Identification of SHRUBBY, a SHORT-ROOT and SCARECROW interacting protein that controls root growth and radial patterning

The root of Arabidopsis thaliana has a simple, consistent and well-defined pattern of cell division and growth that makes it a highly tractable system for studying cellular development (Dolan et al., 1993). The primary root is composed of concentric rings of tissues arranged in a radial axis that surrounds a central cylinder of vascular and procambial cells. From the outside inwards, these rings of tissues are: the epidermis, the cortex, the endodermis (collectively the cortex and endodermis make up the ground tissue) and the pericycle. Overlaid upon the radial organization of the root are three longitudinal zones that describe cellular behavior; classically, these are the meristem, the elongation zone and the differentiation zone. In the Arabidopsis root, the meristem is generally defined as a region of the root apex that extends basally from the quiescent center (QC) to the first set of elongated cells in the cortex. This region includes the initial cells (stem cells) that surround the QC and the transit amplifying cells. Above the meristem, cells begin a phase of rapid cell elongation that precedes cellular differentiation. The boundaries of these domains are plastic and are largely regulated by antagonistic interactions between different phyto-hormones, which regulate the expression of key transcription factors. Here, we describe SHRUBBY (SHBY), a SHORT-ROOT (SHR)- and SCARECROW (SCR)-interacting protein. Mutations in SHBY affect both the longitudinal and radial organization of the root, through regulation of SHR, SCR, PLETHORA1 (PLT1) and PLETHORA 2 (PLT2) levels, and gibberellin signaling.

In the root meristem, cell divisions occur in two groups of cells: in the initials that surround the quiescent center (QC) cells; and in a population of transit amplifying cells that are derivatives of the initials (Dolan et al., 1993). The maintenance of cell division in these populations is dependent upon the presence of a functional QC. Loss of the QC results in premature termination of root growth (van den Berg et al., 1997; Sabatini et al., 2003; Xu et al., 2006). The expression of two sets of partially overlapping transcription factors, SHR and SCR, and the PLT proteins is required for maintaining the QC (Di Laurenzio et al., 1996; Helariutta et al., 2000; Sabatini et al., 2003; Aida et al., 2004). Both SHR and SCR are interacting members of the GRAS family of transcription factors (Cui et al., 2007; Welch et al., 2007). The SHR protein is expressed in the stele, and moves from the stele into the surrounding cells. In the endodermis and the QC, SHR turns on the expression of SCR, which in turn reinforces its own expression specifically in the QC (Helariutta et al., 2000). Loss of either SHR or SCR expression results in the formation of a short root that fails to maintain the QC and meristem.

Also expressed in the QC and root meristem are the PLT proteins. The PLTs are AP2 class transcription factors, the expression of which in the root meristem converges upon the QC and the surrounding initial cells (Aida et al., 2004; Galinha et al., 2007). The PLT1 protein is at its highest level in the QC and in the surrounding initials, with a weaker protein gradient extending basally into the stele. PLT2 also shows strong expression in the QC and the initials, but its expression basally is maintained at higher levels in the cortex and epidermis than in the stele. Loss of both PLT1 and PLT2 expression results in the loss of QC markers, differentiation of the initial cells and formation of root hairs at the root tip, indicating that the PLTs function in both the maintenance and the activity of the QC and initials (Aida et al., 2004). Both PLT1 and PLT2 function in a dose-dependent manner, with high levels of PLT required to maintain stem cell fates (Galinha et al., 2007).

Outside of the QC, SHR and SCR are required for radial patterning of the root (Di Laurenzio et al., 1996; Helariutta et al., 2000). In the meristem, formative divisions in the initials give rise to, or maintain, distinct cell layers. For example, the cortical endodermal initials (CEI) that flank the QC divide asymmetrically to produce daughter (CED) cells that divide periclinally to produce the endodermis and cortex. Periclinal division of the CED is controlled by SHR and SCR. Both of these genes induce the expression of a D-type cyclin, CYCD6;1, in the CED cell, which leads to the asymmetric periclinal cell division that produces the endodermis and cortex (Sozzani et al., 2010). Following this asymmetric division, both SHR and SCR are rapidly degraded in the cortex cell, thus preventing further periclinal cell divisions (Heidstra et al., 2004). Loss-of-function mutations in SHR or SCR result in the production of a single ground tissue layer between the pericycle and the epidermis.

Later in root development, both SHR and SCR control the periclinal cell divisions in the endodermis that produce middle cortex (MC) (Baum et al., 2002; Paquette and Benfey, 2005). Between days 7 and 14 the endodermis begins to divide periclinally and asymmetrically to produce another layer of ground tissue, the MC. In contrast to their similar roles in the formation of the endodermis and cortex, SHR and SCR play different roles in the production of MC. SCR inhibits the divisions in the endodermis that produce the MC, whereas SHR is required for them. The result is that MC forms precociously in plants that lack SCR function (Paquette and Benfey, 2005); it never forms in shr-2 mutants. However, MC forms precociously in shr-2 heterozygotes or in shr hypomorphs (Koizumi et al., 2012). It is likely that SHR promotes formation of MC via activation of CYCD6;1, as Sozzanni et al. (Sozzanni et al., 2010) showed expression of CYCD6;1 in endodermal cells that form MC, and a reduction in MC-generating cell divisions in cycd6;1 mutants. GA signaling also regulates formation of MC. Inhibition of GA signaling in either wild-type or scr roots hastens the formation of MC (Heo et al., 2011; Paquette and Benfey, 2005).

