Dysregulation of cell-to-cell connectivity and stomatal patterning by loss-of-function mutation in Arabidopsis CHORUS (GLUCAN SYNTHASE-LIKE 8)

The generation of tissue patterns in multicellular organisms requires coordinated cell proliferation and differentiation. Encapsulated by cell walls, pattern formation in plants occurs in the absence of cell migration. The last step of plant cell division, cytokinesis, is marked by a cell plate separating the two daughter cells. It is thought that positional information instructs the newly divided plant cells to acquire specific cell fate. One mechanism for transmitting and receiving such information is receptor kinase-mediated signaling, in which secreted ligands diffusing through the apoplast are received by receptors in neighboring cells. Plants possess large numbers of receptor-like kinases (RLKs) to regulate diverse aspects of development (Dievart and Clark, 2004; Shiu and Bleecker, 2001; Torii and Clark, 2000). The second mechanism is direct cell-to-cell movement of regulatory molecules, such as transcription factors and RNAs, through the symplastic continuum via plasmodesmata (Lucas and Lee, 2004; Oparka and Roberts, 2001).

Stomatal patterning has emerged as a model in which to study how positional cues influence functional tissue patterning. In Arabidopsis, stomata differentiate from a subset of protodermal cells called meristemoid mother cells (MMCs). An MMC undergoes an asymmetric entry division that creates a meristemoid and a larger sister cell called a stomatal lineage ground cell (SLGC). The meristemoid reiterates a few rounds of asymmetric cell divisions that amplify surrounding SLGCs. The meristemoid eventually differentiates into a guard mother cell (GMC) that divides symmetrically to form two guard cells constituting a stoma (Bergmann and Sack, 2007; Nadeau and Sack, 2002b). The SLGCs might differentiate into pavement cells or become MMCs, which initiate additional stomatal cell lineages.

Molecular genetic studies have revealed both positive and negative regulators of stomatal development. Three closely-related basic helix-loop-helix (bHLH) proteins, SPEECHLESS (SPCH), MUTE and FAMA, specify the initiation of the stomatal lineage, differentiation of meristemoids to GMCs, and the terminal differentiation of guard cells, respectively, together with SCREAM (SCRM; also known as ICE1) and SCRM2 bHLH proteins (Kanaoka et al., 2008; MacAlister et al., 2007; Ohashi-Ito and Bergmann, 2006; Pillitteri et al., 2008; Pillitteri et al., 2007). Two Myb proteins restrict the division of GMCs (Lai et al., 2005). A ligand-receptor signaling pathway inhibits entry into the stomatal lineage and orients asymmetric division. The components of the pathway include putative ligands, EPIDERMAL PATTERNING FACTOR 1 (EPF1) and EPF2, a putative receptor, TOO MANY MOUTHS (TMM), three redundant receptor-like kinases, ERECTA (ER), ERECTA-LIKE 1 (ERL1) and ERL2, and mitogen-activated protein kinase cascades mediated via YODA, MKK4/5 and MPK3/6 (Bergmann et al., 2004; Hara et al., 2007; Hara et al., 2009; Hunt and Gray, 2009; Nadeau and Sack, 2002a; Shpak et al., 2005; Wang et al., 2007). It remains unclear how the actions of these positive and negative regulators are spatially coordinated during epidermal development.

To better understand the regulation of stomatal patterning, we performed a sensitized genetic screen using er erl1 as a starting background. We report a new stomatal patterning mutant, chorus (chor). CHOR represents a weak allele of the putative callose synthase gene GSL8 (also known as CALLOSE SYNTHASE 10), which was recently shown to be required for cytokinesis (Chen et al., 2009; Thiele et al., 2009). Although Chen et al. also observed that gsl8 mutants show stomatal patterning defects, both the exact nature of the defects and the mechanism by which loss-of-function mutations in a callose synthase gene leads to stomatal clustering remain unclear. On the basis of histochemical callose deposition and cell-to-cell diffusion assays, we suggest a potential role of plasmodesmatal macromolecular trafficking as a mechanism for proper stomatal patterning.

