Regional specification of stomatal production by the putative ligand CHALLAH

To generate and maintain a complex body plan, multicellular organisms must selectively regulate cell fate programs to generate distinct classes, patterns and distributions of cell types in different tissues. Although regional regulation is apparent in the broad brushstrokes of patterning — in the definition of major body axes and establishment of organ domains — it also manifests itself in more subtle cellular patterning events. Regional regulatory cues can influence the production and progression of specialized cell lineages, modulating how many cells enter the lineage, how frequently they divide, and what fate they ultimately adopt. In some instances, regional information simply alters the frequency with which a particular cell type is produced. For example, in Arabidopsis thaliana numerous leaf hair cells are found on the adaxial (top) surface of rosette leaves, but few or no such cells are generated on the abaxial (bottom) surface (reviewed by Hulskamp and Schnittger, 1998). In other cases, regional information alters the division potential of precursors or the ultimate identity of their progeny. For instance, in the arthropod nervous system, segmentally repeated neural precursors give rise to lineages that differ in size, cell type composition and morphology depending on their anterior-posterior position (Prokop et al., 1998; Udolph et al., 1993). Such tissue specificity of cellular behavior implies that there must be regional regulatory factors that can impinge on the core networks that control the production and progression of specialized cell lineages.

In plants, the patterning of the epidermal structures called stomata provides an excellent system in which to study tissue-specific control of specialized cell type behavior. Stomata consist of paired guard cells surrounding a central pore, and these guard cells contract and relax via changes in turgor pressure to modulate gas exchange. In Arabidopsis thaliana, stomata represent the terminal product of a series of stereotyped, yet flexible, asymmetric divisions (Fig. 1A). This relatively sophisticated developmental program provides numerous control points at which regional information can influence stomatal pattern. Initially, naïve epidermal cells known as meristemoid mother cells (MMCs) undergo asymmetric entry divisions to produce a small, triangular meristemoid and a larger sister cell. The meristemoid then undergoes one to three rounds of amplifying divisions, regenerating itself and producing a larger stomatal lineage ground cell (SLGC) each time, much like a mammalian stem cell (Bergmann and Sack, 2007). Eventually, the meristemoid differentiates into a guard mother cell (GMC), which undergoes a symmetric division to produce the paired guard cells of the stoma. Importantly, cells adjacent to an existing stoma or precursor undergo specialized entry divisions known as spacing divisions, which are oriented such that the new precursor forms away from the existing stomatal cell. Spacing divisions prevent stomata from forming in contact and are the primary basis of the ‘one-cell spacing rule’ of stomatal pattern (Geisler et al., 2000). Although this rule is obeyed in all tissues that produce stomata, other aspects of stomatal lineage progression, including the number of entry divisions, the duration of meristemoid self-renewal in the amplifying phase and the incidence of spacing divisions, are more plastic and differ among organs and along organ axes (Bhave et al., 2009; Geisler et al., 1998).

A number of genes encoding elements of classical signal transduction systems have been implicated in the establishment of Arabidopsis stomatal pattern, including plasma membrane receptors, putative ligands and cytoplasmic kinases. Key stomatal receptors include the leucine-rich repeat receptor-like protein TOO MANY MOUTHS (TMM) and the ERECTA (ER) family of leucine-rich repeat receptor-like kinases (Nadeau and Sack, 2002; Shpak et al., 2005). The ER family includes three genes of overlapping function, ERECTA (ER), ERECTA-LIKE1 (ERL1) and ERECTA-LIKE2 (ERL2), and triple mutants produce dense guard cell clusters in all organs that generate stomata (Shpak et al., 2005). Mutations in TOO MANY MOUTHS, although best known for their eponymous production of excess clustered stomata in rosette leaves, confer a suite of other regional phenotypes that highlight tissue-specific rules for stomatal patterning. Most striking among these phenotypes is the elimination of stomata from stems (Bhave et al., 2009; Geisler et al., 1998).

Although upstream components of the TMM and ER pathways remain largely unknown, the peptides EPIDERMAL PATTERNING FACTOR1 (EPF1) and EPF2 are required for correct stomatal patterning and might act as ligands for these receptors (Hara et al., 2009; Hara et al., 2007; Hunt and Gray, 2009). Similarly, although the cytoplasmic targets of TMM and the ER family have not been identified, several mitogen-activated protein kinase (MAP kinase) signaling components, including the MAPKKK YODA (YDA), are required for correct stomatal patterning and display genetic interactions consistent with a function downstream of these receptors (Bergmann et al., 2004; Wang et al., 2007). Unlike tmm, mutations in these other stomatal regulators confer patterning defects that are relatively constant across organs and regions.

