotation database (http://www.gramene.org/), which contained eight exons and encoded a transcriptional regulator containing the MADS domain, I region, K-box domain and C-terminal region (Fig. 6c and d). A single nucleotide mutation of the 80th nucleotide (nt) in the first exon of OsMADS1 was found through DNA sequencing (Fig. 6c). This point mutation from G to A caused a substitution from glycine (Gly) to aspartic acid (Asp) of the 27th amino acid in the MADS domain of the OsMADS1 protein in Oat-like rice (Fig. 6c and d). Some allelic mutations of OsMADS1 have been reported to cause aberrant spikelets and floral organs in rice (Kinoshita et al. 1976; Chen and Zhang 1980; Jeon et al. 2000; Agrawal et al. 2005; Chen et al. 2006; Gao et al. 2010; Hu et al. 2015; Sun et al. 2015; Zhang et al. 2018). We compared the mutation site in the OsMADS1 gene between Oat-like rice and other reported mutants, and we found that the point mutation of Oat-like rice differed from the previously identified mutations in the OsMADS1 gene. Therefore, the Oat-like rice harbors a novel allele of the OsMADS1 gene, which was designated OsMADS1Olr.
The Mutation Site in OsMADS1Olr Is Associated with the Oat-Like Rice Phenotype and the Corresponding Wild-Type Amino Acid Is Highly Conserved in MADS-Box Proteins in Rice
To investigate if the mutation in OsMADS1Olr is responsible for the Oat-like rice phenotype, we initially performed a sequence alignment of exon 1 of the OsMADS1 gene among Oat-like rice and thirteen representatively normal rice varieties including japonica rice NIP and Indica rice 93–11. We found that only Oat-like rice carried the mutation in OsMADS1Olr, while all the other thirteen representative rice varieties with a normal grain shape carried the wild type nucleotide of OsMADS1 (Additional file 1: Figure S4). Besides, We found that the plants showing Oat-like rice phenotype in F2 population only carried the homozygous mutation of OsMADS1Olr, while the normal plants carried the heterozygous OsMADS1Olr or the homozygous OsMADS1. These results indicated that the mutation in OsMADS1Olr was associated with the Oat-like rice phenotype.
Furthermore, the MADS-box domain of OsMADS1 protein was responsible and indispensable for DNA binding between OsMADS1 and its target genes (Arora et al. 2007). To explore the conservation of the substitution site of OsMADS1Olr in the rice MADS-box family, we compared the protein sequence of the MADS-box domain among the 74 MADS-box family members in rice obtained from the Plant Transcription Factor Database (Plant TFDB). By sequence alignment, we found that the amino acid (Gly) and the corresponding nucleotide (corresponding G) in the substitution site was highly conserved in MADS-box family members in rice (Additional file 1: Figure S5 and Fig. 6c and d). Interestingly, this wild-type amino acid (Gly) in the MADS-box domain is not only highly conserved in MADS-box family members in rice, but is also highly conserved in the MADS-box transcription factors from plants, yeast, frog and human, including AP1 (APETALA1) and CAL (CAULIFLOWER) in Arabidopsis, and SRF (Serum Response Factor) in human. And the Gly was important for the DNA binding of MADS-box domain (Pellegrini et al. 1995). AP1 and CAL are two partially redundant MADS-box genes. Both genes are expressed in young flower primordia and involved in regulation of specifying, identity and development of the floral meristem in Arabidopsis. The ap1–2 single mutant of AP1 and ap1-1 cal-2 double mutant of AP1 and CAL exhibited phenotypic alterations in flowers including abnormally proliferous inflorescence, axillary flowers and aberrant floral organs due to a partial conversion of floral meristems into inflorescences. Coincidentally, mutations in AP1ap1–2 and CALcal-3 proteins are exactly the same to those in mutated OsMADS1Olr protein (Mandel et al. 1992; Kempin et al. 1995). Therefore, the wild-type glycine in the substitution site may be essential for the normal DNA binding functions of MADS-box family proteins and amino acid substitution in this site of OsMADS1 may cause the abnormal grain shape in Oat-like rice.
