QUIRKY regulates root epidermal cell patterning through stabilizing SCRAMBLED to control CAPRICE movement in Arabidopsis

SCM, a leucine-rich repeat receptor-like kinase, is required for root epidermal cells to appropriately interpret their location and generate the proper cell-type pattern during Arabidopsis root development. Here, via a screen for scm-like mutants we describe a new allele of the QKY gene. We find that QKY is required for the appropriate spatial expression of several epidermal cell fate regulators in a similar manner as SCM in roots, and that QKY and SCM are necessary for the efficient movement of CPC between epidermal cells. We also show that turnover of SCM is mediated by a vacuolar degradation pathway triggered by ubiquitination, and that QKY prevents this SCM ubiquitination through their physical interaction. These results suggest that QKY stabilizes SCM through interaction, and this complex facilitates CPC movement between the epidermal cells to help establish the cell-type pattern in the Arabidopsis root epidermis.

T he specification of distinct cell fates is a critical process in the development of multicellular organisms. In many cases, cell fate decisions are influenced by the relative position of a cell to its neighbors, indicating that cell−cell communication is crucial [1][2][3] . A simple model system for the study of cell fate specification is found in the Arabidopsis root epidermis, which is composed of two cell types, root hair-bearing cells (hair cells) and non-hair cells, that are patterned in a position-dependent manner 4,5 . The epidermal cells located outside a periclinal cortical cell wall (N position) contacting a single cortical cell differentiate into non-hair cells, while the epidermal cells located over an anticlinal cortical cell wall (H position) contacting two underlying cortical cells differentiate into hair cells.
Many genes are known to influence cell fate specification in the Arabidopsis root epidermis. GLABRA2 (GL2), which encodes a homeodomain-leucine zipper (HD-Zip) protein, is expressed in developing epidermal cells in the N position and specifies the non-hair cell fate 6 . WEREWOLF (WER) and CAPRICE (CPC) encode an R2R3 MYB-type transcription factor and an R3 single MYB-type transcription factor, respectively, and regulate GL2 expression competitively in a dose-dependent manner [7][8][9] . WER is expressed preferentially in the developing N-position cells and directly induces GL2 expression to specify the non-hair cell fate, whereas CPC inhibits GL2 expression in the H-position cells to specify the hair cell fate. Interestingly, WER is a direct positive regulator of CPC in the N-position cells 10 , and CPC protein moves to the neighboring H-position cells 11 to repress the expression of WER and GL2 12 . Further, the H-cell accumulation of CPC is regulated, in part, by the ENHANCER OF GLABRA3 (EGL3), a basic helix-loop-helix protein 13 . Thus, the differentiating epidermal cells communicate with each other to achieve their proper cell-type pattern via the action of a mobile transcription factor (CPC) produced by WER-expressing N-cells to prevent neighboring H-cells from adopting the non-hair cell fate. The intercellular movement of CPC is likely to occur through plasmodesmata (PD), which provide cytoplasmic connections between plant cells 14,15 .
In addition to intercellular trafficking, plant cells communicate with other cells and their environment through receptor-like kinases (RLKs) on the cell surface. A leucine-rich repeat receptorlike kinase (LRR RLK), SCRAMBLED (SCM; also known as STRUBBELIG), is required for proper cell-type patterning in the Arabidopsis root epidermis 16 , as well as outer integument development in the ovule 17 , fruit dehiscence 18 , internode growth 17 , and tissue morphogenesis 17,19 . In the developing root epidermis, SCM accumulates preferentially in the H-position cells through a feedback mechanism 20 , and has been proposed to respond to a positional signal and preferentially inhibit WER expression in the H-position cells 21 . However, it is not yet known how the initial difference in SCM activity between the N-position cell and the Hposition cell is initiated. Furthermore, it is not clear how SCM action leads to inhibition of WER expression in the H-position cell, considering that SCM kinase activity is not required for epidermal cell patterning 17,18 .
To understand how SCM functions in root epidermal cell patterning, we used a genetic approach to search for new regulators acting in the SCM signaling pathway. We identified a mutant with an scm-like root mutant phenotype, and found that it is an allele of the QUIRKY (QKY) gene. QKY, a multiple C2 domain and transmembrane region protein (MCTP), has previously been reported to interact with SCM and affect epidermal cell patterning [22][23][24] , although its role in root epidermal cell patterning is unknown. Here, we show that QKY acts upstream of WER and GL2, and we find that QKY regulates the efficient movement of CPC between epidermal cells through stabilizing SCM protein at the plasma membrane (PM). Furthermore, we demonstrate that QKY and SCM physically interact through their extracellular domain, and that QKY prevents the ubiquitination of SCM which triggers SCM degradation in vacuoles.