To identify tissue-specific regulators of the stomatal development program, we screened for modifiers of tmm-1 tissue-specific phenotypes and identified CHALLAH (CHAL) as a recessive suppressor of stomatal elimination from stems. Here, we show that CHAL represses stomatal production in a subset of tissues and that its expression defines organ regions rather than marking the stomatal lineage. Both the protein sequence and behavior of CHAL suggest that it might act as a ligand for ER family receptors, and, indeed, CHAL is related to putative ligands EPF1 and EPF2. The functional differences between CHAL and the EPFs suggest that these factors belong to a single ligand superfamily that has diversified to include both universal and regional regulators of stomatal development.

Columbia (Col) plants homozygous for tmm-1 and for guard cell-expressed GFP enhancer trap E1728 (Gardner et al., 2008) were mutagenized with ethylmethane sulfonate (EMS). M2 seedlings were screened for enhancement or suppression of tmm-1 stomatal phenotypes. A chal-1 mapping population was generated by outcrossing the original mutant to tmm-3 (C24 ecotype), and a PCR-based approach was used to trace chal-1 to a 106 kb region spanning BACs T9D9 and T6B20 (flanked by CER461306 and CER460870, http://www.arabidopsis.org/browse/Cereon/index.jsp). Genes in this region were sequenced and a C-to-T point mutation was found at position 364 in At2g30370 (numbering derived from cDNA clone DQ446581), which is predicted to confer an amino acid substitution of P122S.

For stomatal counts, tissues were collected from healthy, growing plants, cleared and stored in 70% ethanol, then rehydrated and mounted in Hoyer's medium (Liu and Meinke, 1998) for DIC microscopy. For hypocotyl quantification, stomatal counts reflect analysis of the entire organ. For cotyledon quantification, a field located in the central region of the cotyledon was analyzed, with field size indicated in the text. For qualitative analysis of cotyledon phenotypes in 35S::CHAL plants in various mutant backgrounds, categorization of T1s was based on analysis of whole cotyledons, and each plant was placed in the most ‘rescued’ category to which it belonged. For inflorescence stem quantification, counts reflect analysis of a single microscope field per plant and were taken from epidermal peels of developmentally matched stem segments [those subtending the eighth pedicel below the youngest stage 17 (fully abscised) flower]. Statistical analysis was performed using R statistic software (http://www.R-project.org). Because meristic (count) data are typically non-normal (Sokal and Rohlf, 1995), the Wilcoxon two-sample test, a nonparametric analog of the t-test, was used for analysis. All comparisons were performed as isolated two-sample tests without adjustment of α for multiple comparisons. E1728::GFP, N7::GFP (Cutler et al., 2000) and 35S::YFP-CHAL, as well as select overexpression lines, were imaged on a Leica SP5 confocal microscope with propidium iodide counterstaining to visualize cell outlines.

To assay β-glucuronidase (GUS) expression in CHALpro::GUS lines and enhancer trap line CS100155, tissues (except ovules) were permeabilized in acetone at −20°C and washed twice in PBS, then incubated in GUS staining solution as described (Sessions et al., 1999) with modifications. Following staining, tissues were either fixed in 3:1 ethanol:acetic acid and stored in 70% ethanol or cleared directly in 70% ethanol. For cellular-level analyses, tissue samples were rehydrated and mounted in Hoyer's medium. Reported expression patterns were observed in at least four independent lines unless otherwise noted. For analyses of mRNA expression level in Col, chal-2, and 35S::CHAL, RNA was isolated from seedlings using the RNeasy Plant Mini Kit (Qiagen) with on-column DNase digestion. Total RNA was subjected to additional DNase I (Invitrogen) treatment if necessary and used in a SuperScript III first-strand synthesis reaction with oligo(dT) primers (Invitrogen). In 35S::CHAL analysis, faint genomic contamination was not wholly eliminated, but was comparable among samples. Gene-specific transcripts were amplified from the resulting cDNA using Accuprime Pfx DNA polymerase (Invitrogen). CHAL was amplified for 35 cycles (chal-2 analysis) or 30 cycles (35S::CHAL analysis) with annealing at 54°C. Actin (ACTIN1) was amplified for 35 cycles with annealing at 54°C [primers were either as described previously (Ohashi-Ito and Bergmann, 2006) or Actin1-F, 5′-CGATGAAGCTCAATCCAAACGA-3′ and Actin1-R, 5′-CAGAGTCGAGCACAATACCG-3′].

