The chemical compound bubblin induces stomatal mispatterning in Arabidopsis by disrupting the intrinsic polarity of stomatal lineage cells

Asymmetric cell division produces two cells with different cell fates, thus ensuring cellular diversity in multicellular organisms (Knoblich, 2008). In plants, because of the immobility of their cells, body development strictly relies on spatial and temporal coordination of asymmetric cell divisions (De Smet and Beeckman, 2011). Stomatal development is considered a model system for investigating plant asymmetric division. Stomata are surrounded by guard cells and act as valves for gas exchange between the plant body and atmosphere (Pillitteri and Dong, 2013). Stomata are formed by the postembryonic development of protodermal epidermal cells, and are distributed according to a one-cell spacing rule, with at least one pavement cell between stomata. This stomatal patterning is physiologically important for efficient gas exchange (Dow et al., 2014).

The stomatal lineage originates from an epidermal stem cell that undergoes asymmetric divisions to produce a stomatal precursor cell, called a meristemoid, and a non-stomatal cell, which expands to become a pavement cell (Fig. 1A) (Pillitteri and Dong, 2013). Subsequently, the meristemoid undergoes asymmetric division and eventually differentiates into a guard mother cell (GMC), which divides equally to produce a pair of guard cells. Recent studies have revealed many of the components, including transcription factors, secretory peptides and receptors, that are involved in the regulation of stomatal cell fate transitions and division patterns in Arabidopsis (Lau and Bergmann, 2012; Pillitteri and Torii, 2012). Three paralogous basic helix-loop-helix (bHLH) transcription factors, namely SPEECHLESS (SPCH), MUTE and FAMA, regulate distinct steps in development: entry of stomatal lineage, fate transition from meristemoid to GMC, and terminal division to produce guard cells, respectively (MacAlister et al., 2007; Ohashi-Ito and Bergmann, 2006; Pillitteri et al., 2007). These steps are also regulated by positional cues, including secretory peptides such as the positive regulator STOMAGEN and the negative regulators EPIDERMAL PATTERNING FACTOR 1 (EPF1) and EPF2, and the cell-surface receptors TOO MANY MOUTHS (TMM), ERECTA (ER), ER-LIKE1 (ERL1) and ERL2 (Torii, 2012). Conversely, only two polarity factors have been identified to date: BREAKING OF ASYMMETRY IN THE STOMATAL LINEAGE (BASL) (Dong et al., 2009) and POLAR (Pillitteri et al., 2011). In the pre-mitotic stage of a stomatal stem cell, BASL is first observed in the nucleus and subsequently localizes to one side of the cell cortex or periphery. When the cell divides, peripheral BASL is inherited by only one of the daughter cells – the cell that will differentiate into a non-stomatal cell. Although this much is known, it remains unclear how the pre-mitotic cell polarity of the stomatal precursor is established and how it regulates asymmetric cell division.

Chemical genetics offers an alternative approach to circumvent problems of conventional genetics, such as redundancy and lethality (Schreiber, 1998; Stockwell, 2000). Chemical compounds can be applied in limited time and space, and are thus recognized as ideal probes for investigating developmental processes. In plant developmental biology, chemical screenings have identified a series of small molecules that affect root architecture or shoot regeneration (Serrano et al., 2015); however, no molecule has been found that affects asymmetric cell division. Stomatal development occurs in the outermost cell layer, which is easily accessed by externally applied chemicals. This feature makes stomatal development an ideal system for investigating asymmetric division via a chemical genetic approach. In this study, we used high-throughput phenotypic screening of a chemical library to identify a small pyridine-thiazole derivative capable of perturbing stomatal patterning. This compound, designated bubblin, affected stomatal asymmetric division, resulting in the generation of two identical daughter cells. Our results suggest that bubblin induces stomatal lineage cells to divide without the establishment of stem cell polarity.

A high-throughput screening system for chemical compounds that affect stomatal patterning was established using the Arabidopsis thaliana transgenic line E1728 (Gardner et al., 2009) (Fig. S1). The E1728 line has GFP-labeled guard cells, which enabled us to easily detect abnormal stomatal patterning by inspecting cotyledons under a fluorescence stereomicroscope. From a total of 3650 bioactive small molecules in a chemical library (Library of Active Compounds on Arabidopsis) (Zhao et al., 2007), we isolated a pyridine-thiazole derivative, 4-(4-bromophenyl)-2-pyridin-3-yl-1,3-thiazole (Fig. 1B), which affected stomatal patterning. We named the compound bubblin because it induced the formation of bubble-shaped clusters of stomata on the cotyledon (Fig. 1C,D).

