Timely expression of the Arabidopsis stoma-fate master regulator MUTE is required for specification of other epidermal cell types

Authors


For correspondence (e-mails Montana.Mena@uclm.es or Carmen.Fenoll@uclm.es).

Summary

Epidermal differentiation in Arabidopsis thaliana aerial organs involves stomatal lineage development. Lineages derive from meristemoids, which arise from asymmetric divisions of protodermal cells. Each meristemoid divides repeatedly in an inward spiral before it transits to a guard mother cell (GMC) that produces the stoma, leaving a trail of surrounding stomatal lineage ground cells (SLGCs) that eventually differentiate into endoreplicated pavement cells. MUTE is a bHLH transcription factor that is expressed in late meristemoids and drives their transition to GMCs. Loss-of-function mute mutants are stomata-less dwarf plants with arrested lineages, in which stunted putative SLGCs surround a halted meristemoid. We analysed MUTE functions using a chemically inducible system for mute-3 complementation based on conditional MUTE expression in its normal domain. Continuous induction from germination produced stomata-bearing, normal-sized plants with viable mute-3 seeds. In 2-week-old mute-3 cotyledons, meristemoids appeared to retain their identity and synchronously formed stomata in response to induced MUTE expression. However, arrested SLGCs were not complemented: many produced stomata, leading to stomatal clusters, and others remained unexpanded and diploid. In contrast, non-lineage pavement cells, which are under-endoreplicated in mute-3, expanded and increased their ploidy level upon induction, showing that the lack of response of SLGCs is specific to this arrested cell type. Leaf phenotypic mosaics include wild-type lineages and adjacent mute-3 lineages, whose meristemoids and putative SLGCs remained arrested, indicating that the role of MUTE in SLGC fate is strictly lineage-autonomous. These results show that timely MUTE expression is essential to prevent stomatal fate in SLGCs and to promote their differentiation as pavement cells.

Introduction

The differentiation of the aerial epidermis from the protodermis in Arabidopsis takes place during organ growth through precise coordination of cell division and expansion, with the acquisition of various cell identities (Horiguchi et al., 2006; Harashima and Schnittger, 2010; Gonzalez et al., 2012). Stomata, the bi-cellular epidermal structures that encircle stomatal pores, are the culmination of a reiterative process that produces a stoma and a variable number of clonally related pavement cells (PCs) (Nadeau and Sack, 2002) (Figure S1). Stomatal development begins with the asymmetric division of a protodermal cell (meristemoid mother cell), producing the first recognizable cell of a stomatal lineage, the meristemoid. Meristemoids experience up to three additional asymmetric divisions, each one producing a larger cell (stomatal lineage ground cell, SLGC) and a new meristemoid. The late meristemoid transits to a guard mother cell (GMC), whose symmetric division produces the two guard cells (GCs) that constitute the stoma and terminate the lineage. SLGCs may differentiate into PCs or become meristemoid mother cells, founding satellite lineages (Bergmann and Sack, 2007; Dong and Bergmann, 2010; Pillitteri and Torii, 2012). Three lineage-specific bHLHs positively drive the meristemoid mother cell entry and subsequent amplification divisions (SPEECHLESS, SPCH) (MacAlister et al., 2007), the transit from meristemoid to GMC (MUTE) (Pillitteri et al., 2007) and the transit of GMC to two GCs to form a stoma (FAMA) (Ohashi-Ito and Bergmann, 2006). Two ubiquitous bHLHs, SCREAM/ICE1 and SCREAM2, participate in all three processes (Kanaoka et al., 2008).

Leaf epidermal cell division and expansion have been monitored in great detail (Tisne et al., 2008; Cookson et al., 2010; Kawade et al., 2010; Asl et al., 2011; Pérez-Pérez et al., 2011; Elsner et al., 2012; Gonzalez et al., 2012). Neighboring epidermal cells expand at different rates, and epidermal cell division and expansion appear to occur independently (Asl et al., 2011). Endoreplication and cell division have been proposed to be the result and not the cause of whole-leaf growth (Massonnet et al., 2011; Sterken et al., 2012). PCs derive from protodermal cells or differentiate from SLGCs. SLGCs are often defined in anatomical terms as those cells surrounding a meristemoid/stoma whose arrangement suggests a sequential series of spiral divisions, although their clonal relationships are rarely assessed as is required to consider them SLGCs. Because of their developmental relationship, pavement and stomata cell densities correlate in many natural accessions, with both parameters showing considerable natural variation and partially sharing genetic determinants (Delgado et al., 2011). Although SLGCs contribute to a large extent to PC numbers in the mature epidermis (Geisler et al., 2000), which is composed essentially of stomata and PCs, we know little about how SLGCs exit the stomatal pathway and differentiate. Recent work suggests that local signals between stomata and SLGCs act together with abscisic acid to induce PC enlargement (Tanaka et al., 2013). However, meristemoid and PC proliferation programs appear to be largely independent (Andriankaja et al., 2012). Stomatal lineage-specific peptides (epidermal patterning factors; EPFs) and their receptor/receptor kinases [TOO MANY MOUTHS (TMM)/ERECTA family (ERF)] signal downstream mitogen-activated protein kinase cascades to restrict stomatal fate in SLGCs (Rowe and Bergmann, 2010; Pillitteri and Torii, 2012). These factors, together with other components, such as the mesophyll-borne peptide STOMAGEN (Kondo et al., 2010; Sugano et al., 2010) or the phosphatase AP2C3 (Umbrasaite et al., 2010), help to maintain the balance between stomata and PC differentiation. Recently, brassinosteroids have also been found to be involved in the process via the GSK3-like kinase BR-INSENSITIVE2 (BIN2) (Gudesblat et al., 2012; Kim et al., 2012). The asymmetric distribution of stomatal fate determinants may be part of the mechanism determining SLGC identity; in vivo stomatal lineage tracking revealed that SPCH is maintained only in the meristemoid product of asymmetric divisions (Robinson et al., 2011). Thus, SPCH depletion in SLGCs may promote exit from the stomatal pathway. BASL (BREAKING OF ASYMMETRY IN THE STOMATAL LINEAGE) appears to be involved in SLGC identity, as its polarized peripheral location prior to asymmetric divisions marks the position of future SLGCs in successive meristemoid divisions (Robinson et al., 2011) (Dong et al., 2009). POLAR (POLAR LOCALIZATION DURING ASYMMETRIC DIVISION AND REDISTRIBUTION) (Pillitteri et al., 2011) marks the position of future SLGCs similar to BASL. MUTE is expressed in a subset of meristemoids, probably soon before their transit to GMC (Pillitteri et al., 2007), and is thus regarded as marking a transient cell identity: i.e. the late meristemoid, as opposed to early meristemoids that are not yet committed to form GMCs (Pillitteri et al., 2011). Because loss-of-function MUTE alleles produce arrested meristemoids surrounded by more than three putative SLGCs, it is assumed that MUTE restricts the number of asymmetric divisions of the meristemoid (Pillitteri et al., 2007, 2008) and thus affects SLGC numbers. In mute mutants, SLGCs remain unexpanded, at least for the short developmental times reported (Pillitteri et al., 2007, 2008) (Figure S1). However, MUTE function has not been implicated in SLGC fate or differentiation.