To elucidate the function of SHR in root growth and patterning, we isolated 17 proteins that interact with SHR. Among the isolated proteins were the bona fide SHR interacting proteins SCR, MAGPIE (MPG) and JACKDAW (JKD) (Cui et al., 2007; Welch et al., 2007), and multiple related zinc-finger domain transcription factors. We concentrated our analysis on an uncharacterized protein, SHBY with similarity to the gene responsible for Cohen syndrome in humans (Velayos-Baeza et al., 2004). Mutations in SHBY reduced meristem activity and increased cell divisions in the ground tissue. We show that SHBY interacts with SHR and SCR, and regulates the levels of SHR, SCR, PLT1 and PLT2 proteins. In turn, SHR and SCR also regulate the levels of SHBY, with shr-2 and scr-4 mutants showing considerably reduced SHBY expression. The overall phenotype of the shby mutants suggests roles in protein turnover and in GA signaling. Indeed the shby-1 root phenotype is partially phenocopied by growth of wild-type roots on low doses of MG132 or of the GA inhibitor paclobutrazol (PAC). In addition shby-1 mutants are insensitive to GA and PAC, suggesting that SHBY is required for GA signaling. SHBY therefore provides a link between the SHR/SCR pathway, the PLTs and GA.

Yeast 2-hybrid assays were contracted to Hybrigenics (Paris, France) using a prey library made from cDNAs derived from 1-week-old Arabidopsis seedlings. The SHR bait construct and the second prey library are as described elsewhere (Koizumi et al., 2011).

Arabidopsis thaliana Columbia lines (Col-0) and the heterozygous siblings of the shby-1 allele were used as wild type. Homozygous T-DNA insertions (http://abrc.osu.edu/) were identified using PCR (supplementary material Table S3). Other lines used in this study are described in Table S3. Plants were grown vertically on 0.5× MS medium containing 0.05% (w/v) Mes (pH 5.7), 1% (w/v) sucrose and 0.5% (w/v) Phytagel (Sigma-Aldrich) in growth chambers at 23°C with 16-hour light and 8-hour dark photoperiods.

Gateway pDONR P4P1R plasmid containing the En7 promoter (Heidstra et. al 2004) and pDONR221 with ER-localized GFP (erGFP) were recombined into dpGreen BarT (Lee et al., 2006) using Gateway protocols (Invitrogen/Life Technologies). The resulting binary vector was introduced into Agrobacterium strain GV3101-pSoup-pMP. Standard protocols were used for transformation.

Root cross-sections and confocal images were prepared as described previously (Koizumi et al., 2000; Koizumi et al., 2011). The contrast and brightness of some images were adjusted using Adobe Photoshop 7. For fluorescent quantification, identical settings were used to collect images on the same day for both the wild-type and mutant plants homozygous for the GFP or YFP marker. Intensity levels were measured using ImageJ software (http://rsbweb.nih.gov/ij/) on unmodified root images. SHR movement was examined as described previously (Koizumi et al., 2011). Root meristem cell number and meristem lengths were measured on DIC images of cleared roots using ImageJ. Cell production rate was calculated using cell flux assays as described in Beemster and Baskin (Beemster and Baskin, 1998). Whole-mount in situ hybridization was carried out as described previously (Hejátko et al., 2006). The region between the F3 and R3 primers (supplementary material Fig. S2A; Table S3) was used for making the sense and antisense probes.

For PAC treatment, seeds were germinated for 30 hours on MS plates and individually transferred to MS plates supplemented with 2 μM PAC. Roots were then examined at day 7. For short-term (6 hours) PAC treatment, 5-day-old seedlings were transferred to MS plates supplemented with 10 μM PAC. For GA₃ treatment, seeds were sown on MS plates, including 10 μM GA₃ and examined at day 5. For MG132 treatment, seeds were either sown on MS plates, including 10 μM MG132 or transferred at day 4 for short-term (24 hours) MG132 treatment.

Total RNA was isolated from 5-day-old roots and cDNA was synthesized as described previously (Koizumi et al., 2011). Semi-quantitative (q) RT-PCR was performed using standard PCR techniques. RNA levels were normalized to EIF4A. Real-time qRT-PCR was performed using Power SYBR Green PCR master mix (Applied Biosystems) and Step One Plus Real-time PCR System. RNA levels were normalized using UBQ10 (Sozzani, et al., 2010) as a standard. Primers used are listed in supplementary material Table S3.