Results
Isolation of a new allele of qky. We screened a population of ethylmethanesulfonate (EMS) mutagenized seeds harboring the GL2p:GFP marker and isolated a mutant showing defects in position-dependent root epidermal patterning and expression of the GL2p:GFP marker ( Supplementary Fig. 1a, b and Table 1). We confirmed that this phenotype is caused by a single nuclear recessive mutation by analyzing the F1 and F2 offspring from a cross with wild-type plants. Through a bulk segregant analysis, we found that the mutation is linked to a marker (nga111) on chromosome 1, which is near the QKY gene previously reported to affect root epidermal cell patterning 22 . Allelism testing (by crossing this new mutant with qky-8 22 and with scm-2 16 ) showed that the mutant phenotype was complemented by scm-2 but not complemented by qky-8 ( Supplementary Fig. 1c). We sequenced the QKY coding region in the genomic DNA from this mutant, which revealed a nonsense mutation at the 870th codon (Supplementary Fig. 1d). In addition, we discovered that a QKY genomic DNA fragment including 1.2 kb 5′-and 1 kb 3′-flanking sequences (QKYp:QKY) completely rescued the abnormal root epidermis phenotype of the new mutant (Supplementary Fig. 1e and Table 1). Together, these results indicate that this newly identified mutant is a new allele of the QKY gene, and we named it qky-16.
Expression of cell fate regulators in the qky mutant root. To determine the regulatory relationship between QKY and previously identified transcriptional regulators of the root epidermis pathway, we examined the promoter activity of WER, CPC and EGL3 using transcriptional reporter genes (WERp:GFP, CPCp: GUS, and EGL3p:GUS) in the qky-16 mutant. In the wild-type root, the WERp:GFP and the CPCp:GUS are preferentially expressed in the N-position cells, while EGL3p:GUS is preferentially expressed in the H-position epidermal cells 7,25,26 . In the qky-16 mutant, the position-dependent expression pattern of these three genes was disrupted, causing reporter gene-expressing cells and reporter gene-non-expressing cells to be produced at each position ( Supplementary Fig. 1a), which is similar to the expression patterns in the scm-2 mutant root 16 . These results suggest that, like SCM, QKY acts early during root epidermis development to enable cells to interpret their relative position and appropriately regulate expression of the downstream transcription factor network.
QKY expression pattern in the root epidermis. To evaluate QKY gene expression, we generated transgenic plants bearing a QKYp: GUS transcriptional reporter construct (GUS gene fused to the same QKY regulatory sequences used for the genomic DNA complementation described above). These plants exhibit strong GUS accumulation in the stele cells and the ground tissue in the root meristematic region, as well as a lower GUS level in developing root epidermal cells and root cap cells (Fig. 1a), consistent with previous reports of widespread expression of QKY in all organs 23,24 .
The QKY protein has been reported to accumulate in the PM in most of the tissues in the root using an N-terminal eGFP fusion to QKY (eGFP-QKY) 23 . We generated translational fusions of GFP to the N-terminus or C-terminus of QKY and expressed these under the control of the QKY regulatory sequences (the same 5′ and 3′ regions used in the QKYp:GUS construct), but these constructs failed to complement the qky-16 mutant root phenotype ( Supplementary Fig. 1e). Therefore, we tested several versions of reporter constructs containing GFP inframe with the QKY coding sequence, and found that a translational reporter construct with the GFP sequence inserted between the end of the first transmembrane domain and the beginning of the phosphoribosyltransferase (PRT) domain (QKYp:QKY-GFP) was able to completely rescue the qky-16 mutant phenotype (Supplementary Fig. 1e and Table 1). As expected, the fluorescence signal in this line was detected at the PM of the cells in most of the root tissues including the epidermis, which is similar to the SCM (SCMp:SCM-GFP) localization pattern (Fig. 1b) 20 . The QKY-GFP signal was unevenly distributed in the PM, which is consistent with the previously reported PD-localization of QKY 23 . Interestingly, we discovered greater accumulation of QKY-GFP in the H-position cells than the N-position cells of the root epidermis (Fig. 1b), implying a preferential role for QKY function in the differentiating hair cells during root epidermis patterning.
Due to the broad expression of QKY in root tissues and a report suggesting non-cell autonomous action of QKY in aerial tissues 23 , we tested the long-range action of QKY in regulating root epidermal cell patterning. We expressed QKY-GFP only in the stele, endodermis, cortex, or epidermis/lateral root cap tissues of the qky-16 mutant by using the SHORTROOT (SHR) 27 , SCARECROW (SCR) 28 , CO2 29 , or WER 7 promoter, respectively. We observed the fluorescence signal only in the tissue(s) where each promoter was expected to be active, and we found that only the WERp:QKY-GFP construct was able to rescue the qky-16 mutant phenotype in the root epidermis ( Fig. 1c and Table 1). These results indicate that QKY influences root epidermal cell patterning in a tissue autonomous manner. Notably, SCM has been reported to act in a cell autonomous manner in the root epidermis 20 .