On the epidermis of 10-day-old cotyledons of a transgenic plant expressing the plasma membrane marker GFP-LTI6b (Cutler et al., 2000), bubblin induced formation of stomatal clusters in a dose-dependent manner (Fig. 1E-G). When the epidermis was treated with 10 μM bubblin, over 20% of the stomata were found in clusters composed of more than five stomata (Fig. 1H). The stomatal clustering was induced by 1 μM bubblin, but not by 100 nM bubblin (Fig. S2). Thus, the minimum bubblin concentration required to affect stomatal development is in the range of 100 nM to 1 μM. Smaller clusters composed of two or three stomata were generated in mock-treated epidermis. This observation was consistent with a previous report that the sucrose in growth medium induces clustered stomata (Akita et al., 2013).

We next examined the heat stability of bubblin. HPLC analysis revealed that bubblin was not degraded by heat treatment at 90°C for 1 h (Fig. S3A). The heat-treated bubblin retained stomatal cluster-inducing activity (Fig. S3B). Additionally, stomatal clustering was enhanced by longer exposure to bubblin, and cluster formation stopped when the plants were transferred to regular medium (Fig. S4). These results indicate that we can easily manipulate the strength of the bubblin effect by treatment time. Bubblin also induced the formation of stomatal clusters on juvenile leaves (Fig. S5A) and hypocotyls (Fig. S5B), indicating that stomatal clustering was induced by bubblin in a tissue-independent manner. However, bubblin had less effect on hypocotyls than on cotyledons or leaves. Bubblin also affected cells other than stomata: inhibition of seedling growth (Fig. S5C), hypocotyl cell elongation (Fig. S5B) and elongation of cultured tobacco cells (Fig. S5D) were also observed.

Bubblin is composed of three aromatic rings: a pyridine, a thiazole and a bromophenyl ring. Of the five derivatives of bubblin, A1 to A5 (Fig. 2A), only A1, which differs in pyridine ring conjugation, showed bubblin-like stomatal clustering activity (Fig. 2B). By contrast, A2, in which the pyridine ring is replaced by a phenyl ring, showed no stomatal clustering activity, even at 50 µM (Fig. S6). These results indicate that the pyridine ring of bubblin is essential for its activity. Removal of the bromo group from the bromophenyl ring (A3) or substitution of the ring to a methylphenyl ring (A4) reduced stomatal clustering activity, and substitution to a methoxyphenyl ring (A5) resulted in activity loss (Fig. 2B,C, Fig. S6). Accordingly, these results indicate that the bromo group is required for the high activity of bubblin. Compounds with stomatal clustering activity inhibited seedling growth in a dose-dependent manner, whereas compounds without stomatal clustering activity did not inhibit seedling growth (Fig. S7).

A structurally analogous compound, fatostatin (Fig. S8A), was previously identified as an inhibitor of the sterol biosynthesis regulator SCAP (sterol regulatory element-binding protein cleavage-activating protein) in animal cells (Kamisuki et al., 2009). When cotyledons were treated with fatostatin, stomatal clusters did not form, even at 50 µM (Fig. S8B). Additionally, fatostatin had little effect on seedling growth (Fig. S8C). Fatostatin, which has no effect on stomatal development, has a propyl group conjugated to the pyridine ring, whereas A4, which partially induces stomatal clustering, has the same methylphenyl ring structure as fatostatin. These observations revealed that modifying the pyridine ring of bubblin reduces its stomatal clustering activity. Our results suggest that the potency of bubblin is strictly correlated to its structure.

We found that bubblin-treated younger (7-day-old) cotyledons formed excessive numbers of mispatterned (clustered), small, square cells (Fig. S9). To investigate how this patterning occurred, we examined the divisions of stomatal lineage cells over 36 h. During stomatal stem cell asymmetric divisions, one daughter cell retained stem cell activity and underwent subsequent asymmetric divisions, whereas the other typically ceased mitotic division and expanded to become a pavement cell (Fig. 3A, Movies 1-4). By contrast, in bubblin-treated epidermis, both of the daughter cells divided and the integration of these aberrant divisions generated clusters of small cells (Fig. 3B, Movies 5-8).

