Zhnying Zhng,Xingming Sun,Xioqin M,Bingxi Xu,Yong Zho,Zhiqi M,Gngling Li,Njee Ullh Khn,Yinghu Pn, Yunto Ling, Hongling Zhng,Jinjie Li, Zicho Li,*

aState Key Laboratory of Agrobiotechnology/Beijing Key Laboratory of Crop Genetic Improvement, College of Agronomy and Biotechnology,China Agricultural University,Beijing 100193,China

bGuangxi Key Laboratory of Rice Genetics and Breeding,Rice Research Institute of Guangxi Academy of Agricultural Sciences,Nanning 530007,Guangxi,China

Keywords:Germplasm screening Grain number per-panicle Tiller number Oryza sativa

ABSTRACT The yield of rice is mostly affected by three factors,namely,panicle number,grain number and grain weight. Variation in panicle and grain numbers is mainly caused by tiller and panicle branches generated from axillary meristems(AMs).MOC1 encodes a putative GRAS family nuclear protein that regulates AM formation.Although several alleles of MOC1 have been identified,its variation in germplasm resources remains unclear.In the present study we characterized a novel moc1 allele named gnp6 which has a thymine insertion in the coding sequence of the SAW motif in the GRAS domain. This mutation causes arrested branch formation. The SAW motif is necessary for nuclear localization of GNP6/MOC1 where it functions as a transcription factor or co-regulator.Haplotype analysis showed that the coding region of GNP6/MOC1 was conserved without any non-synonymous mutations in 240 rice accessions. However, variation in the promoter region might affect the expression of it and its downstream genes. Joint haplotype analysis of GNP6/MOC1 and MOC3 showed that haplotype combinations H9, H10 and H11, namely MOC1-Hap1 in combination with MOC3-Hap3,MOC3-Hap4 or MOC3-Hap5 could be bred to promote branch formation. These findings will enrich the genetic resources available for rice breeders.

1.Introduction

Tillering and panicle branching that occur from axillary meristems (AMs) at the vegetative and reproductive stages directly influence panicle number and grain number perpanicle in rice, respectively [1]. Molecular studies of these processes could provide ways to improve yield in rice by different molecular strategies.

Several genes affecting branch formation have been isolated by genetic analysis of induced mutants with reduced tillering or lax panicles, whereas few such genes have been found in germplasm collections.These genes are divided into two groups based on gene function: functional enzymes and transcription factors or transcriptional co-regulators. For example, Gn1a encoding a cytokinin oxidase/dehydrogenase(OsCKX2) degrading cytokinin was the first cloned QTL controlling grain number per-panicle[2]. By contrast, LONELY GUY (LOG) [3]and LONG AND BARBED AWN 1 (LABA1/An-2/OsLOGL6) [4,5]encode cytokinin-activating enzymes that promote cytokinin synthesis. NUMBER OF GRAINS 1 (NOG1)encodes an enoyl-CoA hydratase/isomerase [6], NARROW LEAF1 (NAL1, also known as SPIKE) [7,8]encodes a trypsinlike serine/cysteine protease, GRAIN NUMBER PER PANICLE1(GNP1) [9]encodes GA20ox1, and MONOCULM2 (MOC2) [10]encodes cytosolic fructose-1,6-bisphosphatase. All these enzymes function in the phytohormone, protein stability, or energy supply pipelines.

Transcription factors (TFs)or transcriptional co-regulators are the other group of genes affecting branch formation.MONOCULM1 (MOC1) [11]encodes a GRAS family TF and functions as a co-activator of MONOCULM3(MOC3)to regulate the expression of FON1 in tiller formation [12-14]. LAX PANICLE 1 (LAX1) [15], a basic helix-loop-helix (bHLH) TF,interacts with LAX2/GNP4 which contains a RAWUL domain and might function as transcriptional co-regulator [16-18].FRIZZY PANICLE (FZP) containing an ethylene-responsive factor (ERF) domain prevents the formation of AMs, permitting early establishment of the floral meristem [19]. FZP is recognized and degraded by NAL1 [20]. IDEAL PLANT ARCHITECTURE1 (IPA1) [21,22]encodes a TF SQUAMOSA PROMOTER BINDING PROTEIN-LIKE14 (OsSPL14) that positively regulates TEOSINTE BRANCHED1 (OsTB1) and DENSE AND ERECT PANICLE1 (OsDEP1) to influence tillering and panicle branching.Binding of IPA1 to the promoters of OsTB1 and OsDEP1 can be repressed by the plant-specific SHI family TF OsSHI1 [23,24].These various factors act antagonistically or synergistically with each other in a complex regulatory network.

