Cell wall-associated ROOT HAIR SPECIFIC 10, a proline-rich receptor-like kinase, is a negative modulator of Arabidopsis root hair growth

Highlight RHS10 mediates cell wall-associated signals to maintain proper root hair length, potentially by regulating RNA catabolism and ROS accumulation.


Introduction
A root hair develops in a stepwise manner through cell fate determination, initiation, bulge formation, and tip growth Schiefelbein, 2002, 2008). Three different distribution patterns of hair/non-hair cells (i.e. random, longitudinal, and radial patterning), which derive from different fate-determining mechanisms, have been observed in vascular plants (Dolan, 1996;Clowes, 2000;Schiefelbein, 2000). Although this upstream fate determination step has evolutionarily diverged, the downstream hair-forming morphogenetic steps are likely to be conserved throughout the vascular plant lineage (Kim et al., 2006). ROOT HAIR DEFECTIVE 6 [RHD6; a basic helix-loop-helix (bHLH) transcription factor] has been recognized as a prime positive modulator of root hair initiation in root hair cells of Arabidopsis (Masucci and Schiefelbein, 1996;Menand et al., 2007), and its molecular function is conserved even between moss and Arabidopsis (Menand et al., 2007). RHD6 directly regulates two downstream bHLH transcription factors, RHD6-LIKE2 (RSL2) by Humphrey et al. (2015) demonstrated that the kinase domain of PERKs interacts with AGC family protein kinases. Our previous study demonstrated that loss of function of RHS10/PERK13 enhanced and its overexpression inhibited root hair growth (Won et al., 2009). Arabidopsis IGI1/PERK12 has been implicated in branching and growth of the shoot (Hwang et al., 2010).
Although a few of the aforementioned studies have demonstrated the basic aspects of PERK function, further questions remain to be answered; for example, how the ECD works, what the PERK downstream process is, how PERKs interact with other signaling pathways, and whether cell wallassociated PERK function is evolutionarily conserved. In this study, we characterized these molecular and biological functions of RHS10/PERK13 during root hair growth. Our data demonstrate that minimal proline residues in the ECD are sufficient for RHS10 function, RHS10 affects root hair growth by regulating reactive oxygen species (ROS) levels and RNA metabolism, and RHS10 function during root hair growth has been conserved in angiosperms.

Plant material and growth conditions
Arabidopsis thaliana wild type (WT, Columbia) was used for transformation of transgene constructs unless stated otherwise. Arabidopsis plants were transformed using Agrobacterium tumefaciens strain C58C1 (pMP90). Transformed plants were selected on hygromycin-containing plates (50 µg ml -1 ). All seeds were grown on agarose plates containing 4.3 g ml -1 Murashige and Skoog (MS) nutrient mix (Duchefa), 1% sucrose, 0.5 g ml -1 MES at pH 5.7 with KOH, and 0.8% agarose. Seeds were cold treated before germination at 23 °C under a 16 h/8 h light/dark photoperiod. For observation of root hairs, homozygous transformants were planted on antibiotic-free media, and T 1 and T 2 lines were planted on hygromycincontaining media. Hygromycin did not significantly interfere with root hair development, as shown with the control ProE7:YFP transformants in each experiment. Two control lines were adopted: WT for mutant analysis and ProE7:YFP for transgenic analyses on the medium including hygromycin. Unless specifically mentioned, T 2 or homozygous transformants were used.

Observation of root hair phenotypes and measurement of root hair length
Root hair phenotypes were observed under a stereomicroscope (MZ FLIII, Leica, Heerbrugg, Switzerland). Root hair length was measured as described in Lee and Cho (2006) with modifications. The root was digitally photographed using a stereomicroscope at ×40-50 magnification. The hair length of 8-10 consecutive hairs protruding perpendicularly from each side of the root, for a total of 16-20 hairs from both sides of the root, was calculated using LAS software V2.8.1 (Leica). Root hairs were observed with 3-day-old seedlings.
For the ProE7:RHS10:GFP construct, a genomic fragment of RHS10 was obtained by PCR using the primer sets listed in Supplementary Table S1 at JXB online and Arabidopsis genomic DNA as a template. The PCR product was cloned into SalI/NcoI sites upstream of the GFP (green fluorescent protein) gene, producing a RHS10:GFP fusion. For ProE7:PERK8, ProE7:RNS2, ProE7:OsRHS10 (Os03g37120 and Os06g29080), and ProE7:PtRHS10 constructs, the genomic fragments were obtained by PCR using the primer sets listed in Supplementary Table S1 and A. thaliana, Oryza sativa, and Populus trichocarpa genomic DNA as templates. For ProE7:PERK8 and ProE7:RNS2, the PCR products were cloned into SalI/BamHI sites, and OsRHS10 (Os03g37120 and Os06g29080) and PtRHS10 PCR products were inserted into BglII/NcoI and SalI/KpnI sites, respectively, downstream of ProE7.
For the deletion analysis of the RHS10 ECD, the serial deletion fragments (D1-D5) were generated by PCR using the primers listed in Supplementary Table S1 and inserted into BglII/NcoI sites of the pCAMBIA1300 vector containing ProE7. The nucleotides (RHS10 Sl-MF/RHS10 Bg2-MR in Supplementary Table S1) corresponding to the first four amino acid residues, including the start codon, were inserted into the SalI/BglII sites before the deletion inserts.
For LRX2-Ri and ROL1-Ri-1/2 construction, the RNAi target regions of LRX2 and ROL1 cDNA were amplified using PCR with the primer sets listed in Supplementary Table S1. Each RNAi target region was inserted into XhoI/EcoRI and HindIII/XbaI sites of the pHannibal vector to generate a sense and antisense construct, respectively. For the final RNAi construction in a binary vector, the Cauliflower mosaic virus 35S promoter (Pro35S) was inserted into the HindIII/SalI site of the pCAMBIA1300 vector, and the XhoI/XbaI fragments of the RNAi inserts from the pHannibal vector were transferred into the SalI/XbaI sites downstream of Pro35S.
In order to express RNS2 and RHS10 proteins in Escherichia coli for protein blot analysis and in vitro kinase assay, the cDNA sequences of RNS2 and RHS10 kinase domain were amplified by PCR from the Arabidopsis seedling cDNA library using the primer sets listed in Supplementary Table S1. The PCR products were cloned into EcoRI/SalI restriction sites of the pGEX-4T-1 vector (GE Healthcare, Inchon, Korea), which generated fusion proteins with glutathione S-transferase (GST) at the N-termini of the RNS2 and RHS10 kinase domain.
All constructs were confirmed by nucleotide sequencing and introduced into Arabidopsis plants by the Agrobacterium-mediated floral dipping method. Transgene insertion in the Arabidopsis transformants was confirmed by PCR analysis using transgene-specific primers.