Genetic interaction between QKY and SCM. Given the similar mutant phenotypes, gene expression patterns, protein accumulation, and epidermis action for QKY and SCM, we sought to examine the relationship between these two genes. First, we tested for possible genetic interaction by generating the qky-16 scm-2 double mutant by a genetic cross. The abnormal epidermal cell pattern in this double mutant was not significantly different from the abnormal pattern in each of the single mutants ( Fig. 2a and Table 1). Similarly, the double mutant exhibited disruption in the position-dependent expression of the GL2p:GUS marker that was comparable to the single mutants (Fig. 2b). QKY overexpression using the promoter of the ROOT HAIR DEFECTIVE 3 (RHD3) gene, which shows strong activity ubiquitously in the root (RHD3p:QKY) 30 , was able to rescue the qky-16 mutant root phenotype completely, but it was not able to rescue the scm-2 mutant phenotype and did not affect position-dependent celltype patterning in the wild-type root epidermis ( Fig. 2 and Table 1). Together, these results suggest that QKY and SCM act in a same genetic pathway to regulate root epidermal cell patterning, and that QKY functions upstream of SCM or they function in parallel.
QKY and SCM regulate CPC movement between epidermal cells. The disrupted cell patterning in the qky and scm mutants and the presumed localization of the QKY and SCM proteins at PD 23 led us to examine their possible role in the lateral inhibition between epidermal cells mediated by the mobile transcription factor CPC 12 . First, we examined the root epidermis phenotype of the qky-16 cpc-1 double mutant, and found that it was largely similar to the phenotype of cpc-1 and quite different from the phenotype of the qky-16 mutant (Table 1), implying that the qky-16 mutant phenotype seems to be mediated by CPC at least in part. It was also shown that the scm-2 cpc-1 root phenotype was quite different from the scm-2 mutant phenotype 21 .
Next, we tested whether QKY or SCM affects CPC movement using the SHRp:CPC-GFP line, in which stele-produced CPC-GFP fusion protein is able to move to the epidermis and preferentially accumulate in the H-position cells 13 . In both the qky-16 and the scm-2 mutant roots, the CPC-GFP protein produced by the SHRp:CPC-GFP construct was detected in the epidermal cells (Fig. 3a), indicating that QKY and SCM are not essential for radial trafficking of CPC. However, preferential accumulation of CPC-GFP in the H-position cells was disrupted in these mutants (about 40% of CPC-GFP accumulating epidermal cells were located at the N-position while about 10% in the wild-type root) (Fig. 3b), suggesting that QKY and SCM are required for preferential CPC trafficking or accumulation into the H-position cells.
We used a biolistic particle delivery system to produce GFP or the CPC-GFP fusion protein in a single epidermal cell and monitor its movement. When GFP alone was delivered and expressed in an epidermal cell of wild-type leaves, the GFP signal was also detected in a neighboring cell in more than 70% of the cases after 4 h ( Fig. 3c and Supplementary Fig. 2a). Similar mobility of GFP was observed in the qky-16 or scm-2 mutants, consistent with a previous report, suggesting no general effect of QKY and SCM on GFP movement between cells 23 . The CPC-GFP fusion protein (35Sp:CPC-GFP) was also detected in neighboring cells in about 70% of the cases in wild-type leaves, but only in 46 and 38% of the cases in the qky-16 and scm-2 mutant leaves, respectively (Fig. 3c), implying that QKY and SCM affect CPC-GFP movement between epidermal cells. To determine whether SCM and QKY are involved in export of CPC-GFP from the CPC-GFP expressing cell, we coexpressed SCM-CFP or QKY-CFP together with CPC-GFP in the same leaf epidermal cell (in the scm-2 or qky-16 mutant, respectively) and analyzed CPC-GFP mobility. We found that CPC-GFP protein movement was not enhanced by coexpression of SCM or QKY ( Fig. 3c and Supplementary Fig. 2b), implying that SCM and QKY do not affect export of CPC, so they rather may be involved in the import of CPC. This possibility is consistent with the greater abundance of SCM and QKY in the H-position cells of the root epidermis which preferentially accumulate the mobile CPC protein (Fig. 1b) 20 . It appears unlikely that SCM and QKY affect CPC movement through direct interaction, because we found that neither the cytoplasmic domain of SCM (SCM CD ) nor the three domains of QKY divided by the two transmembrane domains (QKY D1 , QKY D2 , and QKY D3 ) are able to interact with CPC in yeast cells ( Fig. 4a and Supplementary Fig. 3).
Given that CPC mobility is also important for proper trichome spacing in the shoot epidermis 31 , we examined the leaf surface of the qky, scm, and qky scm mutants and discovered instances of adjacent "twin" trichomes (indicating defective CPC-mediated lateral inhibition) in the qky-16 (0.54%, n = 925), the scm-2 (0.55%, n = 900), and the qky-16 scm-2 mutant (0.70%, n = 849), which was never observed in the wild-type leaves (n = 1376) ( Supplementary Fig. 1f). This result further supports the view that QKY and SCM influence CPC movement between epidermal cells.
