Vicia faba (Broad bean; Ryosai Issun) was cultivated hydroponically in a greenhouse as described previously (Shimazaki et al., 1992). GCPs were isolated enzymatically from lower epidermis of 4- to 6-week-old leaves as described elsewhere (Kinoshita and Shimazaki, 1999). Protein concentrations were determined using a Bradford kit according to the manufacturer’s instructions (Bio-Rad Laboratories, Hercules, CA).

Protein-blot analysis was performed using a GST–GF14phi fusion protein as described (Takahashi et al., 2007; Minami et al., 2019). GF14phi is an Arabidopsis 14-3-3 protein (Kinoshita and Shimazaki, 1999). Immunoblotting was performed using an antibody for Vicia 14-3-3 protein (Vf14-3-3a) as described previously (Emi et al., 2001; Takahashi et al., 2007).

To purify 14-3-3 binding proteins, GCPs from V. faba (330–500 μg) in suspension buffer (5 mM MES-NaOH [pH 6.0], 10 mM KCl, 0.4 M mannitol and 1 mM CaCl₂) were treated with 10 μM ABA for 10 min in the dark. The 14-3-3 binding proteins were co-immunoprecipitated from the supernatant using an antibody against vf14-3-3a (Kinoshita and Shimazaki, 1999; Takahashi et al., 2013). The precipitated proteins were separated by sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE). The sample lanes were excised into 7 to 10 segments and subsequently subjected to in-gel digestion according to previous method (Shevchenko et al., 2006). The digested peptides were analysed by nano-liquid chromatography–tandem mass spectrometry (LC-MS/MS). Nano-LC-MS/MS was performed using a Dionex U3000 Gradient Pump (Thermo Fisher Scientific) connected to a Q-Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific). Peptides were loaded into a trap column [L-column ODS (300 μM internal diameter [ID] × 5 mM, 5 μM particle size), CERI] and separated at 500 nl/min using a 5–40% buffer B gradient over 100 min on a nano-HPLC capillary column [NTCC-360 (100 μM I.D. × 125 mM, 3 μM particle size), Nikkyo Technos]. The composition of the LC buffer A was 0.5% (v/v) acetic acid in water and LC buffer B was of 80% (v/v) acetonitrile, 0.5% (v/v) acetic acid. The Xcalibur 3.0.63 system (Thermo) was used to record peptide spectra over the mass range of m/z 350–1800 (70,000 resolution, 3e6 AGC, 60 ms injection time), followed by ten data-dependent high-energy collisional dissociations (HCD) MS/MS spectra generated from ten highest-intensity precursor ions (17,500 resolution, 1e5 AGC, 60 ms injection time, 27 NCE).

Peptides and proteins were identified by means of automated database searching using Proteome Discoverer v. 2.2.0.388 (Thermo Fisher Scientific) against the V. faba expression database. The following search parameters were employed as: peptide mass range (m/z), 350–1,800 Da; enzyme specificity, trypsin or LysC with up to two missed cleavages; and precursor ion and peptide fragment mass tolerances, ± 10 ppm and ± 0.02 Da, respectively. Peptide validation was performed using the Percolator algorithm, and only high-confidence peptides were used for peptide identification and quantification. The identified peptides in each sample lane were summarised (Table 1).

Total RNA was extracted from GCPs and leaves of 4- to 6-week-old V. faba using a TRIzol Plus RNA Purification Kit (Thermo Fisher Scientific). Three independent biological samples of each tissue type were harvested and snap-frozen in liquid nitrogen. Complementary DNA (cDNA) libraries were constructed from 1 μg of total RNA using the TruSeq RNA Sample Prep Kit v. 2 (Illumina) and sequenced on an NextSeq 500 (Illumina), yielding 13.0 to 19.8 million paired-end sequence reads per sample. Adapter sequences were trimmed with bcl2fastq2 (Illumina) and bases with low-quality scores were masked by N with the original script. In total, 288 million reads, which contained >50 non-masked bases, were used for de novo assembly by Trinity (Grabherr et al., 2011), yielding 134,130 contigs. The contigs were functionally annotated by BLASTX analysis against an Arabidopsis database (TAIR10). For MS determination of the amino acid sequences, these contigs were translated in six-frame and used for reference. To calculate expression levels, filtered reads were mapped to these contigs by Bowtie (Langmead et al., 2009).

