Panicle Morphology Mutant 1 (PMM1) determines the inflorescence architecture of rice by controlling brassinosteroid biosynthesis

Background Panicle architecture is one of the main important agronomical traits that determine branch number and grain number in rice. Although a large number of genes involved in panicle development have been identified in recent years, the complex processes of inflorescence patterning need to be further characterized in rice. Brassinosteroids (BRs) are a class of steroid phytohormones. A great understanding of how BRs contribute to plant height and leaf erectness have been reported, however, the molecular and genetic mechanisms of panicle architecture influenced by BRs remain unclear. Results Here, we identified PMM1, encoding a cytochrome P450 protein involved in BRs biosynthesis, and characterized its role in panicle architecture in rice. Three alleles of pmm1 were identified from our T-DNA insertional mutant library. Map-based cloning revealed that a large fragment deletion from the 2nd to 9th exons of PMM1 was responsible for the clustered primary branch morphology in pmm1–1. PMM1 is a new allele of DWARF11 (D11) PMM1 transcripts are preferentially expressed in young panicles, particularly expressed in the primordia of branches and spikelets during inflorescence development. Furthermore, overexpression of OsDWARF4 (D4), another gene encoding cytochrome P450, completely rescued the abnormal panicle phenotype of pmm1–1. Overall, it can be concluded that PMM1 is an important gene involved in BRs biosynthesis and affecting the differentiation of spikelet primordia and patterns of panicle branches in rice. Conclusions PMM1 is a new allele of D11, which encodes a cytochrome P450 protein involved in BRs biosynthesis pathway. Overexpression of D4 could successfully rescue the abnormal panicle architecture of pmm1 plants, indicating that PMM1/D11 and D4 function redundantly in BRs biosynthesis. Thus, our results demonstrated that PMM1 determines the inflorescence architecture by controlling brassinosteroid biosynthesis in rice. Electronic supplementary material The online version of this article (10.1186/s12870-018-1577-x) contains supplementary material, which is available to authorized users.


Background
Inflorescence architecture and spikelet formation are unique features of grasses such as rice, maize and wheat. In rice, inflorescence (also called panicle) architecture can be categorized into nine successive stages according to the morphological dynamic changes [1]. During the transition from vegetative to reproductive phase, inflorescence meristem (IM) firstly initiates from shoot apex meristem (SAM), and subsequently produces primary, secondary, and sometimes a higher order of panicle branch meristem (BM). On lateral panicle branches, spikelet meristem (SM) initiates and subsequently bears florets and finally develops into grain [1][2][3]. As a result, the length of the main axis as well as the number/length of primary and secondary inflorescence branches are two major determinants for the number of spikelets per panicle and yield of rice [4].
In this study, we identified two additional alleles of pmm1, which exhibited a clustered primary branches phenotype during panicle development. Our genetic mapping results revealed that PMM1 is a new allele of D11, which encodes a cytochrome P450 protein involved in BRs biosynthesis pathway. Transgenic analysis indicated that PMM1/D11 and D4 function redundantly in BRs biosynthesis, indicating that BRs biosynthesis is required for the panicle architecture in rice.

Identification of pmm1 mutants
Previously, we have identified a panicle morphology mutant 1 (pmm1) [38], which showed a slight phenotype in vegetative growth (Fig. 1a) but a strong morphological defects in inflorescence architecture (Fig. 1b). We designated this panicle morphology mutant as pmm1-1. The panicle morphology of pmm1-1 was dramatically altered compared with that of wild type (WT), such as clustered primary branching, opposite grains, and small grains (Fig. 1b). In order to collect more panicle morphology mutants, we screened our T-DNA insertional mutant library [60] and identified two additional lines, which showed obvious clustered primary branches similar to pmm1-1 (Fig. 1a, b). Genetic analyses showed that about one-quarter of their progenies of the heterozygous were defect panicle morphology and others showed normal panicle, indicating that the clustered primary branching of each mutant was controlled by a single recessive allele, respectively. We did not detect any phenotypic segregation when crossing each mutant with pmm1-1 (Fig. 1c), suggesting that the mutation loci in these two independent lines were allelic to that in pmm1-1. Thus, these two lines were designated as pmm1-2 and pmm1-3, respectively. We chose pmm1-1 for further examination.

