Expression profiling analysis of RopGEF genes
To identify the pollen-preferred genes of rice, we examined a publicly available Affymetrix rice microarray data [31]. Expression patterns conserved between indica and japonica rice varieties at various developmental stages of anthers and pollen, and anatomical meta-expression data from our established database, Rice Anther Pollen Expression Database (RAPED) (http://ricephylogenomics-khu.org/RAPED/home.php), were analyzed. We then constructed a phylogenic tree by aligning the protein sequences of all OsRopGEFs. Meta-expression data in the context of the phylogenetic tree were combined to check the function of OsRopGEFs with respect to anther and pollen development, compared with other tissues/organs (Fig. 1a). As a result, the rice RopGEF gene family can be divided into two subfamilies, one group to which OsRopGEF2, 3, 4, 6, and 8 belong, and the other to which the remaining six genes belong, namely OsRopGEF1, 5, 7, 9, 10, and 11.
In addition, we identified that OsRopGEF2, OsRopGEF3, OsRopGEF6, and OsRopGEF8 showed selective expression at the stage of late pollen development, i.e., at stages of tricellular pollen grains, mature pollen (MP), and germinated pollen (GP). These four genes were used for further analyzes. Since OsRopGEF3 was also highly expressed in root hairs, we predicted that this gene could affect tip growth (Additional file 1). Genes OsRopGEF4 and OsRopGEF7 showed some expression in the early stages of anther development, although their expression levels decreased during the later stages. Furthermore, because expression is also detected in tissues other than anther and pollen, these two genes cannot be considered to be pollen-preferred expressed genes. In addition, OsRopGEF5 showed weak expression in japonica rice anthers and other tissues, showing no preferred expression in anthers and pollen.
Next, to verify the meta-expression data, we performed quantitative real-time PCR (qPCR) using ten tissues: shoot, root, leaf blade, panicle, flower, seed, tetrad microspore, young microspore, vacuolated pollen, and MP (Fig. 1b). By sampling anthers at various developmental stages, from pollen mother cell to MP and GP, we tried to determine at which stage the target genes were being expressed. As a result, OsRopGEF2, OsRopGEF3, OsRopGEF6, andOsRopGEF8 were shown to be highly expressed in MP. OsRopGEF6 showed the highest level of expression in MP, and OsRopGEF2 and OsRopGEF8 showed similar expression levels in MP. These native pollen-preferred OsRopGEFs were generally not expressed in other tissues, similar to meta-expression profiles using microarray data. Especially in rice, gene expression profiles of mature and germinated pollen were highly correlated [31]. Therefore, these four genes may play key roles in pollen tube growth and other processes that occur after pollen maturation.
Comparative analysis of the conserved domains in RopGEF genomic and protein sequences between rice and Arabidopsis
Genes OsRopGEF2, OsRopGEF3, OsRopGEF6, and OsRopGEF8 encode proteins with similar amino acid sequences. RopGEFs are considered to be a novel gene family with a unique structure containing the PRONE catalytic domain that is exclusively found in the plant RopGEF gene family [9]. The PRONE domain is known to be required for RopGEFs to convert GDP to GTP. The PRONE domain of RopGEFs is highly conserved in all OsRopGEFs. More interestingly, the pollen-preferred OsRopGEFs were found to retain the C-terminus region after the PRONE domain (Fig. 2b), whereas the pollen-non-preferred OsRopGEF1, OsRopGEF9, and OsRopGEF11 lack this C-terminus.
For the comparative analysis of RopGEF gene families, we collected the protein sequences of 14 AtRopGEFs and 11 OsRopGEFs and constructed a neighbor-joining tree (Fig. 2a). Four OsRopGEF genes showing pollen-preferred expression were clustered together with AtRopGEF8 and 9, which also exhibited pollen-preferred expression. In the Arabidopsis RopGEF family, the expression patterns revealed by qPCR were not consistent with those from previous reports: AtRopGEF1, 8, 9, 12, 14 were detected in pollen tissues [8] and AtRopGEF8, 9, 10, 11, and 13 showed pollen-preferred expression, respectively, and, of these, the latter data corresponded more closely with the transcriptome data in Genevestigator (Additional file 2). According to Rice Genome Annotation Project (RGAP, http://rice.plantbiology.msu.edu/), the orthologs for OsRopGEF2 are AtRopGEF8 and AtRopGEF9, and those for OsRopGEF3 are AtRopGEF11, AtRopGEF12, and AtRopGEF13. However, the orthologs for OsRopGEF6 and OsRopGEF8 have not been identified. In the phylogenetic tree, rice members with pollen-preferred expression were closely clustered with those in Arabidopsis and the C-terminal amino acid sequences were more similar than those for other members in rice. The invariant serine residue (S510, numbered as in AtRopGEF12) within the C-terminus, which is important for C-terminal inhibition [13] was also located in pollen-preferred RopGEF members in rice (Additional file 3), supporting the hypothesis that phosphorylation-regulated GEF activity was conserved.