All constructs were generated using the Gateway Cloning System (Invitrogen). To determine the expression pattern of CHAL, a 2.9 kb region 5′ of the translational start site was amplified with Accuprime Pfx using primers 30370pro-ea1F (5′-CACCGCGGCCGCACTTAACTTGGCATGTGACCC-3′) and 30370pro-ea1R (5′-GCGGCCGCTTCTGAAAACCTAAGAAGAGGC-3′). The amplicon was cloned into pENTR (Invitrogen), then transferred into pBGGUS (Kubo et al., 2005) to generate CHALpro::GUS. For overexpression, CHAL cDNA clone DQ446581 [Arabidopsis Biological Resource Center (ABRC)] was inserted downstream of the 35S promoter in pH35GS (Kubo et al., 2005) to generate 35S::CHAL. A mock artificial microRNA, amiR-mock, cloned into pH35GS was used as a negative control for overexpression experiments. For overexpression of the chal-1 cDNA, the chal-1 mutation was introduced into DQ446581 using the QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene), and the resulting sequence was cloned into pH35GS to generate 35S::CHAL_Pro→Ser. For overexpression of N-terminal YFP-tagged CHAL, the DQ446581 cDNA was cloned into pH35YG (Kubo et al., 2005). Constructs were transformed into plants via Agrobacterium-mediated floral dip (Clough and Bent, 1998), and T1s were selected on soil using Finale (1:2000) or on agar plates supplemented with hygromycin (50 μg/ml) as appropriate.

Columbia (Col) ecotype Arabidopsis and E1728 (Col) were treated interchangeably as wild type in all experiments. chal-2 (SALK_072522) was obtained from the ABRC. tmm-1 has been described previously (Nadeau and Sack, 2002). tmm;er;erl1, tmm;er;erl2, tmm;erl1;erl2, tmm;erl1 and er;erl1/+;erl2 (Shpak et al., 2005) were gifts of Dr Keiko Torii. As some lines were in a gl1 (glabrous) background, stomatal phenotypes were scored quantitatively only in tissues lacking trichomes. Multiple mutants were generated by crossing and were identified based on stomatal phenotypes, whole-plant phenotypes (er), sequencing and PCR genotyping. Genotyping primers included Mr_30370-ea1F (5′-TCACGTTGAGGTTGTCCTAGC-3′) and Mr_30370-ea1R (5′-CTCCAAGTGTGGAAGATGCTC-3′) for chal-1 [which create a derived cleaved amplified polymorphic sequence (dCAP) when digested with TasI], 072522s-RP (5′-CATAGAAGTTCCCTAGTGTCCC-3′) and 30370-ea1R (5′-TTTCTCACTCTCAACCACTCCC-3′) for chal-2, and others were based on published sequences (Bergmann et al., 2004; Shpak et al., 2004). Seedlings were grown on 0.5× MS plates in a Percival incubator under constant light. For quantification of inflorescence stem phenotypes, 12 dpg seedlings were transferred to soil in a 22°C growth chamber with a 16-hour light/8-hour dark cycle, and tissue was collected at 50 dpg. Plants not used for quantitative analyses were transferred to soil at 1-3 weeks and grown in a 20-22°C growth room with a 16-hour light/8-hour dark cycle.

The challah-1 (chal-1) mutation was recovered in a suppressor screen performed in a tmm-1 background. Although initially identified for their production of hypocotyl stomata, tmm;chal-1 plants also produce small groups of hypocotyl cells in an arrangement that resembles braided ‘challah’ bread (Fig. 1B,C,G). During characterization and cloning, chal-1 behaved as a recessive allele with incomplete penetrance: we observed hypocotyl stomata in less than a quarter of progeny from tmm;chal-1/+ populations and in approximately two-thirds of tmm;chal-1 homozygotes [n=23/32 at 12 days post-germination (dpg)]. ‘Challah’ clusters of small cells were typically present in homozygous plants that failed to differentiate stomata. As the hypocotyl grows, these cells elongate and may take on a morphology reminiscent of the excess SLGCs observed in er mutants (Fig. 1D) (Shpak et al., 2005). Similar cells are produced in tmm hypocotyls, although in smaller numbers and unaccompanied by stomata, and have been shown to represent arrested, dedifferentiated meristemoids (Bhave et al., 2009).