MOC1, was isolated by genetic analysis of a non-tillering moc1 (moc1-1) mutant that had a 1.9 kb retro-transposon insertion in the SAW motif of the GRAS domain [11]. Several other60Co-γ irradiation or ethyl methanesulfonate (EMS)induced mutations in this gene have changes in other motifs[15,16,25]. The MOC1 protein also interacts with other proteins, such as MIP1, TAD1 and SLR1 in post-translation regulation [25-28]. MOC1 encodes the GRAS family TF;however, its direct targets have not been isolated. Moreover,natural alleles of MOC1 and its co-factors that could be used in molecular design breeding have also not been investigated.In this study, we report the characterization of a new natural allele in another moc1 mutant, with a panicle phenotype similar to the gnp4 mutant and named the gene as Grain Number Per-panicle on chromosome 6 (GNP6). The evidences of natural variations in GNP6/MOC1 together with one regulating downstream gene MOC3 will enrich the genetic resources available for improving plant architecture in rice.

2. Materials and methods

2.1. Plant materials and phenotypic analysis

Lines NIL-GNP6 and NIL-gnp6 were developed as BC3F6nearisogenic lines (NIL) with the mutant allele in TENGXI 138 as the donor parent and 02428 as a recurrent parent (Fig. S1).BC3F3and BC3F4populations were used for primary and fine mapping, respectively. All the rice materials were grown in experimental field of China Agricultural University either in Beijing or at Sanya in Hainan province. Heading date was recorded as days from sowing to the 50% panicle emergence.Plant height was measured from the topmost panicle to the ground at maturity. The diameters of the panicle neck and stem base were measured using digital calipers. Grain length and width were measured using a Wanshen SC-G automatic grain test instrument(Wanshen SC-G,China).

2.2. Transgenic vector construction and transformation

To construct the complementary vector, DNA fragment containing a 1.8 kb promoter region, the coding region and a 0.8 kb 3′-UTR of GNP6 was amplified from genomic DNA extracted from NIL-GNP6 and sub-cloned into the pEASYBlunt Cloning vector (TransGen Biotech, Beijing). Plasmids containing the cloned fragment were confirmed by sequencing, digested with PmeI and AscI, and then cloned into the plant binary vector pMDC163.All the plasmids were introduced into the A. tumefaciens EHA105 using the freeze-thaw method. Transgenic rice plants were produced as previously described[29].

2.3. Gene expression analysis

Total RNA in different tissues was extracted using an RNA extraction kit (Aidlab Biotechnologies Co., Beijing) according to the manual. cDNA was synthesized using a StarScript II Firststrand cDNA Synthesis Kit(GenStar,Beijing).Quantitative RT-PCR was carried out using an ABI7500 analyzer(Applied Biosystems,Foster City,CA).OsActin1 was used as a reference.

2.4. Phylogenetic analysis and three-dimensional model analysis

Homologous protein sequences were identified using the fulllength amino acid sequence of GNP6 as a BLAST query in the Phytozome database (https://www.phytozome.net/), the related. After alignment MEGA 5.0 was used to generate the phylogenetic tree [30]. Three-dimensional (3D) molecular simulation analysis was conducted in SWISS-MODEL(https://swissmodel.expasy.org/).

2.5. Subcellular localization

ORFs were amplified from the cDNA of NIL-GNP6 and NIL-gnp6 to construct ProSuper:GNP6-GFP and ProSuper:gnp6-GFP subcellular localization vectors. These were recombined into linearized pSuper1300-GFP vectors digested with Hind III and Kpn I according to the manufacturer's manual for the Trelief SoSoo Cloning Kit (TSINGKE, Beijing). ProSuper:GNP6-GFP and ProSuper:gnp6-GFP plasmids were extracted and transferred into rice protoplasts together with the OsARF6-mCherry plasmid by the polyethylene glycol mediated method and then observed by confocal microscopy (Olympus FV1000, Japan). The excitation and emission spectra for GFP observation were 488-nm and 500-550 nm,respectively.