Protein structure
Protein sequence analysis and hydropathy plot profiling revealed that AtRopGEF and OsRopGEF, both with pollen-preferred expression, had mostly similar amino acid sequences but differed at the N- and C-termini (Fig. 3). The analyzed rice RopGEF 3-D models exhibit a butterfly-shaped three-dimensional structure in which two PRONE protomers dimerize via their N-terminal groups (Fig. 4). Each protomer consisted of two subdomains, with the subdomain 1 containing the WW-loop, which is a common characteristic in Arabidopsis [7] (Table 1). To more precisely characterize the 3-D model structures, we compared the pollen-preferred rice RopGEF 3-D models with those from Arabidopsis. The subdomains differed in the number of alpha helices and the number of residues in the WW-loop between the OsRopGEFs and the AtRopGEFs. Rice and Arabidopsis RopGEF 3-D models displayed trends in the number of α-helices from 14 to 17, and in the number of residues in the WW-loop from 25 to 50.
OsRopGEFs contained 14 to 15 alpha helices, and 25 to 40 residues in the WW-loop, whereas AtRopGEFs contained 14 to 17 alpha helices, and 38 to 50 residues in the WW-loop. The main difference between OsRopGEFs and the reported AtRopGEF [32] is in the WW-loop, where AtRopGEF displayed α-helices or beta-strands in the WW-loop, whereas OsRopGEFs did not contain α-helices or beta-strands.
Subcellular localization of four RopGEF proteins
The RopGEF is known to regulate ROP signaling in the plasma membrane (PM). Arabidopsis RopGEF8, RopGEF9, and RopGEF14 were shown to be localized to the apical PM of the pollen tube, and RopGEF1 to the entire PM. On the other hand, RopGEF12 was barely detected in the PM, being present in the cytosol of the pollen tube [8], suggesting that the different localizations of RopGEF in pollen tubes represented differences with respect to multiple regulatory ROP signaling actions. Therefore, we examined the subcellular localization of OsRopGEF members to check whether their spatial locations were the same or different. We introduced RopGEF-GFP fusion proteins controlled by the Cauliflower mosaic virus (CaMV) 35S promoter into the epidermal cells of tobacco leaves, and empty GFP protein was used as a control. The GFP signals of four RopGEF genes were observed in the PM. Further, we used FM4-64 staining as a membrane marker to confirm how the membrane marker and the GFP signal of RopGEF protein correlate (Fig. 5). Control GFP signals were present in the nucleus, membrane and cytosol inside membrane (Fig. 5a-5d). Most of the RopGEF protein signals are well merged with the FM4-64 stained RFP signal (Additional file 4). OsRopGEF6 were specifically localized into the PM, consistent with the common RopGEF location (Fig. 5m-5p), while the GFP signals of OsRopGEF2, OsRopGEF3 and OsRopGEF8 were predominantly found in the cytosol as well as in the PM (Fig. 5i-5l). The GFP signal of OsRopGEF2, located in the cytosol, was seems like associated with the endoplasmic reticulum, rather than the nucleus (Fig. 5e-5h) [33], while the OsRopGEF8 signal was associated with the nucleus (Fig. 5q-5t) [34]. The results suggest that the four pollen-preferred members of OsRopGEF could carry out unique functions during pollen germination, as reflected by their distinct subcellular locations.
Promoter analysis of pollen-preferred genes
Up to now, the information about the roles of transcription factors and regulatory motifs in pollen germination in rice has been unclear. Based on the microarray data and the qPCR results, we performed a promoter analysis to identify the cis-elements conserved in the promoters of OsRopGEFs associated with pollen-preferred expression.
We searched the known cis-regulatory elements (CREs) in the promoter based on existing research (Additional file 5). Major pollen-preferred CREs such as POLLEN1LELAT52 (AGAAA), PB Core (CCAC), and GTGANTG10 (GTGA) [35] were identified. In the case of POLLEN1LELAT52, 12 CREs existed in the OsRopGEF3 promoter and seven CREs existed in the OsRopGEF8 promoter. In the case of GTGANTG10, 11 CREs existed in the OsRopGEF8 promoter. In the case of the PB Core, there were five copies in the OsRopGEF2 promoter, and four copies in each of the OsRopGEF6 and OsRopGEF8 promoters. However, in the case of the remaining OsRopGEF family genes, there were also many pollen-preferred CREs present in the promoters of genes which exhibited little or no expression in pollen. On average, there were 4.3 copies of POLLEN1LELAT52, 6.4 copies of GTGANTG10, and 1.6 copies of the PB Core in the promoters of the seven OsRopGEFs compared with the four highly pollen-preferred genes. We calculated the p-values to determine how many of the four highly pollen-expressed OsRopGEF genes had significantly more CREs relative to the rest of the OsRopGEF genes, but all did not exceed 0.01. Based on these results, we assume that known pollen CREs do not regulate the expression of the four pollen-preferred OsRopGEFs.