Expression Analysis Suggests that OsMADS1 Is Involved in Grain Development
Previous reports have indicated that OsMADS1 is mainly expressed in the inflorescence, spikelets and during grain development (Chung et al. 1994; Arora et al. 2007; Liu et al. 2018; Zhang et al. 2018). To investigate whether OsMADS1Olr is expressed and functional, and if OsMADS1Olr has the similar expression pattern of OsMADS1 in NIP, we analyzed the expression patterns of OsMADS1 and OsMADS1Olr in NIP and Oat-like rice respectively during floral and grain development. Our quantitative real time PCR (qRT-PCR) results showed that OsMADS1 and OsMADS1Olr had similar expression patterns, and the average expression of both OsMADS1 and OsMADS1Olr was higher in grains than in inflorescences and floral organs. Additionally, we detected the strongest expression in the early stages of grain development, especially in 1 DAF pistils and hulls and 3 DAF grains (Fig. 7a). Furthermore, expression of OsMADS1 or OsMADS1Olr was also abundant in elongating lemmas (L1), mature lemmas (L3) and paleae (Pa3) (Fig. 7a). OsMADS1 or OsMADS1Olr was expressed in pistils, but expression was not detected in stamens (Fig. 7a). Considering both spikelets and grains of Oat-like rice displayed abnormal morphologies, these results implies that OsMADS1 not only plays a vital role in regulating spikelet development, but also may be functional in regulating grain development.
To further analyze the detailed tissue-specific expression of OsMADS1, we analyzed GUS expression driven by the native OsMADS1 promoter in transgenic grains. The GUS staining showed that OsMADS1 was highly expressed in grains at 0 DAF, 6 DAF and 12 DAF. Then, its expression was significantly reduced and was not detected in fully mature grains (Fig. 7b-n). Overall, OsMADS1 was mainly expressed in the lemma, palea and grain epidermis during the early stage of grain development and in the embryo, endosperm and aleurone layer. Moreover, we subsequently checked and analyzed the tissue-specific expression pattern of OsMADS1 based on the microarray data released in the ePlant Rice database (http://bar.utoronto.ca/eplant_rice/) and the Rice Expression Profile Database (RiceXPro) (https://ricexpro.dna.affrc.go.jp/category-select.php), respectively (Itoh et al. 2005; Waese et al. 2017). OsMADS1 was also mainly expressed in inflorescences, spikelets, floral organs and grains including embryo, endosperm and aleurone layer, whereas expressions of OsMADS1 were not detected or negligible in the vegetative tissues and organs such as roots, stems and leaves (Additional file 1: Figure S8a and c). Taken together, this result was consistent with our qRT-PCR and GUS staining results, and the expression of OsMADS1 in grains by qRT-PCR and GUS staining analysis were generally comparable. Collectively, OsMADS1 was highly expressed during the early stage of grain development and might play a role in the grain development.
OsMADS1Olr Is Responsible for the Oat-Like Rice Phenotype
Partly because of its genetic background of incica rice, the Oat-like rice calli is very hard for genetic transformation to introduce the wild-type OsMADS1 into the Oat-like rice to verify whether OsMADS1Olr was responsible for the Oat-like rice phenotype by complementation experiment. And previous study reported that overexpression of mutated version of the MADS-box gene in transgenetic plants caused dominant-negative effect, such as the mutated AGAMOUS (AG) gene lacking the C-terminal region in Arabidopsis (Mizukami et al. 1996). So, we alternatively selected a dominant negative mutations technique (Herskowitz 1987) by overexpressing the mutated OsMADS1Olr allele in NIP to competitively inhibit the normal function of endogenous wild-type OsMADS1 of NIP to test whether OsMADS1Olr was responsible for the Oat-like rice phenotype. The pUbi::OsMADS1Olr vector was constructed and subsequently introduced into the NIP calli by Agrobacterium-mediated transformation.
And nineteen positive primary transgenic plants (T0) were obtained and showed normal phenotypes indistinguishable from NIP wild-type plants during the vegetative stage. After PCR verification and qRT-PCR analysis, one severe plant (OE-2) and another weak plant (OE-10) which showed the abnormal grain shape, similar to that of Oat-like rice were selected for detailed phenotypic analysis (Fig. 8a-c). The occasional formation of twin brown rice in one grain of OE-2 was also observed (Fig. 8a, indicated by arrowhead). The lemma length, palea length, grain length and brown rice length of both OE-2 and OE-10 were also significantly longer than those of wild-type (Fig. 8g). Conversely, the grain width, lemma width, palea width, grain thickness, brown rice width and brown rice thickness values were all significantly lower than those of wild-type (Fig. 8g). Thus, OE-2 and OE-10 plants had slender grains and brown rice compared with wild type, as indicated by the higher length-to-width ratio of grain and brown rice (Fig. 8a, h and i). In addition, both the 1000-grain weight and 1000-brown rice weight of OE-2 and OE-10 were extremely significantly decreased compared with wild type (Fig. 8j and k). In addition, OE-2 and OE-10 also had abnormal spikelets with various variations in floral organ numbers and morphologies, which was very similar to the spikelets of Oat-like rice (Additional file 1: Figure S6).
