To perform a phylogenetic analysis of the OsRac family, protein sequences were collected from the rice genome annotation project¹ (Moon et al., 2020). Multiple amino acid sequences were aligned using ClustalW, and phylogenetic analysis was performed using MEGA-X under neighbor-joining methods. Using a publicly available rice Affymetrix microarray data in the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO datasets), we normalized the signal intensity with the R language and then transformed them into log2 values. The normalized data, with averaged Affymetrix anatomical meta-expression data, were used for heatmap construction. The TMHMM² and Pfam³ databases were used for domain analysis.

To grow rice (O. sativa japonica cv. Dongjin), the seeds were sterilized with 50% of sodium hypochlorite for 30 min, washed with distilled water for three times, and then germinated on Murashige and Skoog (MS) media under controlled conditions in 7 days (28/25°C day/night, 8-h photoperiod, and 78% relative humidity). The seedlings were grown in the growth chamber or greenhouse for 1 month and then transferred to a paddy field of Kyung Hee University. The tobacco plants (Nicotiana benthamiana) were grown in chambers at controlled conditions (25°C day/night, 16-h photoperiod, and 50% relative humidity) in 3–4 weeks.

Various rice tissues were immediately frozen in liquid nitrogen and ground with Tissue-Lyser II (Qiagen, Hilden, Germany) or mortar and pestle (CoorsTek 60310). The leaf tissues were sampled from 1-month-old plants for mutant analysis, and six other tissues were sampled for real-time quantitative PCR (RT-qPCR) (shoots and roots from 1-week-old plants, leaves and young panicles from 1-month-old plants, developing seeds 5–10 days after pollination, and root hairs from seedling roots 3 days after germination). DNA was extracted using the cetyltrimethylammonium bromide (CTAB)-chloroform method, and the sequence genotype was analyzed in Macrogen Corp.⁴ using BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems). Total RNA was extracted with a TRIzol buffer and purified with an RNeasy plant mini kit (Qiagen); complementary DNA (cDNA) was synthesized using SuPrimeScript RT premix (GeNet Bio). To identify the tissue specific expression by RT-qPCR, we used the Roter-Gene Q instrument system (Qiagen, Hilden, Germany) and the internal control for rice ubiquitin 5 (OsUbi5, LOC_Os01g22490), as previously reported (Kim et al., 2019). RT-qPCR was performed with three independent biological replicates. Relative transcript levels and fold change were calculated by previously reported methods (Schmittgen and Livak, 2008). All RT-qPCR primers used in our experiments are listed in Supplementary Table 1.

To design the guide RNA for CRISPR-Cas9 vector cloning, we selected two target regions using the CRISPRdirect software (Naito et al., 2015). For CRISPR-Cas9 vector cloning, we synthesized oligo dimers with annealed primers and ligated the dimers with the pRGEB32 binary vector (Xie et al., 2015). To clone the OsRopGEF3 promoter, the pOsRopGEF3:OsRopGEF3-GFP plasmid was generated; the 2000 base-pair upstream from the start codon and the open reading frame (ORF) of OsRopGEF3 genes were amplified by PCR and cloned into the binary vector pGA3427, which includes the enhanced green fluorescent protein (eGFP) sequence. To clone the beta-glucuronidase (GUS) fusion vector (pOsRopGEF3: GUS), the 2000 upstream sequence from the stop codon was amplified and ligated into a BamH1-treated pGA3519 vector. Ligated vectors were transformed into Escherichia coli, TOP10. The confirmed plasmid was then transformed into Agrobacterium tumefaciens, LBA4404. The transgenic rice plants were generated via Agrobacterium-mediated cocultivation of the rice callus (Lee et al., 1999). All of the cloning primers we have used in our experiments are listed in Supplementary Table 1. For the protoplast experiments, the ORFs of OsRopGEF3, OsRac3, and OsRBOH5 were each ligated into the Sma1-treated pGA3574 vector, which was tagged with eGFP, mCherry, and N- and C-terminal fragments of Venus.