Characterization of pmm1
Next, we characterized the morphological defects of pmm1 compared with WT. In the paddy field, all the pmm1 plants (pmm1-1, pmm1-2 and pmm1-3) appeared slightly dwarf and produced erected leaves (Fig. 1a). After heading, pmm1 plants showed obvious multiple morphological defects in panicle, such as clustered panicle, shortened axis, clustered primary branches and opposite spikelets (Fig. 1b, d). We further compared the yield traits between pmm1-1 and the WT plants under normal growth conditions. The primary branches number per panicle was increased in pmm1-1 (Fig. 1e), while as for the other traits, no obvious differences concerning secondary branches number per panicle and grain number per panicle in pmm1-1 compared with WT ( Fig. 1f, g). A 30% decrease in 1000-grain weight was decreased in pmm1-1 (Fig. 1h).
To better characterize the panicle developmental defects of pmm1-1, we compared the panicle development between WT and pmm1-1 using scanning electron microscopy (SEM). No remarkable morphological differences in apices were observed between pmm1-1 and WT (Fig. 2a, e). At the reproductive stage, the primary branches were generated normally in WT (Fig. 2b). However, the primary branch primordia in pmm1-1 seemed to generate in a whorl pattern (Fig. 2f). At the secondary branch primordial formation stage, no obvious differences were observed between pmm1-1 and WT (Fig. 2c, g). At the spikelet formation stage, the spikelet primordia were normally developed in WT (Fig. 2d) and pmm1-1 (Fig. 2h). Compared to WT (Fig. 2i, j), histological analysis revealed that such malformation would due to the clustered primary branches pattern developed in pmm1-1 (Fig. 2k, l). These observations suggested that the generation pattern of primary branches were altered in pmm1-1 compared with WT.

Identification of PMM1
Previously, we have already generated a F2 mapping population by crossing pmm1-1 with an indica variety ZS97 [38] and delimited PMM1 to a 147 kb region between the two markers RM3866 and X4 (Fig. 3a). Since no more polymorphic markers could be developed to further narrow down the candidate locus (Additional file 1: Table S1), we sequenced and analyzed all the 21 predicted genes in this region (Additional file 2: Table S2). Finally, we identified a 4124 bp deletion in this region (Fig. 3b), which resulted in the deletion of a large fragment from 2nd to 9th exons of LOC_Os04g39430 (D11) (Fig. 3c). Thus, pmm1-1 mutant might be a knock-out mutant of D11. Further analysis showed that the alteration of panicle architecture in pmm1-1 mutant were not associated with the T-DNA insertion ( Fig. 3d), suggesting that the large fragment deletion might have occurred due to tissue culture process. qRT-PCR results revealed that the transcripts of D11 were significantly blocked in the young panicles of pmm1-1 compared with WT ( Fig. 3e). Therefore, D11 is likely to be a candidate for PMM1.

Genetic complementation of pmm1
To confirm that D11 is responsible for the phenotype of pmm1-1, we generated a vector containing the D11coding region under the control of its native promoter and transformed it into pmm1-1 mutant by Agrobacterium tumefaciens-mediated transformation [60]. Forty-seven independent transgenic plants were generated, and the clustered primary branches of pmm1-1 were completely restored to normal inflorescence architecture in all the transgenic positive plants (Fig. 4a, b). Besides, the grain length and grain width were increased compared to the negative plants (Fig. 4c, d), the grain size also enlarged as much as that in WT. Collectively, these results suggest that the mutation in D11 is responsible for abnormal panicle morphology of pmm1-1.