QKY and SCM interact through their extracellular domains. It has been reported that QKY and SCM physically interact with each other using the FRET-FLIM method and the yeast   two-hybrid assay 23,24 . We further confirmed this interaction using coimmunoprecipitation of epitope tagged full-length proteins (QKY-FLAG and SCM-MYC) expressed in tobacco leaves ( Fig. 4b). In addition, we used the yeast two-hybrid assay to determine that the SCM extracellular domain interacts with the N-terminal domain 1 of QKY (QKY D1 ) ( Fig. 4a-c) rather than the SCM cytoplasmic domain which was previously reported to be a QKY-interacting domain 23 . Further, we found that a mutated version of the SCM extracellular domain (at a prolinerepeat region; SCM Em2 ) was not able to interact with the QKY D1 ( Fig. 4a-d). We also found that this mutated version of SCM (SCMp:SCM Em2 -GFP) was not able to complement the scm-2 mutant phenotype ( Fig. 4e and Table 1). Interestingly, a mutation in a different proline-repeat region in the extracellular domain (SCM Em1 ; SCMp:SCM Em1 -GFP) did not affect the functionality of SCM nor its interaction with QKY ( Fig. 4a-e and Table 1). These results suggest that the QKY D1 could be an extracellular domain that interacts with the SCM extracellular domain. However, it was reported that the cytoplasmic domain of SCM interacts with the QKY D1, leading to the suggestion that this is a cytoplasmic domain 23 . To examine the topology of QKY at the PM, we isolated protoplasts from QKYp:QKY-GFP qky-16 or SCMp:SCM-GFP scm-2 plants, applied proteinase K to digest the exposed extracellular domain of the membrane-spanning proteins, and performed western blot analysis with total protein extracts using anti-GFP antibodies (Fig. 4f). The proteinase K-treated cells yielded a low level of full-length proteins (130 kD for SCM-GFP and 170 kD for QKY-GFP) and a large amount of truncated proteins (70 kD for SCM-GFP and 40 kD for QKY-GFP) in the western blot analysis, implying that the kinase domain of SCM (including the transmembrane domain) and the domain 2 of QKY (including the two transmembrane domains) fused to GFP were protected. We further evaluated the topology of QKY using the bimolecular fluorescence complementation (BiFC) assay (Fig. 4g). We fused the N-terminal half of EYFP (EYFPn) to the C-terminus (cytoplasmic domain) of SCM (SCM-EYFPn-V2), and we fused the C-terminal half of EYFP (EYFPc) between the end of the first transmembrane domain and the beginning of the following PRT domain in QKY-GFP (QKY-EYFPc-V2). We coexpressed these in Nicotiana benthamiana leaf epidermal cells and were able to detect EYFP fluorescence. On the contrary, the SCM-EFYPn-V2 was not able to show the fluorescence complementation with the QKY-EYFPc-V1 or QKY-EYFPc-V3 which has the EYFPc at the end of the Nterminus or the C-terminus of QKY, respectively (Fig. 4g). In a reciprocal experiment, we found that SCM-EYFPn-V1 (EYFPn fused to the N-terminus of SCM) did not show fluorescence complementation with QKY-EYFPc-V2 but did with QKY-EYFPc-V1 or QKY-EYFPc-V3 ( Supplementary Fig. 4).
Together, these results suggest that the domain 1 and domain 3 of QKY are extracellular domains, and that domain 1 is used by QKY to interact with SCM and regulate epidermal cell patterning. To investigate the biological impact of QKY−SCM interaction, we examined SCM-GFP protein accumulation in the SCMp:SCM-GFP qky-16 scm-2 plant root. We discovered that the level of SCM-GFP protein was much lower than in the SCMp:SCM-GFP scm-2 plant root (Fig. 5a, b), while SCM promoter activity and SCM-GFP transcript level were not affected by the qky mutation ( Supplementary Fig. 5). On the contrary, the scm mutation did not significantly affect QKY-GFP protein accumulation and QKY-GFP transcript level ( Fig. 5a and Supplementary Fig. 5). In addition, the fluorescence intensity from the SCM Em2 -GFP (which fails to interact with QKY) was much weaker than the intensity from SCM-GFP and from SCM Em1 -GFP even though their transcript levels are similar ( Fig. 4e and Supplementary  Fig. 6). These results suggest that interaction between QKY and SCM is important for the stabilization of SCM protein.