2.6. Haplotype analysis

A panel of 240 accessions was selected from the rice mini core collection and Rice 3 K Project [31,32]. The SNPs in the accessions were downloaded from Rice 3 K Project RFGB v2.0 and used for haplotype analysis [33]. Tiller numbers at 30 and 45 days post-transplanting were measured in the field. The grain numbers were evaluated after harvest.DnaSP5.10 was used to analyze different haplotypes [34].Details of the 240 accessions are listed in Table S1.

2.7. Primers

Primers details are listed in Table S2.

2.8. Accession numbers

The genes used in this study are available in RAP-DB databases under following accession numbers: GNP6/MOC1(Os06g0610350), Gnp4/LAX2 (Os04g0396500), MOC3(Os04g0663600), FZP (Os07g0669500), FON1 (Os06g0717200)and OsActin1 (Os03g0718100).

3. Results

3.1. Phenotypic analysis of the primary mapping population and near-isogenic lines

In order to isolate the gene conferring grain number perpanicle (GNP), two cultivated varieties, 02428, and TENGXI 138,showing significant difference in GNP were crossed to construct the segregating population. One BC3F3population was clearly different from all others; this population showed a clear segregation of 65 plants with normal panicles and 19 with a sparse panicle architecture indicating the possibility of variation at a single locus (χ23:1= 0.071, P1df= 0.790) with dominance of the normal phenotype. Continued backcrossing and selection confirmed the single locus segregation and led to development of the NIL pair (Fig. S1).

NIL-gnp6 had a reduced number of lower tillers, but a significantly increased number of higher tillers as compared to NIL-GNP6 (Fig. 1 A, B, E; Fig. S2). Panicle length, and primary branch (PB) and secondary branch (SB) numbers for NIL-gnp6 were less than for NIL-GNP6, ultimately a reduction of about 13% in the case of GNP (Fig. 1 C, F, G, H, I). There were also less lateral spikelets on secondary branches in NIL-gnp6 (Fig. 1 D),a phenotype similar to those of the lax1 and lax2 (also known as gnp4) mutants. There was no significant difference in plant height (Fig. 1 J), but NIL-gnp6 had a thinner panicle neck and stem base, and lower seed setting (Fig. 1 K, L, M). These data suggest that GNP6 is a positive regulator of tiller and panicle development.

We also compared grain traits in the NILs grown in in field trials. NIL-gnp6 produced longer and wider grains with an increased thousand grain weight (Fig. 2 A-G), but there was no difference in level of chalkiness (Fig. 1 D, H).

3.2. Fine-mapping of GNP6/MOC1

Bulk segregant analysis (BSA) of 84 BC3F3individuals was used to rapidly locate the locus affecting panicle and tiller numbers and identify linked molecular markers. We initially found recombination of 18.4% with marker RM6446 on chromosome 6. Several more polymorphic markers were then developed and the candidate gene was delimited to a 178 kb region between markers MM2911 and MM2942 (Table S3).

Twelve segregating BC3F3lines were grown; 1460 BC3F4sparse-panicle-plants from a total of 5400 were selected for fine-mapping. Four of these plants were recombinants between ZZY15 and RM20377 thus delimiting GNP6 to a 36.6 kb region (Fig. 3 A). The region between ZZY15 and RM20377 contained five predicated open reading frames(ORFs) (Fig. 3 B). Comparisons of DNA sequences of the five ORFs in the NIL pair revealed no differences in ORF1-ORF4,but ORF5 had a single thymine (T) inserted at the 985 bp position, leading to a shift in amino acid sequence and deferred termination of translation (Fig. 3). Importantly, the thymine insertion polymorphism was not present in the parents and gnp6 change was attributed to a natural mutation in a single plant. According to Rice Genome Annotation Project (http://rice.plantbiology.msu.edu/index.shtml), ORF5 encodes GRAS family transcription factor MOC1. It had been reported previously mutations in this gene caused non-tillering phenotypes [11]. Collectively, our results indicated that ORF5 was the likely candidate locus and that gnp6 could be a novel allele of MOC1.