Microscopic observation of fluorescent proteins
Fluorescence from GFP (green) and FM4-64 (red) was observed using an LSM 510 confocal laser scanning microscope (Carl Zeiss). Localization of the RHS10:GFP and PIN3:GFP fusion proteins was observed in 4-day-old seedlings. To observe brefeldin A (BFA) compartment formation, the transgenic seedlings were co-treated with cycloheximide (50 μM) and BFA (25 μM) for 1-2 h before observation. The PM and endocytosis tracer FM4-64 (2 μM) was applied for <3 min for PM marking and ~15 min for BFA body marking. The control liquid medium included the same amount of DMSO, which was used to dissolve BFA. To estimate the cell wall/ PM ratio of RHS10:GFP and PIN3:GFP, seedling roots were plasmolyzed with 0.45 M mannitol for 5 min before observation. The fluorescence signal intensities from the cell wall and PM were estimated from the same area of a ROI (region of interest) of the cell wall or the PM region after taking confocal micropictographs. The calculation of fluorescence signal intensities was performed using the Adobe Photoshop histogram menu as described previously (Kim et al., 2006;Won et al., 2010).

Preparation of membrane proteins and protein blot analysis
Total cytosolic and membrane proteins were isolated from transgenic Arabidopsis seedlings (ProE7:GFP and ProE7:RHS10:GFP). Fourday-old seedlings were ground in ice-cold homogenization/extraction buffer [250 mM HEPES-KOH (pH7.0), 0.5 M NaCl, 0.1 M EDTA, 50 mM potassium acetate, phenylmethylsulfonyl fluoride, and protease inhibitors]. The resultant total cellular proteins (the supernatant of low speed centrifugation) were subjected to centrifugation at 10 000 g for 1 h to obtain microsomal pellets and 'cytosolic proteins' in the supernatant. Pellets were re-suspended in microsome buffer with 1% Triton X-100 and separated into supernatant and pellet fractions at 10 000 g. This supernatant fraction is denoted as 'microsome proteins'. GFP (from ProE7:GFP) and RHS10:GFP (from ProE7:RHS10:GFP) from cytosolic (C) and microsomal (M) parts were analyzed by western blot analysis using anti-GFP antibody. An equal amount of each protein sample was separated by SDS-PAGE, transferred into a nitrocellulose membrane (Amersham Biosciences, Corston, Bath, UK), and probed with 1/500-diluted anti-GFP rabbit IgG horseradish peroxidase (HRP)-conjugated antibody (Invitrogen, Seoul, Korea). Chemiluminescence detection was performed with Pierce ECL western blotting substrate (Thermo Scientific Inc., Waltham, MA, USA) on a chemiluminescence imaging system (Davinch-Chem, Corebio, Seoul, Korea).

Generation of multiple mutants
Multiple mutants and transformants used to show the genetic relationship with RHS10 were generated by crossing individual lines, and their homozygosity was verified by molecular and physiological genotyping.