To minimize the transformation side effects, at least three generations after the tissue culture of plants were used for all root hair measurements. All root hairs were measured from the primary roots of rice 3 days after germination (DAG). The root hair lengths and widths were quantified 2–8 mm from the root apex. The length was measured only when the starting point of the root hairs was visible from the root, and the width was measured at the center of the root hair. BX61 (Olympus, Tokyo, Japan) and SZX61 (Olympus, Tokyo, Japan) microscopes were used for plant photography, and the root hair lengths and widths were measured using Image J software (Schneider et al., 2012). Data were observed in 10 seedlings each for the control and mutant plants and presented as means and standard deviation.

For GUS staining, transgenic plants were mixed with GUS staining solution and vacuumed for 30 min. After that, samples were incubated for 24 h at 37°C in dark conditions (Moon et al., 2019a). Chlorophyll was gradually removed by 70% and absolute ethanol. For ROS detection, the primary roots from three DAG plants were collected and mixed with 10 μM 2′, 7′-dichlorodihydrofluorescein diacetate (CM-H2DCFDA) (Invitrogen) or 1 mg/ml 3,3′-diaminobenzidine (DAB) solution. Next, the samples were vacuumed for 15 min in dark conditions, followed by incubation for 30 min for CM-H2DCFDA staining and 3 h for DAB staining at room temperature in dark conditions. After staining, the primary roots were washed three times with 1% phosphate-buffered saline (PBS). Root hairs were observed under a BX61 microscope (Olympus) and a laser scanning confocal microscope LSM 510 (Carl Zeiss, Jena, Germany). GFP intensity was calculated using ZEN Blue software.

The full-length coding sequence (CDS) of OsRopGEF3 and the N-terminus region of OsRBOH5 were fused into a pGADT7 vector, and the full CDS of OsRac3 was fused with the pGBKT7 vector. Each construct was transformed into the AH109 strain and plated on SM media lacking leucine, tryptophan, histidine, and alanine and incubated for 3 days at 30°C as previously reported (Fields and Song, 1989). The primer sequences used in Y2H analysis are listed in Supplementary Table 1.

The ORFs of OsRopGEF3 and OsRac3 were amplified from the root hair cDNA and cloned into a pGREEN vector fused with GFP. For BiFC, Venus fluorescence protein, which emits a bright yellow signal, was used for complementation. OsRopGEF3 was fused with the N-terminus of Venus, and OsRac3 was fused with the C-terminus of Venus.

The constructs were transfected into the A. tumefaciens strain GV3101 and infiltrated in tobacco leaf as described previously (Sparkes et al., 2006). All constructs used for tobacco infiltration were expressed under the control of the Cauliflower mosaic virus (CaMV) 35S promoter. After 72 h of infiltration, GFP fluorescence was observed by a confocal laser scanning microscope (Zeiss LSM 510, Jena, Germany) with 500–530 nm for emission and 488 nm for excitation. YFP fluorescence was detected at 530–560 nm for emission and 500–530 nm for excitation. FM4-64 (Thermo Fisher Scientific) was used as a PM marker and observed with red fluorescence protein (RFP) channel in a 558-nm microscope. The empty GFP protein was used as a control (Supplementary Figure 2).

For protoplast transformation, OC cells isolated from rice root cells were cultured in R2S media for at least 1 month. The protoplasts were then extracted using cellulase and macerozyme r-10 (Yakult) as previously reported (Cho et al., 2016; Wong et al., 2018). The construct was transformed using PEG-mediated protoplast transfection. After transformation, the protoplasts were incubated for >16 h in dark conditions. All the primers used in our experiments are listed in Supplementary Table 1.

The meta-analysis of public microarray data based on the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO)⁵ datasets indicated that OsRopGEF1, OsRopGEF3, and OsRopGEF11 of the OsRopGEF family genes are highly expressed in the root hair tissue (Kim et al., 2020). Among them, OsRopGEF3 is also strongly expressed in reproductive tissues, such as anthers and pollen, and has high sequence similarity to pollen specific OsRopGEF genes. Based on the expression pattern and sequence similarity analyses, we assumed that OsRopGEF3 might have a more important function for root hair elongation than the others because pollen tubes and root hairs share tip-focused growth. RT-qPCR and histochemical assay using GUS proteins have verified the expression patterns of OsRopGEF3 (Figure 1). OsRopGEF3 was expressed more in root hair tissues than in other tissues (Figure 1B). In addition, RHE lies upstream of OsROPGEF3 (Figure 1A), indicating possible gene expression regulated by RSL in a root hair-specific manner. In rice seedlings at three DAG, the GUS signal of OsRopGEF3 was strongly observed in the root hair tissues and elongation zone of the root, where the root hairs usually start to grow (Figures 1C,D). GUS signals are mainly observed in the trichoblast cells of the meristematic zone and the root hairs in the elongation zone (Figures 1E,F).