In vitro transcription and translation were performed as described previously (Nomoto and Tada, 2018). First-strand cDNA was synthesised from total RNA of GCPs using SuperScript II reverse transcriptase (Invitrogen, Carlsbad, CA) with oligo(dT)_12–18 as the primer and was used as the template for in vitro transcription. The PCR products were sequenced. To attach an N-terminal FLAG tag to the coding sequences of VfAKSs (VfAKS1-3), two-step PCR was carried out using the primers listed in Supplementary Table S2. In vitro transcription was carried out using a transcription kit (NUProtein). For in vitro translation, the resulting RNA solutions were mixed with wheat germ extract and amino acid mix (NUProtein) and incubated at 16°C for 10 h. The synthesised proteins were solubilised and separated by SDS-PAGE and detected by immunoblotting using an anti-FLAG antibody (Sigma).

GCPs from V. faba (1.3–1.6 mg of proteins) in suspension buffer were treated with 10 μM ABA for 10 min in the dark. The GCPs were disrupted by addition of trichloroacetic acid to a final concentration of 20% (v/v), followed by centrifugation. The precipitated guard-cell proteins were suspended in digestion buffer (8 M urea, 250 mM ammonium bicarbonate, 1× PhosSTOP [Roche]). The suspensions were reduced with Tris (2-carboxyethyl) phosphine hydrochloride, alkylated by iodoacetamide and digested with LysC (FUJIFILM), followed by tryptic digestion. Digestions were performed with the enhancer ProteaseMAX™ Surfactant (Promega). The digested samples were acidified and desalted on MonoSpin C18 columns (GL Sciences). Phosphopeptides were enriched by IMAC (Agilent) from 100 μg of digested peptides and diluted with 0.1% (v/v) TFA, 2% (v/v) AcCN in distilled water for nano-LC–MS/MS. Nano-LC–MS/MS was performed as described above.

Peptides and proteins were identified by means of automated database searching using Proteome Discoverer 2.2.0.388 (Thermo Fisher Scientific) against an expression database of V. faba. The following search parameters were employed as: peptide mass range (m/z), 350–1,800 Da; enzyme specificity, trypsin or LysC with up to two missed cleavages; precursor ion and peptide fragment mass tolerances, ± 10 ppm and ± 0.02 Da, respectively; static modification, carbamidomethyl (Cys); and dynamic modifications, phosphorylation (Ser, Thr and Tyr) and oxidation (Met). Peptide validation was performed using the Percolator algorithm, and only high-confidence peptides were used for peptide identification and quantification. The resulting dataset, which included information on annotated sequences, modifications, master protein accession, peptide spectrum matches (PSMs) and the total number of identified peptide spectra, for each identified peptide, was imported into Microsoft Excel. Using the filter function of Excel, peptides with no phosphorylated residues were excluded from the list. PSMs, the total number of identified peptide spectra matched for the protein, of each protein were compared between the datasets.

Sequence data can be found in the Arabidopsis genome database TAIR10 under the following accession numbers: AKS1 (AT1G51140.1), AKS2 (AT1G05805.1), AKS3 (AT2G42280.1), AKS4 (AT2G43140.1), AKS5 (AT4G09180.1), AKS6 (AT1G35460.1), AtABCG40 (AT1G15520) and CHC1 (AT3G11130).

RNA-seq data supporting the findings of this work have been deposited in the DNA Data Bank of Japan (DDBJ) under accession number DRA012337. The raw MS data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository under accession numbers, PXD027057 for immunoprecipitation and PXD027058 for phosphoproteomics.