To quantify this effect of bubblin, we categorized cell divisions into five types according to the fates of the daughter cells. Divisions resulting in two daughter cells with different cell fates were categorized into three types: type 1, stomatal stem cell and pavement cell; type 2, stoma and pavement cell; type 3, stoma and stomatal stem cell. Divisions resulting in two cells differentiating to the same fate were categorized into two types: type 4, stomatal stem cell; type 5, stoma (Fig. S10). Quantitative analysis revealed that bubblin application increased the occurrence of type 4 divisions over fivefold (Fig. 3C). In bubblin-treated epidermis, over half of the divisions were categorized as type 4 or type 5, resulting in two daughter cells with the same characteristics in stomatal lineage.

In Arabidopsis, three bHLH transcription factor genes, namely SPCH, MUTE and FAMA, are expressed in sequence during stomatal development (Fig. S11A). Using these stomatal-specific markers (ProSPCH:NUC-GFP, ProMUTE:NUC-GFP and ProFAMA:VENUS-N7), we observed that bubblin induced the formation of clusters of small cells that exhibited stomatal lineage activity (Fig. 4A). These observations suggest that bubblin-induced stomatal clusters are formed under the regulation of stomatal-specific bHLH transcription factors. Analysis of endogenous expression levels of stomatal lineage-specific genes also supported ectopic development of stomatal precursors. Quantitative RT-PCR analysis revealed that SPCH, MUTE and FAMA were upregulated after 24 h of bubblin treatment (Fig. 4B), suggesting that bubblin caused the overproliferation of stomatal precursor cells.

To identify the stage of stomatal development affected by bubblin, we investigated its effects in various genetic backgrounds. Bubblin had no effect on spch epidermis, in which no asymmetric divisions occurred and only jigsaw puzzle-shaped epidermal cells were present (Fig. 4C). Similarly, bubblin had no effect on the epidermis of EPF2 overexpressors (Fig. S11B). EPF2 encodes a secretory peptide that represses SPCH expression, so the epidermis of EPF2 overexpressors phenocopies that of spch. These results indicate that bubblin acts on stomatal precursors produced by SPCH. By contrast, bubblin application drastically affected mute epidermis. The epidermis of mute displays arrested triangular meristemoids with inward spiral structures as a consequence of asymmetric divisions. Bubblin-treated mute epidermis exhibited overproliferation of meristemoid-like small cells, which lacked the inward spiral structure (Fig. 4D). These observations suggest that bubblin affects stomatal asymmetric divisions and induces disorganized patterning in meristemoids. Similar overproliferation of arrested meristemoids was observed on the epidermis of bubblin-treated EPF1 overexpressors (Fig. S11B). Overexpression of EPF1 results in arrested meristemoids and the absence of stomata on the epidermis. fama produces stacked narrow cells known as fama tumors. When fama epidermis was treated with bubblin, both fama tumors and bubblin-induced small cell clusters were observed (Fig. 4E). Regarding the fama tumors, there were no significant differences in their dividing patterns between bubblin-treated and mock-treated epidermis (Fig. S12). These results were consistent with the idea that bubblin acts at the early stage of stomatal development, i.e. asymmetric cell division.

Asymmetric cell division occurs in various tissues after embryogenesis. One representative example is the asymmetric division of the cortex/endodermis initial cell, which regulates root radial patterning in Arabidopsis. We applied bubblin to the root apical meristem of the endodermis marker line ProSCR:GFP-γTIP, but found no evidence of excessive cell layers or abnormal cell divisions (Fig. S13). These observations suggest that bubblin does not affect asymmetric cell division in the root apical meristem.

Asymmetric cell divisions are regulated by two fundamental mechanisms: intrinsic cell polarity and extrinsic positional cues (Knoblich, 2008). During plant stomatal development, both mechanisms are well orchestrated and determine the distribution of meristemoids (Facette and Smith, 2012). To investigate whether bubblin affects the extrinsic pathway or intrinsic polarity, we first monitored the expression pattern of STOMATAL DENSITY AND DISTRIBUTION 1 (SDD1), a specific marker of GMCs and early guard cells (Fig. 5A) (Von Groll et al., 2002). In bubblin-treated epidermis, ProSDD1:GFP signals were observed in adjacent cells that subsequently formed a stomatal cluster (Fig. 5B). By contrast, when we applied stomagen, an extrinsic positive regulator of stomatal development, ProSDD1:GFP signals were observed in non-adjacent cells (Fig. 5C, Fig. S14A). Further analysis revealed that ProSDD1:GFP signals were present in a GMC or a pair of guard cells that were often adjacent to other stomatal lineage cells, such as meristemoids or mature guard cells, in stomagen-treated epidermis. These observations suggest that bubblin and stomagen cause stomatal clustering by different mechanisms. Additionally, in tmm epidermis, which has defects in the extrinsic signaling pathway involving stomagen (Lee et al., 2015), ProSDD1:GFP-positive cells were not attached to each other (Fig. 5D). These different expression profiles of ProSDD1:GFP suggest that bubblin affects a pathway that is independent of the extrinsic cell-to-cell signaling pathway involving stomagen and TMM (Fig. S14B).

