Transcriptome analysis of near-isogenic line provides novel insights into genes associated with panicle traits regulation in rice

Panicle traits in rice impact yield and quality. The OsGRF4 gene encodes a growth-regulating factor controlling panicle traits, and was recently cloned. Gene expression profiling analysis can be used to study the molecular mechanisms underlying OsGRF4 regulation. Use of near-isogenic lines (NILs) reduces genetic background noise in omics studies. We compared transcriptome profiling of 7 cm long young panicles of NIL-Osgrf4 and NIL-OsGRF4 using RNAs sequence analyses. Eighty differentially expressed genes (DEGs) were identified. Our target gene OsGRF4 was up-regulated in NIL-OsGRF4 plants, which is consistent with a previous qPCR analysis. Hierarchical cluster analysis showed OsGRF4 is tightly clustered with the up-regulated DEG LOC_Os02g47320. Gene Ontology (GO) and KEGG analysis suggested that DEGs were primarily involved in somatic embryogenesis and chitinase activity. Two up-regulated DEGs, LOC_Os04g41680 and LOC_Os04g41620, were significantly enriched in the top 8 GO terms, and were over_represented in term of seed development, and may play key roles in grain shape regulation. The transcription factor Osmyb1 also exhibited differential expression between NILs, and may be is an important regulator of panicle traits. By searching reported functions of DEGs and by co-localization with previous identified quantitative trait loci (QTL), we determined that the pleiotropic gene OsGRF4 may also be involve in abiotic stress resistance. This study provides new candidates genes for further understanding the molecular mechanisms underlying rice panicle trait regulation.