Taking advantage of a β-estradiol-inducible cell type-specific MUTE expression system for conditional complementation of the loss-of-function mute-3 mutant, we show that MUTE expression at the correct time during lineage progression is essential for SLGC identity. Once this time has passed, SLGCs are unable to differentiate correctly: either they form a stoma or remain diploid and small-sized, instead of differentiating into polyploid PCs. In contrast, arrested mute-3 meristemoids readily respond to late MUTE induction. Use of phenotypic mosaics displaying wild-type and halted mute-3 stomatal lineages within the same epidermis indicates that MUTE action on SLGC fate is strictly lineage-dependent. Protodermal cells, SLGCs and PCs respond differently to ectopic MUTE or FAMA over-expression, highlighting cell-specific differences in the ability to respond to these stomatal bHLHs. All these results indicate that correct SLGC identity requires MUTE expression in distinct developmental windows.

Results

Conditional cell type-specific MUTE expression system

To study the effects of MUTE expression in various genetic backgrounds and organs and at various developmental times, we used a two-component β-estradiol-inducible system (Brand et al., 2006). Activator (proMUTE::XVE) and responder (OlexA::cDNAMUTE; OlexA::GUS) gene fusions were introduced in Arabidopsis Col-0 (see 'β-estradiol-dependent MUTE expression complements the mute-3 phenotype' and Figure S2) using one construct carrying the activator unit plus the GUS reporter responder unit in cis, and another construct carrying the MUTE responder unit. Several lines homozygous for the proMUTE activator plus the GUS responder were crossed with two homozygous MUTE responder lines that were heterozygous for the mute-3 allele, which confers a strong phenotype similar to mute-1 (Pillitteri et al., 2007). Plants homozygous for the two transgenes and for the wild-type MUTE allele (iMUTEwt line) or the recessive mute-3 allele (iMUTEmute line) were selected. In these lines, XVE is expressed in the normal domain of the MUTE promoter (as described by Pillitteri et al., 2007), but only activates GUS and MUTE transcription upon addition of inducer. iMUTEwt plants grown on MS plates containing 0.5–20 μm β-estradiol had a normal size and appearance (Figure S2a). Seedlings grown in 2 μm β-estradiol were further characterized. GUS activity was restricted to a subset of meristemoids and to young GCs (Figure S2c), although occasionally some small cells adjacent to stomata were also GUS-positive, as described for the MUTE promoter fragment that we we used (Pillitteri et al., 2007), and was not observed in other cell types or non-induced iMUTEwt seedlings (Figure S2b). The cotyledon stomatal index was measured for two β-estradiol concentrations at 2 and 16 days after germination (dag), with no significant differences compared with untreated plants (Figure S2d). GUS and MUTE mRNA levels measured 2 days after β-estradiol addition confirmed inducer-dependent expression of the two responders (Figure S3a).

β-estradiol-dependent MUTE expression complements the mute-3 phenotype

Seeds from plants homozygous for the two transgenes and heterozygous for mute-3 were plated on MS with or without 2 μm β-estradiol. In control plates, wild-type and mute-3 dwarf phenotypes segregated as expected (3:1), but on β-estradiol-containing plates, all plants showed a wild-type appearance (Figure 1a). Dwarf plants of control plates displayed a characteristic stomata-less mute-3 epidermis, but on β-estradiol-containing plates, all plants formed stomata (Figure 1c,e). The stomatal index and density in complemented mute-3 plants (iMUTEmute; identified by allele-specific genotyping) were indistinguishable from those in iMUTEwt (Figure 1g). iMUTEmute plants grown in β-estradiol until bolting were similar to iMUTEwt (Figure 1a); in soil, they developed to maturity (Figure 1f) and produced viable homozygous seeds if β-estradiol was continuously applied to developing flowers. Complemented iMUTEmute plants showed GUS activity restricted to the proMUTE domain, as did iMUTEwt, and formed stomata in all expected organs (Figure S4). Only inducer-treated iMUTEmute plants accumulated MUTE transcripts, detectable as early as 4 h after induction, while mute-3 transcripts were present regardless of inducer addition (Figure S3b), as expected because mute-3 is not an RNA-null allele (see 'β-estradiol-dependent MUTE expression complements the mute-3 phenotype').

Figure 1.

Conditional complementation of the mute-3 phenotype by induction of MUTE expression. (a) iMUTEmute and iMUTEwt plants grown with or without 2 μm β-estradiol, 25 days after germination (dag). Both lines carry the constructs shown in Figure S2. (b–e) DIC micrographs of adaxial cotyledons of plants shown in (a). (f) iMUTEwt and iMUTEmute plants continuously treated with β-estradiol and photographed at 53 dag. (g) Adaxial cotyledon stomatal index and density in iMUTEwt and complemented iMUTEmute plants at 25 dag. Values are means ± SD using 10 cotyledon epidermal fields from five individual plants. No statistically significant differences were found between the genotypes. Scale bars = 1 cm (a, f) and 20 µm (b–e).

Arrested mute-3 meristemoids synchronously produce stomata upon MUTE expression

iMUTEmute plants grown for 13 days in control medium formed halted lineages with a central meristemoid and a variable number of surrounding small cells (Figure 2a), like mute-3 (data not shown) or mute-1 (Pillitteri et al., 2007) plants. We added inducer to these 13 dag iMUTEmute plants (upshift treatment; Figure 2), and examined the time course of the meristemoid to stoma transition in the adaxial cotyledon epidermis. Transgene-derived MUTE transcripts were already present 4 h post-induction (hpi) and up to 6 days later (Figure S3b). Faint GUS activity was apparent in meristemoids from the first time point (12 hpi) and increased over time; at 48 hpi, the first stomata were identified, and by 60 hpi, 97.4 ± 2.6% (n = 832) of the meristemoids had synchronously produced stomata (Figure 2d). Therefore, arrested 13 dag mute-3 meristemoids retain all the components necessary for late-meristemoid fate transitions, and resume stoma differentiation upon induction of MUTE expression.

Figure 2.