Measurement of ROS
A ROS-sensitive fluorescent dye [2′,7′-dichlorodihydrofluorescein diacetate (H 2 DCFDA), Invitrogen] was used to estimate ROS levels in the root. Three-day-old seedlings were incubated with 3 μM H 2 DCF-DA (dissolved in DMSO, 0.0045% final concentration) for 60 min at 4 °C, washed with 0.1 mM KCl and 0.1 mM CaCl 2 (pH 6.0), and incubated for 60 min at 22 °C before observation. The green fluorescence images were taken by a fluorescence stereomicroscope (MZ FLIII, Leica), and the ROS levels were estimated by quantifying green fluorescence in the root using the histogram function of Adobe Photoshop CS6 (Adobe Systems) as described previously (Cho and Cosgrove, 2002;Kim et al., 2006). The same area of the ROI was used for different root samples. The experiment was repeated twice, and 11-23 seedlings were observed for each genotype.
Yeast two-hybrid screening and assay A yeast two-hybrid (Y2H) screening and assay was performed using the Matchmaker Yeast Two-Hybrid System (Clontech, USA). The RHS10 kinase domain (amino acids 256-710) was PCR-amplified using the primer sets in Supplementary Table S1 and cloned into the NcoI/PstI sites of the bait pGBKT7 vector, and the Arabidopsis seedling cDNA library was cloned into the pGADT7 library vector. Y2H screening processes were conducted according to the manufacturer's manual. About 79 putative targets of RHS10 kinase were screened from Quadruple Dropout (QDO)/X-α-gal selection medium. Isolated positive plasmids were subjected to nucleotide sequence analyses to confirm gene identities. Root hair-related genes were further screened using cell type-specific expression databases such as Arabidopsis eFP Browser (BAR, http://bar.utoronto.ca/welcome.htm) and Genevestigator (https://www.genevestigator.com). The interaction between the RHS10 kinase domain and several final interactor candidates was re-confirmed with a targeted Y2H assay. The RHS10 N-terminal bait (amino acids 1-236.) was PCR amplified using the primer sets in Supplementary Table S1 and cloned into the NcoI/PstI sites of the bait pGBKT7 vector. In order to suppress background HIS3 activities, 0-70 mM 3-amino-1,2,4-triazole (3-AT) was used.

Quantification of cellular RNA contents
Whole seedlings were stained for 3 h with 5 μM SYTO RNASelect green (Invitrogen) in liquid MS medium at room temperature (Hillwig et al., 2011), washed three times with fresh MS medium, and observed under a stereo-or confocal microscope used to produce green fluorescent root images as described above. Quantification of root RNA content, which was represented by green fluorescence, was performed using the Adobe Photoshop histogram menu with the root fluorescence images as described previously (Kim et al., 2006;Won et al., 2010).

Supplementary methods
The methods for supplementary data are described in the 'Supplementary methods'.

Loss of RHS10 extends the tip-growing period of root hairs
The loss-of-function rhs10 mutant (SALK_075892) seedling grew considerably longer root hairs (50-60%) than the control seedling, and root hair-specific overexpression of RHS10 under the EXPANSIN A7 promoter (ProE7; Cho and Cosgrove, 2002) greatly inhibited root hair growth (Fig. 1A, B; Won et al., 2009). Three other independent insertion mutant lines of RHS10 also showed consistently longer root hair phenotypes ( Supplementary Fig. S1). A complementation of RHS10 under its own promoter in the mutant almost restored WT-level root hair growth (Fig. 1B). Whereas many kinase-related RHS genes are implicated in root hair morphogenetic processes such as hair branching and waving, RHS10 influences only hair tip growth. The longer root hair phenotype of rhs10 seems to be due to prolonged tip growth. Up to 4 h after the initiation of tip growth from a minute bulge, both the WT and the rhs10 mutant grew the hair at a similar speed (Fig. 1C); however, after this occurred and while the WT hair almost stopped growing, the rhs10 hair still maintained a specific growth rate (Fig. 1C), suggesting that the rhs10 mutant grows longer hairs by lengthening the duration of tip growth but not the growth rate.