Full-length cDNAs were used for all BiFC experiments with the exception of SHBY for which part of the cDNA that included the DUF1162 region (which was the region found to interact with SHR by Y2H) was amplified (supplementary material Table S3) from a root cDNA library by PCR and recombined into pDONR221 (Invitrogen/Life Technologies). Assays were performed in onion epidermal cells using pBatTL, BiFC vectors and procedures described elsewhere (Uhrig et al., 2007). A minimum of five fluorescent cells after a single bombardment was regarded as a positive interaction.

Using a modified version of the SHR protein that maintains the ability to move and rescue shr-2 mutants (Koizumi et al., 2011), yeast two-hybrid (Y2H) screens were conducted. The screening of two different Arabidopsis cDNA libraries identified 17 SHR interacting proteins (supplementary material Table S1). These proteins fell broadly into two major classes: regulation of transcription and membrane trafficking. Included in the group of transcriptional regulators were the verified SHR-interacting proteins SCR, JKD and MGP. From this full list of SHR-interacting proteins, we selected candidates for further analysis that had not been extensively characterized and had T-DNA insertion lines available. In addition, we performed bimolecular fluorescence complementation (BiFC) assays on a subset of these new proteins to verify protein interaction; five of the six proteins tested (not including SCR, JKD and MGP) showed an interaction in planta (supplementary material Table S1). Mutations in At1g03860, At2g02080, At3g46620, At4g29950 and At5g12430 caused no obvious root growth or cellular patterning defects, indicating that they are not essential for SHR function. As previously published, sec5a and sec5b single mutants appeared normal; attempts to make double mutants between sec5a and sec5b were unsuccessful (Hála et al., 2008). Therefore, from this screen we narrowed our focus to At5g24740, which had an obvious mutant phenotype (Fig. 1A; supplementary material Fig. S1), and showed interaction with both the SHR and SCR proteins (Fig. 1B).

At5g24740 is a single gene in Arabidopsis (TAIR BLAST AA) and is classified as a vacuolar sorting protein with chorein-N, ATG_C, multiple MRS6 homology and a DUF1162 domain (supplementary material Fig. S2A) (Bowers and Stevens, 2005; Kanehisa and Goto, 2000; Kanehisa et al., 2006; Kanehisa et al., 2010). These domains are found in vacuolar sorting proteins from yeast to human (Marchler-Bauer et al., 2011). Mutations in these proteins are associated with defects in intracellular trafficking, protein turnover and Cohen syndrome in humans (Velayos-Baeza et al., 2004).

Arabidopsis seedlings homozygous for mutations in At5g24740 all had increased branching and a shrubby appearance (supplementary material Fig. S1A,B). We therefore named At5g24740 as SHRUBBY. Plants heterozygous for shby (SHBY/shby-1, SHBY/shby-2, SHBY/shby-3 or SHBY/shby-4) were indistinguishable from wild type. The progeny of self-fertilized SHBY/shby-1 plants segregated 3:1 for the shby phenotype, indicating that the shby mutations are monogenic and recessive. All of the shby mutants developed a significantly shorter root than wild type (Fig. 1A; supplementary material Fig. S3A), with the shby-1 homozygotes being the shortest (Fig. 1A; supplementary material Fig. S3A). shby-1 seedlings were therefore used for a more comprehensive analysis of the shby phenotype. Overall, shby-1 plants were smaller than controls, had dark green curled leaves, late bolting (supplementary material Fig. S1C) and abnormal short-petaled flowers with the gynoecium extending from the unopened flower bud (supplementary material Fig. S1D). shby-1 plants were both male and female sterile. The stamens produced no visible pollen on their surface (supplementary material Fig. S1E), and even when the stigma was manually pollinated with wild-type pollen, no seeds were produced.

There are two predicted splicing variants of SHBY. The first contains the first 15 exons of the full-length cRNA with part of intron 15 (in supplementary material Fig. S2A from the ATG to the first stop codon). The second transcript contains all 46 exons that encode the 3462 amino acid protein. Real-time and semi-quantitative RT-PCR on shby-1 roots revealed no decrease in the signal upstream of the T-DNA insertion, but a significant decrease downstream of the T-DNA, suggesting that the T-DNA insertion in shby-1 resulted in a truncated mRNA (supplementary material Fig. S2B,D). Based upon the position of the insertions in each of the four shby alleles, it is unlikely that the shorter transcript is affected by any of these insertions and therefore that any of the four shby alleles are true nulls. However, using primers homologous to the 5′ end of the transcript and intron 15, which is absent in the longer transcript, no product was detected in either wild-type or shby-1 roots. Likewise, there were no significant differences in the expression of products amplified by primers complementary to the N terminus (1832-1946; supplementary material Fig. S2A,C) or to the C terminus (16668-16776; supplementary material Fig. S2A,C) in the roots of wild-type plants, suggesting that there is little or no expression of the short transcript in the root.