3.3. Functional complementation of GNP6 in transgenic rice

To verify our prediction, a complementation fragment containing the 1.8 kb promoter segment fused with the coding region and 0.8 kb 3′UTR of GNP6 was amplified from NIL-GNP6 and introduced into NIL-gnp6. Expression level of GNP6 increased significantly in the complemented transgenic plants (CM) (Fig. 4 A). All the CM plants clearly showed extreme dwarfing and significantly increased tillering that was almost 4-fold higher than for NIL-gnp6 (Fig. 4 B, C). Panicle size was also significantly decreased relative to NIL-gnp6, and the plants were sterile (Fig. 4 D). It is possible that an appropriate amount of MOC1 expression is necessary for the tiller and panicle development.

Fig.1-Phenotypic analysis of NIL-GNP6 and NIL-gnp6.(A)Morphologies of NIL-GNP6 and NIL-gnp6 plants at heading.(B)Higher tillers in NIL-gnp6 indicated by red arrow.(C-D)Panicle(C)and primary branch(D)phenotypes in NIL-GNP6 and NIL-gnp6.(E-M)Statistical data for tiller number(E),panicle length(F),primary branch(PB)number,secondary branch(SB)number(G-H),grain number per-panicle(I),plant height(J),panicle stem diameter(K),stem base diameter(L),and seed setting rate(M)of NIL-GNP6 and NIL-gnp6 plants.TT,total tiller number;LT,lower tiller number;HT,higher tiller number.Data are means±s.e.m.,n=10 plants.Significant differences were determined by Student's t-tests:*P<0.05,**P<0.01.Scale bars,20 cm for A,5 cm for B and C,1 cm for D.

3.4.The SAW motif is necessary for the nuclear localization of GNP6/MOC1

GNP6 encodes a conserved GRAS transcription factor. Phylogenetic analysis of GNP6 protein and its homologs in a range of plant species showed that GNP6 exists in many species,and that the rice gene is most closely related to a homolog in Triticum aestivum(Fig.S3).The GRAS domain consists of LHRI,VHIID,LHTII,PFYRE and SAW motif.The thymine insertion at 985 bp in gnp6 caused a shift in amino acid sequence in the SAW motif and an extended polypeptide length (Fig. 5 A).These changes mainly affected the structure in three beta sheets and two alpha helices, suggesting a likely disrupted function of the motif(Fig.5 B).To determine whether the SAW motif affected subcellular localization of GNP6, constructs,ProSuper:GNP6-GFP and ProSuper:gnp6-GFP, were transformed into rice protoplasts.The protoplast harboring ProSuper:GNP6-GFP showed clear GFP signals in the nuclei, consistent with subcellular localization of MOC1 with unidentified nuclear localization signals (NLS) as previously reported. However,GFP signals in protoplasts transformed with the ProSuper:gnp6-GFP construct were present in the cytoplasm (Fig. 5-C)suggesting that GNP6 importation into the nucleus is dependent on the SAW motif.

3.5. Haplotype analysis of GNP6/MOC1 and MOC3

In order to explore the variations of GNP6/MOC1 in germplasm, the nucleotide sequences of MOC1 were analyzed in 7 aro,3 aus,127 indica(ind)and 103 japonica(jap)accessions.The coding region of MOC1 was highly conserved without any amino acid changes across the 240 accessions. However, the 240 cultivars were divided into 6 haplotypes based on 21 SNPs in the promoter region, and named MOC1-Hap1 to MOC1-Hap6. Haplotypes MOC1-Hap2, MOC1-Hap3, MOC1-Hap5 and MOC1-Hap6 each included only one accession; MOC1-Hap4 was present only in the aus and ind subpopulations, whereas MOC1-Hap1 was identified in the aro, ind and jap subpopulations (Fig. 6 A). Determination of grain numbers per panicle(GNP)and tiller numbers at 30(DT30)and 45(DT45)days post transplanting showed that the ind and aus groups on average had more grains per-panicle than the jap and aro groups (Fig.S4A), and jap exhibited less tillers than the aus, aro and ind subspecies groups(Fig.S4 B).In order to exclude the effects of genetic backgrounds of each subspecies, we compared the phenotypes of MOC1-Hap1indand MOC1-Hap4indaccessions and found that the tiller number of MOC1-Hap4indwas higher than for MOC1-Hap1ind, but there was no difference in GNP(Fig. 6 B; Fig. S5 A). Expression of MOC1 in MOC1-Hap4 accessions was slightly higher than that in accessions containing MOC1-Hap1 (Fig. S6 A). Moreover, there was a positive correlation between MOC1 expression level and tiller number(Fig.S6 B).