Arabidopsis thaliana Columbia (Col) accession was used as a wild type. Mutants and lines used in this study were in the Col background unless otherwise specified. The chor mutant was derived from an EMS (ethyl methanesulfonic acid)-mutagenized er-105 erl1-2 population and was outcrossed three times to wild type before analysis. The er-105 erl1-2 background possesses an additional gl1 mutation, which confers trichomeless leaves. A T-DNA insertion allele of GSL8 (gsl8-2; GK 851 C04) was obtained from the Nottingham Arabidopsis Stock Centre (NASC). The following transgenic lines and mutants have been described previously: proERL1::GUS, proTMM::GUS, er-105, erl1-2, er-105 erl1-2 and er-105 erl1-2 erl2-1 (Shpak et al., 2004; Shpak et al., 2005); spch-3 (Pillitteri et al., 2007); tmm-1 and proTMM:GUS-GFP (Nadeau and Sack, 2002a); proMUTE::GFP (Pillitteri et al., 2008); proCaMV35S:YFP-PDCB1 (Simpson et al., 2009); atnack1-1 (Nishihama et al., 2002); and scd1-2 (Falbel et al., 2003). The transgenes and known mutant alleles were introduced into chor/+ via genetic crosses. Seedlings were grown in modified Murashige-Skoog media with 1×Gamborg B5 vitamins and 1% w/v sucrose at 21°C under long-day conditions (18/6 light-dark cycle). Plates (90 mm diameter, 25 mm thickness) were sealed with surgical tape (Micropore, 3M) to minimize condensation.

er-105 erl1-2 seeds were treated with 0.3% (v/v) EMS overnight. Seeds were harvested from individual M1 plants, and 528 M2 seed populations were visually screened for seedling epidermal phenotypes.

Mutant genotypes were verified by PCR-based analyses. The chor mutant allele was detected by derived Cleavage Amplified Polymorphic Sequence (dCAPS) (Neff et al., 1998), using the primer pair chorus dCAPS F-NlaIV and chorus dCAPS R-NlaIV, and subsequent digestion with NlaIV. The T-DNA insertion site of gsl8-2 was detected by the primer pair At2g36850 M18F and GABI-K o8409. See Table S1 in the supplementary material for primer sequences. Genotyping of er-105, erl1-2, erl2-1, tmm-1 and spch-3 has been described previously (Lukowitz et al., 2004; Pillitteri et al., 2007; Shpak et al., 2004).

A mapping population of chor was generated by crossing B1F1 chor/+ er erl1 to Ler. Initially, a bulked segregant analysis (Lukowitz et al., 2000) was performed using a combined pool of 33 chor seedlings from the mapping population, which revealed the strong linkage of chor to the mid-bottom of chromosome 2 at SSLP markers MSAT2.36 and MSAT2.9 (Loudet et al., 2002). Further fine mapping was performed within this interval using two SSLP markers (KTL T20F21 and KTL F3G5) and a total of 1831 F2 chor (−/−) seedlings, which located CHOR within the ~145-kb interval encompassing the BACs F13K3 and T1J8. All 36 predicted open-reading frames within this interval were sequenced, and this led to the identification of a single G→A substitution at position 13447 of At2g36850, which encodes GSL8 (CalS10). The SSLP, CAPS, and dCAPS markers for map-based cloning of CHOR were designed based on the information available in Monsanto Arabidopsis Polymorphism and Ler Sequence Collection (http://www.Arabidopsis.org/Cereon/index.jsp). For primer sequences and amplified fragment sizes in Col and Ler, see Table S1 in the supplementary material. For the list of sequencing primers, see Table S2 in the supplementary material.

Total RNA was isolated from 14-day-old wild-type and chor seedlings. cDNA synthesis and reverse-transcription (RT)-PCR were performed as described previously (Bemis and Torii, 2007), using the following primer pairs. For CHOR and chor, chorus cDNA F (5′-CAGCTTCATTACAATGCAATTCC-3′) and chorus cDNA R (5′-TCAAGTTACCAAAATATAGTCAACC-3′); for Actin control, ACT 2-1 (5′-GCCCATCCAAGCTGTTCTCTC-3′) and ACT 2-2 (5′-GCTCGTAGTCAACAGCAACAA-3′); for sequencing RT-PCR products, chorus cDNA Seq F (5′-GCTCTACTGTTGATCATTTACA-3′).