Therefore, we examined the effect of MG132 (an inhibitor of proteasome and cysteine protease), lactacystin (a proteasomespecific inhibitor), and E-64d (a cysteine protease inhibitor) on SCM-GFP protein level. Treatment of MG132 and E-64d increased the SCM-GFP protein level in qky-16 scm-2 and in scm-2 mutant roots, while lactacystin did not affect SCM-GFP accumulation (Fig. 5c), indicating that SCM is rapidly degraded in a cysteine protease-dependent manner rather than by a proteasome in the qky-16 scm-2 mutant. Interestingly, the major protein reacting with the anti-GFP antibodies in the qky-16 scm-2 mutant root had higher molecular weight (MW) (red asterisk) than the SCM-GFP protein in the scm-2 mutant root (black asterisk) (Fig. 5b). Upon treatment with MG132 and E-64d, the high MW SCM-GFP band (red asterisk) showed greater accumulation in the qky-16 scm-2 mutant root compared with the same sized SCM-GFP in the scm-2 mutant root (Fig. 5c). f Proteinase K protection test. Protoplasts from SCMp:SCM-GFP and QKYp:QKY-GFP stable transgenic plants were treated with proteinase K (1 ng/μl). Total protein extracts were subjected to immunoblotting using monoclonal anti-GFP antibodies. g BiFC assay with QKY and SCM. SCM-EYFPn-V2 in which the N-terminal half of EYFP (EYFPn) was fused to the C-terminus of SCM and QKY-EYFPc-Vx in which the C-terminal half of EYFP (EYFPc) was fused to various sites as shown in the upper panel were coexpressed in the Nicotiana benthamiana leaf epidermal cell using a biolistic particle delivery system. Their expression was confirmed by western blot analysis with total protein extracts using polyclonal anti-GFP antibodies. The red and the black asterisks indicate the QKY-EYFPc protein and the SCM-EYFPn protein, respectively. Bar = 25 μm SCM-GFP preferentially accumulated in the PM, and E-64d treatment increased internal accumulation of SCM-GFP in subcellular membranous compartments much more in the qky-16 scm-2 mutant root than in the scm-2 mutant root (Fig. 6a). Because E-64d is known to inhibit the fusion of late endosomes with vacuoles in Arabidopsis 32 , we examined whether SCM-GFP accumulation in late endosomes increased after E-64d treatment. We introduced the UBQ10p:mCherry-RabF2a, a marker of late endosomes 33 , into the SCMp:SCM-GFP qky-16 scm-2 plant by a genetic cross, and discovered that a small portion of the internal SCM-GFP colocalized with mCherry-RabF2a without E-64d treatment (DMSO) (Pearson's correlation coefficient = 0.112727; Fig. 6b). E-64d treatment caused strong colocalization of the internal SCM-GFP with mCherry-RabF2a (Pearson's correlation coefficient = 0.678813), indicating SCM-GFP accumulation in the prevacuolar compartment (Fig. 6b). We also found that treatment of concanamycin A (ConA) (a vacuolar ATPase inhibitor, which prevents the vacuolar degradation of proteins) greatly increased the accumulation of SCM-GFP in the vacuole in the qky-16 scm-2 mutant root, while it caused slightly increased vacuolar accumulation of SCM-GFP in the epidermis of the scm-2 mutant root (Fig. 6c).
Next, we examined the effect of TyrA23, an inhibitor of endocytosis 34 , on the SCM-GFP accumulation and found that TyrA23 treatment greatly increased accumulation of SCM-GFP at the PM in the qky-16 scm-2 mutant root, while it only modestly increased SCM-GFP at the PM in the scm-2 mutant root (Fig. 6d).
These results indicate that SCM is internalized from the PM by endocytosis and degraded in the vacuole, and that QKY prevents this internalization and vacuolar degradation.
QKY prevents SCM protein monoubiquitination. Given the observed size difference between SCM-GFPs from the SCMp: SCM-GFP scm-2 and the SCMp:SCM-GFP qky-16 scm-2 roots (Fig. 5b), the accumulation of high-MW SCM-GFP following MG132 treatment (Fig. 5c), and reports showing that ubiquitination is a signal for internalization and degradation of PM proteins 35,36 , we examined SCM-GFP for possible ubiquitination. Western blot analysis of immunoprecipitated proteins from seedlings treated with MG132 using anti-GFP antibodies and anti-ubiquitin antibodies (P4D1) showed that the high-MW SCM-GFP which accumulated following MG132 treatment was ubiquitinated, while the low-MW SCM-GFP protein was not (Fig. 7a). However, antibodies specific for polyubiquitination (FK1) did not react with this high-MW SCM-GFP (Fig. 7a), while both of the antibodies reacted with the positive control BRI1-GFP (which is known to be polyubiquitinated 37 ). Considering the homogeneous MW of the ubiquitinated SCM-GFP, this result suggests that the SCM-GFP is monoubiquitinated.
To evaluate the importance of ubiquitination for SCM degradation and QKY's potential role in preventing SCM ubiquitination, we generated an SCM construct with all 22 lysine codons in the cytoplasmic domain replaced with arginine codons (SCM m22KR ). This construct was transiently expressed (35Sp: SCM m22KR -GFP) in the leaf epidermis of wild-type and qky-16 mutant plants, and SCM-GFP protein was immunoprecipitated and subjected to western blot analysis (Fig. 7b). Unlike wild-type SCM-GFP, the SCM m22KR -GFP did not yield a high-MW SCM-GFP band in the qky-16 mutant root; rather, the size is similar to the nonubiquitinated form. Indeed, this SCM m22KR -GFP did not react with anti-ubiquitin antibodies (P4D1). Furthermore, the fluorescence intensity from the SCMp:SCM m22KR -GFP qky-16 scm-2 plant root was not significantly weaker than the intensity from the SCMp:SCM m22KR -GFP scm-2 plant root unlike the wildtype version of SCM-GFP, which was confirmed by western blot (Fig. 7c). In addition, the difference in the fluorescence intensity between the N-position cells and the H-position cells was much smaller in the SCMp:SCM m22KR -GFP qky-16 scm-2 root and SCMp:SCM m22KR -GFP scm-2 root than in the SCMp:SCM-GFP scm-2 root (Fig. 7d). Although the fluorescence intensities in the root of those two plant lines were comparable to the SCMp:SCM-GFP scm-2 line, SCMp:SCM m22KR -GFP was not able to completely rescue the abnormal root epidermis phenotype of the scm-2 mutant (Supplementary Fig. 7). This result is reminiscent of the effect of overexpressed SCM in the scm-2 mutant (35Sp:SCM scm-2) 20 . Altogether, these results suggest that monoubiquitination, which is prevented by QKY in the H-position cells, is necessary for differential accumulation of SCM between cells at each position in the root epidermis.