Arrested iMUTEmute meristemoids respond correctly to late induction of MUTE expression. Plants were grown in control plates for 13 days, and 2 μm β-estradiol was added (grey bar). DIC micrographs after GUS staining show adaxial cotyledons at the indicated number of hours post-induction (hpi). Scale bar = 20 µm. (a) iMUTEmute plants at 13 dag (before induction). (b) At 12 hpi, meristemoids show GUS expression. (c) At 48 hpi, some meristemoids have formed stomata (arrowheads) and some cells adjacent to meristemoids or stomata show faint GUS staining (arrows). (d) By 60 hpi, most meristemoids (97.4 ± 2.6%; n = 832) have developed into stomata.

Halted mute-3 SLGCs produce stomata upon late MUTE induction

Unexpectedly, upshifted 13 dag iMUTEmute plants examined at later times post-induction displayed distinct stomatal clusters (Figure 3b). To determine whether the duration of growth as mute-3 prior to induction influenced cluster formation, iMUTEmute plants were grown in control medium for 0–11 dag prior to inducer upshift treatments, and inspected at 16 dag (Figure 3c). When inducer was added at 0, 1 or 2 dag, stomatal pairs appeared only occasionally; stomatal cluster proportion and size increased with the time of growth before induction (Figure 3d). If inducer was added at 4 dag, approximately half of the stomatal units (see 'β-estradiol-dependent MUTE expression complements the mute-3 phenotype' for definition) were clusters of different sizes at the end of the treatment (45.71 ± 3.34%; n = 332 stomatal units from five plants; Figure 3d). Taking into account all stomata, isolated or clustered, this treatment caused that two-thirds of them were clustered (65.99 ± 3.23%; n = 536 stomata from five plants; Figure S5a). Induction at later times resulted in similar clustering. The cluster size ranged from two to five, with stomatal pairs being the most frequent class (Figure 3d) and larger clusters appearing occasionally (Figure S5). iMUTEwt plants treated in parallel did not develop clusters.

Figure 3.

Stomatal clusters appear in iMUTEmute after late induction of MUTE expression. (a, b) DIC micrographs of the adaxial cotyledon epidermis of plants grown for 16 dag in control medium (a), or for 4 days in control medium plus 12 days in β-estradiol (b). Scale bars = 20 µm. (c) Induction treatments used for data presented in (d); gray-filled boxes indicate the presence of inducer, and empty boxes indicate the number of days in control medium. (d) Stomatal cluster size and proportion as a function of the days of growth in control medium prior to induction as shown in (c). Stomata were counted at 16 dag in all treatments.

Upshifted 13 dag iMUTEmute cotyledons showed faint GUS activity in some small cells adjacent to meristemoids/stomata (Figures 2c,d and 4d). Inducer-treated iMUTEwt plants also showed occasional weak GUS expression in equivalent cells (Figure S2c) but never produced clusters. To determine whether the extra stomata in upshifted iMUTEmute came from meristemoid divisions or from small GUS-expressing cells, cell division was prevented by adding colchicine to the agar plates at the time of induction, thus inducing metaphase arrest. This treatment produced GUS-positive rounded cells, isolated or clustered (Figure 4c,e), that we interpret as GMCs derived from the meristemoid and its adjacent small cells. These plants showed a uniform cotyledon response whereby virtually all meristemoids were replaced by rounded GUS-positive cells. In 13 dag iMUTEwt, colchicine + inducer treatment produced no observable phenotype, as epidermal cell divisions had long ceased.

Figure 4.

Stomatal clusters in iMUTEmute derive from putative SLGCs. iMUTEmute plants grown in control plates for 13 dag were treated with β-estradiol alone (dark gray) or together with colchicine (light gray) for 10 days. Images are DIC micrographs of adaxial cotyledon epidermis after GUS staining. (a) iMUTEmute plants in control medium at 13 dag. (b–e) iMUTEmute plants at 23 dag, 10 days after addition of β-estradiol alone (b, d) or plus colchicine (c, e). The arrows show cells adjacent to a stoma (d) and a GMC-like cell (e). Scale bars = 20 µm.

To unequivocally assess cell ontogeny in stomatal clusters, we tracked their cell division histories through serial epidermal imprints in living cotyledons during upshift treatments (Figure 5a–c). iMUTEmute plants were grown without inducer for 13 days, taking epidermal imprints at 2, 6 and 13 dag. Plants were then transferred to inducer-containing plates, grown for 3 or 10 more days, and stained for GUS activity. Individual GUS-positive stomata or stomatal clusters were identified, and their cell history was tracked back in the serial imprints to their founder cells, scoring at least ten lineages corresponding to clusters or isolated stomata per genotype. We found that all stomata (n = 42) within a given cluster were the final developmental product of a stomatal lineage initiated by an individual cell at 2 dag (Figure 5c). Therefore, clustered stomata derived from a meristemoid and its SLGCs, as suggested by cell anatomy. Isolated stomata in iMUTEmute and the iMUTEwt control also formed through the canonical spiral divisions of meristemoids (Figure 5a,b); when a small cell adjacent to a stoma was GUS-positive, we were able to identify it as a true SLGC in all cases and genotypes (n = 23 for iMUTEwt; n = 13 for iMUTEmute).

Figure 5.

Cell histories and transcriptional profiles during stomatal cluster development. (a–c) Representative cell division and differentiation histories of stomatal lineages in inducer-treated iMUTEwt (a) and iMUTEmute (b, c). Serial resin impressions of adaxial cotyledon epidermis were taken at the indicated intervals, and plants were stained for GUS activity at the end of the experiment. (a) iMUTEwt plants grown on inducer since germination. (b, c) iMUTEmute plants grown in control medium for 13 days, then transferred to β-estradiol-containing plates and tested for GUS activity 3 (b) or 10 (c) days later. GUS-positive lineages were identified, and their history was tracked back by inspecting the serial resin imprints. The drawings show lineage reconstructions. The number of lineages inspected was 23 (a), 13 (b) and 42 (c). Micrographs were taken under DIC optics, and all are shown at the same magnification. Scale bar = 20 μm. (d) Evolution of transcript profiles for stomatal development marker genes after MUTE induction at 13 dag in iMUTEmute plants. Relative gene expression measured as inline image by quantitative PCR is presented in a heatmap format. Each row represents a gene and each column represents a treatment. Red indicates high expression, and white indicates low expression. iMUTEmute and iMUTEwt cotyledons were collected after the treatments as indicated. hpi, hours post-induction; dpi, days post-induction; c, colchicine. Relative expression values are shown in Figure S6.

These results indicate that mute-3 SLGCs respond to MUTE, and suggest that, in normal lineages, developmentally scheduled MUTE expression in its normal spatial and temporal domain prevents stomatal fate in SLGCs; this blockage is not imposed when SLGCs developed in contact with the arrested MUTE-less meristemoids typical of mute-3.