We next investigated whether the chal-1 mutation affected tmm phenotypes in organs other than the hypocotyl. In the seedling, chal-1 did not significantly alter stomatal density in the adaxial cotyledon (although a subsequently identified stronger allele, chal-2, had a slight effect in this organ) (Fig. 1G and see Fig. S1G in the supplementary material; see below). In the inflorescence stem, by contrast, chal-1 increased stomatal density more than 20-fold from near-zero starting levels, much as it did in the hypocotyl (Fig. 1E-G). The chal-1 mutation thus appears to derepress stomatal production in a specific subset of tissues, conferring little or no effect in other regions.

Using a PCR-based positional cloning approach (see Materials and methods), we found that chal-1 was a missense mutation in At2g30370, an uncharacterized gene predicted to encode a small, allergen-like protein (TAIR, http://www.arabidopsis.org). At2g30370 will subsequently be referred to as CHAL. Although the annotation indicates a gene structure encoding four exons and yielding a 593 bp mRNA coding sequence, we did not detect this splice variant in seedling tissues. Instead, we observed a transcript consistent in size with a 471 bp cDNA sequence deposited in GenBank (DQ446581), which corresponds to a gene structure with three exons and a large second intron (Fig. 2A). This mRNA encodes a 156 amino acid protein with a series of hydrophobic residues near the N-terminus that is predicted to act as a transmembrane domain or signal peptide (SignalP 3.0, http://www.cbs.dtu.dk/services/SignalP/; Aramemnon, http://aramemnon.botanik.uni-koeln.de/).

As CHAL contains no known functional domains, we performed BLAST searches to identify related molecules. Putative orthologs of CHAL are found in diverse plant species, including grape, rice, poplar and Physcomitrella patens (GI:157336978, GI:57900264, GI:169118872, GI:168054544), and the gene family appears to be plant specific. In the Arabidopsis genome, eleven loci are predicted to encode small proteins homologous to CHAL (Fig. 2B,C). Intriguingly, these homologous genes include EPF1 and EPF2, which act as negative regulators of stomatal production and have been proposed to encode ligands for TMM and/or ER family receptors (Fig. 2B,C) (Hara et al., 2009; Hara et al., 2007; Hunt and Gray, 2009). CHAL and its relatives display sequence similarity primarily in a C-terminal region that is characterized by six cysteine residues with conserved spacing (Fig. 2C). In the chal-1 mutation, the proline adjacent to the fourth cysteine, which is common to CHAL and the EPFs and is generally conserved among CHAL homologs (9/11), is converted to a serine (P122S; Fig. 2A,C, arrow).

The replacement of proline, a sterically constrained amino acid that generates turns and kinks, with serine, a more labile and less specialized residue, would tend to cause disruptions in secondary structure and might prevent or perturb the formation of disulfide bonds. Given the molecular nature of chal-1, we hypothesized that this allele might encode either a defective protein or a protein with novel functions, perhaps similar to the stabilizing proline-to-serine alleles of IAA proteins (e.g. bodenlos-1) (Hamann et al., 2002; Nagpal et al., 2000; Tian and Reed, 1999). To confirm that the lesion identified in chal-1 mutants was responsible for the chal phenotype and to determine whether chal-1 was a loss-of-function allele, we obtained and characterized a line with a T-DNA insertion in the first intron of the CHAL locus (chal-2, SALK_072522). In RT-PCR analysis, no wild-type full-length CHAL transcript could be detected in chal-2 homozygous plants (Fig. 2D). When introgressed into a tmm;chal-1 background, this mutation appeared to be allelic to chal-1: 17/21 F2 plants produced hypocotyl stomata in a tmm background, including both chal-1/chal-2 and chal-2 homozygous individuals. The two alleles displayed qualitatively similar behavior in a tmm background, restoring stomata to hypocotyls and generating clusters of small cells, but chal-2 restored stomata at higher penetrance (n=24/25) and in larger numbers (Fig. 2D,F). These results indicate that chal-1 is a reduction-of-function allele, largely abolishing CHAL activity rather than producing a protein with novel activities. This conclusion was confirmed by overexpression analysis of the defective chal-1 cDNA (see below). Notably, the phenotypes associated with loss of CHAL function are dependent on the absence of functional TMM. In a wild-type background, chal does not appear to alter stomatal density in the hypocotyl or cotyledons, nor does it confer production of ‘challah’ clusters of small cells (see Fig. S1A-F in the supplementary material).