ARTICLE
To identify the ubiquitination sites in the SCM protein, we analyzed the gel mobility of mutant versions of SCM-GFP with single or multiple lysine residues replaced by arginine residues and expressed transiently in qky-16 mutant leaves ( Supplementary  Fig. 8). The SCM-GFPs with a substitution of Lys-520, Lys-525 or Lys-526, and two substitutions at Lys-525 and Lys-526 showed intermediate gel mobility (between the mobility of wild-type SCM-GFP in the qky-16 mutant and the wild type), and SCM-GFPs with two substitutions at Lys-520 and Lys-526 showed similar mobility to that of nonubiquitinated SCM-GFP. Therefore, we changed all three of these lysine residues to alanine (SCM K520/525/526R -GFP), expressed this construct transiently in qky-16 mutant leaves, immunoprecipitated with anti-GFP antibodies, and analyzed its gel mobility and ubiquitination (Fig. 7e). This SCM K520/525/526R -GFP showed similar mobility to that of nonubiquitinated SCM-GFP and did not react with antiubiquitin antibodies (P4D1) (Fig. 7e). These results suggest that SCM is modified by multi-monoubiquitination and that these three residues (Lys-520, and at least one from Lys-525 and Lys-526) represent the ubiquitination site.

Discussion
SCM, an atypical LRR RLK, is a critical regulator of cell-type patterning in the Arabidopsis root epidermis 16 . However, the mechanism and regulation of SCM action had been unclear prior  test (a, c, d). Values represent means ± SD of five individual 4-day-old seedlings for (a, c) and of ten seedlings for (d).
In each seedling, ten cells were examined to this study. Here, we showed that SCM function in root epidermal patterning is dependent on QKY. Like scm mutants, the qky mutant lacked position-dependent cell-type specification ( Supplementary Fig. 1). The QKY gene displayed a similar root expression pattern as the SCM gene, and the QKY and SCM proteins both accumulated in the PM, preferentially in the Hposition cells (Fig. 1). Furthermore, we found that SCM and QKY are likely to act in the same genetic pathway (Fig. 2).
We have also shown that QKY and SCM are required for movement and preferential accumulation of CPC in the H-position epidermal cells (Fig. 3). PD are intercellular channels conferring cytoplasmic continuity between plant cells and controlling movement of various molecules by targeted and nontargeted mechanisms. Some plant RLKs including SCM appear to associate with PD [38][39][40][41] , and some of these are suggested to regulate movement through PD 39,40 . Interestingly, a QKY-related protein (one of the 16 members of the multiple C2 domain and transmembrane region proteins (MCTP) family in Arabidopsis) called FT-INTERACTING PROTEIN1 (FTIP1) shows preferential localization to PD and regulation of specific protein In vivo ubiquitination analysis of SCM-GFP and SCM m22KR -GFP. Ubiquitination assay was performed as in (a) with the immunoprecipitates from Arabidopsis leaves transiently expressing SCM-GFP or SCM m22KR -GFP for 12 h with or without MG132 (20 μM) after introducing 35Sp:SCM-GFP or 35Sp:SCM m22KR -GFP by the Biolistic PDS-1000/He system. c GFP accumulation in the root epidermal cells of SCMp:SCM-GFP scm-2, SCMp:SCM-GFP qky-16 scm-2, SCMp:SCM m22KR -GFP scm-2 and SCMp:SCM m22KR -GFP qky-16 scm-2 (left). SCM m22KR indicates that all the lysine residues in the cytoplasmic domain of SCM protein were mutated to arginine residue. Same confocal detection setting was used to compare the fluorescence intensities from different plants. Bar = 25 μm. The protein level was also analyzed by western blot (right).  42 . FTIP was shown to directly interact with FT in companion cells through its third C2 domain and to control FT export from companion cells to the neighboring sieve elements 42 . In contrast, we discovered that QKY directly interacts with SCM through its N-terminal region containing C2 domains (Fig. 4) rather than CPC (Supplementary Fig. 3) and seems to control CPC import rather than export between epidermal cells (Fig. 3c). Furthermore, QKY seems to regulate the movement of CPC indirectly, as QKY is a regulator of SCM protein accumulation (Figs. 5 and 6).