Transcript profiling of stomatal lineage marker genes upon MUTE expression

MUTE expression is required for transition from the meristemoid, a stem cell, to a GMC, a cell type already committed to stoma morphogenesis (Pillitteri et al., 2007). To study MUTE-dependent molecular transitions in arrested 13 dag iMUTEmute lineages, we measured the expression levels of stomatal development landmark genes during upshift treatments. All of them (except STOMAGEN) have been previously described as strictly expressed in stomatal lineage cells, and thus their expression changes detected in whole cotyledons may be ascribed to stomatal lineage development. The results are consistent with the anatomical phenotypes observed (Figure 5d; see Figure S6 for relative expression profiles). Untreated 13 or 23 dag iMUTEmute cotyledons expressed only immature-lineage markers, including the mute-3 allele. Inducer-elicited gene expression changes in iMUTEmute were detected as early as 4 hpi and rapidly evolved during the following days. Transcripts of early lineage markers (SPCH, POLAR and BASL, and also EPF2 and TMM) immediately and steadily decreased after induction, reaching levels similar to mature iMUTEwt (23 dag) cotyledons. Markers for later stages (such as MUTE and EPF1) followed a different pattern as their expression increased transiently, starting at 4 h post-induction, peaked at 2 days post-induction, and decreased at 10 days post-induction. Guard cell markers such as FAMA, which is almost absent in untreated iMUTEmute, increased as guard cells formed; KAT1 showed a similar but less pronounced trend. Untreated 13 or 23 dag iMUTEwt cotyledons used as controls expressed mature stomata markers (such as FAMA or KAT1), but not MUTE, as they had stomata but no developing lineages. The expression time courses of lineage marker genes correlate well with the course of stomatal differentiation observed during the treatments, suggesting that, upon MUTE expression, halted mute-3 lineages synchronously undertake developmental pathways that involve molecular components similar to those of wild-type plants. Transcript levels for most genes were almost unaffected by simultaneous treatment with colchicine, with the exception of guard cell markers (notably KAT1, but also FLP), whose increases were lower than in induced non-colchicine-treated samples. Round GUS-expressing cells in these treatments did not possess pore-like structures or showed other features suggesting GC identity, and appeared to be blocked at a GMC-like stage, even 10 days after induction (Figure 4c).

Late MUTE induction differentially affects pavement cells and SLGC ploidy

The epidermal mute-3 phenotype is accompanied by compromised growth that leads to extremely dwarfed plants with tiny cotyledons and leaves. Organ growth in Arabidopsis is accompanied by cell expansion and endoreplication (Dewitte et al., 2007; Massonnet et al., 2011), and most PCs in the mature adaxial leaf epidermis are endopolyploid (70%; Melaragno et al., 1993) while GCs remain diploid. To determine how organ size and epidermal ploidy levels were influenced by MUTE expression, we inspected the area and relative DNA cell content of iMUTEwt and iMUTEmute cotyledons during upshift treatments (Figure 6).

Figure 6.

Late MUTE expression in iMUTEmute plants has distinct cell ploidy effects on putative SLGCs and pavement cells. Plants were grown in control medium for 13 days, inducer was added where indicated, and cotyledons were examined 10 days later. (a) Cotyledon area of iMUTEwt and iMUTEmute at various times and after various treatments. Values are means ± SD. (b–f) Ploidy levels. (b) Nuclear ploidy levels in whole iMUTEwt and iMUTEmute cotyledons measured by flow cytometry. (c) Adaxial epidermal cell ploidy distribution (excluding stomata and meristemoids). (d–f) Micrograph-based drawings of representative adaxial cotyledon epidermis of (d) iMUTEwt at 23 dag, (e) non-induced iMUTEmute at 23 dag, and (f) iMUTEmute grown for 13 days in control medium plus 10 days in inducer. Approximately 360 (iMUTEwt) or 670 (iMUTEmute) nuclei from at least eight individual plants were scored. Scale bar = 20 µm.

In non-induced iMUTEmute, cotyledon area increased marginally from 13 to 23 dag, but had doubled 10 days after induction of 13 dag plants (Figure 6a). Whole-cotyledon ploidy levels, which were lower in untreated 13 dag iMUTEmute than in iMUTEwt, also increased in upshifted iMUTEmute (Figure 6b), reaching a distribution similar to that of iMUTEwt.

We then quantified the DNA content of individual cells in the adaxial cotyledon epidermis. At 5 dag, nuclei from GCs (in iMUTEwt) and meristemoids (in non-induced iMUTEmute) had a similar size, which was established as the 2C DNA content and remained unchanged at 13 dag in both genotypes (Figure S7a). The remaining epidermal cells showed various ploidy levels. In 5 dag iMUTEwt, many were 4, 8 or 16C (Figure S8a,b), and appeared large and profusely lobed; only some cells in contact with stomata (most likely SLGCs) were small and diploid. Non-induced 5 dag iMUTEmute cotyledons showed a much higher 2C-cell proportion (Figures S7b and S8a); epidermal cells that were not in contact with meristemoids often had a higher DNA content and puzzle piece-like contours (Figure S8c), although most appeared smaller than in the wild-type. Cells surrounding meristemoids were always small and 2C, showing the characteristic morphology of mute-3 SLGCs.

By 23 days, iMUTEwt ploidy had increased and the non-stomatal 2C cell proportion was small (Figure 6c,d). At this time, the distribution of epidermal cell ploidy levels in untreated iMUTEmute had barely changed, showing mostly 2C cells (77.1%; n = 134 from eight cotyledons) and a small proportion of 4C cells, with infrequent 8C cells (Figure 6c,e). In contrast, 10 days after upshifting 13 dag iMUTEmute plants, the 2C cell proportion decreased and more 8C and even some 16C cells appeared (Figure 6c,f). By this time after the upshift, stomata, isolated or clustered, had formed. Some PCs away from stomata were large and inter-digitized; however, most cells adjacent to stomata/clusters remained diploid and unexpanded (Figure 6f). Thus, late MUTE expression in halted 13 dag iMUTEmute plants increased overall organ and PC ploidy, but putative SLGCs remained stunted and diploid.

Phenotypic mosaics illustrate a lineage-restricted role for MUTE in SLGCs and stomata-less leaf growth

To discriminate between physiological effects derived from stomata presence and differentiation signals linked to stomata lineage development, we generated phenotypic mosaics with mixed wild-type and mute-3 phenotypes (Figure 7). iMUTEmute plants were sown on inducer-containing plates and grown for 13 dag, before transfer to soil for 10 more days for further growth.

Figure 7.