Introduction
Rice is a crucial agronomic crop worldwide, especially in Asians. By the year 2030, a 40% increase in rice production will be necessary to accommodate the rapidly increasing world population [1]. Panicle traits such as grain size, panicle shape, seed shattering, and seed germination, are the main determinants of yields and quality in rice. Recent developments in molecular genetics approaches have led to cloning of many panicle trait genes in rice. Plant hormones such as cytokinin, auxin, and brassinosteroid (BR) play an importance role in rice a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 panicle development. Grain number 1a (Gn1a) was the first cloned major QTL responsible for grain number in rice. Gn1a encodes a cytokinin oxidase/dehydrogenase (OsCKX2), which can reduce the content of phytohormone cytokinin. Low expression level of Gn1a causes cytokinin accumulation in the inflorescence meristem, which results in increased grain number and yield in rice [2]. GAD1 encodes a small signal peptide, and loss of function in GAD1 leads to increases grain number and grain length in cultivated rice. It was speculated that GAD1 activates DST and CKX2 to degrade the phytohormone cytokinin, and results in reduced grain number in wild rice [3]. Auxin is another key plant hormone in rice panicle development. The PAY1 gene improves secondary branching, grains per panicle, and grain yield per plant, by influencing the activity of polar auxin transport and altering distribution of indole-3-acetic acid [4]. In recent years more attention was concentrated on BR research in rice. GNS4, a novel allele of DWARF11, significantly enhances grain number and grain size by regulating the expression levels of BR synthesis and BR response related genes [5]. MRG702 also regulates grains per panicle and grain shape by affecting BR biosynthesis and signal transduction [6]. OsRLCK57 interacts with OsBRI1 (a rice BR receptor) and negatively regulates BR signaling, which affects rice panicle secondary branching [7].
Among cloned rice panicle trait genes, some show pleiotropic effects on tiller or heading date and plant height. In some cases, tillers and panicle branches are consistently regulated by a single gene. MONOCULM 1 (MOC1) is an important example in rice, the moc1 mutant plants have fewer tillers and panicle branches [8]. Similarly, LAX1 regulates both tillers and panicle branches in coordination [9]. However, in some cases, the number of tillers or panicle branches changes in the opposite way. For example, IPA1 (Ideal Plant Architecture 1) encodes OsSPL14 and is regulated by OsmiR156. OsSPL14 mutation perturbs OsmiR156 direct regulation of OsSPL14, which results in reduced tiller number and increased panicle branches [10]. This is similar to PAY1, where the PAY1 mutant displays reduced tiller numbers and increased panicle branches [4]. These findings suggest that rice tillers and panicle branches may be regulated by distinct mechanisms. Some grain number regulation genes demonstrate pleiotropic roles in regulating heading date and plant height. Ghd7 encodes a CCT domain protein, and has a significance effect on grain number. By regulating Ehd1 and Hd3a, Ghd7 affects heading date under long-day conditions. Ghd7 can also enhance plant height by increasing cell numbers [11]. Ghd8/DTH8 and Ghd7.1/DTH7 also have pleiotropic effects on grain number, heading date, and plant height [12][13][14][15].
OsGRF4, an important pleiotropic gene can significantly enhance rice yield, and was recently cloned. OsGRF4 encodes the growth-regulating factor 4, a transcription activator, which improve grain size, panicle length, and seed shattering. A rare nucleobase polymorphism in OsGRF4 in the target site of OsmiRNA396 is correlated to high expression of OsGRF4, which leads to the phenotypic changes in panicle traits [16][17][18]. OsGSK2 negatively regulates transcription activation activity of OsGRF4 by direct interaction; OsGRF4 also interacts with GRF-interacting factor 1 (OsGIF1), which results in bigger grain [19]. The OsmiR396c-OsGRF4-OsGIF1 regulatory pattern influences grain shape and improves rice yield [20]. Field studies show that OsGRF4 can increase yield of hybrid rice (13.7%-28.0% increase) [16,20]. Despite this, the molecular mechanisms of OsGRF4 in regulating panicle traits remain elusive.
RNA sequencing technology (RNA-Seq) is a promising means for genome-wide transcriptomic analysis. Recently, RNA-Seq has been widely applied to transcriptional analyses associated with rice fertility, heterosis, drought tolerance, salinity stress, heat stress, cold tolerance and biotic stress [21][22][23][24]. In consideration of OsGRF4 is a notable panicle trait gene and has broad application prospects in rice breeding. In this study, to illuminate gene expression regulatory networks involved in panicle trait development, we performed a comprehensive transcriptomic analysis between NIL-OsGRF4 and NIL-Osgrf4 in young panicles that were 7 cm long using RNA-seq. We detected over 331 million raw reads, of which roughly 317 million reads were clean. The error rate of all samples was very low (0.02%). In total, we identified 80 significant differentially expressed genes (DEGs). Hierarchical Cluster analysis showed that OsGRF4 was tightly clustered with the up-regulated DEG LOC_Os02g47320, those two genes have highly similar gene expression pattern and LOC_Os02g47320 may play a crucial part in panicle trait regulation. We searched the reported functions of DEGs and compared them with previous identified quantitative trait loci (QTL) influencing panicle traits and stress tolerance. Out result suggest that the pleiotropic gene OsGRF4 may be involved in abiotic resistance. The novel candidate genes identified here may play a significant role in panicle traits regulation, and our results provide valuable information for understanding the molecular mechanisms underlying rice panicle trait development.

Plant materials
The big-grain rice variety CDL (contains OsGRF4) was crossed with the medium-grain variety R1126 (contains Osgrf4, an allele of OsGRF4) to obtain F 1 plants, and F 10 population of recombined inbred lines (RIL) were obtained by single-seed descent. Grain shape difference was observed in one F 10 line (L28), and L28 showing heterozygous at OsGRF4 locus was selfed to yield NIL-OsGRF4 and NIL-Osgrf4.

RNA isolation
Young panicles 7 cm in length were collected from NIL-OsGRF4 and NIL-Osgrf4 for RNA isolation and total RNA was extracted using TRlzol Reagent according to the manual instruction (Life technologies, California, USA). Each sample had three biological replicates. The concentration and purity of each RNA sample were assessed by gel electrophoresis and a NanoPhotometer 1 spectrophotometer (IMPLEN, CA, USA). RNA integrity was assessed with an Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA, USA). RNA was used for RNA-seq and qRT-PCR.