Our discovery of QKY-dependent stabilization of the SCM protein provides an explanation for the close relationship between QKY and SCM. We show that QKY directly interacts with SCM and generates high levels of nonubiquitinated SCM, rather than ubiquitinated SCM which is internalized and degraded in the vacuole (Figs. 4-6). The precise mechanism by which QKY generates nonubiquitinated SCM is unclear. QKY may promote SCM deubiquitination or it may prevent SCM ubiquitination. The former possibility appears unlikely, because although QKY directly interacts with SCM through their extracellular domains (Fig. 4), QKY has no discernible protease domain nor shows any similarity to known deubiquitinating enzymes. The effect of QKY on SCM stability may also explain the previously reported preferential accumulation of SCM in the H-position cells of the root epidermis 20 . Notably, a nonubiquitinated mutant version of SCM (SCM m22KR ) expressed under the control of the SCM promoter accumulated highly in both epidermal cell positions (H cell and N cell), even in the absence of QKY (Fig. 7c, d). This implies that the observed preferential H-cell accumulation of wild-type SCM is due to preferential SCM stabilization in the H-cell position as a result of the greater QKY accumulation in the H-position cells. Accordingly, an important future goal is to explore the mechanism responsible for preferential H-position cell accumulation of QKY.
In mammals and yeast, monoubiquitination and K63polyubquination generally govern trafficking of membrane proteins, whereas K48-polyubiquitination generally regulates proteasome-dependent protein degradation 43,44 . In plants, diverse forms of ubiquitination appear to be used for the degradation of different membrane proteins. Polyubiquitination has been reported to be important for internalization and degradation of LRR RLKs, such as BRI1 and FLAGELLIN SENSING2 (FLS2) 37,45 . Other membrane proteins have also been reported to be ubiquitinated. The multi-monoubiquitination of ION-REGULATED TRANS-PORTER 1 (IRT1) drives its internalization to early endosomes (EE) 46 and subsequent K63-polyubiquitination of the multimonoubiquitin moieties is required for its vacuolar degradaion 47 . BOR1, a boron transporter protein, is mono-or diubiquitinated and degraded in the vacuole 48 . We found that SCM is multimonoubiquitinated, and this promotes its internalization and vacuolar degradation (Figs. 6c, d and 7). Taken together, monoubiquitination seems to be sufficient for the internalization of membrane proteins in plants, as observed in nonplant organisms 43,44 , but multi-monoubiquitination and K63polyubiquitination may increase the efficiency of internalization, as also shown in yeast 49,50 , and may be responsible for the vacuolar targeting. It is possible that SCM is also polyubiquitinated, which may have gone undetected due to its low level. Here, however, we showed that blocking SCM degradation in vacuoles resulted in greater accumulation of multi-monoubiquitinated SCM, implying that multi-monoubiquitination plays a major role in the vacuolar degradation of SCM even though we cannot rule out the possibility of polyubiquitination.
Together, these results provide an updated model for cell fate specification in the Arabidopsis root epidermis. In this model, QKY protein preferentially accumulates in the H-position cells of the root epidermis and stabilizes SCM protein in these cells by preventing its ubiquitination. The preferential accumulation of SCM at PD in the H-position cells facilitates import of CPC protein from the neighboring N-position cells. The accumulation of CPC in the H-position cells inhibits WER function 9 . In turn, feedback regulation (negative regulation by WER and positive regulation by CPC on SCM expression 20 ) further establishes preferential accumulation of SCM protein in H-position cells, which ultimately results in the cell-type pattern. This model is consistent with the inability of QKY overexpression to rescue the scm-2 mutant phenotype (Table 1) and the report that increasing SCM gene copy number shows a mild reduction of the qky mutant phenotype 23 . It is also consistent with the finding that the scm mutation and the qky mutation did not cause a significant increase in hair cell specification in the N-position cells in the absence of CPC (Table 1) 21 . However, we also found that ubiquitous expression of QKY in the wild type did not cause ectopic hair cell fate specification in the N-position cells (Table 1), which indicates that an additional H-position factor might be required for effective stabilization of SCM by QKY.
For plant growth, seeds were sterilized, germinated and grown vertically on agarose-solidified medium containing mineral nutrients 52 .
Analysis of epidermal cell patterning. For each line, we analyzed at least ten 4day-old seedlings and performed three independent experimental repeats. We scored the ten H-position cells and the ten N-position cells for each seedling using differential interference contrast microscopy.