To examine the root growth phenotype in the shby-1 mutants, we measured meristem size, meristem cell number, and both cell size and cell production in shby-1 and wild-type roots (Fig. 1C,D; supplementary material Fig. S3B). Both the size of the meristem and the number of cortex cells in the meristem of shby-1 roots was reduced compared with wild type. The growth curves for the shby-1 and wild-type roots and root meristems were also different (supplementary material Fig. S3A,B). Over a 4-day period (from days 3-7) the wild-type meristem doubled in size and the number of cortex cells increased by 70%. This means that in wild type, new and longer cells were added to the meristem during the first week of root growth. In the shby-1 roots, the size of the meristem increased by only 10% over the same 4-day period and the number of cortex cells was increased by 33%. In contrast to wild-type roots, the average size of the cortex cells in the shby-1 meristem decreased over the first week of growth. To investigate whether the short root phenotype of shby-1 was caused by reduced cell expansion or cell division, or both, we conducted cell flux assays (Beemster and Baskin, 1998). Cell production in shby-1 was approximately half of what we measured in wild type (supplementary material Fig. S3B). The length of the mature cortex cells in shby-1 was also reduced (wild type=212.3±5.6 μm, n=10; shby-1=168.0±8.7 μm, n=10; t-test, P=0.005), but not to the same extent as cell production (supplementary material Fig. S3B). These results suggest that impaired root growth in shby-1 is the product of both reduced cell production and reduced cell expansion.

As the QC in shby-1 mutants appeared disorganized and defects in root growth often correlate with abnormal QC function, we examined expression of the QC markers, WUSCHEL RELATED HOMEOBOX 5 (WOX5), PLT1, PLT2, SHR and SCR in wild-type and shby-1 roots. To examine the QC cells, we counted the number of pWOX5:GFP-expressing cells in serial confocal sections in the meristem (Fig. 2A,B). In 5-day-old wild-type roots, we counted 4.6 (±0.2, n=18) GFP-positive cells on average, whereas in shby-1, there were 5.8 (±0.3, n=17, t-test, P=0.004). These results are consistent with cell divisions in the QC leading to a weak expansion of the WOX5-expression domain, which is associated with a reduction in QC identity (Sarkar et al., 2007). These results suggest QC defects in the shby-1 mutants.

We next examined the expression of pPLT1:PLT1-YFP and pPLT2:PLT2-YFP in wild-type and shby-1 mutants. Both PLT1 and PLT2 act in a concentration-dependent manner in the root meristem to promote stem cell fate and cell divisions (Galinha et al., 2007). The levels of pPLT1:PLT1-YFP and pPLT2:PLT2-YFP were markedly reduced in shby-1 compared with wild type (Fig. 2C-F). These results suggest that loss of SHBY activity affects either the expression and/or stability of PLT1 and PLT2. However, the overall domain of PLT1 and PLT2 expression was not altered in shby-1. We also examined both SCR and SHR expression specifically in the QC and found that the levels were reduced (Fig. 2G-J). This was more pronounced for SCR-GFP, than for SHR-GFP, but in both cases the reduction was statistically significant compared with the levels in wild-type roots. Collectively, these results show that SHBY is required to maintain SHR, SCR, PLT1 and PLT2 at wild-type levels in the QC, and therefore suggest a role for SHBY in meristem maintenance.

In addition to regulating the meristematic activity of the root, both SHR and SCR play roles in root patterning (Di Laurenzio et al., 1996; Helariutta et al., 2000). We therefore examined the patterning of the ground tissue in the shby-1 root compared with wild type Wild-type roots contain single layers of endodermis and cortex, each of which is composed of eight cells (Fig. 3A; Fig. 6A) (Dolan et al., 1993). The shby-1 mutant root contained additional ground tissue layers and an increase in the number of cells in each layer (Fig. 3B,C; Fig. 6B). This increase in the number of cell layers was not uniform throughout the root, leading to different numbers of layers in different regions of the root (Fig. 3B,C). In addition, based upon morphology, the extra cell layers in shby-1 arose from either the cortex (Fig. 3B) or the endodermis (Fig. 3C). Expression analysis of early markers of cortex (pCo2:RE3xYFP) and endodermis (pEn7:erGFP) showed that expression of pEn7:erGFP was not maintained outside of the endodermis, whereas pCo2:RE3xYFP was often detected strongly in more than one cell layer. These results indicate that the extra cell layers in shby-1 adopt a cortex cell identity (Fig. 3D-G) (Heidstra et al., 2004).

In the root meristem, the SHR protein is expressed in the stele and moves into the endodermis. In the stele, the SHR protein localizes both to the nucleus and cytoplasm, whereas in the endodermis the protein is exclusively localized in the nucleus (Fig. 4A) (Nakajima et al., 2001; Gallagher et al., 2004; Cui et al., 2007; Gallagher and Benfey, 2009). To examine SHR-GFP in the shby-1 mutants, we examined levels of GFP fluorescence in both the pSHR:SHR-GFP and pSHR:erGFP fusions in wild type and shby-1 mutants. Overall, the patterns of pSHR:SHR-GFP and pSHR:erGFP expression did not differ between wild-type and shby-1 roots (Fig. 4A-D). However, compared with wild type, the expression levels of pSHR:erGFP and the pSHR:SHR-GFP in shby-1 were 80% (t-test, P=0.005; wild type, n=15; shby-1, n=17) and 60% (t-test, P=0.0008; wild type, n=20; shby-1, n=18) of wild-type levels, respectively, for the same transgenes. We saw no obvious changes in the subcellular localization of SHR-GFP in either the stele tissue or the endodermis in the shby-1 roots compared with wild type. These results suggest that SHBY affects SHR both at the level of the protein and at the level of the transcript, but not protein localization.