Differential expression analysis
The reads numbers mapped to each gene were counted using HTSeq v0.6.1 software, and gene expression levels were calculated using the reads per kb per million reads (RPKM) values. Differential gene expression analysis between NIL-OsGRF4 and NIL-Osgrf4 plants was performed by the DESeq R package, filtering DEGs with |log2 (Fold Change)| > 1 and corrected p-value (padj) < 0.05. Gene ontology (GO) enrichment analysis of DEGs was implemented by the GOseq R package. GO terms with corrected p-value < 0.05 were considered significantly enriched. KOBAS software was used to test the statistical enrichment of DEGs in the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. RNA-seq data were deposited in the National Center for Biotechnology Information Sequence Read Archive under accession number SRP131560.

RNA-seq validation by quantitative real-time PCR
To validate RNA-seq DEG data, quantitative real-time PCR (qRT-PCR) was conducted on 24 randomly selected genes (12 down regulated and 12 up-regulated DEGs). Total RNA of NIL-OsGRF4 and NIL-Osgrf4 was reverse transcribed using the TransScript All-in-One First-Strand cDNA Synthesis SuperMix for quantitative PCR (qPCR) kit (TransGen Biotech, Beijing, China). The relative expression was analyzed from cycle threshold values using the 2 −ΔΔCt method with the rice Ubiquitin 5 gene used as a reference, and all of the results had three biological replicates. The 24 DEGs and primer sequences are listed in S1 Table.

The effect of OsGRF4 on panicle traits phenotype of rice
Rice grains of NIL-OsGRF4 were significant bigger and heavier compared to that of NIL-Osgrf4. This difference resulted in an increased storage capacity (Grain weight × Spikelet number per panicle × Tiller number) in NIL-OsGRF4 plants by 30.37%. OsGRF4 also positively regulated panicle length, but resulted in lower seed setting percentage. Interestingly, OsGRF4 also significantly improve shattering degree [18]. The two NILs had similar germination rates just after harvest, but NIL-OsGRF4 had very low germination after storing at room temperature for one year (S1 Fig). This result suggests that OsGRF4 may negative regulate germination.

Whole genome transcriptome profiles of NIL-Osgrf4 and NIL-OsGRF4
The young panicle stage is critical for panicle traits determination. OsGRF4 had the highest expression levels in young panicles measuring 7 cm long, and the relative expression was different between NIL-Osgrf4 and NIL-OsGRF4 [18]. Therefore, we sequenced the transcriptome profiles of 7 cm young panicle from the two NILs. We detected over 331 million raw reads, and obtained roughly 317 million clean reads, accounting for 95.83% of the raw reads. The error rate was low for all samples (0.02%). The average Q20 and Q30 were 96.35% and 91.02%, respectively. After aligning clean reads to the rice indica varieties 9311 reference genome, a total of 80.24-81.37% of reads were mapped, 78.39% of mapped reads were unique, and 2.07%-2.65% of reads were mapped to multiple loci (Table 1). Significant correlation was detected among transcriptome data across all biological replicates of both from NIL-OsGRF4 and NIL-Osgrf4 with a correlation coefficient greater than 0.97. The correlation analysis also showed that NIL-OsGRF4 and NIL-Osgrf4 had similar genetic background (S2 Fig).

Identification of DEGs between NILs
We detected 24807 and 25370 expressed genes in NIL-OsGRF4 and NIL-Osgrf4, respectively. A total of 24280 genes were common to both genotypes, with 527 and 1090 genes uniquely expressed in NIL-OsGRF4 and NIL-Osgrf4, respectively (S3 Fig). In total, 80 significant DEGs were detected between NIL-OsGRF4 and NIL-Osgrf4, including 23 up-regulated genes and 57 down-regulated genes (Fig 1a; S2 Table). RNA-seq showed the expression level of the target gene OsGRF4 (LOC_Os02g47280 / BGIOSGA005785) was up-regulated in NIL-OsGRF4 plant, which agrees with our previously reported qPCR analysis [18]. Hierarchical cluster analysis of DEGs showed that OsGRF4 was tightly clustered with LOC_Os02g47320 (S4 Fig), showing that these two genes have similar gene expression patterns. NIL-OsGRF4 and NIL-Osgrf4 clustered together (Fig 1b), which revealed analogous genetic backgrounds.