Stomatal asymmetric divisions are accompanied by asymmetric degradation of SPCH, a stomatal lineage determinant, after cell division (Robinson et al., 2011). SPCH is expressed in both daughter cells after cytokinesis, but is immediately degraded in one cell, which subsequently grows in size to form a pavement cell. In mock-treated epidermis, SPCH-GFP signals were observed in a pair of cells and a single cell after division (Fig. 5E). By contrast, in bubblin-treated epidermis, SPCH-GFP signals were observed in attached small cells, suggesting that SPCH remained in daughter cells following cell divisions (Fig. 5F).

Recently, SPCH degradation was suggested to be regulated by the polarity factor BASL, which is specifically localized at the cell periphery (Zhang et al., 2015). In the pre-mitotic stomatal precursor cell, BASL is observed in the nucleus and at the cell periphery. After cell division, nuclear BASL is seen in both daughter cells, but peripherally localized BASL is maintained in only one of the daughter cells (Fig. S15A, Movie 9) (Dong et al., 2009). Thus, the peripheral BASL determines the polarity of the stomatal stem cell. In bubblin-treated epidermis, BASL was localized to the nuclei of the clustered small cells but did not exhibit polar localization at the periphery (Fig. S15B, Movies 10 and 11). To quantify this effect of bubblin, we examined cell pairs expressing nuclear BASL and assessed whether they maintained peripheral BASL (Fig. 5G). Bubblin application drastically increased the proportion of cell divisions without peripheral BASL. These results suggest that bubblin disrupts pre-mitotic cell polarity in stomatal lineage cells to produce stomatal clusters. BASL-deficient mutants have also been shown to exhibit stomatal clusters (Dong et al., 2009). We found that bubblin enhanced the stomatal clustering phenotype in basl mutants (Fig. S16). These results suggest that bubblin affects one or more factors other than BASL that regulate stomatal patterning, although the possibility that bubblin could directly inhibit peripheral BASL accumulation is not excluded.

Asymmetric division is a fundamental mechanism to generate cellular diversity. To our knowledge, this is the first study in which a large-scale chemical library has been screened to isolate compounds affecting asymmetric division in stomatal patterning. We established a high-throughput screening system using intact Arabidopsis seedlings that identified bubblin as a stomatal clustering molecule. Bubblin, or a product of its metabolism, perturbed stomatal asymmetric division and induced the formation of mispatterned stomatal precursors. Bubblin affected cell elongation throughout Arabidopsis seedlings and in cultured tobacco cells, but showed no effects on asymmetric division in root apical meristem. From these data, we hypothesize that bubblin has a dual effect on cell elongation and stomatal asymmetric cell division.

Stomatal clusters are also generated in TMM-deficient mutants or by excessive application of stomagen (Sugano et al., 2010; Yang and Sack, 1995); however, in this study we revealed that GMC clusters are rarely formed in these cases. These observations are consistent with previous reports that extrinsic signals in stomatal development mainly regulate the number of stem cells and the orientation of their divisions, but not intrinsic cell polarity (Geisler et al., 2000). By contrast, bubblin abolishes the asymmetry of cell division and causes stomatal precursor clustering, suggesting that bubblin acts on cell polarity independently of extrinsic signaling. Meristemoid cell polarity is generated by BASL that localizes to the cell periphery and is predicted to regulate SPCH degradation (Zhang et al., 2015). Bubblin induced stomatal lineage cells to divide without establishing BASL polarity, and this accounted for the retention of SPCH, which resulted in stomatal clustering (Fig. 5H). Our results suggest that a possible target of bubblin is involved in regulating stem cell polarity establishment prior to cell division in stomatal development.