To test whether shby-1 has any affect on the movement of SHR from the stele into the endodermis, we measured the level of the SHR-GFP signal in the endodermis as a percentage of the stele signal, as this is a good indication of SHR movement (Gallagher and Benfey, 2009; Koizumi et al., 2011). In wild-type roots, the SHR-GFP signal in the endodermis is generally 1.2-1.5, depending on the age of the root and the growth conditions. When we examined the ratio in shby-1 roots and their wild-type siblings, we saw no significant differences between these two groups. The average ratio of endodermal signal to stele signal was 1.2 in wild type (n=27) and 1.29 in shby-1 (n=31), indicating that SHBY does not play an essential role in SHR movement.

To examine SCR expression in shby-1 mutants, we crossed the pSCR:erGFP and pSCR:SCR-GFP markers lines into shby-1 mutants. In wild type, SCR:erGFP is highly expressed in the endodermis (Fig. 4E). SCR-GFP expressed from the same promoter is found at high levels in the endodermis and to a lesser extent in the cortex (Fig. 4G) (Gallagher et al., 2004; Wysocka-Diller et al., 2000). In shby-1 roots, SCR:erGFP was also expressed in the endodermis, but at only 55% of what we saw in wild type (Fig. 4F) (t-test, P=0.009; wild type, n=8; shby-1, n=8). By contrast, the levels of pSCR:SCR-GFP in the endodermis of shby-1 were not different than wild type (Fig. 4G,H) (t-test, P=0.596; wild type, n=15; shby-1, n=19), so although expression levels were decreased based upon the pSCR:erGFP results, the protein levels were not. In addition, in shby-1 roots, SCR-GFP was frequently detected at higher levels in the cortex relative to the endodermis when compared with wild-type roots (t-test, P=0.02; shby-1, n=18; wild type, n=10; Fig. 4H shows an example of this). This effect on SCR-GFP was particularly true in regions of the root where abnormally oriented or ectopic divisions had occurred. In pSCR:erGFP lines, erGFP was present in the middle ground tissue layer immediately following division of the endodermis, but was not maintained (Fig. 4F). Likewise following the division that produces the endodermis and cortex layers, pSCR:erGFP was not maintained in the cortex. As the SCR protein was found at higher than wild-type levels in the cortex and middle ground tissue layer, these results suggest that SCR is not efficiently lost following the asymmetric division that generates the endodermis and cortex or the middle ground tissue layer in the shby-1 mutants.

Sozzani et al. (Sozzani et al., 2010) reported that the Arabidopsis D-type cyclin CYCD6;1 was a direct target of SHR and SCR, and that this cyclin promoted the formative periclinal cell division that generates the endodermis and cortex, and the divisions that pattern the MC. In 5-day-old roots, pCYCD6;1:GFP-GUS was specifically expressed in the CEI and the CEI-daughter cells (CEDs). By day 10, the expression of CYCD6:1 was expanded into the endodermis (Sozzani et al., 2010). To understand the relationship between SHBY and CYCD6;1, we examined the expression pattern of pCYCD6;1:GFP-GUS in shby-1. Under our growth conditions, pCYCD6;1:GFP-GUS was expressed in the CEI and CED cells of 5-day-old seedlings (Fig. 5A), and weakly and sporadically expressed in some cells of the endodermis in 7-day-old seedlings (Fig. 5G). In the root meristem of 5-day-old shby-1 mutants, GFP was weakly detected in the CEI and CED cells, moderately expressed in the endodermis, and strongly expressed in regions of the endodermis that showed amplification of the ground tissue (Fig. 5B). These result indicate that SHBY plays a role in inhibiting CYCD6;1 expression in the endodermis.