GCPs from V. faba were treated with ABA at 10 μM for 10 min in the dark. A 61 kDa protein was confirmed as a 14-3-3 protein-binding protein in response to ABA by protein blot using a recombinant GST-14-3-3 fusion protein (Figure 1; Takahashi et al., 2007). GST alone did not show a prominent 61 kDa band (Kinoshita et al., 2003; Takahashi et al., 2007). Therefore, 14-3-3 protein specifically binds to the 61 kDa protein in an ABA-dependent manner.

In Arabidopsis, ABA-dependent phosphorylated proteins were identified as bHLH transcription factors, named ABA-responsive kinase substrate 1 (AKS1) to AKS3 in GCPs (Takahashi et al., 2013). AKSs possess two 14-3-3 protein-binding sites. However, there is no sequence information regarding AKSs in V. faba. Therefore, we performed RNA-seq analysis using GCPs and leaves from V. faba and constructed an expression database of V. faba by de novo assembly. At least seven AKS-like proteins, which possessed 14-3-3 binding sites and a bHLH domain, as well as Arabidopsis AKSs were expressed in V. faba. We named these proteins VfAKS1 to VfAKS7 (Figure 2A; Table 1). In addition, we identified AT2G43140, AT4G09180 and AT1G35460 as AKS homologs in A. thaliana, and designated them AKS4, AKS5 and AKS6, respectively (Figure 2A). Figure 2B shows a phylogenetic analysis based on the full-length amino acid sequences of AKSs from V. faba and A. thaliana. VfAKS1 and VfAKS2 were similar to AKS1, and VfAKS3 to AKS3. All possessed two 14-3-3 protein-binding motifs at the N-terminus and middle of the sequence. Previous in vitro experiments suggested that phosphorylation of AKS1 at Ser-284, −288 and − 290 induces monomerisation of AKS1 and inhibits its transactivation activity (Takahashi et al., 2017). VfAKS1, VfAKS2, VfAKS3 and AKS3 also possess Ser or Thr upstream of the bHLH domain (Figures 2A,B). In contrast, VfAKS5, AKS5 and AKS6 lack the corresponding Ser or Thr and their sequences were shorter than other AKSs. Of these, VfAKS5 was the shortest and lacked the middle 14-3-3 binding motif. VfAKS4 was similar to AKS2 and AKS4. In contrast, no ortholog of VfAKS6 and VfAKS7 lacking the 14-3-3 protein-binding motif at the N-terminus was found in A. thaliana. AKS1, AKS3, AKS5 and AKS6 are identical to FLOWERING BHLH3 (FBH3), FBH4, FBH2 and FBH1, respectively (Ito et al., 2012).

Next, we investigated the expression of VfAKS1–VfAKS7 in GCPs and leaves based on RNA-seq data. The expression levels of VfAKS1, VfAKS3, VfAKS4, VfAKS6 and VfAKS7 were higher in GCPs compared to leaves (Figure 2C). This is consistent with a report that the response is specific to guard cells and is not found in other cell types, such as mesophyll cell protoplasts and root and leaf cells (Takahashi et al., 2007). In addition, previous promoter GUS assay of AKS1 in Arabidopsis thaliana also revealed preferential expression of AKS1 in guard cells and vascular tissues (Takahashi et al., 2013).

ABA-dependent 14-3-3 binding proteins were isolated by co-immunoprecipitation using anti-14-3-3 protein antibodies from Arabidopsis GCPs (Takahashi et al., 2013). Next, we performed immunoprecipitation of ABA-treated Vicia GCPs using anti-14-3-3 protein antibodies against Vicia 14-3-3a (Emi et al., 2001) and LC–MS/MS using gel segments after resolving the immunoprecipitate by SDS-PAGE. Immunoprecipitates from control and ABA-treated GCPs contained >507 proteins, including those annotated as 14-3-3 proteins, plasma membrane H⁺-ATPases, vacuolar ATP synthase subunit A, pleiotropic drug resistance 12 and clathrin heavy chain (Supplementary Table S1). The number of peptides from VfAKS1–VfAKS5 increased in response to ABA (Table 1).