MOC1 interacts with MOC3 and functions as a co-activator to promote transcription of targets [14]. A haplotype analysis of MOC3 was also performed on the germplasm panel. The coding region of MOC3 was similarly highly conserved with only synonymous variations. Based on the 31 SNPs in the promoter region we identified five haplotypes and named them MOC3-Hap1 to MOC3-Hap5. MOC3-Hap1 was present mainly in the ind subpopulation, MOC3-Hap2, MOC3-Hap3,and MOC3-Hap4 were mainly present in the jap subpopulation,and MOC3-Hap5 was identified in all four subpopulations(Fig. 6 C). Accessions with MOC3-Hap5 had higher grain numbers per panicle than that those with MOC3-Hap1 in the ind sub-population, and higher tiller number in the jap subpopulation(Fig.6 D; Fig.S5 B).

In order to identify elite haplotypes regarding grain number and tiller number, a joint haplotype analysis of MOC1 and MOC3 variants divided the 240 accessions into 11 haplotypes named H1 to H11.Haplotypes H3,H4 and H7 were present mainly in the ind subpopulation and H9,H10 and H11 mainly occurred in the jap subpopulation(Fig.6 E).There were no obvious differences in grain number per-panicle among varieties with different joint haplotypes (Fig. S5 C). However,H9jap, H10japand H11japhad more tillers than H7japin the jap subpopulation(Fig.6 F).To determine if MOC1 and MOC3 were selected in breeding the proportions of each haplotype in landrace accessions (LAN) and improved accessions (IMP)were compared. The proportions of H3indand H4indwere higher in the IMP group than in LAN, whereas the proportion of H7indwas lower in the IMP group than in LAN (Fig. 6 G).These results indicated that H3indand H4indmay have used in breeding.The proportion of H9japin japonica was higher in IMP than in the LAN group, indicating possible application in breeding (Fig.6 H).

3.6. GNP6/MOC1 acts as upstream regulator in axillary meristem development

The gnp6 mutant has a similar phenotype to lax1, gnp4/lax2 and FZP-OE plants. In order to examine the relationship between GNP6 and these genes, their expression levels were compared in NIL-GNP6 and NIL-gnp6. The expression level of Gnp4 in NIL-gnp6 versus NIL-Gnp6 was significantly lower whereas the expression level of FZP was significantly increased, but there was no significant difference in the expression levels of LAX1.Recently,it was reported that MOC1 interacts with MOC3 and function as a coactivator to activate FON1 expression and thus regulate tiller bud outgrowths[14].We found that the expression level of FON1 was significantly lower in NIL-gnp6, consistent with the function of FON1 (Fig.7). Collectively, our results imply that GNP6 acts upstream of several regulators of axillary branching, such as Gnp4, FON1 and FZP.

4.Discussion

4.1. GNP6 is a novel allele of MOC1

MOC1 is an important regulator of tillering in rice and is required for the axillary meristem (AM) formation and subsequent bud outgrowths. It was first isolated as the normal allele of a moc1 mutant (moc1-1) with a single main tiller. The moc1-1 allele had a 1.9 kb retrotransposon in the ORF,causing premature termination of translation and loss of the SAW motif [11]. Another mutant, moc1-3, with a point mutation causing a premature termination at position Trp 370 in the SAW motif had little effect on tillering, but caused reduced numbers of panicle branches and spikelets [15,35].Mutant moc1-4 with a nonsense mutation within the GRAS domain caused defective panicle branching but no significant effect on tillering [16]. Mutant moc1-5 involved deletion of a cytosine 38 bp downstream from the ATG start codon,causing disruption of the GRAS domain significantly reduced numbers of tillers[25].In this study,we identified a mutant novel allele of MOC1, and named it gnp6 (moc1-6). This allele had a single thymine inserted at 985 bp causing disruption of the SAW motif and a longer polypeptide (Fig. 3 C; Fig. 5 A). The mutation affected both tiller and architectures(Fig.1).