To identify the subcellular localization of the OsRopGEF3 protein, we generate gain-of-function transgenic lines that express OsRopGEF3 protein fused with GFP under the native promoter (pOsRopGEF3:OsRopGEF3-GFP) in a wild-type (O. sativa japonica cv. Dongjin) background (Figure 2A). The OsRopGEF3 tagged with GFP protein localizes strongly at root hairs. When the development of root hair cells initiates, the GFP signal is mainly clustered and located in the cytoplasm of the tip region (Figures 2B,C). In the growing root hairs, the GFP signal was spread near the PM but was mainly located at the apical tip region (Figures 2D,E). The signal is getting stronger as the root hairs keep growing (Figures 2F,G). When the root hairs were fully grown, the fluorescent signals were distributed throughout the PM (Figures 2H,I). To verify this localization pattern, we stained the root hairs with FM4-64 dye, which has been used to track the PM. As a result, it was confirmed that the GFP signal appears PM-related in the root hairs through colocalization with FM4-64 signals (Figures 2J–M). It was colocalized in the PM, strongly observed in the root hair tip, and identified in the root hair body. However, when predicting protein domains through transmembrane helices hidden Markov models (TMHMM) and Pfam databases (hidden Markov models), no transmembrane domain was observed in OsRopGEF3 (Supplementary Figure 1). Of the 514 total amino acids in OsRopGEF3, S71 to K416 were aligned to the plant-specific ROP nucleotide exchanger (PRONE) domain.

To know the functional roles of the gene, we generated the loss-of-function mutants of OsRopGEF3 via targeted mutagenesis using the CRISPR-Cas9 system. We then identified two homozygous lines carrying mutations in different exon regions. The second and third exon of the OsRopGEF3 were targeted for gene editing, and corresponding mutants are named, osropgef3-1 and osropgef3-2 (Figure 3A). The osropgef3-1 line has one base (adenine) insertion in the second exon, and the osropgef3-2 line has one base (adenine) insertion in the third exon. Both lines have abnormal protein sequences, which change the ORFs but do not create a stop codon. All loss-of-function mutant lines produce defective root hair development phenotypes. Differences in root hair development between mutant and wild types can be clearly distinguished from three DAG (Figures 3B–D). Phenotypic differences were the greatest in plants on 3–5 DAG. To compare the phenotypes in detail, we measured the length and width of root hairs between the wild-type and osropgef3 mutants at three DAG. On average, wild-type root hairs were 290 μm long and 7 μm thick (Figures 3H,I). In contrast to wild-type root hairs, the length of the root hairs of osropgef3-1 and osropgef3-2 decreased by 180 μm, while the width increased up to 16–18 μm and showed a wavy phenotype (Figures 3E–I), clearly presenting defects in root hair elongation.

To check the intracellular levels of ROS generation in root hair cells, primary roots were stained with CM-H2DCFDA and DAB at three DAG (Figure 4). After CM-H2DCFDA staining, the wild-type plants were strongly stained throughout the roots (Figures 4A,B). However, the roots of osropgef3-1 and osropgef3-2 mutants were less stained, with negligible staining in the root hairs (Figures 4C–F). We then observed the root hairs using a confocal microscope to detect differences in ROS levels. Using the same fluorescence intensity, ROS was detected throughout the entire root hairs of the wild-type plants, but osropgef3 mutants had weak signals in the apex or were not stained at all (Figures 4G–O). The GFP intensity of the root hair apex was approximately 3,500 for the wild-type plants and 0–500 for the mutants (Supplementary Figure 2). Next, to confirm the variation in ROS intensity in the OsRopGEF3 gain-of-function transgenic line (pOsRopGEF3:OsRopGEF3-GFP), we used DAB dye for staining. CM-H2DCFDA dye could not be used because this transgenic line was tagged with GFP. The wild-type root hairs were strongly stained at the apex region (Figure 4P), whereas osropgef3 root hairs were weakly stained (Figures 4Q,R). The gain-of-function line showed staining similar to the wild-type plants (Figure 4S). These results demonstrate that OsRopGEF3 plays a role in ROS production in root hairs.