Recently, Cruz-Ramirez et al. (Cruz-Ramirez et al., 2012) reported that treatment of roots with the proteosome inhibitor MG132 resulted in increased SCR, ectopic expression of CYCD6;1 and an increase in the frequency of periclinal cell divisions in the ground tissue. As this phenotype is very similar to shby-1, and SHBY is a putative vacuolar-sorting protein, we tested whether we could phenocopy the loss of SHBY by growing seedlings on media containing low levels (10 μM) of MG132 for 5 days, or short-term for 24 hours. Similar to what was reported by Cruz-Ramirez et al. (Cruz-Ramirez et al., 2012), both treatments showed a significant increase in the number of ground tissue layers (Fig. 5C; Fig. 6C; supplementary material Table S2), which correlated with activation of CYCD6;1 expression in the endodermis. No further increase in ground tissue layers was seen in the shby-1 mutants treated with MG132 (supplementary material Table S2). In contrast to Cruz-Ramirez et al. (Cruz-Ramirez et al., 2012), we saw no significant effects of either the 5-day or 24-hour treatments with MG132 on the level of SCR-GFP in the endodermis (P=0.93 and P=0.30, respectively, t-test; control, n=10; 10 μM MG132, n=10). However, in endodermal cells that had undergone MG132-induced periclinal cell division, SCR-GFP was often retained in the middle ground tissue layer (examples shown in Fig. 5D-F), indicating that SCR is not efficiently degraded. MG132 also decreased root growth and caused a general disorganization of the cellular patterning of the QC (boxed region in Fig. 5C). This correlated with an increase in the domain of pWOX5:GFP expression (supplementary material Fig. S4A,B) and a reduction in the levels of both PLT1 and PLT2 in the root meristem (supplementary material Fig. S4C-F) (reduced by 36% and 32%, respectively, at 24 hours; 12 roots were analyzed for each treatment; and by 39% and 45%, respectively, at day 5; PLT1 control, n=10; 10 μM MG132, n=10; PLT2 control, n=10; 10 μM MG132, n=10; P<0.01). These results show that many of the shby-1 phenotypes can be reproduced by reducing protein turnover.

As shby-1 mutants have an increase both in the number of ground tissue layers and changes in SHR and SCR levels, we examined genetic interactions between shby-1 (Fig. 6B), shr-2 (Fig. 6D) and scr-4 (Fig. 6H). For most of the shby-1 double mutant combinations, the roots were marginally but significantly shorter than the single mutants alone (supplementary material Fig. S3E), indicating additive effects on root growth. For the shby-1 shr-2, this correlated with a reduced meristem cell number compared with the corresponding single mutants (Fig. 1D-H; supplementary material Fig. S3C,D).

shr-2 and scr-4 roots make only one ground tissue layer of approximately eight cells. shr-2 fails to make MC, whereas scr-4 does so precociously. The shby-1 shr-2 double mutants showed three different phenotypes in roughly equal proportions. A third of roots had a shr-2 phenotype (Fig. 6E), whereas the other two-thirds showed either an increase in anticlinal (Fig. 6F) or an increase in both anticlinal and periclinal cell divisions (Fig. 6G) compared with the shr-2 single mutant (Fig. 6D), but still significantly fewer cell divisions (particularly periclinal cell divisions) than in shby-1 (supplementary material Fig. S3F). In the shby-1 scr-4 double mutants, the shby-1 mutation increased the number of cell layers in the scr-4 mutants (Fig. 6H,I). shby-1 scr-4 mutants had two layers of ground tissue instead of one, indicating that SCR is not required for the ectopic periclinal cell division in shby-1 mutant; however, the number of ground tissue cells in the shby-1 scr-4 double mutant was significantly reduced compared with the shby-1 single mutant (supplementary material Fig. S3F). These results suggest that both SHR and SCR promote ectopic or precocious cell divisions in shby-1, but that neither alone is sufficient for these cell divisions.

As shby-1, shr-2 and scr-4 do not show clear epistatic relationships, we examined SHBY expression in shr-2 and scr-4 mutants. Whole-mount in situ hybridization with the SHBY sense and antisense probes in 5-day-old wild-type seedlings showed strong expression of SHBY in all the cells of the root meristem extending into the elongation and differentiation zones (supplementary material Fig. S5C). In shr-2 and scr-4 roots, the expression pattern of SHBY was fairly similar to wild type; however, the levels of expression were reduced with scr-4 roots showing levels of SHBY expression that were intermediate between wild type and shr-2 (supplementary material Fig. S5C,D). If SHBY levels are decreased in shr-2 and scr-4 backgrounds, then based upon our results, PLT1 and PLT2 should also be decreased in shr-2 and scr-4 backgrounds. Indeed, both mutants showed reduced PLT1-GFP (49% and 53% of wild type in shr-2 and scr-4, respectively) and PLT2-GFP (47% and 37% of wild type in shr-2 and scr-4, respectively) expression (supplementary material Fig. S4G-J). These results suggest that SHBY provides a link between the SHR/SCR and PLT pathways.