To achieve successful asymmetric cell division, cell polarity establishment and cell cycle progression must be tightly coupled. In animals, polarized cells only divide after polarity is fully established, so that polarized elements are correctly segregated into the daughter cells (Noatynska et al., 2013). In stomatal lineage cells, however, how the cell cycle is coordinated with specific cell polarity factors has not been clarified. Robinson et al. (2011) suggested that the re-establishment of peripherally localized BASL is essential for achieving the stomatal one-cell spacing pattern. Our findings suggest that bubblin acts on the fine-tuned coordination between the re-establishment of peripheral BASL and cell cycle progression in stomatal lineage cells. One possible mode of action of bubblin is that it accelerates cell division in meristemoids, even in the absence of cell polarity, by abolishing the activity of some negative regulator of the cell cycle. Our time-lapse analysis supports this idea because stomatal lineage cells treated with bubblin divided more often than control cells: bubblin-treated cells underwent three rounds of asymmetric division in 36 h (Fig. 3B), whereas asymmetric division occurred only once in control cells (Fig. 3A). Arabidopsis broadly expresses a protein called RETINOBLASTOMA-RELATED (RBR), which restricts cell cycle progression (Gutzat et al., 2012). Reducing RBR function results in excessive division of stomatal lineage cells at both early and terminal stages (Borghi et al., 2010; Lee et al., 2014; Matos et al., 2014). Bubblin affects early stomatal lineage cells that undergo asymmetric divisions; however, it has little effect on mature guard cells. Hence, bubblin appears not to act on RBR itself, but on cell cycle regulators that function in a narrower range of stomatal development.

Another possibility is that bubblin inhibits polarity factors, including BASL. Peripheral localization of BASL requires MPK3/6-mediated phosphorylation (Zhang et al., 2015). Bubblin might act as a kinase inhibitor to perturb BASL phosphorylation or regulate the targeting of phosphorylated BASL to the cell periphery. Bubblin abolished the peripheral accumulation of BASL in stomatal lineage cells, but still enhanced the stomatal clustering phenotype when it was applied to basl mutants (Fig. 5G, Fig. S16). Previous reports suggest that an alternative mechanism, acting in parallel to the BASL pathway, regulates stomatal stem cell polarity because the stomatal clustering defects in basl mutants are milder than those observed in transformants with defects in SPCH degradation (Lampard et al., 2008; Zhang et al., 2015). Bubblin might inhibit the multiple determinants of cell polarity by affecting protein accumulation at the periphery of stomatal cells.

At present, any direct target of bubblin and its inhibitory mechanism remain unknown. Previous studies in animals show that some pyridine-thiazole molecules exhibit bioactivity by binding to their target proteins. Fatostatin directly interacts with the endoplasmic reticulum (ER)-localized protein SCAP and inhibits the ER-Golgi translocation of sterol regulatory element-binding proteins (SREBPs) (Kamisuki et al., 2009). Pyridine-thiazole analogs of resveratrol bind to the active site of aromatase, an estrogen synthase (Mayhoub et al., 2012). Bubblin might interact with a target protein, either membrane-bound or soluble, and perturb its enzymatic activity or capacity to engage in protein-protein interactions.

Although it is important to establish cell polarity prior to asymmetric division, there are few studies to support this event in plant development. Recently, there have been a number of reports on the use of chemical probes to investigate plant development (Serrano et al., 2015); however, no molecule had been found that affects stem cell polarity. Bubblin is the first chemical tool for investigating asymmetric division and cell polarity in plants. Furthermore, bubblin caused stomatal lineage cells to divide symmetrically, generating two cells with stomatal cell fates. The spatiotemporal application of bubblin enables us to artificially direct stomatal stem cells to divide symmetrically, and to regulate the number of stomatal precursors. Bubblin is a potentially valuable tool for clarifying the mechanism underlying cell polarity establishment during asymmetric cell division in plants.

We screened the 96-well format chemical library LATCA (Library of Active Compounds on Arabidopsis) (Zhao et al., 2007). This library was assembled from three commercial libraries: Diverset (ChemBridge), LOPAC (Sigma-Aldrich) and Spectrum (Microsource Discovery Systems). The LATCA library consists of 3650 small molecules and all compounds were dissolved in dimethyl sulfoxide (DMSO) at 2.5 or 5 mM. For screening, each chemical compound was added at 1:255 dilution to liquid medium containing 1% (w/v) sucrose and Murashige and Skoog (MS) salts (Wako) and incubated with a few E1728 seeds (Gardner et al., 2009). Black-well plates and cotton balls in each well helped E1728 seedlings to grow straight, enabling easy observation of whole cotyledons. After incubation at 22°C for 7 or 8 days, stomatal patterning in cotyledons was observed under an Olympus MVX10 fluorescence stereomicroscope.