RHS10 localizes to the PM and exhibits association with the cell wall in root hair cells
To determine the subcellular localization of RHS10, the RHS10:GFP fusion protein was expressed under ProE7 and analyzed in the root hair cell. Root hairs specifically expressed the RHS10:GFP fusion protein, which could inhibit root hair growth, but with a somewhat lower efficiency than that of native RHS10, which could be due to partial functional hindrance of the kinase domain where GFP fused ( Supplementary Fig. S2). By using confocal microscopy, the RHS10:GFP signal was observed along the root hair cell boundary, and it overlapped with the FM4-64 dye signal ( Fig. 2A). The endocytotic tracer FM4-64 stains the PM when applied for a short duration (i.e. <3 min), indicating that RHS10 localizes to the PM. Internal accumulation of a protein in the BFA-induced compartment is cytological evidence supporting PM localization of the protein (Ganguly et al., 2012). Similarly, to what has been shown for PIN-FORMED proteins (Geldner et al., 2001), RHS10 obviously accumulated into the FM4-64-overlapping BFA compartments, and the RHS10 signal in the PM greatly decreased following prolonged BFA treatment ( Fig. 2B), indicating that RHS10 is localized in the PM, and recycles between the PM and endosomes.
The protein blot analysis with proteins from the membrane and cytosolic fractions further supports PM localization of RHS10. Whereas GFP was detected mostly in the cytosolic fraction, RHS10:GFP was detected predominantly in the membrane fraction (Fig. 2C).
Because RHS10 includes extensin-like motifs in its putative ECD, we examined whether RHS10 shows a cell wall association, using the plasmolysis method. When the root hair cell of the RHS10:GFP-expressing transformant seedling was plasmolyzed by 0.45 M mannitol, RHS10:GFP signals were observed not only in the PM, but also in the cell wall (Fig. 2D). Furthermore, even after plasmolysis, the RHS10:GFP signal exhibited connections (i.e. Hechtian strands) between the PM and cell wall (Fig. 2D). Many speckle-like RHS10:GFP signals were also observed in the cell wall, which potentially represent RHS10:GFP proteins left behind in the cell wall after plasmolysis and/or the strong RHS10-cell wall adhesion spots. PIN3 showed no detectable polarity in the root hair cell, and PIN3 consistently exhibited a very weak cell wall association ( Supplementary  Fig. S3). When quantified, the degree of cell wall association (i.e. the relative signal ratio of the protein in the cell wall over the PM protein) was considerably higher for RHS10 than for PIN3 in the root hair cell (Fig. 2E). This observation provides further evidence for the strong association of RHS10 with the cell wall, most probably through its proline-rich ECD. The protein blot analysis with the anti-GFP antibody shows that the RHS10:GFP fusion protein is mostly found in the total membrane fraction (M), whereas soluble GFP was found in the cytosolic fraction (C). (D) Three independent trichoblast cells showing RHS10:GFP signals, which are associated with the cell wall during plasmolysis. RHS10 remains in (bracket and asterisk), or forms Hechtian strands (arrow heads) with, the cell wall after plasmolysis using 0.45 M mannitol. The scale bar is 10 μm in all images. (E) RHS10 tends to localize more frequently in the cell wall than PIN3. Relative signal intensities of RHS10:GFP (ProE7:RHS10:GFP) and PIN3:GFP (ProE7:PIN3:GFP) were observed in the cell wall and PM after plasmolysis, and the intensity ratio between the cell wall and PM (CW/PM) was calculated. Data represent the mean ±SE from 98 ROIs (from 50 cells of 10 seedlings) for each.
Minimal proline residues of the ECD are sufficient for RHS10 function in root hair inhibition Due to the proline-rich extensin-like domain, PERKs have been suggested to be associated with the cell wall. The extensin-like putative ECD (composed of 236 residues before the TM domain) of RHS10 includes 88 proline residues (37.3% of the 236 residues) and seven extensin motifs (SP 3-5 ), where the proline-rich region stretches up to the 224th residue from the N-terminus (Fig. 3A). In order to determine which region of the ECD is necessary for RHS10 function in terms of root hair inhibition, we generated serial N-terminal deletion constructs (D1-D5) of RHS10 (Fig. 3A, B), expressed these truncated forms in the root hair cell using ProE7, and estimated root hair length of independent transformant lines to reflect RHS10 activity.
In the first round deletion, considerable RHS10-mediated root hair inhibition [~14% of the control (ProE7:YFP) level] was observed up to the D3 (amino acids 173) deletion, which excluded all seven extensin-like motifs and 88% of proline residues (Fig. 3C, D). On the other hand, the D4 (amino acids 226) deletion carrying no proline residue greatly restored root hair growth (68% that of the control). The transformant only expressing the soluble kinase domain (deletion D5) grew root hairs the same size as the control. Because the region between D3 and D4 turned out to be critical for root hair inhibition activity by RHS10, we made one more deletion in this region (D3-1) that excludes 16 more residues from D3; however, the D3-1 deletion also showed the same intensity of root hair inhibition as D2 (Fig. 3C, D). D3-1 still includes eight proline residues (Fig. 3A). This deletion analysis of the ECD suggests that the extensin-like (SP 3-5 ) repeats are not essential, but a few proline residues or other motifs probably could be critical for the activity of RHS10. The D4 deletion of RHS10, which includes no proline residue in its ECD, still had ~30% root hair inhibition activity (Fig. 3C). This level of activity might reflect basal RHS10 kinase activity without proline-involved extracellular signaling. Almost no root hair inhibition by the deletion of D5 (only containing the kinase domain) indicates that kinase activity on the downstream effectors requires ECD-mediated signaling from the cell wall.

RHS10 function is likely to be conserved in angiosperms
Although cell wall compositions vary by species (Carpita and Gibeaut, 1993), and thus the cell wall-associated events could be different among species, root hair-specific gene regulation and root hair morphogenetic processes seem to be conserved in the angiosperm lineage (Kim et al., 2006). In this context, we wondered whether there are RHS10-homologous proteins in other angiosperms and if they function like Arabidopsis RHS10 in the root hair cell. Therefore, we screened RHS10homologous protein sequences from a monocotyledonous grass family and other eudicot members, and analyzed the phylogenetic relationship to identify RHS10-orthologous sequences ( Supplementary Fig. S4). In this study, we focused on rice and poplar orthologs. The AtRHS10 clade included two Arabidopsis paralogs, and two rice and two poplar orthologous sequences (Supplementary Fig. S4). Here, we use the terms orthologs and paralogs following the definition of Koonin (2005). We chose two rice and one poplar RHS10orthologous sequences; Os03g37120 and Os06g29080 for rice and PtEEE90055 (PtRHS10 hereafter) for poplar. These RHS10-orthologous sequences possessed all three prolinerich (with multiple extensin motifs), TM, and kinase domains, which are similar to AtRHS10 (Supplementary Figs S5, S6).
The AtRHS10 gene is expressed specifically in root hairs with at least three RHE motifs in the proximal promoter region ( Fig. 4A; Won et al., 2009). First, we examined whether RHS10-orthologous genes also include the RHE in their promoter regions. The promoters from rice and poplar orthologs included multiple RHEs; two RHEs for PtRHS10 and six or four RHEs for rice RHS10 within 1400 bp from the start codon (Fig. 4A). All RHE motifs from RHS10-orthologous sequences possessed functionally essential nucleotides such as 'T' on the left part (LP), 'CACG' on the right part (RP), and other positions with limited flexibility as defined by Kim et al. (2006) (Fig. 4B, C). This suggests that rice and poplar RHS10-orthologous genes could also be root hair specific and be genuine RHS10 orthologs.
To test the orthologous relationship functionally, we introduced these orthologous sequences into the Arabidopsis root hair cell using ProE7 and observed whether they also inhibit root hair growth. Both rice and poplar RHS10 genes considerably inhibited root hair growth in Arabidopsis (53% for PtRHS10 and 52-69% for rice orthologs) (Fig. 4D, E). Although the strength of hair inhibition by these two orthologs was weaker than that by AtRHS10, a result potentially due to different cell wall compositions or kinase activities among species, their root hair-inhibitory activities strongly suggest that RHS10 orthologs play a general role in root hair growth in angiosperms.
In addition to orthologs, we also tested if RHS10 paralogs have a root hair-inhibitory function because they share a similar protein structure. We chose four representative RHS10 paralogs from different Arabidopsis PERK clades ( Supplementary Fig. S4), and analyzed root hair phenotypes of mutants and overexpression transformants. Lossof-function mutants (Supplementary Fig. S7) for PERK1, PERK5, PERK8, and PERK10 all showed significantly longer root hair phenotypes, whereas the phenotypes of perk5, perk8, and perk10 were comparable with that of rhs10 (perk13), and perk1's phenotype was relatively weak (Fig. 5A, B). ProE7driven overexpression of PERK5 and PERK8 greatly inhibited root hair growth (Fig. 5C, D). This result suggests that PERK family members transduce similar extracellular signals and target similar effectors, at least in the root hair cell.