Phenotypic mosaics generated in iMUTEmute plants. Seeds were plated on β-estradiol and plants were transferred to soil 13 days later; the fading gray color represents the induction level during growth in plates (see Figure S9 for induction level time course). The scheme of plant development during treatment shows the third leaf (yellow/green) as it develops exclusively under conditions of little or no or induction, while development of the first leaf (dark-green) is initiated under conditions of high induction and completed without induction. Under this treatment, the iMUTEmute plants are phenotypic mosaics. (a–d) Epidermal characterization of phenotypic mosaics. (a) DIC micrograph of a resin imprint of the adaxial surface of a 40-day-old iMUTEmute first leaf, showing wild-type (blue) and adjacent mute-3 stomatal lineages; note the large pavement cells. Scale bar = 100 µm. (b) Micrograph-based drawing of the adaxial epidermis of a representative 3rd leaf 10 days after transfer to soil. Scale bar = 20 µm. (c) Adaxial epidermal cell ploidy distribution and (d) area in third leaves of phenotypic mosaics and untreated plants.

We measured the degree of induction elicited by β-estradiol during seedling growth on the inducer-containing plates through GUS transcript level determination (Figure S9). Induction was maximal at 2 dag, had dropped to less than 50% at 6 dag, and was residual by 9 dag, in agreement with the described gradual β-estradiol inactivation under light (Zuo et al., 2000). Therefore, in iMUTEmute plants (Figure 7), organs or organ parts formed during the early days were phenotypically wild-type but those formed after 9 days were phenotypically mute-3.

iMUTEmute first-pair leaves, whose growth started under high induction levels (from 2–6 dag) but was completed under conditions of little or no induction (from 9 days onwards), were phenotypic mosaics in which wild-type and mute-3 lineages co-existed in close proximity (Figure 7a). Stomata surrounded by large lobed cells represent wild type-like lineages, presumably formed early during leaf development when the level of β-estradiol-dependent transcription was high; mute-3 rosette-like lineages presumably formed after the inducer effect (and thus MUTE expression) had faded. In these leaves, wild-type lineages did not seem to affect the small cells surrounding meristemoids in rosette-like arrangements (putative SLGCs) of nearby halted mute-3 lineages. Thus, these putative SLGCs showed distinct behavior as compared with PCs, and their differentiation appears to be strictly dependent on the correct timing of MUTE expression in their adjacent meristemoid.

As expected from their developmental timing, the third leaves in downshifted iMUTEmute plants had only mute-3-type stomatal lineages (Figure 7b) and an epidermal cell ploidy pattern similar to non-induced iMUTEmute (Figure 7b,c). However, the area of the third leaf was significantly larger than in untreated iMUTEmute (Figure 7d), although both lacked stomata and showed arrested lineages.

Developmental windows for cell-specific ability to respond to MUTE

The above results reveal distinct time-dependent responses of various lineage cell types to MUTE expression in its own promoter domain. To further examine the ability of epidermal cells to respond to ectopic expression of stomatal regulators depending on their identity and previous history, we constructed lines that express MUTE (iMUTEoe) or FAMA (iFAMAoe), in a β-estradiol-dependent manner, in the expression domain of the constitutive G10-90 promoter (Zuo et al., 2000) in Col-0, and confirmed inducer-dependent transgene expression by quantitative PCR (Table S3). Constitutive MUTE or FAMA over-expression transforms all aerial protodermal cells into either stomata (Pillitteri et al., 2007, 2008) or GCs (Ohashi-Ito and Bergmann, 2006), but to what extent differentiating epidermal cells undergo fate transitions normaly led by MUTE or FAMA has not been examined. iMUTEoe and iFAMAoe plants were grown in control medium for 0, 2, 4 or 10 days; upshift treatments were performed by inducer addition at various time points, the plants were grown for 4 more days, and their adaxial cotyledon epidermis was inspected (Figure 8).

Figure 8.

Timing of ectopic MUTE and FAMA over-expression affects the resulting epidermal phenotypes. iMUTEoe and iFAMAoe seeds were plated in control medium, inducer was added at the indicated days after germination (dag), and cotyledons were examined 4 days later. Confocal images show the adaxial cotyledon epidermis of iMUTEoe (a–d) and iFAMAoe (e–h). Note the presence of stomatal clusters (brackets) and aberrant stomata with distinct pores (arrowheads) in iMUTEoe. Arrows show large pavement cells with symmetric divisions in iMUTEoe (b, c) or pore-like structures in iFAMAoe (g, h). Note the absent or mild effects in plants upshifted after 10 (iMUTEoe) or 4 (iFAMAoe) dag. Cell contours were stained with propidium iodide. Scale bars = 50 μm.

In iMUTEoe, all treatments (except for induction after 10 dag) induced extra stomata. MUTE over-expression since germination reproduced the previously reported stomata-only phenotype (Figure 8a); induction after 2 dag produced stomatal clusters and large, abnormally shaped, stomata-like structures, suggestive of PC transdifferentiation (Figure 8b). Similarly, induction after 4 dag induced clusters and stomata-like structures; this treatment also produced ectopic symmetric divisions, which were particularly striking in large cells that divided along their longest axis instead of the shortest axis (Figure 8c). Thus, as growth time prior to induction increased, the effects of MUTE were milder and more restricted to clustered stomata, in spiral cell arrangements suggestive of mis-spaced lineages. PCs gradually lost the ability to respond to MUTE as they differentiated.

Ectopic FAMA over-expression in iFAMAoe since germination also produced the expected phenotype (Figure 8e). If inducer was added after 2 dag (Figure 8f), most epidermal cells underwent unpaired GC-like differentiation, but some stomata were also produced. Induction after 4 dag (Figure 8g) barely altered the epidermis, which contained apparently normal stomata and PCs, although some of the latter formed ectopic pore-like cell-wall differentiations (Figure 8g). Therefore, FAMA expression does not over-ride endogenous controls in most stomatal-lineage cells; in PCs, it elicits limited transdifferentiation compared with the striking conversion of protodermal cells to GCs.

Discussion

A versatile system for conditional cell type-specific MUTE expression

The conditional MUTE expression system developed in this work provides a useful tool to delineate the role of MUTE in the aerial epidermis at various stages of development (Figure 1 and Figure S2). In a wild-type MUTE background and over a wide range of concentrations, inducer did not produce aberrant epidermal development or altered stomatal abundance/distribution; induced plants maintained the cell-specific MUTE expression pattern. In the mute-3 background, β-estradiol treatment since germination fully complemented the mutant phenotype (Figure 1), allowing production of homozygous mute-3 seeds and thus circumventing the tedious propagation of mute alleles in segregating heterozygous lines and their experimental limitations. Using various iMUTEmute induction protocols, we obtained new information on stomatal lineage development. For instance, arrested mute-3 stomatal lineages responded to inducer quasi-synchronically, revealing features of early and late anatomical and transcriptional evolution of stomatal lineages, as well as the impact of these transitions on organ size and ploidy levels, and allowed mapping of changes in the DNA content of epidermal cells. We also produced phenotypic mosaics in iMUTEmute, with older wild-type photosynthetic organs and stomata-less young leaves. Such stomata-less leaves displayed the arrested lineages characteristic of mute-3 but showed increased growth, probably at the expense of stomata-bearing organs. Therefore, although normally stomatal lineages are major contributors to the PC population that constitutes most of the leaf surface, relatively large leaves may be built with mute-3 lineages, providing material to address physiology and development of stomata-less photosynthetic organs.