Bubblin [4-(4-bromophenyl)-2-pyridin-3-yl-1,3-thiazole] was purchased from Key Organics (MS-6628). Additional analogous compounds were obtained from other sources: A1 [4-(4-bromophenyl)-2-pyridin-4-yl-1,3-thiazole;hydrobromide] from Maybridge (ML00034); A2 [4-(4-bromophenyl)-2-phenyl-1,3-thiazole] and A5 [4-(4-methoxyphenyl)-2-pyridin-3-yl-1,3-thiazole] from ENAMINE (Z48854171 and Z48847277); A3 (4-phenyl-2-pyridin-3-yl-1,3-thiazole) from Alfa Aesar (H51748); A4 [4-(4-methylphenyl)-2-pyridin-3-yl-1,3-thiazole] from Vitas-M Laboratories (STK723938); and fatostatin hydrobromide from Sigma-Aldrich (F8932).

Bubblin was dissolved in DMSO (Nacalai Tesque) at 10 mM and incubated at 22°C or 90°C for 1 h, then these samples were diluted 1:10 in DMSO and subjected to HPLC. The HPLC analysis was performed on a PU-2089 Plus series system (JASCO) using a CHEMCOBOND 4.6 mm×150 mm 5-ODS-H column (Chemco Plus Scientific) in 0.1% (v/v) trifluoroacetic acid in water, with a linear elution gradient of 0-100% acetonitrile at a flow rate of 1.0 ml/min for 20 min with detection at 254 nm.

Arabidopsis thaliana Columbia-0 (Col-0, CS60000) ecotype was used as wild type except for mute-2 where Ws-4 was used. T-DNA insertion mutants and an enhancer trap line were obtained from the following sources: SAIL_36_B06 (spch-3), SALK_100073 (fama-1) and SALK_011958 (tmm) from the Arabidopsis Biological Resource Center (ABRC) at Ohio State University; FLAG_225D03 (mute-2) from the French National Institute for Agricultural Research (INRA); and E1728 from the Nottingham Arabidopsis Stock Centre (NASC). Transgenic plants used were: Pro35S:GFP-LTI6b (Cutler et al., 2000) and ProSCR:GFP-γTIP (Saito et al., 2005). Seeds of SPCH:SPCH-GFP (ProSPCH:SPCH-GFP), SPCH:nucGFP (ProSPCH:NUC-GFP), MUTE:nucGFP (ProMUTE:NUC-GFP), BASL:GFP-BASL (ProBASL:GFP-BASL) and basl-2 were kindly provided by Dominique C. Bergmann (Stanford University). The Gateway cloning system (Life Technologies) was used to construct pGWB1 ProFAMA:VENUS-N7, pB2GW7 Pro35S:EPF1, pB2GW7 Pro35S:EPF2 and pGWB250 ProSDD1:GFP. The sequence of VENUS fused to N7 was amplified by PCR using primers 5′-CAAAGGGTGGGCGCGATGGTGAGCAAGGGCGAG-3′ and 5′-AGCTGGGTCGGCGCGTTACTCTTCTTCTTGATC-3′ from a VENUS-N7 sequence cloned in Pro35S:DII-VENUS-N7 (Brunoud et al., 2012). The In-Fusion HD Cloning Kit (Takara Bio) was used to introduce the DNA sequence of VENUS-N7 into the AscI site of the pENTR/D-TOPO FAMApro plasmid (Shirakawa et al., 2014). The pENTR/D-TOPO ProFAMA:VENUS-N7 plasmid was introduced into the binary vector pGWB1 (Nakagawa et al., 2007) using the LR reaction. The coding regions of EPF1 and EPF2 were amplified by PCR using primers 5′-CACCATGAAGTCTCTTCTTCTCCTT-3′ and 5′-GCTTATTTCCTTTCAAGGGACAGGGTAGGA-3′ for EPF1; and 5′-CACCATGACGAAGTTTGTACGCAA-3′ and 5′-GGGCCAATAGCATTTAATTTAATATCCC-3′ for EPF2. These fragments were cloned into pENTR/D-TOPO and introduced into pB2GW7 (Karimi et al., 2002) using LR reactions. The promoter region of SDD1 spanning −2039 bp to +3 bp relative to the start codon (+1) was prepared by PCR with primers 5′-CACCATGGGGTGAGCTGTGTG-3′ and 5′-CATTGGAGAGAGTTAAAAAAAGGAG-3′. The amplified fragment was cloned into pENTR/D-TOPO to produce pENTR/D-TOPO ProSDD1. Then, the plasmid was subjected to the LR reaction with the destination vector pGWB250 (Keerthisinghe et al., 2015). The constructs were introduced into plants using Agrobacterium tumefaciens-mediated transformation via the floral-dip method (Clough and Bent, 1998). Plants expressing ProSDD1:GFP were crossed with tmm to observe ProSDD1:GFP patterning in the tmm background. The genotype of this line was confirmed by PCR amplification using the gene-specific primers 5′-GGGCATGTGCTTTTCGTATGCACAGAC-3′ and 5′-AAAGACTTGGAAGTGCTTGACACG-3′, and the T-DNA-specific primer 5′-TGGTTCACGTAGTGGGCCATCG-3′.