Differential interference contrast (DIC) microscopy was performed according to Bemis and Torii (Bemis and Torii, 2007). For confocal microscopy, seedlings were treated with either propidium iodide (PI) or FM4-64 (Molecular Probes, Invitrogen). Fluorescence of GFP, YFP and PI/FM4-64 were simultaneously captured using a Zeiss LSM510 Meta or a Zeiss LSM700 (excitation, 488 nm and 543 nm; emission, 505-530 nm, 545-585 nm and 570-610 nm, respectively) with a 20× objective lens. For histochemical callose staining, seedlings were infiltrated with 0.1 mg/ml Aniline Blue fluorochrome (Biosupplies Australia) in 0.1% Silwett L77 (Lehle Seeds, Round Rock, TX) for 60 minutes and, prior to mounting, wild type and chor were counterstained briefly with FM4-64 to highlight the cell periphery. The abaxial side of rosette leaves was observed with a Zeiss LSM700 (excitation, 405 nm; emission, 475-500 nm). Aniline Blue-stained root hairs were observed with an Olympus BX51 epifluorescence microscope with a UV/B filter.

Twelve-day-old seedlings of wild type, chor, er erl1 and er erl1 chor were subjected to DIC microscopy. Four non-overlapping frames of cotyledons, excluding the mid-vein region, and numbers of stomata and their cluster size were quantified using ImageJ.

Arabidopsis 10-day-old seedling rosette leaves were transiently transformed as previously described (Kanaoka et al., 2008). Briefly, microcarriers were prepared with 100 ng of plasmid containing proCaMV35S::eCFP (von Arnim et al., 1998), a combination of proCaMV35S::eCFP and proCaMV35S::3xGFP (J.S.L. and K.U.T., unpublished), or a combination of proCaMV35S::eCFP and proCaMV35S::SPCH-YFP (Lynn Pillitteri and K.U.T., unpublished). These plasmids were coated on 1.0-μm gold particles and bombarded using the PDS-1000/He particle delivery system (Bio-Rad, Hercules, CA, USA). Specimens were observed under a Zeiss LSM 510 Meta or an LSM700 confocal microscope. CFP, GFP or YFP fluorescence was monitored 6 hours after the bombardment.

The GenBank Accession number for the GSL8 (CalS10) mRNA sequence is GQ373182.

We conducted a genetic screen using a sensitized er erl1 background, in which two of the three ER-family genes are absent. er erl1 plants develop an epidermis with an elevated stomatal density but are otherwise normal (Fig. 1) (Shpak et al., 2005). On rare occasions, paired stomata were observed both in wild-type and er erl1 cotyledons (0.11% in wild type; 0.57% in er erl1; Fig. 1E). Neither genotype developed stomatal clusters with three or more adjacent stomata (n=0/1771 and 0/1921 for wild type and er erl1, respectively; Fig. 1E). By contrast, er erl1 erl2 triple loss-of-function mutants develop excessive numbers of clustered stomata (Shpak et al., 2005). We reasoned that in the absence of two of the three receptors, plants would be more sensitive to the loss of any additional regulators of stomatal patterning.

We isolated a line (#67). In the mutant leaves, multiple stomata are often formed in a compact aggregate with stomatal pairs (15.8%), trios (7.26%), quartets (2.19%), and 5-10 adjacent stomata (1.95%; Fig. 1B,E). In addition, aggregates of small epidermal cells, many resembling meristemoids, are formed in sectors across the epidermis (Fig. 1B). Segregation analysis showed that the mutation is recessive and monogenic (207:602, χ²=0.149, P=0.700). The mutant was named chorus (chor) owing to its aggregated stomatal phenotype.

We introduced chor into the wild-type background to test whether the observed phenotype requires er erl1. The epidermal phenotype of chor leaves in the wild-type background was much milder than that of chor er erl1. Numbers of stomatal pairs (9.4%), trios (1.93%) and quartets (0.44%) were diminished. No stomatal clusters larger than quartets were found in chor (n=0/715; Fig. 1D,E). Similarly, fewer small epidermal cells were observed (Fig. 1D). Thus, chor phenotypes are enhanced by the loss of function in ER and ERL1. chor seedlings showed pleiotropic phenotypes, including severe dwarfism, short roots with disrupted cell files, short and branched root hairs, and seedling lethality (Fig. 1F; see also Figs S1 and S2 in the supplementary material). Furthermore, incomplete cytokinesis was frequently observed in chor (see Fig. S3 in the supplementary material). These findings suggest that CHOR is required for proper cell division and tissue patterning throughout plant organs, including stomatal patterning.