Halted mute-3 meristemoids fully retain their competence to respond to MUTE

Understanding of the meristemoid to GMC transition is limited; no mutant halted at the GMC stage is available, and asynchronous stomatal lineage development hinders identification of GMC-specific molecular signatures. To date, MUTE is the only gene that has been described as essential for this transition (MacAlister et al., 2007; Pillitteri et al., 2007). Arrested iMUTEmute meristemoids in 13 dag cotyledons were fully able to transit to GMCs upon induction, and produced normal stomata (Figure 2), in contrast with the inability of PCs in 10-day-old plants to respond to MUTE expression (Figure 8d) and the aberrant stomata induced by MUTE in protodermal cells (Figure 8a). Thus, halted meristemoids contain all the components needed for normal MUTE function, and/or produce them upon MUTE expression. These meristemoids rapidly respond to MUTE: using GUS as a cell-specific marker for MUTE expression, we found that the halted meristemoid to stomata transitions take place within 48 h, as in wild-type contexts (Robinson et al., 2011). This strongly supports previous hypotheses that halted mute meristemoids are indeed late meristemoids (Pillitteri et al., 2007), and demonstrates that this cell type retains its identity regardless of organ age or cell context.

Untreated 13 dag iMUTEmute plants accumulated early-lineage marker transcripts. Due to the synchronic stomatal lineage development of these plants upon inducer adition, we determined that early marker levels steadily decreased upon MUTE expression, in agreement with the observed course of anatomical transitions of meristemoids to stomata (Figure 5). All the markers used (except STOMAGEN) are lineage cell-specific, and their expression is restricted to the epidermal cell types that gradually develop during treatment. MUTE and EPF1 mark a later lineage stage (Hara et al., 2007; Pillitteri et al., 2007); unexpectedly, EPF1 expression responded to MUTE induction, although it was also expressed in untreated iMUTEmute, albeit at a low level and only in relatively young mutant plants. Thus, although MUTE is not strictly required for EPF1 expression (Pillitteri et al., 2011), it may reinforce it. Our data suggest that halted mute-3 lineages respond to MUTE through well-known regulators of normal stomata differentiation. Most gene expression trends were unaffected by colchicine-promoted metaphase arrest at the time of induction, indicating that cell division may be uncoupled from cell differentiation markers in young developing lineages. Notable exceptions are the guard cell-specific genes KAT1 and FLP, and to some extent FAMA, whose expression increased only slightly in the presence of colchicine; this agrees with the apparent GMC morphology of GUS-expressing cells (Figure 4c) and their lack of GC-like features, such as the crescent cell shape or pore-like wall structures as observed in iFAMAoe (Figure 8e). Colchicine may also interfere with the cytoskeleton dynamics required for cell-wall morphogenesis.

Ectopic late MUTE expression using iMUTEoe (Figure 8) demonstrated a remarkable ability of MUTE to induce symmetric divisions in non-lineage cells; upon induction, PCs in 2- or 4-day-old cotyledons divided along their longest axis rather than their shortest axis, a rare feature that is characteristic of GMCs. Abnormal PC divisions were observed after constitutive expression of E2FA (De Veylder et al., 2002) and inactivation of the RETINOBLASTOMA-RELATED protein (RBR) (Desvoyes et al., 2006), but most were along the shortest cell axis. Our observations add to the recently described uniqueness of the GMC division, the only one in the Arabidopsis epidermis where the phragmoplast position is not polarized towards the inner side of the epidermis (Lucas and Sack, 2012).

Timely MUTE expression prevents stomatal fate in SLGCs

In contrast with the apparently normal response of 13 dag mute-3 meristemoids to MUTE, many of their sister SLGCs differentiated into stomata, producing stomatal clusters (Figure 3). Treatment with colchicine resulted in clustered GUS-expressing cells with GMC-like morphology (Figure 4), ruling out the possibility that clusters originate from meristemoid or GMC divisions after induction; GUS-expressing clusters arise from previously adjacent cells that show GUS activity, i.e. the arrested meristemoid and its surrounding cells. Serial epidermal imprints of living cotyledons confirmed the clonal relationships of stomata and their adjacent GUS-expressing cells and of all stomata constituting a cluster, unequivocally demonstrating SLGC identity (Figure 5). Therefore, clusters appear because of the altered fate of halted 13 dag SLGCs. In wild-type plants, stomatal fate in SLGCs is prevented by patterning mechanisms (Dong and Bergmann, 2010) that seem dysfunctional in arrested mute-3 lineages. Halted meristemoids do not restrict stomatal fate in SLGCs, or halted SLGCs fail to perceive or interpret inhibitory meristemoid-borne signal(s). In these situations, even residual MUTE expression triggers stomata formation, an event that is perhaps favored because SLGCs remain 2C for a long time (Figure 6). Although more than 65% of the stomata were clustered, many remained isolated. This heterogeneity may stem from the asynchrony in stomatal development (Nadeau and Sack, 2002), with late-borne lineages being less prone to form clusters.

Growth as mute-3 for just 2-3 days elicits cluster formation upon MUTE induction; growth for 4 days without MUTE expression saturates this response (Figure 3d). By 4 dag, most adaxial cotyledon lineage divisions are completed (Geisler and Sack, 2002) and the whole organ shifts to the cell expansion phase; however, we found that pre-formed halted SLGCs and meristemoids can nevertheless divide symmetrically to form stomata. This fits with the reported uncoordinated proliferation of PCs and meristemoids (Andriankaja et al., 2012), and suggests a partly independent control of stomatal lineage-related divisions.

SLGC fate and differentiation require MUTE expression in a strict developmental- and lineage-dependent manner

mute-3 cotyledons develop some 8C cells, even as early as 5 dag (this study); thus, lack of MUTE does not compromise epidermal cell endoreplication per se. The low ploidy levels and small size of non-induced iMUTEmute cotyledons may stem from the compromised physiology and growth of stomata-less plants, as organ ploidy level and area were complemented by MUTE induction, which also produced stomata (Figure 6). PCs away from stomata in upshifted 13 dag iMUTEmute also showed increased ploidy 10 days after induction (Figure 6c,f). All these results indicate that MUTE-promoted stoma formation (and thus CO2 uptake) indirectly sustained organ and PC growth and endoreplication.