Overall, the OE-2 plant had a more severe Oat-like phenotype than OE-10, including a much lower seed setting rate, which may be ascribed to the possibly stronger expression level of OsMADS1Olr in OE-2 than in OE-10 as indicating by much higher expression level of total mutated OsMADS1Olr and endogenous wild-type OsMADS1 in OE-2 than in OE-10 (Fig. 8a-c). However, we were not able to detect the amount of solo OsMADS1Olr transcript in neither OE-2 nor OE-10 only by qRT-PCR because it is unlikely to differentiate the mutated OsMADS1Olr transcript from the endogenous wild-type OsMADS1 transcript by this method due to single SNP between OsMADS1Olr and OsMADS1 transcripts. Thus, to further detect the amount of solo OsMADS1Olr and the relative ratio of OsMADS1Olr transcript to OsMADS1 transcript in OE-2 and OE-10, we performed RT-PCR amplification with a designed OsMADS1-RT primer pair (Additional file 1: Table S5) and subsequently sequenced the cDNA fragments spanning the mutation site in OsMADS1Olr or the corresponding wild-type site in OsMADS1 from NIP (Fig. 8d), OE-2 (Fig. 8e) and OE-10 (Fig. 8f).
And we were surprised to find that the signal peak of mutated A at the mutation site in OsMADS1Olr cDNA derived from Oat-like rice was extremely higher than the weak signal peak of wild-type G at the corresponding wild-type site in OsMADS1 cDNA derived from NIP in both OE-2 and OE-10 (Fig. 8e and f). Considering the weak signal peak of wild-type G at the mutation site but strong signal of the another adjacent G ahead of the mutation site, there also might be the possibility that the weak signal peak may contains the residual background signal of the another adjacent G ahead of the mutation site in OE-2 and OE-10. This result suggested that the overexpressed mutated OsMADS1Olr transcript is extremely higher than the endogenous wild-type OsMADS1 transcript in both OE-2 and OE-10 by using the dominant negative mutations technique. Thus, the Oat-like rice phenotype of OE-2 and OE-10 could be caused by their overexpression of mutated OsMADS1Olr and relatively trace expression of endogenous wild-type OsMADS1. Taken together, we concluded that the OsMADS1Olr was indeed responsible for the Oat-like rice phenotype, including the aberrant grain shape.
Suppression of OsMADS1 Expression by RNAi Produced Oat-Like Rice Phenotype
To further confirm the roles of the OsMADS1 gene in regulating grain shape in rice, the pUbi::OsMADS1-RNAi vector was constructed and transformed into NIP to suppress the expression of OsMADS1. Twenty-three independent T0 transformants were obtained, and all the OsMADS1-RNAi transgenic plants developed normally during the vegetative stage. Ultimately, three independent transgenic plants (Ri-2, Ri-4 and Ri-12) with the lowest OsMADS1 expression were chosen for phenotypic analysis (Fig. 9a and b). Compared with NIP, the relative expression of OsMADS1 in spikelets and grains at 12 DAF was negligible in the three OsMADS1-RNAi transgenic lines (Fig. 9b). Additionally, all three OsMADS1-RNAi plants had not only abnormal spikelets harboring variations in floral organ numbers and morphologies but also aberrant grains similar to those of Oat-like rice (Additional file 1: Figure S7 and Fig. 9a). Furthermore, the seed setting rate of Ri-2, Ri-4 and Ri-12 plants was extremely reduced compared with wild type but comparable to that of Oat-like rice (Figs. 5j and 9c). In addition, the occasional formation of twin brown rice in one grain of Ri-4 was also observed, as indicated by the white arrowheads (Fig. 9a). Moreover, the lemma length, palea length and grain length of Ri-2, Ri-4 and Ri-12 plants were significantly longer than those of wild type (Fig. 9a and d). Conversely, apart from the increased grain width in Ri-2 plant, the lemma width, palea width, grain thickness, brown rice width, brown rice thickness, 1000-grain weight and 1000-brown rice weight values in Ri-2, Ri-4 and Ri-12 plants were all significantly reduced compared with of wild-type (Fig. 9a, d, g and h). The three OsMADS1-RNAi plants had slender grains, slender brown rice, and smaller and lighter brown rice compared with wild type, as indicated by the higher length-to-width ratio of the grain and brown rice, lower 1000-grain weight and 1000-brown rice weight (Fig. 9e-h). In summary, Ri-2, Ri-4 and Ri-12 plants had negligible expression of OsMADS1 in the spikelets and grains, indicating that they were severe OsMADS1-RNAi plants. Additionally, the phenotypes of the three OsMADS1-RNAi plants were highly similar to OsMADS1Olr-overexpressing plants and Oat-like rice, indicating that the severe phenotypes of Oat-like rice in grain shape and floral organs were caused by OsMADS1Olr.