The chor leaf epidermis exhibited incomplete cytokinesis and excessive epidermal cell proliferation (Fig. 1). We hypothesized that the latter defect is due to excessive entry into the stomatal lineage. To test this, we examined the expression patterns of stomatal-lineage markers in chor. TMM marks early stomatal-lineage cells, including MMCs, meristemoids and SLGCs (Fig. 2A). Clusters of small, highly divided epidermal cells in chor seedlings exhibited high reporter activity of proTMM::GUS-GFP (Fig. 2B,C), suggesting that the chor mutation causes a vast increase in stomatal lineage divisions in a disorganized manner. MUTE is specifically expressed in late meristemoids that are differentiating into GMC (Fig. 2D) (Pillitteri et al., 2008; Pillitteri et al., 2007). Occasionally, multiple adjacent cells showed high MUTE promoter activity (Fig. 2F, asterisks), which is likely to predict the eventual formation of stomatal clusters.

The excessive formation of small stomatal-lineage cells is a hallmark of SPCH overexpression (MacAlister et al., 2007; Pillitteri et al., 2007). We therefore tested whether SPCH is ectopically expressed in chor, and whether SPCH is required for the chor epidermal phenotype. SPCH was expressed in early stomatal-lineage cells (Fig. 2G), and the spch mutant was devoid of stomatal cell lineages (Fig. 2J). Numerous SPCH-positive cells were observed in the young cotyledon epidermis of chor (Fig. 2H,I). Introduction of spch into chor eliminated all small, stomatal-lineage cells (Fig. 2K,L). By contrast, incomplete cytokinesis of pavement cells still occurred in the chor spch double mutant (Fig. 2K,L, arrowheads). These findings point to two independent roles of CHOR in regulating leaf epidermal development: SPCH-mediated initiation and division of stomatal cell lineage; and SPCH-independent cytokinesis. In support of this idea, the root epidermis of spch chor exhibited disrupted cell files with incomplete cytokinesis (see Fig. S1 in the supplementary material).

The stomatal clustering phenotype of chor in the wild-type background remained relatively mild (Fig. 1). Therefore, the TMM/ER-family signaling pathway might still be able to largely enforce stomatal spacing in chor. To test this, we investigated the genetic interactions of CHOR with TMM and ER-family genes (Fig. 3). Quantitative analysis of stomatal cluster size was performed for key genotypes (see Fig. S4 in the supplementary material). In tmm, adjacent stomata are formed (Fig. 3C) (Nadeau and Sack, 2002a). The tmm chor double mutant developed massive clusters of stomata far more severe than those in either the tmm or the chor single mutant (Fig. 3B-D; see also Fig. S4 in the supplementary material). Similarly, the chor mutation intensified the stomatal cluster phenotype of the er erl1 erl2 triple mutant (Fig. 3E,F).

We sought to understand whether CHOR acts in an independent pathway from (i.e. additive interaction) or an overlapping pathway with (i.e. synergistic interaction) the TMM/ER family. For this purpose, we genetically dissected the contribution of each ER-family member. The three ER-family genes exhibit unequal redundancy, and a stomatal cluster phenotype is manifested only when the entire family is lost (Figs 1, 3) (Shpak et al., 2005). er single mutants occasionally produce SLGCs that do not accompany differentiated stomata (Fig. 3G, dotted lines) (Shpak et al., 2005). The er chor epidermis formed severely divided patches of small cells (Fig. 3H, dotted brackets), but otherwise appeared similar to that of chor (Fig. 3H, arrowheads), thereby suggesting additive effects.

In contrast to er chor, er erl1 chor and er erl2 chor triple mutants exhibited severe stomatal clustering phenotypes (Fig. 1B; Fig. 3H,L; see also Fig. S4 in the supplementary material). We further tested whether such synergistic effects can be observed even when only a single copy of ERL1 is eliminated. Indeed, er erl1/+ chor also showed a strong stomatal clustering phenotype (Fig. 3J). We conclude that chor intensifies stomatal cell-cell signaling defects and breaks down the functional redundancies among ER-family genes, with sensitivity to the dosage of ERLs.

To understand the molecular nature of CHOR, we performed map-based cloning. Our fine mapping delineated the CHOR locus within a 145-kb interval on chromosome 2 (Fig. 4A). Sequencing the interval led to the identification of a single G-to-A substitution at position 13447 of At2g36850, which encodes a putative callose synthase, GSL8 (CALS10) (Fig. 4B). Because the original chor mapping population contained a mixed genome of Col/WS and Ler, wild-type GSL8 alleles from WS and Ler were also sequenced. Both alleles possess G at this position, confirming that CHOR is not a naturally occurring polymorphism.