PMM1 encodes a cytochrome P450 involved in BRs biosynthesis
D11 encodes a cytochrome P450 superfamily protein CYP724B1, facilitating the supply of 6-DeoxoTY and TY during BRs biosynthesis [55]. As PMM1 was identified as a new allele of D11, we speculated that pmm1 would be deficient in BRs biosynthesis and sensitive to BRs. To confirm that, we carried out morphological examination of pmm1-1 in the complete darkness. The emergence rate of mesocotyls and the length of coleoptile were and pmm1-1 (k, l) at the primary branch primordial formation stage (i, k) and the second primordial formation stage (j, l). Bars = 100 μm investigated in pmm1-1 and WT plants. In the dark, the WT plants showed an obvious skotomorphogenic phenotype including elongated mesocotyls and coleoptile (Fig. 5a). However, the mesocotyls were not elongated and the length of coleoptile was short in pmm1-1 (Fig. 5b). Then, we treated pmm1-1 plants with the most bioactive BRs compound BL in darkness. The emergence rate of mesocotyls in pmm1-1 was increased with improving the concentration of BL (Fig. 5c). The length of coleoptile in pmm1-1 was almost the same as that in WT under the  (Fig. 5d). Our result confirmed that pmm1-1 is a BR-sensitive mutant and PMM1 might be involved in BRs biosynthesis.

PMM1 transcripts are abundant in young panicle
To examine the PMM1 expression pattern, we firstly conducted RT-PCR assays on the RNA samples from root, stem, leaf blade, leaf sheath and developing panicles. PMM1 transcripts were detected at a very low level in the examined vegetative organs, but highly accumulated in the developing panicles (Fig. 6a). Subsequently, in situ hybridization was carried out to determine the precise expression pattern in the young panicles. The RNA in situ hybridization signals was obviously detected in the branch primodia and the spikelet primodia in the developing young panicle (Fig. 6b). However, the PMM1 signals were absent in pmm1-1 (Fig. 6c). Using the sense-strand probes, no signal was detected either in pmm1-1 (Fig. 6d) or in WT (Fig. 6e). Because PMM1 preferentially expressed in young panicle and its mutant showed panicle defects, we deduced that PMM1 would be required for the panicle architecture in rice.
Overexpression of OsDWARF4 could rescue the abnormal panicle architecture of pmm1 OsDWARF4 (D4) has been shown to function redundantly with D11 in BRs biosynthesis [54]. In order to examine the roles of D4 in the panicle architecture in rice, we firstly conducted the sequences comparison between D4 and D11. Sequence alignment showed that the amino acid sequence of D11 shared more than 40% identity with D4 (Additional file 3: Figure S1a). Then, we performed qRT-PCR analysis to examine the expression patterns of D4 in different tissues of both pmm1-1 and WT. The results showed that D4 was highly expressed in the root, culm, leaf and leaf sheath but lowly expressed in the developing panicle (Additional file 3: Figure S1b). The transcript levels of D4 were increased in pmm1-1 plants, suggesting that D4 expression might be feedbackregulated when loss of function of PMM1/D11. To confirm the functional redundancy between D4 and PMM1/D11, we generated 30 plants overexpressing D4 in pmm1-1 and in Zhonghua 11 (control), respectively. qRT-PCR analysis confirmed the positive transgenic lines (D4-OE, D4-OE/ pmm1-1) had overexpression of D4 in young panicles compared with the negative transgenic lines (Fig. 7a). D4-overexpressing plants (e.g. D4-OE #8) exhibited larger leaf angles, a typical phenotypic characteristic of excess BRs (Fig. 7b). In the plants of D4-OE/pmm1-1 (#13), plant height and the abnormal panicle morphology could be completely recovered by overexpression of D4 expression in pmm1-1 (Fig. 7b, c). We have also generated 48 plants overexpressing PMM1/ D11. qRT-PCR analysis confirmed the overexpression of  (Fig. 7d). The plants overexpressing PMM1 (#12) displayed large leaf angles (Fig. 7e, g) and increase in plant height (Fig. 7e, h), but a decrease in panicle length (Fig. 7f, i). We have detected a slight increase grain number per panicle and 1000-grain weight in PMM1 overexpression lines compared with control (Fig. 7j). Taken together, our result suggested that D4 and PMM1/D11 function redundantly in biosynthesis of BRs, which is crucial for the normal panicle architecture development.