Auxin and ethylene cannot rescue RHS10-inhibited root hair growth
The root hairless phenotype of rhd6 can be rescued by the addition of auxin or ethylene, suggesting that these two hormones act downstream of RHD6 Schiefelbein, 1994, 1996). To determine the regulatory hierarchy between these hormones and RHS10 for root hair growth, we tested whether RHS10-mediated root hair inhibition can be rescued by auxin and ethylene. The WT and rhs10 mutants grew longer root hairs in response to 20 nM indole-3-acetic acid (IAA) and 5 μM 1-aminocyclopropane-1-carboxylic acid (ACC; the ethylene precursor), whereas root hair defects of RHS10ox transformants could not be rescued by these two hormones ( Supplementary Fig. S8A). A similar result was obtained with root hair-specific PERK8-overexpression transformants ( Supplementary Fig. S8B). PERK8 shares a similar structure with RHS10/PERK13 and belongs to a neighboring clade ( Supplementary Figs S4-S6). These results indicate that PERK family members may generally function as repressors of root hair growth downstream of hormonal action.
Root hair-specific expression of axr2-1 (ProE7:axr2-1), the dominant mutant of IAA7, in rhs10 completely inhibited root hair growth (Fig. 6), most probably by inhibiting auxin signaling in the root hair cell. Root hair-specific overexpression of PIN3 (ProE7:PIN3), the auxin efflux carrier, also greatly suppressed root hair growth of rhs10 by depleting auxin in the root hair cell. These data indicate that auxin is necessary for root hair growth. In these transgenic backgrounds, the rhs10 mutant has lost its capacity to grow longer hairs, resulting in the original transgenic phenotypes (suppressed root hair growth). The ethylene signaling-defective ein2 mutant also grew shorter root hairs than the WT, and the rhs10 ein2 double mutant also showed ein2-like short root hair phenotypes. Defects in RHD2 (encoding a NADPH oxidase; Foreman et al., 2003) and COW1 (CAN OF WORMS 1; encoding a phosphatidylinositol phosphate transfer protein; Böhme et al., 2004) also suppressed the enhanced root hair growth phenotype of rhs10. These results collectively suggest that these genes are epistatic to rhs10 during root hair growth.
The cross lines of rhs10 and several cell wall-related mutants showed interesting phenotypes. LRX1 and LRX2 encode leucine-rich repeat extensin proteins whose defects cause aborted, swollen, and/or branched root hair phenotypes (Baumberger et al., 2001(Baumberger et al., , 2003Fig. 6). The single lrx2 mutant did not show any detectable root hair phenotype but synergistically functioned with the lrx1 mutant (Baumberger et al., 2003). In this study, we introduced double loss-of-function mutations in LRX genes and RHS10 to determine their relationship. Because the genetic loci for RHS10 and LRX2 are close to each other on chromosome 1, we generated an RNAi line of LRX2 (Pro35S:LRX2-Ri) to cross with rhs10. The lrx1-rhs10 double mutant expressed more severe lrx1 phenotypes (i.e. more aborted root hairs) than the single lrx1 mutant (Fig. 6). Furthermore, rhs10 in the LRX2-Ri background expressed the aborted root hair phenotype, although the LRX2-Ri transformant itself did not express such a phenotype (Fig. 6). These results indicate that the loss of RHS10 enhances LRX defects in the root hair.
On the other hand, ROL1 [repressor of lrx1, encoding a rhamnose (a pectin unit) biosynthetic enzyme] behaved in such a way that its loss in the lrx1 background restored the normal root hair phenotype (Diet et al., 2006). Because the genetic loci for RHS10 and ROL1 are close to each other on chromosome 1, we generated two independent RNAi lines of ROL1 (Pro35S:ROL1-Ri1 and 2) and crossed them with rhs10. The suppression of ROL1 shortened the root hair, and the rhs10 lrx1 ROL1-Ri triple mutant still harbored aborted root hairs, but showed restoration of root hair growth compared with the lrx1 rhs10 double mutant or ROL1-Ri lines (Fig. 6), suggesting that suppression of ROL1 can rescue the lrx mutants as previously reported (Diet et al., 2006), even under the rhs10 background. It is an intriguing that the rhs10 lrx1 ROL1-Ri triple mutant expresses a phenotype with longer root hairs than ROL1-Ri lines. The enhanced lrx root hair phenotypes caused by the loss of RHS10 might result from overall increases to the hair-expanding capacity in the rhs10 background, which should lead to the weakened root hair cell walls of lrx mutants breaking more readily.