In contrast to PCs, most SLGCs (defined in anatomical terms) in upshifted 13 dag iMUTEmute stayed diploid despite the fact that they were in contact with isolated or clustered stomata and had thus been exposed to MUTE-expressing meristemoids. These SLGCs remain stunted or produce stomata, indicating that they cannot respond correctly to late MUTE expression. As SLGCs differentiate normally when MUTE is expressed at the onset of or very soon after germination, they gradually lose this ability in a mute-3 epidermis. The ability of leaf epidermal cells to divide in response to RBR inactivation was also partially lost as the epidermis matured (Desvoyes et al., 2006), and conditional interference with RBR expression resulted in over-proliferation of TMM::GFP-expressing small cells that did not resume stomatal development after RBR restoration (Borghi et al., 2010); thus, stomatal lineage cells lost their ability to respond to RBR because of transient disturbances in scheduled RBR levels, much as we found for mute-3 SLGCs with respect to MUTE expression.

Additional support for developmental and cell-specific constraints in the ability to respond to MUTE comes from conditional over-expression time courses (Figure 8a–d). While MUTE over-expression under the control of the 35S promoter (Pillitteri et al., 2007, 2008) or the G10-90 promoter (this study) transforms all protodermal cells into stomata, MUTE over-expression in phenotypically wild-type plants at 2 or 4 dag elicits distinct cell responses that include stomatal clusters (suggesting conversion of all lineage cells to stomata), but also symmetric divisions in very large PCs; however, many of these cells were no longer able to differentiate into stomata or did so to a limited extent, suggesting differential sensitivity to distinct MUTE functions. Similarly, the ability to respond to FAMA, a GC-fate driver (Ohashi-Ito and Bergmann, 2006; Hachez et al., 2011), has a developmental window. Constitutive FAMA over-expression converts all protodermal cells to GCs (Ohashi-Ito and Bergmann, 2006) (Figure 8e), but in 4-day-old plants it allowed stomatal lineage progression (Figure 8) and only promoted ectopic pores in some PCs (see also Hachez et al., 2011). Cotyledons at 10 dag appeared unable to respond to MUTE or FAMA ectopic expression. RBR protein inactivation at late developmental stages also had limited epidermal effects and only in some leaf cells (Desvoyes et al., 2006). It is notable that pavement and lineage cells retain their ability to respond to MUTE for longer than their ability to respond to FAMA. Their anatomy in upshifted 2 dag iMUTEoe also suggests that MUTE-driven divisions carry architectural information regarding differentiation of the cell wall common to the two sister guard cells, as this wall forms a correctly oriented stomatal pore.

Because mute mutants over-produce putative SLGCs, it is accepted that MUTE terminates the meristemoid divisions that produce SLGCs (Pillitteri et al., 2007). SLGC identity is marked by SPCH disappearance (Robinson et al., 2011) and transient peripheral location of BASL (Dong et al., 2009) and POLAR (Pillitteri et al., 2011). Although MUTE has not been shown to be involved in these processes, our results demonstrate that it plays an essential role in SLGC fate. This role operates in a narrow developmental window and probably depends on meristemoid-borne signals; transient lack of MUTE permanently disturbs the mechanisms that prevent stomatal fate, establish SLGC fate and allow exit from the stomatal lineage program/entry into PC differentiation. A cell-cycle blockage, as indicated by the fact that many halted SLGCs remained 2C and thus had not gone through S phase, may counteract endoreplication-related PC identity determinants and facilitate the acquisition of stomatal fate. Similarly, communication between stomata and SLGCs has been recently proposed to play a role in PC expansion (Tanaka et al., 2013).

In leaf phenotypic mosaics (Figure 7a), cells adjacent to stomata differentiated correctly into PCs, but these wild-type lineages did not rescue putative SLGCs in nearby arrested mute-3 lineages, suggesting that MUTE-dependent SLGC fate is acquired through strictly lineage-autonomous mechanisms. MUTE-expressing meristemoids may signal to their clonally related, adjacent SLGCs through apoplast or plasmodesmata-diffusible molecules. EPF1 is a candidate for an apoplast-diffusible SLGC identity determinant because it is expressed in the late meristemoids and its over-expression results in proliferation of SLG-like cells (Hara et al., 2007, 2009; Hunt and Gray, 2009). However, EPF1 expression increased upon late MUTE induction (Figure 5 and Figure S6), but this did not release halted SLGCs or prevent stomatal clusters. Symplast-transported determinants may also be involved in SLGC fate, as suggested for the formation of satellite lineages involving expression of miR824 and its target AGL16 in different cell lineage types (Kutter et al., 2007); loss-of-function chorus mutants show altered plasmodesmata traffic between meristemoids and their SLGCs, producing stomatal clusters, probably because of SPCH or MUTE leakage (Guseman et al., 2010).

Our results show that the identity of SLGCs and their ability to differentiate into a PC is established in a narrow developmental window that requires MUTE expression in their clonally related, short-lived late meristemoid and/or guard mother cell. Determining the molecular and cellular mechanisms for this function of MUTE on stomatal lineage development will provide additional insight into how epidermal cell differentiation is orchestrated.

Experimental Procedures

Plant material and growth conditions

Arabidopsis thaliana (L.) Heynh Columbia-0 seeds were obtained from the European Arabidopsis Stock Centre (http://arabidopsis.info/) (N1092). mute-3 was identified in our laboratory from an EMS-mutagenized Col-0 population on the basis of its stomata-less phenotype, similar to previously described loss-of-function mute mutants (MacAlister et al., 2007; Pillitteri et al., 2007). The recessive mute-3 allele carries a C→T change at position 923 that generates a premature stop codon, potentially producing a MUTE protein truncated at position 134. Plants were grown in soil or MS plates (Sigma, http://www.sigmaaldrich.com/) plus 1% sucrose and 0.8% plant agar, pH 5.7, under long-day conditions (16 h light/8 h dark) at 21°C, as described previously (Delgado et al., 2011) unless otherwise stated. AGI codes for genes described here are listed in Table S1.

Generation of transgenic plants

The β-estradiol-inducible, cell type-specific expression system for MUTE was generated using pLB12 and pMDC160 vectors (described by Brand et al., 2006) (Figure S2). For the activator unit, in which proMUTE drives XVE expression, 2 kb of sequence upstream of the MUTE start codon were amplified as described previously (Pillitteri et al., 2007) and cloned into pENTR™ D-TOPO® using the pENTR™ D-TOPO® cloning kit (Invitrogen, http://www.invitrogen.com/). The fragment was transferred to pLB12 (Brand et al., 2006), which carries the responder OlexA::GUS reporter fusion. For the MUTE responder unit, we used pDONR221 (Invitrogen™) containing 609 bp from the MUTE coding sequence (Arabidopsis Biological Resource Center, https://abrc.osu.edu/; stock number DQ446639), which was transferred into pMDC160 (Brand et al., 2006) under the control of the OlexA promoter. The constructs were used to transform Arabidopsis Col-0 (pLB12) or mute-3/+ (pMDC160) plants by the floral-dip method (Clough and Bent, 1998). Homozygous pLB12 T-DNA Col-0 and pMDC160 T-DNA mute-3/+ plants were crossed, and homozygous pLB12 and pMDC160 T-DNA plants identified as described previously (Kihara et al., 2006). Genotypes were confirmed by plating the F3 progeny on selective media, and checking for transgenes and the mute-3 allele using PCR and dCAPS, respectively (see Tables S1 and S2 for primers and enzymes used). Several combinations of proMUTE activator and MUTE responder lines were obtained and examined, selecting one line homozygous for pLB12 and pMDC160 T-DNAs and heterozygous for mute-3. By selfing this line, plants homozygous for either the mute-3 allele (iMUTEmute plants) or the wild-type allele (iMUTEwt) were obtained.