Expression patterns of DEGs were validated by qRT-PCR
To confirm DEGs identified in our transcriptome analysis, 24 randomly selected DEGs were analyzed by qRT-PCR. The qRT-PCR analysis showed that the 12 up-regulated DEGs (such as LOC_Os02g47320, LOC_Os04g55159 and LOC_Os04g41680, which involved in vacuolar ATPase G subunit, seed storage and chitinase family protein precursor) demonstrated higher  expression levels in NIL-OsGRF4 than in NIL-Osgrf4 (Fig 2). The 12 down regulated DEGs (such as LOC_Os05g35500, LOC_Os04g28620 and LOC_Os09g25850, which involved in MYB transcription factor, male sterility and uncharacterised domain Wax2) showed lower expression levels in NIL-OsGRF4 than in NIL-Osgrf4. Expression trends were consistent for all transcripts in qRT-PCR and RNA-Seq analyses.

Functional classification of DEGs by GO analyzing
Gene ontology analysis of up-regulated genes showed 10 DEGs were enriched in 230 GO terms, including 63 functional terms in molecular function, 46 in cellular components, and 121 in biological processes. Our target gene OsGRF4 was enriched in 83 functional terms, and over-represented (P < 0.05) in 2 GO terms, hydrolase activity (GO:0016787) and macromolecule metabolic process (GO:0043170) (S3 Table). Eight significantly enriched GO terms (corrected P-value < 0.05) were detected in up-regulated DEGs ( Table 2, S3 Table, Table).

KEGG pathway analysis of DEGs
KEGG analysis showed 10 out of 80 DEGs were enriched across 17 functional categories. Among the up-regulated genes, one showed notable enrichment (corrected P-value < 0.05) in the pathways for amino sugar and nucleotide sugar metabolism (osa00520), suggesting this pathway may perform an important function in rice panicle development (S5 Table). Among the down-regulated genes, three over-represented KEGG terms (P < 0.05) were identified, including cutin, suberine and wax biosynthesis (osa00073), pantothenate and CoA biosynthesis (osa00770), and beta-Alanine metabolism (osa00410) (S6 Table). Subsequent functional analysis of DEGs involved in the KEGG pathway may help elucidate the complex regulatory networks of panicle traits in rice.

Chromosome co-localization of DEGs with previously identified QTL responsible for panicle traits
To identify the potential function of DEGs, we co-localized previously reported QTLs (http:// qtaro.abr.affrc.go.jp/, http://archive.gramene.org/) and DEGs onto rice chromosomes. A total of 54 DEGs (18 up-regulated and 36 down-regulated) were co-localized within 50 QTLs intervals. Among them, 22 QTLs were responsible for grain size or grain weight, 4 for panicle length, 2 for seed shattering, 2 for seed set percent, 9 for seed germination/dormancy, and 11 for cold/drought tolerance. Chromosome 2 had the greatest number of co-localized QTLs with 11 QTLs, followed by chromosome 4 with 8 QTLs (Fig 4, S7 Table). Three up-regulated DEGs (LOC_Os02g55670, LOC_Os03g26350, and LOC_Os04g53550) co-localized with 3 panicle traits QTLs, respectively. The pleiotropic QTLs SLCHL4 (NAL1) and OsPTR9 controlling grain weight and panicle size [25,26] were co-localized with LOC_Os04g53550 and LOC_Os06g50980, respectively. The up-regulated DEG LOC_Os04g41680 belonged to the significantly enriched KEGG pathway (S5 Table) and colocalized with yd4 responsible for grain yield [27], which suggests that LOC_Os04g41680 may be a candidate panicle trait QTL, playing an important role in the regulatory network of panicle traits by differential expression levels. We also found some DEGs were located in cold/  (Fig 4, S7 Table). For instance, pleiotropic QTLs OsHsp23.7 and qLTG-7 were not only responsible for seed germination/dormancy but also for cold/drought and salt tolerance [28,29]. Interestingly, our target gene OsGRF4 (LOC_Os02g47280) was located in a cold tolerance QTL qSDW2 [30] and a drought tolerance QTL qGY-2b [31]. This suggests that the pleiotropic gene OsGRF4 may be involved in abiotic stress resistance. A previous study showed that OslecRK was responsible for both plant innate immunity and seed germination, and knocking out OslecRK reduced seed germination and resistance to bacterial and fungal pathogens as well as insects in rice [32]. We found a down-regulated gene (LOC_Os05g35500) encoding the Myb transcription factor Osmyb1, which may be involved in trans-regulation, was co-localized with 3 QTLs responsible for grain weight (QTL gw5b and gw5.1) and seed set percent (QTL ssp-5) [33][34][35]. In summary, the co-localized DEGs identified in this study provide a starting point for elucidating the molecular regulation mechanisms of rice panicle traits.