VfAKS1–VfAKS5 immunoprecipitated with GCPs, suggesting that one may be the 61 kDa protein. However, the estimated molecular masses of VfAKS1–VfAKS5 based on the amino acid sequences were < 61 kDa (Table 1). Next, we synthesised FLAG-VfAKS1, VfAKS2 and VfAKS3, which had high estimated molecular masses (Table 1), by in vitro translation assay and determined their molecular masses by SDS-PAGE. As shown in Figure 3A, full-length CDSs of VfAKSs were amplified by RT-PCR from cDNAs of Vicia GCPs. Finally, we detected FLAG-VfAKS1 (61.4 kDa), FLAG-VfAKS2 (54.3 kDa) and FLAG-VfAKS3 (43.9 kDa) proteins by Western blotting (Figure 3B). The FLAG tag is around 1.0 kDa. These results indicated that the 61 kDa protein is VfAKS1.

To confirm ABA-dependent phosphorylation of VfAKSs, we performed phosphoproteomic analyses using GCPs treated with ABA. We detected 7,459 phosphopeptides belonging to 2,938 proteins. Of these, 929 phosphopeptides (12.5%) showed at least 2-fold increase in ABA-treated GCPs compared to ABA-untreated GCPs, whereas 489 phosphopeptides (6.6%) showed at least 2-fold decrease.

We detected multiple phosphorylation sites in VfAKS1, VfAKS2, VfAKS3, VfAKS4 and VfAKS6 (Table 2). The conserved Ser residues in the 14-3-3 protein-binding motifs (Figure 2A; Table 2; single asterisk) were phosphorylated in response to ABA (in the case of VfAKS1, Ser-23 and Ser-177). Moreover, ABA-dependent phosphorylation of the conserved Ser or Thr leading to monomerisation and inactivation of transactivation in AKS1 (Takahashi et al., 2016) was observed (Table 2; double asterisks; in the case of VfAKS2, Ser-341 and Thr-345). Ser-182 in VfAKS3 was phosphorylated in response to ABA. Interestingly, Ser-47 of VfAKS1 and Ser-57 of VfAKS2, corresponding to Ser-52 of AKS1, were highly phosphorylated in untreated and ABA-treated GCPs.

Interestingly, we detected ABA-dependent phosphorylation of proteins annotated as pleiotropic drug resistance 12 (PDR12) and clathrin heavy chain, both of which immunoprecipitated with 14-3-3 protein (Table 3; Supplementary Table S1). Arabidopsis AtPDR12 encodes ATP-binding cassette (ABC) transporter (ABCG40), which is an ABA importer expressed in guard cells. We designated the ortholog in V. faba as VfABCG40. VfABCG40 showed high sequence similarity (72%) with AtABCG40. The phosphorylation sites of VfABCG40 detected by phosphoproteomics were located at the N- and C-termini (Figure 4A). Most phospho-Ser and -Thr at the N-terminus were conserved in Arabidopsis ABCG40, but those at the C-terminus were not. At the N-terminus sites, phosphorylation of Ser-30/ Ser-32/ Ser-33 and Thr-62/Ser-64 was slightly upregulated in response to ABA (Table 3).

Clathrin heavy chain protein showed high sequence similarity (91%) with Arabidopsis clathrin heavy chain 1 (CHC1; Blackbourn and Jackson, 1996), and we designated its ortholog in V. faba as VfCHC1. The phosphorylation site at Thr-67 in VfCHC1 is conserved in AtCHC1 (Figure 4B). ABA induced phosphorylation of Thr-67 (Table 3).