RHS10 suppresses ROS accumulation
ROS production by RHD2 is necessary for normal root hair growth (Foreman et al., 2003). Here, we tested if RHS10 regulates ROS production to affect root hair growth. The ROS level in the root was visualized by H 2 DCFDA, an ROS probe. The long-haired rhs10 mutant showed significantly higher ROS accumulation than the WT, and complementation of rhs10 with RHS10 (by its own promoter) lowered the ROS level almost to that of the WT (Fig. 7). Conversely, overexpression of RHS10 under ProE7 in the WT background (RHS10ox) considerably suppressed ROS accumulation, and hairless rhd6 accumulated much fewer ROS than the WT (Fig. 7). This result demonstrates that RHS10-mediated regulation of root hair growth is correlated with the ROS level.

RHS10 affects RNA contents probably by modulating RNS2
In order to identify the direct molecular targets of RHS10, we performed a Y2H screening. We obtained 79 positive clones from the screening and focused on one gene called RIBONUCLEASE 2 (RNS2, AT2G39780). Four (two identical) out of the 79 positive clones represented RNS2, all of which included most of the RNS2 coding region; 86-963, 85-1024, and 95-1018 bp of the 1074-bp full-length cDNA sequence (start to stop codon). In a repeated Y2H assay, RNS2 interacted with the C-terminal kinase domain but not with the N-terminal ECD of RHS10 (Fig. 8A).
If RHS10 positively regulates RNS2, RHS10 would work positively to degrade RNA. To test this idea, we examined cellular RNA levels along regions of root development, root tips, root hair growing, and root hair maturation. To stain RNA in the root tissue, SYTO RNASelect Green Fluorescent Cell Stain (Hillwig et al., 2011) was used, and fluorescence intensity was quantified to compare RNA levels of different genotypes. In all root regions, RNA levels were considerably increased in the rns2 mutant background by 57, 30, and 27% compared with the control, in tips and growing and maturation regions of root hairs (Fig. 8F). Regarding the rhs10 mutant, although a significant increase (24%) in RNA content was observed in the root tip region, a greater increase (~38%) occurred in growing and maturation regions. Conversely, root hairs specifically overexpressing RHS10 (ProE7:RHS10) and RNS2 (ProE7:RNS2) RNA significantly decreased but only in the hair-growing region. These results imply that RHS10 inhibits root hair growth at least by modulating RNS2-mediated RNA degradation.
Next, we tested whether the RHS10 kinase domain can phosphorylate RNS2. Heterologously expressed GST-fused proteins were used in the kinase assay. The fusion proteins were detected at their expected molecular sizes in the protein blot analysis using anti-GST antibody (Supplementary Fig.  S9A). In the kinase assay including [γ-32 P]ATP as a phosphate source, the RHS10 kinase domain did not appear to phosphorylate RNS2, whereas the kinase was autophosphorylated ( Supplementary Fig. S9B). According to the deletion analysis of the RHS10 ECD, the D5 deletion, which includes only the kinase domain, failed to inhibit root hair growth (Fig. 3C). These results, when analyzed together, suggest that the RHS10 kinase domain is not able to target the downstream effector to inhibit root hair growth when acting alone, although it is capable of autophosphorylation ( Supplementary Fig. S10).

Discussion
The molecular function of PERKs has been poorly understood. In this study, by analyzing the molecular function of RHS10 and its PERK homologs in root hair growth, we infer their putative cell wall association motif, evolutionary conservation, and downstream mechanism.