Discussion
Panicle traits such as panicle shape, grain size, seed germination/dormancy, and seed shattering are important to rice yield and quality. Although a number of genes controlling panicle traits have been isolated, the genetic regulation mechanisms underlying them are still uncertain. Gene expression profiling analysis is a good method for understanding the molecular mechanisms involved in panicle traits regulation. At the same time, current sequencing technologies provide high throughput, accurate, and economical methods to study whole transcriptomes. In recent years, RNA-seq has been widely applied to investigate the transcriptome of numerous plants species across a range of environment conditions, including rice [22], maize [36], sorghum [37], tea plant [38], bamboo [39] and wild species as well [40,41]. NILs are ideal material for detecting DEGs, having a single gene difference between two genotypes, which eliminates most genetic background noise. For example, Raorane et al analyzed the proteome of spikelets, roots, and flag leaves between NILs with different drought resistant phenotypes and identified DEGs under drought conditions [42]. Similarly, we performed a transcriptomics sequence analysis of young rice panicles between NIL-Osgrf4 and NIL-OsGRF4 to study molecular mechanisms of panicle trait regulation. By comprehensively analysis of RNAseq data, we provide new insights into panicle trait regulation in rice.
Our RNA-Seq analysis identified 80 significant DEGs, including 23 up-regulated DEGs and 57 down-regulated DEGs. The target gene OsGRF4 (LOC_Os02g47280) was up-regulated according to RNA-Seq results, which is consistent with our previous qPCR results [18]. To evaluate the potential role of identified DEGs, we searched reported functions in existing databases. Eighteen DEGs were reported to be involved in panicle trait regulation in rice. The  down-regulated DEG BSG1/TH1 (LOC_Os02g56610) encodes a DUF640 domain protein and determines grain width, thickness, and yield by regulating expression levels of cell expansion and division genes [43,44]. BSG1/TH1 also co-localized with 100-grain weight QTL gw2.1 and panicle length QTL pl2.1 [45]. Rice grain production is also regulated by cytokinin content [2], and we identified four cytokinins significant differences between NIL-OsGRF4 and NIL-Osgrf4 [18]. The purine permease family (OsPUP) is involved in cytokinin transporter and has 12 members in rice. OsPUP7 mutants had high cytokinin content, which resulted in larger seeds and increasing sensitivity to salt and drought stresses [46]. The down-regulated DEG OsPUP11 (LOC_Os02g46380), is also a cytokinin transporter and may influence grain shape. Polyamine is important for normal seed development in Arabidopsis [47] and polyamine oxidase (PAO) is a FAD-dependent enzyme associated with polyamine catabolism. The rice genome has 7 PAOs [48], and of these, the down-regulated DEG OsPAO2 (LOC_Os03g09810) co-localized with seed weight QTL gw3 [49], thus, OsPAO2 may influence grain weight. DEG LOC_Os02g50490 was mainly detected during seed and panicle developmental (stages −5, −6), and therefore may be involved in panicle trait regulation [50]. Transcription factors (TFs) are regulatory proteins and plays a crucial role in panicle trait development [51,52]. The seed shattering gene sh4 encodes a nuclear TF that is a member of trihelix TF family, which may be derived from the MYB gene family [53]. Four up-regulated DEGs: LOC_Os02g49420, LOC_Os04g41680, LOC_Os04g41620, and LOC_Os08g39330, are up-regulated by sh4 [54]. In plants, MYB TFs have been isolated from many species, and serve various functions including involvement in development, primary and secondary metabolism, flowering time, and stresses response [55,56]. The Ant28 gene encodes a R2R3 MYB protein which influences barley seed dormancy [57]. MYB56 encodes a R2R3 MYB TF, which positively controls seed size in A. thaliana in an unknown pathway [58]. Osmyb1 (LOC_Os05g35500) was down-regulated in this study and was found to be co-localized with 3 QTLs which determine grain weight (QTL gw5b and gw5.1) and seed set percent (and QTL ssp-5) [33][34][35]. Phylogeny analysis showed that AtMYB66 (encoding the R2R3 MYB protein) had the closest genetic relationship with Osmyb1 [59]. The expression levels of Osmyb1 and Osmyb4 peaked in seeds at 14 days after flowering, which suggests that these two genes may function in seed maturation [60]. Other studies have shown that Osmyb1 is associated with cold and drought tolerance [61,62]. Therefore, the Osmyb1 TF may act as a key regulator of panicle traits and abiotic resistance.
The pleiotropic gene OsGRF4 may also affect other phenotype besides panicle trait. It was reported that OsGRF4 (LOC_Os02g47280) was a target gene of osa-MIR396c, and overexpression of osa-MIR396c was shown to reduce alkali and salt stress tolerance. OsGRF4 was up-regulated when treated by saline, anoxia, ABA, and gibberellins, and down-regulated by arsenate [63]. Another study found that LOC_Os02g47280 showed differential expression in root knot nematode infection [64]. Interestingly, we found that OsGRF4 co-localized with both a cold tolerance QTL (qSDW2) [30] and a drought tolerance QTL (qGY-2b) [31]. Many DEGs detected in this study, such as LOC_Os04g55159, LOC_Os04g41680, LOC_Os04g41620, LOC_Os04g53550, OsGL1-1 (LOC_Os09g25850), LOC_Os10g31330 and LOC_Os04g33990, were found to be regulated by biotic and abiotic stress such as cold, salt, and drought stress as well as blast disease infection [65][66][67][68][69]. The Osmyb1 TF not only influences seed maturation [60], but was also found to be associated with cold and drought tolerance [61,62]. Therefore, OsGRF4 and some DEGs detected in this study may be involved in abiotic and biotic stress responses, which need further investigation.
In our GO enrichment analysis, two up-regulated DEGs (LOC_Os04g41680 and LOC_ Os04g41620) were significantly enriched in the top 8 GO terms, and were Over_represented in term of seed development (GO: 0048316) (S3 Table). Therefore, LOC_Os04g41680 and LOC_Os04g41620 may be important genes regulating grain shape. Along with LOC_ Os04g41680 and LOC_Os04g41620, three up-regulated genes (OsGRF4, LOC_Os02g47320, and LOC_Os02g46260) and 9 down-regulated genes (LOC_Os02g50490, LOC_Os03g01800, LOC_Os02g48900, LOC_Os12g04320, LOC_Os10g05910, LOC_Os06g13830, LOC_ Os06g06250, LOC_Os05g06720, and LOC_Os07g46350) were enriched for hydrolase activity (GO: 0016787) (S3 and S4 Tables). TGW6 encodes a novel protein that has indole-3-acetic acid-glucose hydrolase activity, and enhances rice seed weight and yield [70]. Therefore, DEGs enriched in hydrolase activity GO term may be potential candidate genes for grain shape. OsGRF4 and LOC_Os02g47320 were collectively enriched in 14 GO terms (S3 Table), and LOC_Os02g47320 was also involved in 3 KEGG pathways (S5 Table). In addition, OsGRF4 was tightly clustered with LOC_Os02g47320 (S4 Fig), which shows that these two genes have highly similar gene expression pattern and LOC_Os02g47320 may play a crucial role in panicle trait regulation, however further research is needed.
In summary, our transcriptome sequence analyses provide some important candidate genes involving in panicle trait regulation, which may be used in future gene cloning studies. Functional analysis of the DEGs and pathways identified here will serve to further our understanding of the transcriptional network and molecular mechanisms underlying panicle trait regulation in rice.