Seeds were surface-sterilized with 70% ethanol and then sown onto 0.5% (w/v) gellan gum (Wako) containing 1% (w/v) sucrose and MS salts. The seeds were incubated at 22°C under continuous light, and then transferred onto vermiculite for growth in 16 h light/8 h dark cycles at 22°C. In chemical assays, sterilized seeds were sown onto solid medium as described above and incubated at 4°C for 2-3 days to break seed dormancy. Then, the seeds were transferred into liquid MS medium containing 1% (w/v) sucrose with 0.2% (v/v) DMSO or each compound dissolved in DMSO, and were germinated and grown for the indicated days under continuous light at 22°C.

Suspension cultured cells of tobacco BY-2 (Nicotiana tabacum L. cv. Bright Yellow 2) were maintained as described previously (Mitsuhashi et al., 2000). BY-2 cells were subcultured in MS medium once a week at 26°C in the dark with an orbital shaker. For chemical assays, 2-day-old BY-2 culture was diluted 1:2 in MS medium and co-incubated with 0.01% (v/v) DMSO or bubblin dissolved in DMSO for 3 days.

Confocal images were obtained using a Zeiss LSM 780 laser scanning microscope equipped with a 488 nm 40 mW Ar/Kr laser or a 544 nm 1 mW He/Ne laser, and a 40×0.95 N.A. or 20×0.80 N.A. dry objective (Plan-Apochromat, 440654-9902-000 or 440640-9902-000, Zeiss). Outlines of cells were visualized by staining with 0.5 mM FM4-64 or 50 μg/ml propidium iodide (PI) as necessary.

Images were analyzed with ZEN2010 software (Zeiss) and processed with ImageJ software (NIH). We extracted the leaf epidermis from stacked images to show epidermal structure at low magnification. Images were scanned from top to bottom (along the z-axis), and the signal intensity at each voxel was compared with a specified threshold intensity. If the intensity of the voxel exceeded the manually determined threshold, we discriminated that the surface was located in the voxel. We extracted voxels in a manually determined range further from the surface. To generate the final 2D images of the leaf epidermis, the extracted regions were processed using maximum intensity projection along the z-axis.

Sterilized wild-type seeds were sown onto solid medium as described above and incubated at 22°C. Then, 3-day-old germinated seeds were transferred into liquid MS medium with 0.2% (v/v) DMSO or 10 μM bubblin dissolved in DMSO. After 24 h, total RNA was isolated from aerial tissue of seedlings using the RNeasy Plant Mini Kit (Qiagen). Total RNA was used to synthesize cDNA with Ready-To-Go RT-PCR beads (GE Healthcare) and oligo(dT)_12-18 primer (Invitrogen). Quantitative real-time PCR (RT-PCR) was performed with gene-specific primer sets (SPCH, At02321069_m1; MUTE, At02235996_g1; FAMA, At02279294_g1; ACTIN2, At02335270_gH; Applied Biosystems) and a TaqMan gene expression assay kit (Applied Biosystems) with a StepOnePlus RT-PCR system (Applied Biosystems). The relative quantity of target cDNA was calculated using ACTIN2 as a control.

We thank Dominique C. Bergmann (Stanford University, USA) for critical analysis of the manuscript and the kind donation of the seeds; and Keiko Kuwata (Nagoya University, Japan) for discussions.