Sequence analysis of the wild-type GSL8 transcripts revealed mis-annotation of GSL8 in the existing database (see Materials and methods). Based on the experimental data, GSL8 is a large gene consisting of 50 exons with a coding region of 5712 bp (GenBank Accession number GQ373182). In chor, the G-to-A substitution disrupted the splice acceptor site at intron 46 (Fig. 4B), which resulted in three major splicing variants (Fig. 4C). Variant 1 adopted the AG at +1 created by the chor G-to-A substitution and was one base pair shorter than the CHOR mRNA (see Fig. S5 in the supplementary material). Variants 2 and 3 were 60 and 130 bp longer than the CHOR mRNA, respectively. Variant 2 was produced by a cryptic splice acceptor site at −60 bp from the correct splicing acceptor site, whereas variant 3 was unspliced (see Fig. S5 in the supplementary material). All splicing variants caused frameshifts, each of which truncated the coding sequence starting from exon 47 and added instead a short stretch of 19, 10 and 16 out-of-frame amino acids, respectively (see Fig. S5 in the supplementary material).

GSL8 encodes a large integral membrane protein of 1905 amino acids with sixteen predicted transmembrane helices. The three splicing variants all result in truncation at the sixth extracytoplasmic hydrophilic loop and removal of five C-terminal transmembrane helices (Fig. 4D). To confirm that CHOR is GSL8, we characterized the phenotype of the available GSL8 T-DNA insertion allele, gsl8-2 (Toller et al., 2008; Chen et al., 2009). It was originally reported that gsl8 confers male gametophytic lethality (Toller et al., 2008). However, data obtained by us and recently reported by other groups (Chen et al., 2009; Thiele et al., 2009) show that gsl8 mutations result in seedling lethality with severe dwarfism (see Fig. S6 in the supplementary material). Cotyledons and leaves of gsl8-2 plants exhibited similar stomatal patterning defects to those in chor, indicating that the chor phenotypes are due to loss of function in GSL8 (Fig. 4E; data not shown). The fact that the degree of stomatal clustering was less severe in chor than in gsl8-2 (Fig. 4E) suggests that chor might represent a partial loss-of-function allele. GSL8 is expressed broadly throughout plant organs/tissues, based on the available expression data (Winter et al., 2007).

Callose, a linear 1,3-β-glucan polymer, accumulates in various tissues/cells/organelles during normal development (Stone and Clarke, 1992). The incomplete cytokinesis observed in chor pavement cells (see Fig. S3 in the supplementary material) is consistent with the recent finding that GSL8 acts as a cell-plate callose synthase (Thiele et al., 2009). In addition to its role in the cell plate, callose deposition imposes physical constraints to the symplastic channel (Levy et al., 2007b; Radford et al., 1998). To address the role of GSL8 during stomatal development, we examined callose deposition in wild-type and chor seedlings using a callose-specific dye, Aniline Blue fluorochrome (Fig. 5) (Stone et al., 1984). In the wild-type rosette leaf epidermis, bright, linear signals marked cell plates in dividing cells (Fig. 5A, asterisks and dots). In addition, strong, punctate fluorescent signals were detected on the walls of epidermal cells, most notably in pavement cells and SLGCs (Fig. 5A), which is characteristic of plasmodesmatal callose deposition (Levy et al., 2007a; Simpson et al., 2009). Consistently, these punctate signals co-localized with the plasmodesmatal marker YFP-PDCB1 (Fig. 5E) (Simpson et al., 2009).

In contrast to the wild type, almost no Aniline Blue signals were detected in the chor leaf epidermis (Fig. 5B). We also noted the absence of callose deposition in chor root hairs and abnormal root hair morphology (see Fig. S2 in the supplementary material). On the basis of these findings, we conclude that GSL8 is required for the proper accumulation of callose at cell plates, cell walls and plasmodesmata.

Because GSL8 is required for both cell-plate and plasmodesmatal callose deposition (Fig. 5), we postulated that the overproliferation of stomatal-lineage cells might be due to the reduction in plasmodesmatal callose deposition. As a first step towards gaining mechanistic insight, we investigated whether symplastic cell-to-cell connections are altered in the chor epidermis. Monomeric soluble CFP was used as a reporter to monitor passive diffusion from cell to cell. Specifically, CaMV35S::CFP was bombarded to the leaf epidermis, and diffusion of CFP signals from the point of initial CFP expression was assayed. A similar method has been recently used to assess cell-to-cell trafficking (Levy et al., 2007a; Simpson et al., 2009). In nearly all cases in wild type (n=19/21), a single cell at the site of transformation showed a bright CFP signal (Fig. 6; see also Fig. S7 in the supplementary material). A few surrounding cells showed a dim CFP signal, indicating diffusion (Fig. 6; see Fig. S7 in the supplementary material). chor epidermal cells without obvious cytokinesis defects were examined in diffusion assays. An average of four to five adjacent cells in the chor epidermis showed strong CFP fluorescence with many more surrounding cells showing dim CFP signals (Fig. 6; see Fig. S7 in the supplementary material). Total numbers of CFP-expressing cells per site in the chor epidermis (16.8±2.68, mean ± s.e.m.) were significantly higher than those in the wild-type epidermis (5.7±1.24; P<0.0006).

