Plants of Commelina benghalensis ssp. were cultured in soil at 25 ± 3°C in a greenhouse for 4 weeks. Arabidopsis thaliana (ecotype Col-0) plants were grown in soil for 4–6 weeks under controlled conditions (20–24°C, 55–70% humidity, 16 h light/8 h dark) in growth chambers. Seeds were vernalized at 4°C in the dark for 2 days before being transferred to soil. Oat (Avena sativa) seedlings were grown in growth chambers with same conditions of Arabidopsis for 10 days.

A protease inhibitor (PI) library (130 PIs dissolved in DMSO at 10 mM) was purchased from APExBIO Company. Repurchased chemicals were SJB3-019A (CAS No. 2070015-29-9, >99.00% purity, Medchem Express), NSC 405020 (CAS No. 7497-07-6, >99.07% purity, Sellek Chemicals), SB-3CT (CAS No. 292605-14-2, >98.00% purity, TCI). All chemicals were stored at −20°C.

Screening of chemicals was performed as described previously (Toh et al., 2018) with some modifications. Four-week-old C. benghalensis plants were transferred from a greenhouse to a dark room for incubation overnight to ensure complete stomatal closure on day 2. Leaf discs of 4 mm diameter were excised from dark-adapted C. benghalensis using a hole punch under dim light. The leaf discs were floated on basal reaction buffer [5 mM MES-BTP (Bis-trispropane), pH 6.5, 50 mM KCl and 0.1 mM CaCl₂] with chemical compounds and incubated in light (150 μmol m^–2 s^–1 red light and 50 μmol m^–2 s^–1 BL) or in the dark for 3 h. Leaf discs were incubated with compounds for 30 min before light exposure. Stomatal apertures of the abaxial epidermis were measured as described previously (Toda et al., 2018). Inhibition by PI1, PI2, and PI3 of stomatal opening was identified. To assay viability, abaxial epidermis was removed from leaf discs using forceps after chemical and light treatments and incubated in fresh basal reaction buffer containing 1 μg/mL fluorescein diacetate (FDA) for 30 min in the dark. Epidermis was washed three times with Milli-Q water to remove FDA. Fluorescence microscopy was used to detect fluorescence.

Phosphorylation of PM H⁺-ATPase was determined by immunohistochemical analysis following previous methods using ImageJ (Hayashi et al., 2011; Ando and Kinoshita, 2018).

Leaf discs excised from dark-adapted 5-week-old Arabidopsis were incubated in basal reaction buffer with PIs (100 μM) or an equal volume of DMSO in the dark for 1 h in advance. Leaf discs were irradiated with BL (50 μmol m^–2 s^–1) or kept in the dark for 1 h. Immunoblot analysis was performed as described previously (Kinoshita et al., 2003; Inoue et al., 2008). Proteins were separated by 7.5% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes. The primary and secondary antibodies were anti-phot1 (Emi et al., 2005; Inoue et al., 2008) and anti-rabbit IgG HRP (Bio-Rad, CA) respectively.

We determined the expression of the ABA-maker genes RAB18 (At5g66400) and RD29B (At5g52300) by quantitative RT-PCR in Arabidopsis (Tomiyama et al., 2014). Two-week-old seedlings cultured in solid 1/2 MS medium were transferred to liquid 1/2 MS medium containing 50 μM ABA or 100 μM PIs or an equal volume of DMSO and irradiated with white fluorescent light for 3 h at 24°C. Total RNA was isolated from seedlings and first-strand cDNA was synthesized using the Prime Script II First Strand cDNA Synthesis Kits (TaKaRa, Tokyo, Japan). Quantitative RT-PCR was performed using Power SYBR Green PCR Master Mix and the Step One Real-Time PCR System (Applied Biosystems, Foster City, CA). RAB18, RD29B, and TUB2 cDNAs were amplified using specific primers (Supplementary Table 1).

Arabidopsis seeds were incubated in MQ water containing 50 μM ABA or 100 μM PIs or an equal volume of DMSO in 96-well plates sealed with surgical tapes. The space between wells was filled with water. The plates were kept at 4°C in darkness for 2 days and transferred to an illumination incubator at 24°C with a 16 h light/8 h dark cycle for 7 days.

Drought tolerance assays were performed as described previously (Toh et al., 2018) with slight modifications. Rose bouquets purchased from a local shop were cultured in a light incubator for 2 days for environmental adaption. Rose leaves and 10-day-old oat seedlings were sprayed with 100 μM PIs or an equal volume of DMSO in 0.05% Approach BI (Maruwa Biochemical) and 0.02% Silwet L77 (Biomedical Science) under white light (50 μmol m^–2 s^–1) and 70% humidity for 3 h at 24°C. Rose and oat leaves were removed and illuminated with 50 μmol m^–2 s^–1 white light for 8 h and 30 min, respectively, at 27°C with 40% humidity.