PMM1 may participate in multiple biological processes
Previous works have elucidated that PMM1/D11 encodes a cytochrome P450 involved in BRs biosynthesis process [54,55]. In order to investigate the possible roles of PMM1 in BR biosynthesis, we examined the expression levels of one BR biosynthesis gene (D4) and seven BR signal transduction genes (OsMDP1, BU1, ILI1, RAVL1, OsBRI1, OsLIC and SG1) in the young panicles of pmm1-1 by qRT-PCR. Compared to that of WT, the expression levels of BR-signaling genes did not altered significantly in pmm1-1 except for MDP1, which was significantly reduced in pmm1-1 (Fig. 8a). As reported in previous literature, the expression level of the BRs biosynthesis gene D4 was significantly increased in d11/pmm1 (Fig. 8a), indicating that a feedback regulation of BRs biosynthesis genes may be trigged in d11/pmm1 [54]. In the D4-OE/WT plants, the expression levels of some BR signal transduction genes (such as OsBRI1, OsMDP1, and BU1) showed dramatically increased compared to the WT (Fig. 8b). These results further indicated that D11 and D4 might be responsible for BR biosynthesis in different organs. Although overexpression of D4 in the young panicles could rescue the clustered panicles in pmm1, the dosage effect of BRs might cause the different expression patterns of BR signal transduction genes in D4-OE/ pmm1 plants (Fig. 8b).
LAX1, LAX2 and FZP are classic genes, which determine the inflorescence architecture by controlling rachis-branch meristem development [6,9,17]. We have examined whether the panicle architecture regulatory pathway is associated with the PMM1/D11 using qRT-PCR analysis in young panicle. The expression levels of FZP and LAX2 were not significantly changed in pmm1-1, suggesting that PMM1 does not affect the expression of FZP and LAX2 (Fig. 8c). In contrast, the expression of LAX1 was significantly up-regulated in pmm1-1, indicating that BRs might be involved in LAX1 regulatory pathways which influence panicle formation in rice.
Previous studies have suggested that BRs can influence seed sizes by affecting cell cycling genes in rice [49]. In this study, mutation in PMM1 also caused small and round grains compared with WT (Fig. 1b). Then, we  (Fig. 8d). These results suggested that PMM1, a BRs biosynthesis gene, might contribute a possible connection between BRs and cell cycle regulation in controlling grain size in rice.
In this study, we identified 3 independent lines with mutations in pmm1 exhibiting whorled primary branches and clustered spikelets (Fig. 1c, d). SEM results showed that the clustered branch precordia occurred during the young panicle development in pmm1-1 (Fig. 1). Therefore, we hypothesized that PMM1 plays an important role in determination of branch and spikelet primordium formation in inflorescences. In addition, expression data showed that PMM1 was predominantly expressed in developing young inflorescence, particularly in branches and spikelet primordia (Fig. 6b). We speculate that PMM1 largely affects the panicle growth and development in the early stages of axillary meristem initiation. Considering pmm1-2 and pmm1-3 are allelic to pmm1-1, we have not further characterized their panicle morphology in detail. Actually, by screening our T-DNA insertion mutant library, we have collected more than 50 independent mutant lines that showed obvious altered panicle architecture. However, only these 3 independent lines (pmm1-1, pmm1-2, and pmm1-3) caused an obvious clustered primary branches in panicle. Previous literature have reported that CPB1 gene, an allele of DWRAF11, also controls panicle architecture and seed size in rice [39]. Mutation in DWRAF11 resulted in remarkable clustered branches in panicles, indicating that BR biosynthesis in panicle is essential for normal inflorescence architecture in rice.