To address whether the size exclusion limit (SEL) was increased in chor, we simultaneously introduced trimeric GFP (3×GFP) and control CFP via co-bombardment and monitored their diffusion (Fig. 6). In wild type, all cells expressing 3×GFP (n=22/22) were solitary (Fig. 6A,C). Thus, unlike with CFP (Fig. 6A,B), there was no diffusion of 3×GFP, indicating that 3×GFP is beyond the SEL in the wild-type epidermis. By contrast, in the chor epidermis, a cluster of epidermal cells expressed 3×GFP (Fig. 6A,C). In some cases, a cell adjacent to the site of transformation exhibited a dim GFP signal (Fig. 6A, asterisk), indicating that 3×GFP was able to spread into neighboring cells in the chor epidermis. These results suggest that the reduced plasmodesmatal callose deposition caused by the chor mutation is associated with an increase in symplastic connectivity between leaf epidermal cells.

In chor, SPCH-mediated stomatal cell-lineage divisions become excessive and dysregulated (Fig. 2). To explore the mechanistic link between increased cell-to-cell connections and stomatal patterning, we tested the diffusion of functional SPCH-YFP in wild-type and chor epidermis. For this purpose, SPCH-YFP and control CFP constructs were co-bombarded (Fig. 7). In wild type, YFP signal was detected solely in the nucleus of a single cell at the site of transformation (n=20/20; Fig. 7A,C), indicating that SPCH-YFP does not diffuse from cell to cell. By contrast, in chor epidermis multiple adjacent cells expressed SPCH-YFP (Fig. 7A,C). In some cases, neighboring cells exhibited a dim YFP signal in the nucleus (Fig. 7A, asterisk). The presence of a continuous cell periphery indicates that these YFP-positive nuclei are not likely from a single multinucleate cell. The results demonstrate that SPCH-YFP diffuses from cell-to-cell in the chor epidermis and suggests that the failure to properly contain cell-fate determinants in chor causes the stomatal patterning defects.

Because GSL8 is required for cytokinesis, incomplete cell division within the epidermis might also contribute to the dysregulated partitioning of regulatory macromolecules. To clarify this, we examined the epidermal phenotype and cell-to-cell connectivity in two well-described cytokinesis mutants, atnack1 (hinkel) (Nishihama et al., 2002; Strompen et al., 2002) and scd1 (stomatal cytokinesis-defective 1) (Falbel et al., 2003). Like chor, atnack1 and scd1 seedlings exhibit dwarfism but produce rosette leaves and survive for 2-3 weeks after germination (data not shown).

Consistent with the previous descriptions in tobacco leaves (Nishihama et al., 2002), cotyledon and rosette leaf epidermis of atnack1 showed occasional incomplete cytokinesis of pavement cells (Fig. 8A, arrowheads) and GMCs (Fig. 8A, asterisks), the latter of which resulted in abnormal stomata with incomplete separation of the guard cells. As reported previously (Falbel et al., 2003), incomplete cytokinesis and reduced lobing of pavement cells are also observed in the scd1 epidermis (Fig. 8A, plus signs). scd1 GMCs failed to divide and expand (Fig. 8A). In spite of these defects, neither atnack1 nor scd1 epidermis showed excessive production of stomatal-lineage cells or severe stomatal clusters (Fig. 8A), indicating that stomatal patterning phenotypes in chor are not simply due to defective cytokinesis.

Next we investigated cell-to-cell diffusion of SPCH-YFP and soluble CFP in atnack1 and scd1 leaf epidermis (Fig. 8B-D), and compared the results with wild type and chor (Fig. 7). Among the four genotypes, only chor exhibited a significant difference in CFP diffusion (Tukey's HSD test with ANOVA, P<0.01). As in wild type, SPCH-YFP was detected solely in the nucleus of a single cell at the site of transformation in atnack1 (n=16/16) and scd1 (n=20/20; Fig. 8B,D). The results are inconsistent with the hypothesis that leakiness of cells is a common consequence of incomplete cytokinesis.