RHS10-mediated root hair inhibition requires arabinogalactan protein-like motifs in its ECD
The deletion analysis of the ECD of RHS10 revealed that only 47 amino acid residues (of 236 residues for the whole ECD) were sufficient to cause considerable RHS10-mediated root hair inhibition (Fig. 3). This is comparable with the result from serial deletion analysis of LRX1 extensin motifs, where deleting most extensin did not affect LRX function (Ringli, 2010). The short stretch of the RHS10 ECD region includes eight proline residues. Because proline residues are critical for the function of hydroxyproline-rich glycoproteins (HRGPs), this small number of proline residues may play a key role for the RHS10 ECD, which would mediate some extracellular signals, thereby allowing the RHS10 kinase domain to trigger the downstream root hair-inhibitory processes.
In terms of motifs, the ECD of PERK members includes not only extensin-like motifs, but also AGP motifs. RHS10 orthologs and paralogs from rice, poplar, Arabidopsis, and even Physcomitrella all have multiple (28-42) AGP motifs ( Fig. 3A; Supplementary Fig. S6). In this context, the ECD of RHS10 orthologs and other PERKs can be considered as hybrid HRGPs of extensins and AGPs. Although the ECD of PERKs carry extensin-like SPx motifs, their ECD region almost completely lacks the YXY/VYK motif or even a tyrosine residue ( Fig. 3A; Supplementary Fig. S6), which functions for peroxidase-mediated inter-and intramolecular cross-linking in canonical extensins (Epstein and Lamport, 1984;Lamport et al., 2011). Interestingly, the tyrosine residue in the extensin part of LRX1 was also reported not to affect cell wall cross-links (Ringli, 2010). The weak extensin nature of the PERK ECD suggests that the PERK ECD may behave like AGPs at the molecular level. The nature of AGP for the function of the PERK ECD can be conceived by our result demonstrating that RHS10-mediated root hair inhibition required the minimal AGP motif-containing ECD region (D3-1 deletion), but not the extensin-like motif (Fig. 3). A recent study has demonstrated that some root hair-specific prolyl 4-hydroxylases are required for normal root hair growth (Velasquez et al., 2011). Prolyl 4-hydroxylases hydroxylate the proline residues of extensins and AGPs, which will subsequently be glycosylated (Showalter et al., 2010). It would be interesting to determine whether these root hair-specific prolyl 4-hydroxylases target proline residues of the RHS10 ECD.
Glycophosphatidylinositol (GPI)-anchored AGPs can reside in the PM, and some AGPs were shown to bind to pectins (Iraki et al., 1989;Serpe and Nothnagel, 1995), indicating the possibility that AGPs act as sensors for cell wall dynamics (Humphrey et al., 2007). WAKs (cell wall receptors) were shown to bind directly to pectins (Wagner and Kohorn, 2001), and PERK4 seems to associate with pectins (Bai  et al., 2009a). The loss of RHS10 restored root hair growth, which was inhibited by loss of ROL1 in the lrx background (Fig. 6), implying that the function of RHS10 requires a ROL1-mediated process for pectin organization. These results prompted us to hypothesize that the AGP motifs of the RHS10 ECD associate with certain cell wall components, causing the RHS10 kinase domain to transduce signals from wall structure changes into the cytoplasm. Cell wall-loosening proteins, such as expansins, can cause changes in cell wall structure (Cosgrove, 2000;Choi et al., 2006), which then might be sensed by wall-associated kinases including WAKs and PERKs. Because root hair-specific expansins (Cho and Cosgrove, 2002;Kim et al., 2006;Won et al., 2009Won et al., , 2010Lin et al., 2011;Yu et al., 2011) and PERKs (this study) have been functionally well conserved in angiosperms, root hair growth would provide a model system to characterize the relationship between cell wall dynamics and PERK functions.
In terms of motif organization and molecular function, the ECD of PERKs is likely to be conserved in land plants. As mentioned previously, extensin-like and AGP motifs are conserved in RHS10 homologs from Arabidopsis, rice, and poplar, which commonly exhibit root hair-inhibitory effects, even in a moss homolog (Figs 4, 5; Supplementary Fig. S6). This implies that AGP motifs could be commonly operational for the molecular function of PERKs in land plants.