Although the shby-1 phenotype was partially phenocopied by treating roots with MG132, the overall phenotype of the shby-1 mutants [dwarfed, bushy plants with dark green leaves and delayed flowering (supplementary material Fig. S1)] indicates defects in GA signaling. Likewise, the phenotype of the shby-1 scr-4 roots (Fig. 6I) is reminiscent of the scarecrow-like 3 (scl3) scr-5 double mutant phenotype reported by Heo et al. (Heo et al., 2011). SCL3 promotes GA signaling in the regulation of ground tissue patterning. To determine whether SHBY functions in the GA pathway, we examined the shby-1 phenotype under three different conditions that affect GA signaling. First, we made double mutants between scl3-1 and shby-1, and then tested the sensitivity of the shby-1 mutants to PAC (an inhibitor of GA biosynthesis) treatment and addition of GA₃ to the growth medium. In otherwise wild-type roots, mutations in SCL3 increase the frequency of periclinal cell divisions in the ground tissue (Heo et al., 2011); however, not quite to the same degree as shby-1 (Fig. 6J; supplementary material Fig. S3F). In shby-1 scl3-1 double mutants, the number of ground tissue cells was moderately increased compared with scl3 single mutants, but decreased compared with shby-1 (Fig. 6K; supplementary material Fig. S3F), suggesting that SHBY pathways partially overlap with SCL3. To test the effects of inhibition of GA, we treated wild-type and shby-1 mutants with PAC, and examined ground tissue patterning. PAC had no effect on shby-1 (supplementary material Table S2), indicating that shby-1 is insensitive to PAC. Likewise, treatment with GA₃ also failed to elicit a response in shby-1 mutants, indicating that SHBY is required for GA₃-induced signaling.

The above results suggest that shby-1 mutants have reduced GA signaling. To test whether we could phenocopy the effects of shby-1 by inhibiting GA, we treated wild-type seedlings with PAC. The expression of pPLT1:PLT1-YFP and pPLT2:PLT2-YFP were not repressed by PAC as they are in the shby-1 mutants; however, treatment with GA₃ caused moderate upregulation of pPLT2:PLT2- YFP (supplementary material Fig. S4K,L), suggesting that at least PLT2 is responsive to GA. As previously published, the shr-2 mutants were insensitive to PAC (Paquette and Benfey, 2005). Interestingly, treatment of scr-4 seedlings with 2.0 μM PAC phenocopied the shby-1 scr-4 double mutant (Fig. 6L; supplementary material Table S2): there were now two ground tissue layers in the PAC-treated scr-4 roots. In addition, treatment of the CYCD6;1:GFP-GUS roots with PAC caused an increase in expression of the transgene (Fig. 5H). PAC treatment had no effect on cell divisions in the cycd6;1 lines (supplementary material Table S2), indicating that PAC exerts its effect on ground tissue patterning via CYCD6;1.

As the shby-1 mutants resemble GA-deficient mutants and are insensitive to GA₃ application (supplementary material Table S2), we attempted to rescue the shby-1 phenotype downstream of GA by crossing into the shby-1 background the recessive GA INSENSITIVE (gai-2) allele, which allows GA signaling in the absence of GA. We examined 72 F2 seedlings and recovered the expected numbers of gai-2 and shby-1 single mutants, but never recovered the gai-2 shby-1 double mutant. In an F2 population of 471, there was a total of 28 ungerminated seeds, which could represent the gai-2 shby-1 double mutants; however, attempts to genotype these seeds were unsuccessful. Collectively, these results show that PAC can partially phenocopy the shby-1 patterning phenotype and suggest that SHBY promotes GA signaling downstream of GA synthesis, as shby-1 mutants are insensitive to application of GA₃.

In the present study, we identified 17 SHR-interacting proteins, including the verified SHR-interacting proteins SCR, JKD and MGP (Cui et al., 2007; Welch et al., 2007). In the group of other proteins, we were able to verify the interaction of SHR with five of the six tested. However, none of the single insertion lines for these genes (with the exception of SHBY) had a single mutant phenotype. Among the SHR-interacting proteins was SHBY, a vacuolar sorting protein with similarity to VPS13 (COH1) (Marchler-Bauer et al., 2011). Velayos-Baeza et al. (Velayos-Baeza et al., 2004) reported homology in the N terminus between SHBY and yeast VPS13p. The human homologs of VPS13p are mutated in chorea-acanthocytosis and Cohen syndrome (Kolehmainen et al., 2003; Rampoldi et al., 2001). Functional analysis of VPS13b in yeast showed that it is involved in the trafficking of membrane proteins between the trans-Golgi network and the prevacuolar compartment/late endosome (Brickner et al., 2001; Redding et al., 1996), suggesting roles in protein recycling.

We isolated and characterized four different alleles of shby and found that SHBY is required for normal root growth. The roots of shby mutants were consistently shorter than wild type. Analysis of the shby-1 mutant showed that the reduced root growth was the result of reduced cell elongation and reduced cell production. Part of the meristem defect in shby-1 can be explained by a decrease in the levels of SHR and SCR in the QC. However, as the shr-2 and scr-4 mutants are not clearly epistatic to shby-1 with respect to root growth and meristem cell number, other pathways must also be affected in shby-1. Consistent with this, we see a strong decrease in PLT1 and PLT2 levels in the root meristem of shby-1 mutants. The SHR/SCR and PLT pathways represent two partially independent pathways that are required to maintain the function of the stem cell niche in the Arabidopsis root. The reduction of PLT1 and PLT2, and of SHR and SCR in the QC of the shby-1 mutants suggests that SHBY acts in both pathways to maintain expression at wild-type levels. Consistent with its role in root growth and meristem maintenance, SHBY shows high expression in the root meristem in a domain that overlaps with PLT1, PLT2, SHR and SCR.