From a sensitized genetic screen, we have identified CHOR, which is required for organized initiation of the stomatal cell lineage and proper stomatal patterns. The molecular identity of CHOR as a putative callose synthase, GSL8, and seedling callose staining data demonstrate that GSL8 is required for cell-plate and plasmodesmatal callose deposition. A series of cell-to-cell diffusion assays illuminates the link between altered symplastic connectivity and stomatal fate specification.

Among the twelve putative callose synthase genes [GSLs (CalSs)] (Hong et al., 2001), GSL1 (CalS11), GSL5 (CalS12), GSL2 (CalS5), GSL8 (CalS10) and GSL10 (CalS9) have been shown to be required for microgametogenesis and pollen development (Dong et al., 2005; Enns et al., 2005; Nishikawa et al., 2005; Toller et al., 2008). Recent reverse-genetic studies by Chen et al. showed that most gsl mutants exhibited a nearly normal seedling growth phenotype (Chen et al., 2009), suggesting that many callose synthase genes act redundantly during sporophytic development. Our findings extend the recent reports (Chen et al., 2009; Thiele et al., 2009), and reveal roles of GSL8 in addition to its role in cytokinesis. Swollen, branched root hairs in chor and gsl8-2 (see Fig. S2 in the supplementary material) indicate that root-hair callose is necessary for proper root-hair-cell polarity and morphogenesis. Thiele et al. (Thiele et al., 2009) reported that seedlings of massue, which is allelic to chor, exhibit normal root-hair morphology. The difference between our observations and those of Thiele et al. (Thiele et al., 2009) might be due to specific growth conditions, such as a difference in mineral ion concentrations in the growth media, which affect root hair morphology and patterning (Muller and Schmidt, 2004; Yang et al., 2008).

Callose turnover at the plasmodesmata is likely to be determined by an intricate balance of callose synthase- and β-1,3 glucanase activities. Recent proteomic analyses of plasmodesmatal proteins led to the identification of a β-1,3 glucanase, AtBG_ppap (Levy et al., 2007a), as well as of GPI-anchored plasmodesmatal neck proteins that anchor callose via direct binding (Simpson et al., 2009). A loss-of-function mutation in AtBG_ppap conferred increased plasmodesmatal callose deposition and reduced cell-to-cell diffusion of free GFP (Levy et al., 2007a), a phenotype that is opposite to that of chor.

The fact that GSL8 is required for both cell-plate and plasmodesmatal callose deposition complicates the interpretation of chor epidermal phenotypes, as incomplete cytokinesis could also contribute to cell leakiness. Our cell-to-cell diffusion assays using atnack1 and scd1 demonstrate that this is not likely to be the case. AtNACK1 encodes a kinesin-related protein that is required for cell-plate expansion (Nishihama et al., 2002; Strompen et al., 2002). SCD1 encodes a protein with a DENN domain and WD-40 repeats, which is implicated in the trafficking of secretory vesicles during cell-plate formation and polar cell growth (Falbel et al., 2003). Although atnack1 and scd1 produced aberrant stomata with incomplete GMC divisions, neither mutant showed overproliferation of stomatal-lineage cells or severe stomatal clusters. Thus, the overproduction of stomatal-lineage cells in chor is likely to be a consequence of reduced plasmodesmatal callose deposition.

chor conferred massive stomatal cluster formation in plants lacking TMM or partially lacking members of the ER family (Figs 1, 3), suggesting that CHOR is required for full manifestation of inhibitory signaling via TMM and the ER family. How do cell leakiness and loss/reduction in a transmembrane signaling system together result in severe stomatal cluster formation? We propose the following model based on our findings (Fig. 9). In wild type, stomatal cell-fate determinants are tightly contained in stomatal initials (i.e. MMC) and precursors (i.e. meristemoids). Inhibitory signals, such as EPF peptides, further prevent the neighboring cells form adopting a stomatal-lineage fate (Fig. 9A). In the absence of CHOR, cell-fate determinants leak from the stomatal initials/precursors to their neighbors, causing excessive entry divisions. However, functional TMM/ER-family signaling can target these cell-fate determinants and enforce stomatal patterning to some extent (Fig. 9B). The loss of CHOR in combination with the loss of TMM/ER-family signaling (or with even a slight reduction in signaling, such as a heterozygous loss of ERL1 in er chor) results in catastrophic and excessive stomatal differentiation (Fig. 9C).