Because chal mutations primarily affect the hypocotyl and inflorescence stem, we hypothesized that CHAL expression might be restricted to, or elevated in, these organs. Furthermore, we anticipated that CHAL, like EPF1 and EPF2, might be expressed specifically in the developing stomatal lineage (Hara et al., 2009; Hara et al., 2007; Hunt and Gray, 2009). To visualize CHAL expression, we generated a transcriptional reporter comprising 2.9 kb of CHAL upstream sequence fused to the β-glucuronidase gene (CHALpro::GUS) and assessed its activity in seedlings and the developing inflorescence. To confirm that the activity of CHALpro::GUS accurately reflected expression of the CHAL gene, we also obtained and characterized a GUS enhancer trap line (CS100155) bearing an insertion in the first exon of CHAL (see Fig. S2I-M in the supplementary material).

We found that CHAL expression is regionally specific, consistent with observed chal phenotypes, but that it is not specific to the stomatal lineage or epidermis. At 36 hours post-germination (hpg), before hypocotyl stomata are present, we observed moderate but distinct expression of the CHALpro::GUS reporter in the internal layers of the root and hypocotyl, with staining typically stronger in the basal hypocotyl region (Fig. 3A). At a cellular level, signal was strongest in a ring of cells surrounding the vascular elements and was sometimes observed in a subset of cells above the root meristem (see Fig. S2A,B in the supplementary material). At 3 dpg, when both stomata and precursors are present in the hypocotyl, CHALpro::GUS activity continued in the internal tissues, with a maximum at the top of the hypocotyl (Fig. 3B). Expression of CHALpro::GUS was not typically observed in cotyledons at either 36 hpg or 3 dpg (Fig. 3A,B), but strong staining was detected in the midrib of developing rosette leaves in older (14 dpg) plants (Fig. 3D and see Fig. S2C,D in the supplementary material). Interestingly, we did not observe elevated expression of CHALpro::GUS in stomatal precursors or stomata at any stage, even in plants with strong subepidermal GUS signal (Fig. 3C).

In the inflorescence, CHALpro::GUS displayed a more complex pattern of activity, a predominant feature of which was strong expression in young stem tissue. We observed particularly high activity in immature axillary stems (Fig. 3E) and in young tissue near the apex of more mature stems (e.g. Fig. 3F, bracket). As in the hypocotyl, CHALpro::GUS was expressed strongly in the internal tissues of the inflorescence stem, with expression strongest in (but not exclusive to) a ring of cells subtending the cortex and encircling the vascular core (see Fig. S2E-G in the supplementary material). Expression was concurrent with stomatal production, and a faint epidermal signal was detectable in the strongest lines, but we did not observe specific GUS signal in either precursors or guard cells during stomatal lineage progression (see Fig. S2H in the supplementary material). Strong staining was also present in a ring of cells at the base of the developing silique, at the point of floral organ attachment (Fig. 3F, arrow), and in the chalazal endosperm of developing seeds (Fig. 3G). Although no expression was observed in early embryos, essentially mature embryos (as found in ovules with browning seed coats) showed clear GUS activity in the hypocotyl and root, with the signal highest in, but not exclusive to, the internal tissues (Fig. 3H; faint expression is observed in the hypocotyl epidermis and in the cotyledons of the strongest line).

The CS100155 enhancer trap line recapitulated the primary features of the CHALpro::GUS expression pattern, displaying strong, consistent activity in the embryonic hypocotyl (see Fig. S2L in the supplementary material) and in growing stems (see Fig. S2J,K,M in the supplementary material). Interestingly, signal was observed more broadly in cotyledons, leaves and floral organs of CS100155 (see Fig. S2I-L in the supplementary material), although strong stem activity remained the most striking feature of the expression pattern. The elevated transcriptional activity in young hypocotyls and stems observed in both CHAL reporter lines is consistent with the stomatal phenotypes observed in chal mutants.

CHAL is expressed in a relatively limited domain, and its domain of expression correlates with its loss-of-function stomatal lineage phenotypes. We thus hypothesized that spatially constrained expression of CHAL, rather than limited availability of a receptor or modifier, might underlie its regionally specific effects. To assess whether widespread expression of CHAL could confer ubiquitous stomatal phenotypes, we generated stable transgenic lines expressing CHAL cDNA under control of the constitutive 35S promoter (35S::CHAL). Consistent with a failure to identify CHAL in a previous large-scale screen for stomatal-regulatory ligands (Hara et al., 2009), most 35S::CHAL T1 plants did not display obvious stomatal phenotypes, and stomatal density in T1s did not consistently differ overall from that in controls (Fig. 5A). In analyzing larger numbers of T1s, however, we repeatedly observed a fraction of plants that displayed distinct reductions in stomatal number (n>8; Fig. 4B,C). Moderately affected individuals produced some normal stomata in the hypocotyl and cotyledons but also displayed arrested stomatal lineage cells or small pavement cells suggestive of stomatal lineage transdifferentiation (Fig. 4B, arrows), and some also produced paired stomata at elevated frequency (see Fig. S3A in the supplementary material). Severely affected plants produced few or no stomata, with the most dramatic individuals resembling speechless (spch) mutants, although a few entry divisions were typically observed (Fig. 4C) (MacAlister et al., 2007). Although these phenotypes were scored primarily in T1 seedlings, similar effects were apparent in mature tissues (see Fig. S3B in the supplementary material) and in the T2 and T3 generations.