Previous results by Aida et al. (Aida et al., 2004) suggested that the expression of PLT1 mRNA was moderately decreased in both the shr and scr mutants. We were able to confirm these results at the level of the protein for both PLT1 and PLT2 using the PLT-GFP marker lines. Interestingly, we find that the levels of SHBY expression are also reduced in both the shr-2 and scr-4 mutants. As PLT1 and PLT2 expression were both reduced in shby-1, a decrease in SHBY in the shr and scr mutants may explain the reduced PLT1 and PLT2 expression. In this scenario, the reduced levels of SHBY in the shr-2 and scr-4 mutant backgrounds result in the observed decrease in PLT1 and PLT2. Another possibility is that the reduction in SHR in shby-1 directly affects PLT levels. However, we think that this is unlikely, as the level of SHR in the shby-1 mutant was 60% of the wild type and plants heterozygous for shr-2 have no obvious meristem maintenance phenotypes (Koizumi et al., 2012). Therefore, SHBY may provide a link between the SHR/SCR and the PLT pathways in maintenance of root growth.

The other obvious phenotype of the shby-1 roots is the production of extra ground tissue layers, as well as extra cells in each layer. The ectopic periclinal cell divisions seen in shby-1 mutants could point to a misregulation of MC formation. In wild-type plants, roots begin making MC at around days 7-14. In these roots, MC forms first in endodermal cells that lie in the xylem axis and then proceeds circumferentially until all of the endodermal cells have divided (Baum et al., 2002; Paquette and Benfey, 2005). The periclinal cell divisions in the shby mutants are present by day 3 and they do not follow a regular pattern; instead, they form randomly, which indicates that in shby-1 roots the patterning of MC may be affected. However, the effects of shby-1 on the patterning of the ground tissue are not limited to the endodermis; they also occur in the cortex and are either periclinal or anticlinal, which is inconsistent with formation of MC. In addition the formation of MC is entirely dependent upon functional SHR. This was not true for the cell divisions in shby-1. Based upon the double mutant analysis, neither SHR nor SCR was essential for the divisions in shby-1, although both the shr-2 and scr-4 mutations reduced the number of ground tissue layers in shby-1. Collectively, these results suggest that the role of SHBY in patterning of the ground tissue extends beyond only the regulation of MC and beyond SHR and SCR.

Although SHR was restricted to the endodermis and we saw no expression of pSCR:erGFP outside of the endodermis, there was an increase in SCR-GFP in the cortex cell layer, particularly in the middle ground tissue layer following periclinal cell division. As SHBY is a putative vacuolar-sorting protein that interacts with SCR, this may reflect a reduction in the normal degradation of SCR following periclinal cell divisions in the ground tissue of shby-1 roots. Consistent with a role for SHBY in protein turnover, there was an overall decrease in SCR transcription in shby-1 without a concomitant decrease in SCR protein levels. Likewise, we could phenocopy the effects of shby-1 on SCR-GFP levels outside of the endodermis by treating roots with the proteosome inhibitor MG132. Interestingly, in the MG132-treated roots there were ectopic divisions in the ground tissue, and expression of the pCYCD6;1:GFP-GUS marker in the ground tissue beyond the initial cells and their immediate daughters; this reflects what we observed in shby-1 roots. Treatment of wild-type roots with MG132 also phenocopied the effects of shby-1 on PLT1 and PLT2 levels, and the QC defects. These results are consistent with shby-1 playing a role in protein degradation in the root meristem and in the regulation of cell division and QC maintenance.

Both the aerial and the root phenotypes of shby seedlings suggested defects in GA synthesis, perception or signal transduction (Ariizumi and Steber, 2007; Ariizumi et al., 2008; Heo et al., 2011; Tyler et al., 2004; Wen and Chang, 2002; Zhang et al., 2007). In the root, GA acts in a partially overlapping pathway with SCR to inhibit asymmetric cell divisions in the ground tissue. In the meristem, GA promotes auxin signaling which maintains the meristem. To test whether shby-1 seedlings are defective in GA signaling, we treated the plants with PAC or GA and found that shby-1 is insensitive to both treatments and that aspects of the shby phenotype could be induced in wild-type roots by treatment with PAC. These results suggest that SHBY acts downstream of GA biosynthesis either in GA perception or in GA signaling. The precise function of SHBY in the GA pathway remains to be elucidated. However, SHBY provides a connection between GA, SHR and PLT signaling in the root meristem.

We thank R. Schlarp, K. Miyahara and J. Ugochukwu for technical assistance; R. Sozzani for providing the pCYCD6;1:GUS-GFP seeds; T. P. Sun for providing the gai-2 seeds; and R. Sozzani, C. Perin and D. Wagner for critical comments on the manuscript. The Salk Institute Genomic Analysis Laboratory provided the sequence-indexed Arabidopsis T-DNA insertion mutants. N. Matsumoto made the pSHR:erGFP line.