PMM1 encodes a cytochrome P450 related to BR biosynthesis
Physiological studies have demonstrated that most BRbiosynthesis genes are involved in diverse processes during plant development, such as stem elongation, leaf angle, tiller number, plant height, male fertility, senescence and biotic/abiotic stresses [63,64]. In this study, loss function of PMM1 also resulted in a typical phenotype of BR biosynthesis deficiency, such as compact plant type with reduced leaf angle, small and round grains (Fig. 1b, d). Actually, we also observed the reduced leaf angles and shortened internodes in pmm1-1 compared to WT (Additional file 4: Figure S2 a-g). Additionally, the expression levels of genes associated with lamina joint inclination and BR-related genes associated with internode elongation have been changed in pmm1-1 plants (Additional file 4: Figure S2 h, i). The pmm1-1 plants showed a highly sensitivity to BLs treatment (Fig. 5). These results strongly suggest that PMM1 encodes a cytochrome P450 and participates in BR biosynthesis pathway [55].
Besides the typical characteristic phenotypes of BR deficiency, the pmm1-1 mutant, with the deletion of a large fragment from 2 to 9 exons, displayed a novel phenotype of whorled branches and clustered spikelets, with 2-3 abnormal spikelets clustering on each panicle branch (Fig. 1d). PMM1 is a novel allele of DWARF11. Previous studies found that some allelic D11 mutations were generated because of single nucleotide deletion (d11-1), insertion (d11-2) or substitution (d11-3, d11-4 and cpb1) [39,55]. Nevertheless, none of these mutants exhibited whorled primary branches and clustered spikelets, except for that cpb1 showed clustered primary branches but no obvious clustered spikelets [39]. These results suggest that Oryza sativa cytochrome P450 family member CYP724B1 affects plant architecture, especially inflorescence architecture, might be in a transcript dosage-dependent manner, or depending on rice varieties.
Many other kinds of BR-related genes have also been demonstrated to affect rice plant architecture, such as BRD1, BRD2 and XIAO [49,51,53,56]. Among them, XIAO was reported to be involved in the control of organ size by cell cycling. We have observed the small and round grains in three pmm1 mutants (Fig. 1b). Expression analysis showed that there were no significant changes in gene expression between WT and pmm1-1, except for MCM2 and MCM3 (Fig. 8d). These results suggest that PMM1/D11 controls the seed sizes possibly through cell cycling as well.

The relationship between panicle architecture and genes involved in BRs signaling and biosynthesis
It is well known that there is an interplay between the signaling and biosynthesis of BRs in rice [49]. However, some of these tested BR-related genes, including RAVL1, OsBRI1, OsLIC and SG1, did not show conspicuous alterations in pmm1-1 except for MDP1 and DWARF4 (Fig. 8a). A previous study suggested that OsMDP1 function in BR signal transduction and act as a negative regulator of floral organ development and floral identity in rice [65]. To investigate whether panicle architecture in rice requires OsMDP1, a closer examination of panicle morphology with osmdp1 need to be carried out.
As previously reported, D11 and D4 may function redundantly in C-22 hydroxylation for BR biosynthesis [54]. We have verified that overexpression of D4 under the background of pmm1-1 mutant could completely rescue the abnormal inflorescence in pmm1-1 (Fig. 7c). These results indicate that enhancing the expression level of D4 could complement the BR deficit in pmm1-1. However, in plants with gain or loss function of D4, no significant variations could be observed in branch and spikelet morphology (Fig. 7c), indicating that D4 is not crucial to the establishment of panicle morphology in rice. These results suggest that PMM1 acts on the establishment of rice panicle architecture by regulating BR accumulation, while D4 might contribute to vegetative development preferentially.