As shown in Fig. S7, compared with moc1-1 and moc1-5,moc1-3 and moc1-6 are weak alleles of MOC1,especially moc1-3 showing no significant difference in tillering.The D-box in the LHRI motif of MOC1 is required for interaction with TE/TAD1 and the interaction mediates its degradation in a cell cycledependent manner thus reducing tiller formation [25]. Truncation of any single motif in MOC1 was not sufficient to eliminate its interaction with SLR1 for being protected from degradation [28]. It is possible that the LHRI, VHIID, LHTII,PFYRE, and SAW motifs in MOC1 and SLR were complementary for each other. Additionally, the 370-441 amino acids SAW motif in the conserved GRAS domain is more essential for panicle branch formation and is possibly needed for interaction with unknown panicle development regulators.

Fig.5- Amino acids analyses and subcellular localization of GNP6/MOC1.(A) Schematics of the conserved motifs in GRAS domain of GNP6 and gnp6.(B)Predicted protein model of GNP6/MOC1 in NIL-GNP6 and NIL-gnp6.The black and red dotted frames indicate possible structural alterations in beta sheets and alpha helices,respectively.(C)Subcellular localization of GNP6 and gnp6 in rice protoplasts.Scale bars,10 μm for C.

4.2. Utilization of natural variation of MOC1 and MOC3 in breeding

Through haplotype analysis, some valuable variations of MOC1 and MOC3 were found in panel rice genotypes. These variations in the promoter regions vary in frequency among the different taxonomic groups. Their caused phenotypic change has only been shown in ind or jap subpopulations,but not in full population. It may be caused by diversified genetic background between ind and jap. As ind varieties possess more tillers and grain numbers per-panicle than that of jap varieties, it is very different to utilize alleles of MOC1 and MOC3 in order to improve tiller number and/or grain number per-panicle between ind and jap.

Our results showed that MOC1-Hap4 promotes more tillers than MOC1-Hap1 in ind population(Fig.6 B),and although the combined genotype of MOC1-Hap4 and MOC3 (H3 + H4) had a tendency to increase tiller numbers there was no difference in ind compared with MOC1-Hap1(H7+H11)(Fig.S5A).Moreover,varieties with MOC3-Hap5 in ind had more grains than those with MOC3-Hap1. Although there was no significant difference in grain number per panicle among joint haplotypes H3,H4, H7 and H11 (Fig. S5B and C), H4indand H11indapparently had higher grain numbers than H3indand H7ind. It also suggested that the combined haplotype MOC1-Hap4 + MOC3-Hap5 had a minor effect in increasing tillers and grain number per panicle in the ind subpopulation. MOC1 had only one haplotype (MOC1-Hap1) in the jap subpopulation, Therefore,to increase the tiller number or grain number per panicle in jap accessions only superior haplotypes of MOC3 need be selected. In this regard MOC3-Hap3/4/5 accessions had more tillers than MOC3-Hap1. However, there was no difference in grain number per panicle among MOC3 haplotypes limiting their value increasing tiller number in jap.

Since panicle number per plant and grain number per panicle are both important components of grain yield it is possible to increase yield by improving either trait.Haplotype analyses of MOC1 and MOC3 indicated that it was difficult to identify a single haplotype of MOC1 or MOC3 to simultaneously increase panicle number and grain number per panicle in ind or jap possibly due to ‘yield component compensation'. Due to the genetic complexity tiller number and grain number and sensitivity to environmental influences the only way to understand the regulatory role of individual haplotypes and haplotype combinations might be to evaluate them by genetic transformation into at least two different common genetic background.