We then searched to find CREs that existed in the promoters of only pollen-preferred OsRopGEFs. Firstly, we searched for motifs that were common to the promoters of the four pollen-preferred genes, using MEME (Fig. 6). We downloaded the 2000-base pair upstream sequences of the four OsRopGEF genes and found ten conserved elements. Among them, three CREs were absent from the promoters of the other, pollen-non-preferred OsRopGEF genes. These three CREs were present at −2000 to −1500 bp from the transcription start sites of OsRopGEF3 and OsRopGEF6, and mainly existed within −1000 bp of the transcription start sites of OsRopGEF2 and OsRopGEF8. All CREs were present once in the promoter of each gene, except for the first CRE (presented as a dark blue box; Fig. 6), which was present twice in the promoter of OsRopGEF2. We predicted that these three CREs will be involved in pollen expression, indicating their functional similarity in the transcriptional regulation process. We analyzed whether these three CREs are conserved in the promoter of AtRopGEF, which is highly expressed in pollen. Although ATRopGEF2 has the first and the third CREs, and ATRopGEF12 has the third CRE, the other pollen-expressed AtRopGEF genes do not contain, indicating that the different regulating system between rice and Arabidopsis might exist. However, OsCrRLK1L13 (LOC_Os06g03610), also known as RUPO (Rupture Pollen tube), possesses these three CREs in promoter regions. RUPO is highly specifically expressed in pollen and is known to regulate pollen-tube growth and integrity in rice [36]. Our results indicate that these three CREs could be candidate for rice pollen-specific gene regulation, and further examination such as deletion assay remains for the clarification of our estimation. Next, using TOMTOM, we revealed that these elements were present in transcription factor genomic sequences such as MYB and bHLH, indicating that the four OsRopGEFs could be regulated in pollen by transcription factors such as MYB, and bHLH, for example.
Plant phenotype of RopGEF mutant
To determine the function of RopGEF in rice pollen tube growth, we identified a T-DNA insertional lines at T2 generation: 2B-00114 and 3A-12915, having T-DNA within RopGEF6, 3A-00157 and 2A-10752, having T-DNA in RopGEF8 and RopGEF2, respectively. Flaking region of about 10 kb T-DNA insertion was identified into exon region near to 5’ region of genes, indicating their functional gene loss. Therefore, we considered these T-DNA lines as knock-out lines and confirmed the seed segregation ratio. All the single T-DNA insertional lines exhibit normal growth and produce homozygous seed production (Fig. 7a), indicating the male gametic transmission is not disturbed by single gene knock-out, probably due to functional redundancy as shown in Arabidopsis [37]. To confirm it, we used CRISPR/Cas9 system to generate homozygous multiple knock-out mutants at T0 generation [38]. In the knock-out homozygous mutants of RopGEF2, the fertility ratio was slightly reduced but comparable to wild-type. Pollen germination and tube growth did not show difference compared with wild-type plants. In addition, single knock-out homozygous mutants of RopGEF3 did not show any significant changes for growth and seed production, indicating that the male-gametic transmission is normal (Additional file 6).
In contrast, RopGEF2 and RopGEF8 double homozygous bi-allelic plants (ropgef2/8) (Fig. 7a), reduced seed fertility up to 20-30% compared with wild-type plants (Fig. 7b). There was no significant difference in vegetative and reproductive development between wild type and mutants (Fig. 7c-e). To investigate the possibility of any defect in pollen grain formation, mature pollen grains were examined. The starch formation of mature pollen grains assayed by iodine staining in double knock-out mutant did not differ from that of the wild type (Fig. 7f). To check whether pollen grain formed intact pollen walls, we also stained mature pollen grains with auramine O and calcofluor white and confirmed no differences between mutant and wild type (Fig. 7g-h). Next, we examined pollen germination behavior of double mutant pollens. Under in vitro pollen germination condition, about 77% wild-type pollen germinated among 83% hydrated pollen grains. In contrast, only about 20% of ropgef2/8 double mutant pollens germinated, even though 75% of the mutant pollens were hydrated (Fig. 7i-k). These data indicate that OsRopGEF genes are functionally redundant for pollen germination in rice. Further different combination of multiple gene mutants could explain functional relationship between four pollen-specific RopGEF genes.