Conclusion
We have identified a new allele of D11, PMM1, encoding a cytochrome P450 protein involved in BR biosynthesis pathway. A large fragment deletion in pmm1-1 caused an obvious change in rice panicle morphology. Our research revealed that BR biosynthesis is required for the primordial initiation of branches and spikelets during the young panicle development. Thus, PMM1 determines the inflorescence architecture of rice by controlling brassinosteroid biosynthesis.

Plant materials and growth conditions
All rice plants used in this study were japonica (O. sativa ssp. geng) variety Zhonghua 11 (ZH11). Three pmm1 mutants were obtained by screening the enhancer trap mutant library [60]. Rice plants were cultivated in the experimental field of Huazhong Agriculture University in the normal growing season in Wuhan, China (latitude 30.5°N, 15 m above sea level; average daily temperature approximately 28°C).

Plasmid construction and rice transformation
To prepare the complementation vector, we digested ZH11 BAC clone OSJNBa0020J04 (kindly provided by Luo Meizhong) with EcoRI and XbaI, and a 6.2-kb genomic DNA fragment containing the entire PMM1 coding region and the 1600-bp upstream and 800-bp downstream sequences was inserted into pCAMBIA2301. The empty vector pCAMBIA2301 (http://www.cambia.org/daisy/cambia/585) was used as a negative control. For overexpression of PMM1/D11 or OsDWARF4, full-length cDNA of PMM1/D11 or DWARF4 was cloned into the pU1301 vector, which was then electroporated into the Agrobacterium tumefaciens strain EHA105, and finally transformed into rice callus to generate transgenic plants.

Scanning electron microscopy
For scanning electron microscopy, young panicles at different developmental stages from WT and pmm1-1 mutants were dissected, subsequently fixed with a blade, and immediately placed in 70% ethanol, 5% acetic acid, and 3.7% formaldehyde for 24 h at 4°C overnight. Tissues were dehydrated through an ethanol series of 25 to 100% and dried. Following ethanol dehydration, the samples were critical point dried, sputter coated with gold in an E-100 ion sputter, and then observed under a scanning electron microscope (S570, Hitachi, Tokyo, Japan).

RNA extraction and gene expression analysis
Total RNA was extracted from frozen samples using TRIZOL reagent (Invitrogen). After RNase-free DNase I treatment, the first-strand cDNA was synthesized from 4 μg of total RNA with oligo (dT) 15 as the primer, using a reverse transcription kit (Invitrogen). qRT-PCR was carried out using ABI7500 real-time PCR system with the SYBR Premix Ex Taq (TaKaRa) following the manufacturer's instructions. The rice UBIQUITIN5 gene was used as an internal control. Gene expression level was determined from three independent replicates, each consisting of three plants, and three technical replicates per tissue sample were analyzed. The primers for real-time PCR are listed in Additional file 5: Table S3.

BR test
The WT and pmm1-1 mutant were used for BR induction experiments. Rice seeds were sterilized with 0.1% HgCl 2 and then grown on 1% agar medium containing half-strength MS medium in complete darkness at 28°C for 10 days. For BL (Wuhan DingGuo Biotech. Co. Ltd) induction of shoot elongation, rice seeds were grown on 1% agar medium containing half-strength MS medium and various concentrations (10 − 7 M and 10 − 6 M) of BL and incubated at 28°C under continuous light. After 2 weeks, the length of coleoptile and the appearance of hypocotyl were measured. A total of 45 plants were used for each treatment.
In situ hybridization RNA in situ hybridization was performed as described previously [66]. The probe was PCR-amplified from ZH11 using primers in situ-PMM1 F/R (Additional file 5: Table S3). The sense and antisense probes were then transcribed in vitro from the T7 or SP6 promoter with polymerase using a digoxigenin RNA labeling kit (Roche).