Several lines of evidence support the hypothesis that SPCH might be such a cell-fate determinant. First, excessive entry into stomatal cell lineages, as seen in chor, is characteristic of SPCH overexpression. Second, spch loss-of-function mutation eliminates all of the small, highly divided cells in chor cotyledons and leaves, whereas incomplete cytokinesis defects in pavement cells remain unaffected. Third, more cells in the chor protoderm than in wild type express SPCH-GFP protein driven by its own promoter. Lastly, SPCH-YFP spreads from cell to cell in the chor epidermis. It has been proposed that the inhibitory cell-cell signal mediated via TMM/ER-family receptors targets and inhibits SPCH activity via YODA-dependent phosphorylation (Lampard et al., 2008). In this regard, it is worth mentioning that the expression of a phosphorylation-resistant version of SPCH also conferred a breakdown of redundancy among ER-family genes (Lampard et al., 2008).

ERLs were more sensitive than ER to the effects of chor (see Fig. 3; see Fig. S4 in the supplementary material). This further supports our hypothesis that SPCH activity becomes excessive and dysregulated in chor. ER and ERLs have redundant yet unique roles during stomatal development. ER is expressed throughout protodermal tissues, while ERLs are specifically expressed in stomatal-lineage cells (meristemoids and SLGCs) (Shpak et al., 2005). Consistently, ERLs inhibit the differentiation of meristemoids, while ER only has an earlier role (Shpak et al., 2005). It is possible that while the TMM/ER-family-mediated signaling pathway targets SPCH, SPCH in turn promotes the expression and hence the function of ERLs.

Our studies reveal an exciting link between the regulation of intracellular macromolecule trafficking and stomatal patterning. Selective trafficking of regulatory molecules, such as transcription factors and small RNAs, influences cell-fate specification in plants, such as root tissue patterning and coordinated growth within the shoot apical meristem (Lucas and Lee, 2004; Oparka and Roberts, 2001). Thus far, intercellular movements of key transcription factors have not been observed during stomatal development. Rather, all five key stomatal bHLH proteins are expressed solely in specific stomatal precursors (SPCH, MUTE and FAMA), or in stomatal precursors throughout stomatal development (SCRM and SCRM2), but not in their neighbors, consistent with their intrinsic, cell-autonomous actions (MacAlister et al., 2007; Ohashi-Ito and Bergmann, 2006; Pillitteri et al., 2008; Pillitteri et al., 2007).

Owing to their nature as turgor-driven valves, mature guard cells constituting stomata are symplastically isolated and are consistently devoid of functional plasmodesmata (Willmer and Saxton, 1979). However, ultrastructural studies in some plant species have shown that developing stomatal precursors and their neighbors are connected via plasmodesmata (Kaufman et al., 1970; Willmer and Saxton, 1979). Imposing restrictions in cell-to-cell trafficking by callose deposition therefore could be a crucial means of segregating key cell-fate determinants while maintaining integrity between neighboring cells. In addition to regulatory proteins, movement of small RNAs might influence stomatal patterning. It has been reported that the microRNA miR824 and its target AGL16 are localized in different cell types within stomatal cell lineages, and this observation led to the hypothesis that they might act non-cell autonomously via plasmodesmata to regulate satellite meristemoid divisions (Kutter et al., 2007). Future studies to determine the distribution and cell-to-cell trafficking of these regulators in chor, and a precise delineation of cytokinesis and the plasmodesmatal functions of CHOR (GSL8), might reveal the exact mechanism by which cell-to-cell trafficking enforces proper stomatal patterning.

We thank Lynn Pillitteri and Jennifer Nemhauser for comments; Lynn Pillitteri, Michiko Sasabe, Yasunori Machida, Sebastian Bednarek, Andrew Maule, and NASC for plasmids and mutants/transgenic lines; Bruce Godfrey for assisting particle bombardment; Greg Martin and Amanda Rychel for assisting confocal microsopy; Tetsuya Higashiyama and Narie Sasaki for providing facilities to M.M.K. This work was supported by NSF (IOB0744892), DOE (DE-FG02-03ER15448) and JST-PREST awards to K.U.T., NSF (MCB0548525 and IOB0543923) awards to Z.H., and Japan NEXT Grant-in-Aid for Young Scientists (Start-up:20870020 and B:21770041) to M.M.K. K.M.P. is an NSF graduate research fellow.