RHS10-mediated signaling for root hair tip growth
We have a paucity of evidence regarding downstream targets of RLKs. There is no downstream process identified for PERKs. In this study, we proposed a downstream process of RHS10 action. In an effort to identify the direct interactor of RHS10 using a Y2H screening, an RNase (RNS2) was identified as an RHS10 interactor (Fig. 8A). Although phosphorylation of RNS2 by the RHS10 kinase domain was not observed in vitro ( Supplementary Fig. S9), RNS2 displayed the root hair-related function in accordance with RHS10 function; namely, negative effects on root hair growth ( Fig. 8B-E). Furthermore, the loss of RHS10 increased, and RHS10 overexpression decreased the level of RNA in the root hair-growing region of the root in co-ordination with RNS2 activity (Fig. 8F). These consistent phenotypic effects of RHS10 and RNS2 on root hair growth and RNA levels suggest that RNS2 is one of the downstream targets of RHS10. Recently, Humphrey et al. (2015) showed that the kinase domain of PERKs, including RHS10/PERK13, interacts with AGC family protein kinases. These results together suggest that PERKs have multiple targets to modulate cellular processes. It would be interesting to know if the RHS10 targets are cell-type specific.
Although the RHS10 kinase domain alone is able to phosphorylate itself, full or modulated kinase activity may require its ECD. The cell wall association and signaling through the RHS10 ECD might be the key modulating process for RHS10 to phosphorylate downstream targets and inhibit root hair growth. Considering the ECD deletion result, a short stretch of the ECD might be enough to perform this activity. The failure of root hair inhibition by the D5 deletion of RHS10, which includes only the kinase domain, also supports the idea that the ECD is necessary to phosphorylate downstream targets ( Supplementary Fig. S10); however, we do not exclude other possibilities, such that the mislocalization of the RHS10 kinase domain, owing to the loss of a TM domain, results in failure to identify proper targets.
RNS2 belongs to class II RNase T2, whose molecular and biological roles have rarely been characterized (MacIntosh et al., 2010). Because RNS2 is highly expressed throughout diverse tissues, it was suggested to play a housekeeping role by recycling RNA (Taylor et al., 1993;Bariola et al., 1999;Hillwig et al., 2011). RNS2 activity has neutral pH optima and is very low under acidic conditions (Hillwig et al., 2011). Although RNS2 was known to localize mainly to the endoplasmic reticulum and vacuoles, where the pH is acidic, a Fig. 9. A model illustrating balanced root hair tip growth. RHE-containing root hair-specific (RHS) genes are under the control of the putative root hair-specific transcription factor (RHF). RHS genes are divided into two subgroups, where RHS + (blue) enhances, but RHS − (red) inhibits root hair elongation. RHS10, a proline-rich RLK, may play a key negative role in root hair elongation by the global control of RNA degradation via RNS2, an RNase. On the other hand, RHS1, a calmodulin-like (CML) protein, may negatively affect root hair elongation through an unknown negative Ca 2+ signaling pathway. The balanced activities between RHS + and RHS − will generate a specific root hair size under certain cellular and environmental conditions. Blunt end, negative action; arrow, positive action; solid line, experimentally supported; broken line, hypothetical.
minor portion of RNS2 appears to be present in the cytosol where pH is neutral, many RNA substrates exist, and RNS2 activity is optimal (Hillwig et al., 2011). PM-localized RHS10 might target this cytosolic RNS2. While rRNAs are the most prominent substrates owing to their abundance in the cell, there is a possibility that RNS2 also targets other RNA species (Hillwig et al., 2011). In addition to its role in cellular homeostasis by its RNA-recycling activity (Hillwig et al., 2011), RNS2 may degrade RNAs that are required for various cellular functions. In root hair cells, an overall decrease in the cellular RNA level should be inhibitory to root hair growth (Fig. 8).
It is conceivable that RHS10 has multiple targets. Although the mechanistic process has yet to be characterized, RHS10 obviously showed an inhibitory effect on ROS accumulation in the root (Fig. 7). RHD2 (an NADPH oxidase)-mediated ROS generation is required to maintain root hair tip growth, most probably by modulating Ca 2+ -mediated vesicle trafficking toward the growing hair tip (Foreman et al., 2003). Including RHS10, RHS genes are positively regulated by RHD2 (Jones et al., 2006;Won et al., 2009), suggesting that ROS not only affect the accumulation of materials for hair tip growth, but also modulate the production of protein tools for root hair growth. Whereas RHD2 regulates downstream RHS genes, the RHD2 gene itself is likely to be regulated by an unidentified RHE-binding transcription factor (RHF; Kim et al., 2006) since its proximal promoter region includes an RHE (Won et al., 2009;Cho, 2013), and its expression in the root epidermis is hair cell specific (Brady et al., 2007). Overall, these relationships would lead to feedback and feed-forward pathways among RHF, RHD2, and RHS10, as shown in Supplementary Fig. S11, which will result in balanced root hair elongation.

Positive versus negative players in root hair elongation
The function of RHS10, together with RHS1, in root hair development is unique in that it negatively regulates root hair tip growth. Although the loss of AKT1 (encoding a potassium transporter), IPK2α (encoding an inositol polyphosphate kinase), or SIZ1 (encoding a SUMO E3 ligase) results in longer root hair growth, this occurs only when potassium, calcium, or phosphate are deficient, respectively (Desbrosses et al., 2003;Miura et al., 2005;Xu et al., 2005). In addition to RHS10, the loss of RHS1 (encoding a calmodulin-like protein) also enhances, and its overexpression inhibits, root hair elongation (Won et al., 2009). While numerous genes positively contribute to root hair elongation without affecting hair morphology, such as branching, waving, and swelling (Grierson and Schiefelbein, 2008;Won et al., 2009), these two RHS gene products are, so far, the only known negative players in root hair elongation under normal conditions, except for the proteins that negatively regulate cellular auxin activity (Grierson and Schiefelbein, 2008).
Among the known RHE-containing root hair-specific genes, RHD2 (Schiefelbein and Somerville, 1990), RHS2 (Won et al., 2009), and EXPA7 (Lin et al., 2011) have been shown to affect root hair elongation positively in Arabidopsis. Therefore, in terms of their effect on root hair elongation, RHS genes can be divided into two subgroups: root hairenhancing RHS (RHS + ) and root hair-inhibiting RHS (RHS − ) ( Supplementary Fig. S11). RHS genes are thought to be controlled by RHF, which appears to be conserved, at least in the angiosperm lineage (Kim et al., 2006). Because there is no noticeable structural difference in the RHE consensus sequence between RHS + and RHS − (Won et al., 2009), these two RHS subgroups are likely to be equally controlled by RHF. This suggests that both positive and negative RHS genes would be equally required for normal root hair elongation, leading to the hypothesis that diverse environmental factors affecting the final root hair length might directly regulate either RHS + or RHS − proteins rather than their transcription.

Supplementary data
Supplementary data are available at JXB online. Figure S1. Root hair phenotypes of different insertion mutant lines of RHS10. Figure S2. The effect of RHS10:GFP fusion protein overexpression on root hair growth. Figure S3. Confocal microscopic images of PIN3:GFP in a root hair cell after plasmolysis. Figure S4. Phylogenetic relationship of PERKs. Figure S5. Alignment of RHS10 homologs. Figure S6. The N-terminal protein structure of RHS10 homologs. Figure S7. T-DNA insertion positions in rns2 and perk mutants. Figure S8. Effects of auxin and an ethylene precursor on root hair restoration of RHS10ox or PERK8ox transformants. Figure S9. Protein blot analysis and an in vitro kinase assay of RNS2 and the RHS10 kinase domain. Figure S10. A model for the role of the extracellular domain (ECD) of RHS10. Table S1. Primer list. Supplementary methods.