Reduced stomatal production in 35S::CHAL plants is consistent with the tmm;chal phenotype, which indicates an endogenous role for CHAL as a stomatal inhibitor. To correlate stomatal phenotypes with construct expression level, we screened T3 seedlings from multiple lines and collected individuals that fell into either of two phenotypic categories: moderate (arrested cells and at least ten stomata in the adaxial cotyledon) and severe (fewer than five stomata in the adaxial cotyledon). Bulked individuals from each category were split into two groups, one of which was used for quantification and the other for RT-PCR analysis. As expected, both overexpression lines produced more CHAL mRNA than wild-type plants, and CHAL expression was negatively correlated with cotyledon stomatal production (Fig. 4D,E). To confirm that high levels of CHAL protein conferred stomatal underproduction phenotypes, we also generated transformants overexpressing a YFP-tagged CHAL variant (35S::YFP-CHAL). T2 transformants with bright fluorescence (n=4 independent lines) frequently showed phenotypes similar to those of 35S::CHAL plants, including transdifferentiated or arrested stomatal lineage cells (see Fig. S3C in the supplementary material). These results confirm that high levels of CHAL protein confer reduced stomatal production.

Taking advantage of the stomatal reduction phenotypes conferred by CHAL overexpression, we analyzed the chal-1 missense allele and the functionality of its protein product in greater detail. If chal-1 were a reduction-of-function allele, its overexpression would be likely to confer a weak phenotype resembling that of 35S::CHAL. If it were a gain-of-function allele, overexpression would be expected to generate a chal-like phenotype, perhaps restoring stomata to hypocotyls or ubiquitously increasing stomatal production in tmm. To distinguish between these possibilities, we overexpressed the chal-1 mutant cDNA under control of the 35S promoter (35S::CHAL_Pro→Ser) in a tmm background. T1 transformants displayed stomatal underproduction phenotypes resembling those of 35S::CHAL in tmm (see Fig. S3E in the supplementary material; see below), with some plants failing to produce stomata in the cotyledons (n=4) and others producing markedly reduced numbers (n=4). 35S::CHAL_Pro→Ser phenotypes were less severe than those conferred by wild-type CHAL overexpression in a tmm background, indicating that the chal-1 protein product retains some activity but acts less efficiently than wild-type CHAL.

The molecular features of CHAL, together with its ability to confer overexpression phenotypes in diverse tissues, suggest that it, like EPF1/2, might serve as a ligand for a broadly expressed stomatal receptor. CHAL signaling, therefore, might feed into known stomatal-inhibitory pathways, perhaps acting through TMM or ER family receptors. The identification of chal as a genetic suppressor of tmm, however, argues against it being a ligand for TMM itself. To test the relationships between CHAL and its potential receptors, we first overexpressed CHAL in tmm, a background that blocks both EPF1 and EPF2 overexpression phenotypes (Hara et al., 2009; Hara et al., 2007). In striking contrast to the EPF1/2 results, the CHAL overexpression phenotype was dramatically enhanced in tmm, such that virtually all T1 transformants failed to produce stomata and resembled spch mutants (Fig. 5B,D). This finding is consistent with our hypothesis that CHAL is unlikely to be a TMM ligand.