β-estradiol-inducible over-expression lines that express MUTE or FAMA under the control of the constitutive pG10-90 promoter (Ishige et al., 1999) were derived from pER8 (Zuo et al., 2000) within the TRANSPLANTA consortium (http://bioinfogp.cnb.csic.es/transplanta_dev/) , and termed iMUTEoe and iFAMAoe, respectively.

β-estradiol treatments

Phenotype complementation, selection of triple homozygous pLB12 T-DNA/pMDC160 T-DNA/mute-3 plants and initial experiments were performed by germination on MS plates containing 2 μm 17-β-estradiol (Sigma) unless otherwise indicated. For inducer upshift treatments, β-estradiol was added to plates on which plants were growing, to a final concentration of 2 μm; colchicine was added to the plates similarly, to a final concentration of 0.005%. Plants transferred to soil were treated by applying 2 μm 17-β-estradiol in 0.01% Silwet-77 (Lehle Seeds, http://www.lehleseeds.com/) using a paintbrush every 2 days.

GUS staining, microscopy and quantitative phenotyping

In situ GUS activity was detected as described previously (Jefferson et al., 1987). Epidermal observations of GUS-stained specimens and dental resin imprints were performed as described previously (Delgado et al., 2011, 2012). For differential interference contrast (DIC), we used a Nikon Eclipse 90i microscope (http://www.nikon.com/). Images were recorded using a DXM 1200C camera (Nikon) and Nis-Elements BR software (Nikon). Confocal microscopy after propidium iodide staining was performed as described previously (Delgado et al., 2012) using a Leica TCS SP2 confocal microscope (http://www.leica-microsystems.com/). Organ areas, stomatal index and density were determined as described previously (Delgado et al., 2011). A stomatal unit consists of either a normally spaced stoma or an entire cluster of stomata (Yang and Sack, 1995; Delgado et al., 2012). Cell counts were performed using imagej (http://rsbweb.nih.gov/ij/) or Geographic Information System and Arcview mapping software from ESRI (http://www.esri.com/). At least ten independent plants were used for all determinations. Statistical treatments were performed using spss version 17.0 (http://www-01.ibm.com/software/analytics/spss/products/statistics/). The index and density data comply with normality and variance homogeneity criteria. Phenotypic differences were tested statistically using Student's t test.

In vivo epidermal imprints

iMUTEmute seedlings were grown for 13 days on MS plates. Adaxial cotyledon epidermal imprints were obtained using dental resin as described previously (Delgado et al., 2012) at 2, 6 and 13 dag. Plants were then transferred to β-estradiol-containing plates, grown for 3 or 10 more days, and stained for GUS activity. Individual GUS-positive stomatal units were identified, and their cell division history was tracked back in nail-polish replicas of resin imprints from 13 dag to 2 dag by bright-field microscopy. iMUTEwt plants were germinated on β-estradiol-containing plates, imprints were taken at 1 and 2 dag, and GUS staining was performed at day 3. At least ten lineages corresponding to clusters or isolated stomata were scored per genotype.

Ploidy measurements

Cotyledons were chopped with a blade in nuclei isolation buffer (Galbraith et al., 1991), and the suspension was filtered through 30 μm nylon mesh, RNase A-treated (200 mg ml−1), and stained with propidium iodide (50 μg ml−1). Nuclear DNA content was analyzed using an XL-MCL flow cytometer (Beckman Coulter Inc., https://www.beckmancoulter.com/). For in situ relative DNA content quantification, cotyledons were clarified overnight in 100% ethanol, rehydrated for 30 min, stained with 0.5 mg ml−1 4,6-diamidino-2-phenylindole dihydrochloride for 5 sec at room temperature, rinsed, and observed using a Nikon Eclipse 90i microscope equipped with a fluorescence unit. Fluorescence images used for DNA quantification were the maximum projection of at least seven focal planes covering the whole adaxial epidermis thickness; nuclei from at least eight independent plants were counted for each treatment. Nuclei size was measured using imagej. Mean nuclei sizes from guard cells in Col-0 and meristemoids in mute-3 at two time points established the 2C DNA content (Figure S7). 4C, 8C and 16C ploidy levels were defined as 2C multiples.

RNA isolation and quantitative RT-PCR

Total RNA was extracted from frozen cotyledons using Trizol (Invitrogen) and cleaned using RNeasy mini-columns (Qiagen, http://www.qiagen.com/). cDNA was obtained from 2 μg total RNA using a High-Capacity cDNA Archive Kit (Applied Biosystems, http://www.appliedbiosystems.com/). Quantitative RT-PCR was performed with an ABI Prism 7500 real-time PCR system (Applied Biosystems), using Power SYBR® Green PCR Master Mix (Applied Biosystems). The primer sets used are given in Table S1. Three biological replicates were analyzed. CT values were obtained using the 7500 system sds software version 1.3 (Applied Biosystems). Relative expression changes were determined by the comparative CT method, where the fold change is calculated as inline image. UBC10 (At5 g53300) was used as the reference gene.

Acknowledgements

We thank Ana Rapp for technical assistance, José León (Instituto de Biología Molecular y Celular de Plantas, Consejo Superior de Investigaciones Científicas, Valencia, Spain) and the TRANSPLANTA consortium for contributing to obtain MUTE and FAMA over-expressing lines, Mark Curtis (Institute of Plant Biology and Zürich-Basel Plant Science Centre, University of Zürich, Switzerland) for the vectors used in construction of the dual-inducible system, and the Nottingham Arabidopsis Stock Centre for providing Col-0 seeds. This work was supported by grants from the Spanish Government (BIO2007-60276 and BIO2012-33952 to C.F. and M.M., BFU2009-9783 to C.G., and CSD2007-00057 TRANSPLANTA to C.F.), the Castilla la Mancha Government (PAI07-0036-3278 to M.M. and pre- and postdoctoral grants to M.T. and I.B.), and an institutional grant from the Fundación Ramón Areces to the Centro de Biología Molecular Severo Ochoa.

Ancillary