Altered Expression of Representative Genes Related to Grain Shape in Oat-Like Rice and Transgenic Plants
To explore the correlation of OsMADS1 gene between its function and the regulatory network in controlling grain shape, we analyzed the expression levels of ten representative genes regulating grain shape in Oat-like rice, OsMADS1Olr-overexpressing and OsMADS1-RNAi plants which all showed altered and abnormal grain shape compared with control plants NIP (Figs. 1, 8 and 9). Compared with control plants, in general, we found that each gene had relatively similar expression patterns in Oat-like rice, OsMADS1Olr-overexpressing and OsMADS1-RNAi plants in grains at 12 DAF by qRT-PCR assays. The results indicated that the expression levels of DEP1, GGC2 and RGB1 were significantly decreased in Oat-like rice, OsMADS1Olr-overexpressing and OsMADS1-RNAi plants, except that DEP1 showed no significant difference in expression levels between Oat-like rice and NIP. And transcript levels of GS3 were also reduced significantly, which is consistent with the increased grain length of Oat-like rice, OsMADS1Olr-overexpressing and OsMADS1-RNAi plants. Additionally, although GS5 and GW8 showed decreased or increased transcript levels respectively in Oat-like rice, transcript levels of both GS5 and GW8 were elevated in OsMADS1Olr-overexpressing and OsMADS1-RNAi plants. Furthermore, both GW2 and GW5 showed significantly upregulated expression levels in Oat-like rice, OsMADS1Olr-overexpressing and OsMADS1-RNAi plants, which is consistent with the decreased lemma width and palea width of these plants. Additionally, transcript levels of OsBC1 were significantly downregulated in Oat-like rice, OsMADS1Olr-overexpressing and OsMADS1-RNAi plants. Additionally, expression levels of OsBU1 was decreased in Oat-like rice but significantly increased in OsMADS1Olr-overexpressing and OsMADS1-RNAi plants (Fig. 10a-c). These results indicated that the differentially-expressed DEP1, GGC2, RGB1, GS3, GW2, GW5, GW8 and OsBC1 were shared in Oat-like rice and the transgenic lines, suggesting that OsMADS1 regulates grain shape possibly by affecting the expression levels of these genes.
OsMADS1 May Has a Direct Role in Seed Development
The OsMADS1Olr-overexpression and OsMADS1-RNAi plants showed abnormal grain shape similar to that of Oat-like rice, which indicated that OsMADS1Olr is responsible for phenotypes in grain shape of Oat-like rice. Nevertheless, it is unlikely to exclude the secondary effect brought by the aberrant lemma and palea development to the final grain shape of Oat-like rice. To solve the problem whether OsMADS1 has a direct role in seed development independent of lemma and palea development, we used seed-specific OsMADS1-RNAi technique to specifically inhibit the expression of OsMADS1 during grain development.