We then examined whether one or more of the ER family receptors might mediate CHAL activity. For this assay, we used the unequivocal overexpression phenotype observed in tmm as a baseline and assayed the mitigating effects of ER family mutations, overexpressing CHAL in tmm plants bearing ER family double-mutant combinations (tmm;er;erl1, tmm;er;erl2 and tmm;erl1;erl2). Substantial suppression of the 35S::CHAL phenotype was observed in all three backgrounds, such that more than 20% of transformants in each background produced one or more stomata in the cotyledon (Fig. 5C). In tmm;er;erl1 and tmm;er;erl2 backgrounds, stomatal differentiation was not extensively rescued (typically, stomata were not produced or were produced only in the hydathode; Fig. 5C), but asymmetric divisions were observed in the cotyledons of transformants at high frequency (Fig. 5E,F). In tmm;erl1;erl2, stomatal differentiation was more broadly restored, with more than 90% of transformants producing a stoma and more than 20% displaying stomata in the peripheral or central cotyledon (Fig. 5C,G). Importantly, although suppression of 35S::CHAL phenotypes was observed in each ER family double-mutant background, the majority of transformants in all backgrounds still produced stomata at greatly reduced density, suggesting that any individual ER family receptor is sufficient to transduce CHAL activity.

Using a similar approach, we examined the relationship between CHAL and the protease STOMATAL DENSITY AND DISTRIBUTION1 (SDD1) (Berger and Altmann, 2000). As with the ER family receptors, SDD1 prevents stomatal overproduction and clustering, but is believed to act independently of the TMM/ER family pathway (Hara et al., 2007; Lampard et al., 2008). When the CHAL overexpression assay was repeated in a tmm;sdd1 background, almost all transformants failed to produce stomata in the cotyledon (44/45 T1s), much as in a tmm background (Fig. 5C,H). Although asymmetric divisions were observed in many T1s, such divisions were relatively uncommon (as compared with tmm;er;erl1 or tmm;er;erl2 T1s) and only rarely gave rise to small, precursor-like cells. These observations indicate that CHAL is unlikely to be processed by SDD1, but, equally importantly, they suggest that suppression of 35S::CHAL phenotypes by ER family mutations might be specific to this family (rather than a non-specific, additive effect conferred by the loss of any stomatal inhibitor). Together, the overexpression assays indicate that TMM acts to restrict the stomatal-inhibitory effects of CHAL, and that ER family receptors are likely to mediate these inhibitory effects.

Because the results of our overexpression assays suggested that ER family receptors might mediate CHAL activity, we generated multiple mutant combinations to assess the genetic relationships between CHAL and ER family members. As the chal phenotype is detectable only in a tmm background, we generated triple mutants bearing tmm, chal-1, and an additional er or erl2 mutation and assessed the effect of each on hypocotyl stomatal production. The two mutations had distinct effects on the chal-1 phenotype, perhaps reflecting ligand-receptor specificity or complex combinatorial interactions among ER family receptors. An erl2 mutation moderately, but significantly, enhanced hypocotyl stomatal production relative to tmm;chal-1 (Fig. 6D,F). An er mutation, on the contrary, did not enhance chal-1 restoration of stomata to hypocotyls, but instead suppressed this phenotype, such that tmm;er;chal-1 plants seldom produced hypocotyl stomata (n=2/29; Fig. 6E,F). A similar epistatic relationship was recently reported between er and epf2 in control of stomatal density (Hunt and Gray, 2009). Because erl1, like chal, can restore stomata to tmm stems (Shpak et al., 2005), we also compared hypocotyl stomatal production in tmm;chal-1 and tmm;erl1 plants. tmm;erl1 produced more hypocotyl stomata than tmm;chal-1, consistent with a role for CHAL as one of several partially redundant ERL1 ligands (Fig. 6C,F).

CHAL encodes a small, potentially secreted protein that is homologous to the known stomatal regulators EPF1 and EPF2 (Hara et al., 2009; Hara et al., 2007; Hunt and Gray, 2009). Consistent with their shared sequence motifs, both CHAL and the EPFs regulate stomatal production, and both act as repressors in this process. Nonetheless, the distinct loss-of-function phenotypes and genetic interactions of these factors suggest fundamental differences in their endogenous functions. First, whereas epf1 and epf2 confer similar patterning defects in all tissues examined (Hara et al., 2007) (E.B.A., unpublished observations), chal confers strong phenotypes only in select organs, notably the hypocotyl and inflorescence stem. Thus, unlike EPF1 and EPF2, which universally regulate specific stages of stomatal development, CHAL displays regional specificity. Second, and perhaps more intriguingly, CHAL and EPF1/2 display fundamentally divergent genetic interactions with TMM. Whereas a tmm mutation is fully epistatic to epf1 and can block overexpression phenotypes of both EPF1 and EPF2 (Hara et al., 2009; Hara et al., 2007), chal was isolated as a regional suppressor of tmm and its phenotypes are apparent only in a tmm background. The relationship between TMM and CHAL, unlike that between TMM and EPF1/2, does not suggest a stimulatory ligand-receptor interaction, but rather indicates that CHAL must act through one or more receptors other than TMM.