To find the functional interaction partner for OsRopGEF3 among the OsRac family members in the rice genome, we first analyzed the expression patterns of the entire OsRac genes. Heatmap expression data based on the public Affymetrix microarray data consisting of various rice tissues and organs were generated and integrated into a phylogenetic tree (Figure 5A). The meta-expression analysis suggests some candidate OsRac that have the potential to interact with OsRopGEF3 in root hairs. Of the seven OsRac genes, OsRac6 has the highest expression in the root hair tissue, followed by OsRac3. OsRac4 and OsRac7 have weak expression in root hair. However, none of the OsRac genes showed strong expression only in the root hairs. Besides root hairs, OsRac6 was also highly expressed in the stem, root, palea, lemma, embryo, endosperm, and all reproductive tissues. OsRac3 was evenly expressed in all analyzed tissues and organs, and OsRac4 was expressed in the stem, palea, and lemma. RT-qPCR results for six tissues, i.e., shoot, root, leaf, panicle, seed, and root hair, tended to be similar to meta-expression data patterns (Figures 5B–D). As expected, three genes, OsRac3, OsRac6, and OsRac7, showed a significant level of expression in root hairs. In particular, OsRac6 showed the highest expression in root hair tissues. The expression of the remaining genes, i.e., OsRac1, OsRac2, OsRac4, and OsRac5, was negligible in the root hairs (Supplementary Figure 3).

Using Y2H, we confirmed the interaction between OsRopGEF3 and these three OsRac proteins (Figure 5E). The full CDS of each protein was fused with the prey and bait domain. As a result, OsRac3 interacted with OsRopGEF3. However, although OsRac6 was most strongly expressed in root hairs, it did not show any interaction with OsRopGEF3. The interaction between OsRopGEF3 and OsRac3 has also been verified in tobacco leaf cells. The empty GFP vector was used as a positive control (Supplementary Figure 4). As shown in Figure 6, when expressed individually, OsRopGEF3 was located near the PM (Figures 6A–D), and OsRac3 was located in the PM and the cytoplasm (Figures 6E–H). FM4-64 dye was used as a PM marker. When 1 M NaCl treatment was used to cause plasmolysis, the GFP signals were merged well with FM4-64 (Supplementary Figure 5). Next, the interaction of these two proteins was verified within tobacco cells using BiFC. Similar to the Y2H results, we confirmed that OsRopGEF3 and OsRac3 bind near the PM (Figures 6I–L). Interestingly, OsRac3 was spread out in the cytoplasm when expressed alone, but with its binding partner OsRopGEF3, it was also localized near the PM. Intracellular colocalization of the proteins was confirmed through the rice root protoplast system. OsRopGEF3 was mainly located in the PM, whereas OsRac3 was located in the PM and the cytoplasm (Figures 7A–C). Next, the interaction of the two proteins was verified through BiFC in rice root cells. The Venus signal was observed in the PM, as shown by the BiFC results in tobacco (Figures 7D–F). Although we did not confirm the exact position of the two proteins in the rice root hair cells, our data suggest that the two proteins were colocalized in the root cells and interacted in the PM.

Significant ROS decrease in the osropgef3 root hairs may be due to the functional failure of OsRac3. ROP/Rac binds to multiple proteins and is involved in various mechanisms. Among these roles, ROP/Rac has been reported to bind to RBOH, which produces ROS inside plant cells. Three RBOH genes are highly expressed in the root hairs of rice: OsRBOH2, OsRBOH3, and OsRBOH5 (Kim et al., 2019). Accordingly, we aimed to detect OsRBOH, an interaction partner of OsRac3, using Y2H. The full CDS of OsRac3 and N-terminal region of OsRBOH were combined with the prey and bait domain. As a result, OsRac3 weakly interacted with the N-terminus of OsRBOH5 in the yeast system (Figure 5F). Furthermore, to confirm their interaction inside the rice cells, BiFC was performed in root protoplast cells. When both constructs were coexpressed within the protoplast, a strong clear signal was observed in the PM (Figures 7G–I). These results indicate that OsRac3 interacts with the N-terminus of OsRBOH5 in the PM of rice root cells.