We cloned the promoter of OsTip3 gene, which is specifically expressed in grains including embryo and endosperm, but not in vegetative tissues of rice plants such as root and shoot (Takahashi et al. 2004). The public microarray data from ePlant Rice database (http://bar.utoronto.ca/eplant_rice/) show the OsTip3 gene is only expressed in seed (starting from seed S2 stage) but not in inflorescence (Additional file 1: Figure S8b) (Itoh et al. 2005; Waese et al. 2017). And the RiceXPro results (https://ricexpro.dna.affrc.go.jp/category-select.php) confirmed the expression pattern of OsTip3 and also indicated that its expression level is much higher than that of OsMADS1 during seed development in general (Additional file 1: Figure S8c and d) (Sato et al. 2013). Moreover, expression analysis of both OsTip3 and OsMADS1 in representative stages of inflorescences, spikelets and grains by our transcriptome analysis between Oat-like rice and NIP also supported the above results (data not shown). Thus, these results suggested that OsTip3 promoter (pOsTip3) could be used as seed-specific promoter to silence OsMADS1 gene.
The pOsTip3::OsMADS1-RNAi vector was constructed and subsequently transformed into NIP. Thirty-one positive primary transgenic plants (T0) were subsequently obtained and all displayed undifferentiated phenotypes from wild-type plants during the vegetative and reproductive stages. As shown in Additional file 1: Figure S9, T0 plants eventually produced normal and mature spikelets with closed lemma and palea. However, mature grains of these plants showed mild defect with a narrow crack in either side or both sides of a grain, and expression levels of OsMADS1 in the developing grains of these grains were extremely (T3-Ri-3 (T0), T3-Ri-4 (T0) and T3-Ri-7 (T0)) or significantly (T3-Ri-6 (T0)) reduced (Additional file 1: Figure S10).
To obtain more accurate result, we performed further detailed statistical analysis by using another three homozygous pOsTip3::OsMADS1-RNAi lines. Plants of T2 lines (T3-Ri-5, T3-Ri-9 and T3-Ri-12) also developed normal and mature spikelets with closed lemma and palea. However, although T3-Ri-5, T3-Ri-9 and T3-Ri-12 exhibited seemingly similar grain shape with the negative control line, the mild defect of a narrow crack in either side (63.91 ~ 69.00%) or both sides (20.36 ~ 23.67%) of a grain can differentiate them from the negative control line. Nevertheless, the grain shape of T3-Ri-5, T3-Ri-9 and T3-Ri-12 was apparently different from that of OsMADS1Olr-overexpression and OsMADS1-RNAi plants with open lemma and palea which affected inner brown rice (Figs. 8a, 9a, 11a, e-g, Additional file 1: Figure S11).
Although T3-Ri-5, T3-Ri-9 and T3-Ri-12 didn’t exhibit consistent difference in the length and width of grain as compared with negative control (Fig. 11a-l), the length and width of brown rice in these lines were significantly decreased (Fig. 11h, i, and l). However, OsMADS1Olr-overexpression and OsMADS1-RNAi plants exhibited more decrease in brown rice width (Figs. 8g, 9d). Besides, the three lines showed relatively comparable grain length-to-width ratio and brown rice length-to-width ratio to that of the negative control line. It is different from OsMADS1Olr-overexpression and OsMADS1-RNAi plants with slender grains and brown rice as partly indicating by significantly increased grain length-to-width ratio and brown rice length-to-width ratio. In addition, unlike the remarkably decreased 1000-grain and 1000-brown rice weight of both OsMADS1Olr-overexpression and OsMADS1-RNAi plants, the three lines only exhibited slight or mild decrease in both 1000-grain and 1000-brown rice weight compared with the negative control line (Figs. 8a and h-k, 9a and e-h, 11a-j and l-p).
Interestingly, the thickness of grain and brown rice in these lines were both significantly increased (increased by 2.73 ~ 5.21%, 4.02 ~ 6.29%, respectively) (Fig. 11c, j and l), which are different from that of OsMADS1Olr-overexpressing and OsMADS1-RNAi plants (Figs. 8 a and g, 9 a and d). Besides, more than 90% grains of these lines displayed mild defect with a narrow crack in either side or both sides of a grain (Fig. 11a,e-g and k, Additional file 1: Figure S11), possibly resulting from the push of inner brown rice which are thicker.
Compared with OsMADS1Olr-overexpressing and OsMADS1-RNAi plants, seed-specific OsMADS1-RNAi lines produced another kind of grain shape with mild defect with a narrow crack in either side or both sides of a grain, our results indicated that OsMADS1 may directly regulate seed development, independent of lemma and palea development. Therefore, our results also indicated that the dramatic slender grain phenotypes of Oat-like rice, OsMADS1Olr-overexpressing, and OsMADS1-RNAi plants are majorly caused by the abnormal lemma and palea development.