The distinct properties of CHAL and EPF1/2 may be ascribed to differences in both expression pattern and protein activity. Specifically, the transcriptional patterns of these factors can explain their respective regional and general effects, whereas differences in biochemical activity appear to underlie their opposite interactions with TMM. Unlike EPF1 and EPF2, which are expressed specifically and universally in particular stomatal precursors (Hara et al., 2009; Hara et al., 2007; Hunt and Gray, 2009), CHAL does not show elevated expression in the stomatal lineage and is instead expressed in a region-specific manner consistent with its loss-of-function phenotypes. Differences in expression, however, cannot account for the opposite interactions of these factors with TMM, as these interactions are observable in constitutive expression assays. CHAL, EPF1 and EPF2 all reduce stomatal production when overexpressed in a wild-type background, yet their effects diverge in tmm: CHAL overexpression confers an enhanced phenotype, such that plants almost invariably fail to produce stomata, whereas EPF1/2 overexpression no longer reduces stomatal number (Hara et al., 2009; Hara et al., 2007). The opposing genetic interactions of these factors with TMM are thus inherent in their respective coding sequences, pointing to differences in protein activity that might reflect distinct ligand-receptor interactions.

Jointly, the expression pattern and genetic interactions of CHAL offer a straightforward explanation for one of the more striking regional phenotypes of tmm: the loss of stomata from stems. CHAL not only shows elevated expression in young stem tissue (Fig. 7A), but also displays dramatically enhanced stomatal repressor activity in a tmm background. Thus, elimination of stomata from tmm stems is likely to reflect overactivity of endogenous CHAL due to loss of a TMM-dependent buffering mechanism, with overactivity conferring meristemoid arrest and dedifferentiation (Bhave et al., 2009). This explanation raises the intriguing question of how such a mechanism might function at the molecular level. In previous models of stem stomatal patterning, TMM has been proposed to dampen signaling through ERL1 and other ER family receptors via ligand titration or physical inhibition, such that loss of TMM confers ER family overactivity and eliminates stomatal production (Shpak et al., 2005). These models point to the existence of an ER family ligand, the effects of which, unlike those of EPF1 and EPF2, are suppressed rather than enhanced by TMM.

We propose that CHAL corresponds to this hypothesized ligand, repressing stomatal production through multiple ER family receptors in a pathway that is inhibited by TMM. Several lines of evidence support such a model. First, loss of ER family receptors can suppress CHAL overexpression phenotypes, suggesting that CHAL signaling passes through one or more members of this receptor family. Second, CHAL overexpression can still confer phenotypes in the absence of any two ER family receptors (in tmm;er;erl1, tmm;er;erl2 and tmm;erl1;erl2 backgrounds), indicating that each of the three receptors is individually capable of mediating CHAL activity. We thus propose a model in which endogenous CHAL interacts with multiple ER family members (Fig. 7B), such that the chal phenotype reflects reduced activity of multiple receptors. To explain the unexpected epistasis of er to chal-1, we further propose that ER may serve as both a mediator of CHAL activity and a buffer that protects its sister receptors from excess CHAL signal. The CHAL pathway appears to be independent of the protease SDD1, as an sdd1 mutation fails to mitigate CHAL overexpression phenotypes, adding to a growing body of evidence (Hara et al., 2009; Hara et al., 2007; Lampard et al., 2008) that places SDD1 in a pathway separate from TMM and the ER family.

The activity of CHAL and the EPFs alone cannot account for the full spectrum of tmm phenotypes, raising the intriguing possibility that other EPF/CHAL superfamily members might also participate in stomatal patterning. Such additional factors might universally modulate stomatal production, similar to EPF1/2, or provide a layer of regional refinement, similar to CHAL. Further analysis of this superfamily might not only identify new stomatal regulators, but also define the sequence motifs that confer EPF-like and CHAL-like behaviors, laying the groundwork for biochemical and mechanistic studies of the TMM/ER family system.

We thank members of our laboratory for helpful comments on the manuscript; the S. R. Long laboratory (Stanford) and Carnegie Institute, Department of Plant Biology, for the use of microscopes; and Dr Keiko Torii (University of Washington) and Dr Julie Gray (University of Sheffield) for discussion and communication of results prior to publication. This work was supported by NSF grants IOS-0544895 and IOS-0844521. E.B.A. was supported in part by an ASPB SURF Fellowship, by a Stanford VPUE Grant (Biology Department) and by a Stanford Major Grant.