Gene constructs and plant transformation. For qky-16 complementation experiments, the genomic DNA fragment containing a 1210-bp 5′ flanking region DNA fragment and a 576-bp 3′ flanking region DNA fragment from QKY were used. To analyze the promoter activity of QKY, the β-glucuronidase gene or the GFP gene was inserted between the same 5′ and 3′ flanking DNA fragments that were used for the complementation test (QKYp:GUS and QKYp:GFP). To visualize the QKY protein, the GFP coding region was inserted at the end of the first transmembrane domain of QKY (QKY-GFP). For the QKYp:QKY-GFP construct, the chimeric QKY-GFP was inserted between the same flanking sequences of QKY. To express QKY-GFP in various tissues, the chimeric gene was fused to each tissuespecific promoter. For the RHD3p:QKY construct, QKY was fused to the 3′ end of the RHD3 promoter 30 . SCM Em1 -GFP and SCM Em2 -GFP fusions were fused to the 3′ end of the SCM promoter 20 . For the BiFC assay, QKY-EYFPc-V1, QKY-EYFPc-V2, QKY-EYFPc-V3, SCM-EYFPn-V1 and SCM-EYFn-V2 constructs were made by PCR cloning using Gibson assembly technology and expressed under the control of the 35S promoter. To identify ubiquitination sites, SCM, SCMm 22KR and SCMm K520/525/526R were fused to 5′ end of the GFP gene, and transiently expressed under the control of the 35S promoter in Arabidopsis leaves. Plant transformation was achieved by electroporating constructs into the Agrobacterium strain GV3101 followed by introduction into Arabidopsis using the floral dip method 53 . The primers used for the constructions are described in Supplementary Table 1.
Confocal microscopy. To analyze GFP expression, seedlings were counterstained with 5 μg/ml propidium iodide (PI) or 5 μM FM4-64 (Invitrogen) and examined using a Zeiss LSM880 confocal microscope 7 . The PI signal was detected with a 543 nm excitation mirror and a 560 nm long-path filter. The mCherry-RabF2a signal was detected with a 561 nm excitation mirror and a 661 nm long-path filter. Laser intensity settings were kept constant across samples in each experiment.
Transient gene expression with the particle gun. Rosette leaves were bombarded with DNA-coated tungsten particles using the Biolistic PDS-1000/He system (Bio-Rad) according to the manufacturer's instructions. Tungsten (M10) particles were coated with 1 μg of the DNA, or with 1 μg of each DNA for co-bombardment. The analysis was carried out after 4−12 h of incubation.
Yeast two-hybrid assay. The C2-domain containing region of QKY (QKY D1 ), second domain of QKY (QKY D2 ), and third domain of QKY (QKY D3 ) were joined as a C-terminal fusion to the yeast GAL4 DNA-binding domain in pAS2-1. Likewise, the extracellular domain (SCM ED ), the cytosolic domain (SCM CD ) of SCM, and the mutated extracellular domains (SCM Em1 , SCM Em2 ) were fused to the GAL4 activation domain in pACT2. Transformation was performed using yeast strain AH109 and assays were examined in selective media. The gene-specific primers used are listed in Supplementary Table 1.
Quantitative real-time RT-PCR. To measure the steady-state level of transcripts, total RNA was extracted from the root tips of 4-day-old seedlings and treated with RNase-free DNase I. One microgram of total RNA was used for the reverse transcription using the AccuScript High Fidelity 1st Strand cDNA synthesis kit (Agilent Technologies, 200820), and 1 µl of resulting first-strand cDNA (20 µl) was used as a PCR template for the quantitative real-time RT-PCR. Quantitative PCR analysis was performed using SYBR Premix EX Taq (Takara, #RR82LR) with an AriaMx real-time PCR machine (Agilent Technologies). EF1 was used as an internal reference to normalize the relative level of each transcript. Each experiment was repeated three times, and each time the experiment included triplicate samples. The raw data underlying the averages are shown in Source Data file.
Quantification of SCM-GFP accumulation. Confocal images were obtained from ten 4-day-old seedlings of the corresponding plant line. Mean fluorescence intensities of SCM-GFP at the PM and in the inner area were measured in the ten H-position epidermal cells from each seedling image by creating a region of interest covering the PM area and entire cytoplasm area, respectively using ImageJ program (Fiji version). The raw data underlying the averages are shown in Source Data file.
Analysis of colocalization of internalized SCM and RabF2a. Confocal images of 4-day-old seedling roots were analyzed using the Image-pro 10 software to produce scatterplots and to calculate colocalization of SCM-GFP and mCherry-RabF2a. The scatterplots that were obtained by analyzing all pixels inside five different cells in one seedling produced the Pearson's correlation coefficient based on the automated Otsu threshold 54 . The raw data underlying the averages of the Pearson's correlation coefficient are shown in Source Data file.
Proteinase K protection test. Protoplasts were isolated from SCMp:SCM-GFP and QKYp:QKY-GFP stable transgenic plants. The protoplasts were incubated in the presence or absence of proteinase K (1 ng/μl). Then, PMSF and SDS sample buffer were added and denatured at 95°C to inactivate protease activity.
Bimolecular fluorescence complementation. Fifty microliters of Agrobacterium overnight culture was centrifuged, washed, and suspended in a solution containing 10 mM MES, 10 mM MgCl 2 , and 200 µM acetosyringone. Coinfiltration of Agrobacterium strains containing the BiFC constructs and the p19 silencing plasmid was carried out. The leaves of 4-week-old Nicotiana benthamiana were infiltrated and subsequently analyzed at day 3 after infiltration using a Zeiss LSM 880 confocal microscope with ZEN software. EYFP signal was detected with a 514 nm excitation mirror and a 520−600 nm band-path filter.
Reporting summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.