4.3. GNP6/MOC1 is involved in multiple pathways

Several regulators related to GNP6/MOC1 have been identified to control tiller bud formation and growth.SLR1 and TAD1/TE interact with MOC1 and act upstream to suppress or promote its degradation [25,26,28]. MOC1 encodes a GRAS family transcription factor but its direct targets have not been identified. However, it was found that MOC1 functions as a co-activator with MOC3 in activating transcription of FON1.MOC1 and MOC3 independently regulate tiller bud formation and outgrowth, while FON1 only affects tiller bud outgrowth[14]. moc1 (spa)/lax1 and moc1/gnp4 (lax2) double mutants produced only single main tillers and no panicle branching[15-17]. Tiller bud formation is arrested once the activity of MOC1 is disrupted.This means that MOC1 is vital for tiller bud formation and outgrowth. The lax1, gnp4 and fon1 single mutants showed slight variations in tiller numbers due to differing effects in reducing tiller outgrowth. The lax1/gnp4(lax2) double mutant showed one main tiller but without secondary panicle branch.These imply that there exist Gnp4-dependent and GNP6/MOC1-dependent pathways for tiller formation. Additionally, a complex cross may exist among them to regulate tiller formation and unknown regulatory mechanism to regulate panicle branch formation, especially primary branch as shown in Fig.S8.

Fig.6- Haplotype analysis of GNP6/MOC1 and MOC3.(A).Haplotype frequencies of GNP6/MOC1.(B). Tiller numbers in GNP6/MOC1 30 and 45 days post anthesis.Data are means±s.d.,n =87, 103,and 36 for MOC1-Hap1ind,MOC1-Hap1jap and MOC1-Hap4ind,respectively.(C).Haplotype frequencies for MOC3.(D).Tiller numbers of MOC3 haplotypes.Data are means±s.d.,n=89,7,23,5,38 and 67 for MOC3-Hap1ind,MOC3-Hap1jap,MOC3-Hap3jap,MOC3-Hap4jap,MOC3-Hap5ind,and MOC3-Hap5jap,respectively.(E).Joint haplotypes of GNP6/MOC1 and MOC3.(F).Tiller numbers of different joint haplotypes.Data are means±s.d.,n=18,18,67,7,23,5,20 and 67 for H3ind,H4ind,H7ind,H7jap,H9jap,H10jap,H11ind and H11jap,respectively.Different lowercase letters above the error bar indicate significant differences between the means(P>0.05,Student's t-test).(G-H).Frequencies of joint haplotypes in landrace(LAN)and improved(IMP)accessions within the ind (G)and jap(H)subpopulations.

5. Conclusions

We identified gnp6, a novel mutant allele of moc1 showing a similar arrested branch formation phenotype to mutant gnp4.GNP6 was fine-mapped to a 36.6 kb region on chromosome 6 and the genomic sequence showed a single thymine insertion in the coding sequence in the SAW motif of the GRAS domain.The SAW motif is essential for nuclear localization of GNP6/MOC1 where it function as a transcription factor or coregulator. Haplotype analysis of a panel 240 rice accessions provided evidence that combinations of MOC1-Hap1 with MOC3-Hap3, MOC3-Hap4 or MOC3-Hap5 had potential for yield improvement.

Fig.7- GNP6 acts upstream of several regulators of branching.Total RNAs extracted from the young panicles were used for cDNA synthesis.Data are means±s.e.m.,n=3 replicates.Significant differences were determined by Student's t-test:*P<0.05,**P< 0.01.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2020.04.011.

Declaration of competing interest

Authors declare that there are no conflicts of interest.

Acknowledgments

This work was supported by the National Natural Science Foundation of China(31801324 31171521),the Open Project of Guangxi Key Laboratory of Rice Genetics and Breeding (2018-05-Z06-KF08) and China Postdoctoral Science Foundation(2017T100117 and 2019M650902).

Author contributions

Zhanying Zhang designed the research, performed most of experiments, and wrote the manuscript. Xingming Sun conducted haplotype analysis. Xiaoqian Ma was responsible for the phenotypic analysis of the germplasm panel. Bingxia Xu, Yong Zhao, Zhiqi Ma, and Gangling Li performed part of the work.Najeeb Ullah Khan revised the paper.Yinghua Pan,Yuntao Liang, Hongliang Zhang, and Jinjie Li provided technical assistance. Zichao Li conceived the project and revised the manuscript.