We evaluated the ability of 130 PIs on stomatal movements (Supplementary Figure 1). C. benghalensis is useful for determination of stomatal aperture due to its larger stomata compared to A. thaliana. Seventeen inhibitors suppressed light-induced stomatal opening by >50% compared to the control (Supplementary Figure 2). No PI induced stomatal opening in the dark (Supplementary Figure 3). Staining with the fluorescent dye FDA showed that none of the PIs affected cell viability (Supplementary Figure 4).

Figure 1A shows inhibition of light-induced stomatal opening by 17 chemicals selected through screenings. These inhibitors were designated PI1–PI17 and their structures are shown in Figure 1C. The targets of PIs in mammals are shown in Supplementary Table 2. The fungal phytotoxin FC leads to irreversible stomatal opening in the dark by binding of 14-3-3 protein to phosphorylated PM H⁺-ATPase, preventing dephosphorylation of PM H⁺-ATPase (Kinoshita and Shimazaki, 2001). To investigate whether the 17 PIs function upstream or downstream of PM H⁺-ATPase activation, we examined their effects on FC-induced stomatal opening (Figure 1B). PI4, PI5 and PI11 did not inhibit FC-induced stomatal opening, indicating that they may have effects upstream of PM H⁺-ATPase. The other PIs significantly inhibited stomatal opening induced by FC to varying degrees, suggesting that they regulate downstream of PM H⁺-ATPase, but not the BHP or PP1, or inhibit PM H⁺-ATPase phosphorylation.

Next, we examined the concentration-dependency of the top three inhibitors (PI1, an inhibitor of ubiquitin-specific protease 1; PI2 and PI3, inhibitors of membrane type-1 matrix metalloproteinase and matrix metalloproteinase-2, respectively) for light-induced stomatal opening (Figure 2). Leaf discs were treated with the indicated compounds and exposed to light for 3 h. The 50% inhibitory concentration (IC₅₀) values of PI1, PI2, and PI3 were 12.28, 35.08, and 33.41 μM, respectively, indicating that PI1 inhibited light-induced stomatal opening more efficiently than PI2 and PI3 at low concentrations. The IC₅₀ value of ABA was 2.903 μM.

To investigate whether PI1, PI2, and PI3 influence BL-dependent phosphorylation of PM H⁺-ATPase, we performed immunohistochemical analysis of guard cells using epidermis isolated from Arabidopsis thaliana (Hayashi et al., 2011). BL-induced phosphorylation was suppressed completely by these three PIs (Figure 3A). We next determined the inhibitory effects of PI1, PI2, and PI3 on FC-induced phosphorylation of the PM H⁺-ATPase. FC-induced phosphorylation was significantly inhibited by PI1, and partially by PI2 and PI3 (Figure 3B). The different inhibitory effects of these inhibitors are consistent with their IC₅₀ values for light-induced stomatal opening (Figure 2).

Next, we determined whether PI1, PI2, and PI3 inhibit the kinase activity of phototropin. Autophosphorylation of phot1 induced by BL in leaves was detected by mobility shift assay using an antibody against phot1 in Western blotting (Emi et al., 2005; Inoue et al., 2008). PI1, PI2, and PI3 did not affect the BL-induced mobility shift of phot1 (Figure 3C), suggesting that PI1, PI2, and PI3 inhibit light-induced stomatal opening without affecting phototropin activity.

To investigate whether PI1, PI2, and PI3 inhibit BL-induced stomatal opening in a manner similar to ABA, we analyzed the expression levels of the ABA-responsive genes, RAB18 and RD29B, by quantitative RT-PCR in Arabidopsis seedlings and germination of Arabidopsis seeds (Toh et al., 2018). Incubation of seedlings with 50 μM ABA for 3 h induced expression of RAB18 and RD29B (Figure 4A). By contrast, PI1, PI2, and PI3 did not induce the expression of ABA-responsive genes (Figure 4A). Moreover, ABA at 50 μM inhibited germination of Arabidopsis seeds (Figure 4B). However, PI1, PI2, and PI3 had no effect on germination (Figure 4B). The results suggest that the mechanisms underlying the inhibitory effects of the PIs and ABA on light-induced stomatal opening are different and that the PIs have more specific roles in regulating stomatal movement.

SCL1, which reportedly induces stomatal closure, prevents wilting of detached oat and rose leaves, which is likely to be by inhibiting light-induced stomatal opening (Toh et al., 2018). We investigated the effect of PI1 on drought resistance using the same method with slight modifications. Oat and rose leaves were sprayed with PI1 at 100 μM and incubated for 3 h, and the leaves were detached. PI1 inhibited leaf withering in oat plants compared with the control (Figure 5A). Rose leaves treated with PI1 wilted slower than that of control (Figure 5B). Therefore, PI1 improves drought resistance in dicot and monocot plants.

We screened for compounds that affect light-induced stomatal opening using a commercially available PI library (APExBIO) and identified 17 PIs that significantly suppressed light-induced stomatal opening by >50% (Figure 1A). PI1, which had the greatest inhibitory effect on stomatal opening, inhibited both light- and FC-induced stomatal opening (Figures 1A,B) and both BL- and FC-induced phosphorylation of PM H⁺-ATPase in guard cells (Figures 3A,B). By contrast, PI1 had no effect on phototropin activity (Figure 3C) and did not induce ABA-dependent responses, such as inhibition of seed germination and expression of ABA-responsive genes (Figure 4). These results suggest that PI1-sensitive proteases mediate phosphorylation of PM H⁺-ATPase (Figure 6) by positively regulating unidentified kinases that directly phosphorylate PM H⁺-ATPase or down-regulating protein phosphatases that directly dephosphorylate PM H⁺-ATPase (Inoue and Kinoshita, 2017). PI2 and PI3 markedly suppressed light-induced stomatal opening (Figure 1A), and partially inhibited FC-induced stomatal opening (Figure 1B) and FC-induced phosphorylation of PM H⁺-ATPase in guard cells (Figure 3B). Therefore, the targets of PI2 and PI3 are likely to include PM H⁺-ATPase and downstream components of light-induced stomatal opening (Figure 6).

It is worth noting that PI4, PI5 and PI11 suppressed light-induced stomatal opening and didn’t inhibit FC-induced stomatal opening significantly (Figures 1A, B), suggesting that the targets of these inhibitors are components upstream of BL-signaling pathway, such as phototropins, BLUS1, BHP, and PP1.

In mammalian cells, PI1 is a specific inhibitor of ubiquitin-specific protease 1 (USP1) (Mistry et al., 2013; Das et al., 2017). To predict the target of PI1 in Arabidopsis, we performed a BLAST search (The National Center for Biotechnology Information)¹ using the amino acid sequence of USP1 from Homo sapiens and found two ubiquitin specific processing proteases (UBP3 and UBP4) with relatively high identity (around 33%) to USP1 (Supplementary Table 3). The UBP family in Arabidopsis has a similar sequence to human USPs and plays a critical role in de-ubiquitination of proteins (March and Farrona, 2018). In Arabidopsis, the UBP family has 27 members (Supplementary Figure 5A), which are divided into 14 subfamilies based on their domains (Yan et al., 2000; Liu et al., 2008; Zhou et al., 2017). All Arabidopsis UBPs contain a UBP domain, and each subfamily shares other conserved domains that are likely to mediate protein-protein interactions (Liu et al., 2008; Komander et al., 2009; March and Farrona, 2018). To our knowledge, the binding site of USP1 with PI1 is unknown.

We hypothesized that some of the 27 UBPs participate in the stomatal-opening pathway in response to BL. These UBPs may function separately or redundantly and up-regulate the unidentified kinases that directly phosphorylate or down-regulate phosphatases that dephosphorylate the PM H⁺-ATPase. We also investigated their expression levels in guard cells using Arabidopsis eFP Browser² (Supplementary Figure 5B). UBP6 and UBP13 exhibited the highest expression levels. Stomatal closure in the mutant of ubp24 was less sensitive to ABA than the WT and the stomata of complemented transgenic plants showed sensitivity to ABA similar to the WT, indicating that UBP24 is involved in ABA-mediated stomatal closure (Zhao et al., 2016). However, PI1 is less likely to modulate the factors in the ABA signaling pathway (Figure 4), so UBP24 may not be the target of PI1.

In mammals, the targets of PI2 and PI3 are membrane type-1 matrix metalloproteinase (MT1-MMP/MMP-14) and matrix metalloproteinase-2 (MMP-2), respectively (Kleifeld et al., 2001; Remacle et al., 2012). MMPs are a large family of zinc- and calcium-dependent endopeptidases (Rawlings et al., 2006). In human, MT1-MMP and MMP-2 have a common domain structure that includes a signal peptide, a propeptide, a catalytic domain with the active zinc binding site, a hinge region, and a hemopexin-like domain. MT1-MMP has two domains absent in MMP-2—membrane linker and cytoplasmic domains (Overall and López-Otín, 2002). PI2 directly targets the hemopexin-like domain of MT1-MMP (Remacle et al., 2012). PI3 is an MMP-2 inhibitor that specifically binds the catalytic zinc ion (Kleifeld et al., 2001). In Arabidopsis, five genes named At1-MMP to At5-MMP encode MMPs with common structural features with mammalian MMPs but lacking a hemopexin-like domain (Maidment et al., 1999; Marino and Funk, 2012; Supplementary Figures 6A,B). We evaluated the expression levels of five MMPs in Arabidopsis guard cells using an eFP Browser; the expression of AT5-MMP was highest (Supplementary Figure 6C). The identities of At-MMPs aligned with human MMP-2 are shown in Supplementary Table 3. In soybean, glycine max1-matrix metalloproteinase (GM1-MMP) and glycine max2-matrix metalloproteinase (GM2-MMP) are PM proteins, and the stomatal aperture of transgenic Arabidopsis plants overexpressing GM1-MMP and GM2-MMP was significantly larger than the WT (Liu et al., 2017, 2018), suggesting that GM1-MMP and GM2-MMP regulate stomatal opening. Therefore, AtMMPs are candidate targets of PI3 in Arabidopsis.