Proteome-wide cellular thermal shift assay reveals unexpected cross-talk between brassinosteroid and auxin signaling.

SignificanceChemical genetics, which investigates biological processes using small molecules, is gaining interest in plant research. However, a major challenge is to uncover the mode of action of the small molecules. Here, we applied the cellular thermal shift assay coupled with mass spectrometry (CETSA MS) to intact Arabidopsis cells and showed that bikinin, the plant-specific glycogen synthase kinase 3 (GSK3) inhibitor, changed the thermal stability of some of its direct targets and putative GSK3-interacting proteins. In combination with phosphoproteomics, we also revealed that GSK3s phosphorylated the auxin carrier PIN-FORMED1 and regulated its polarity that is required for the vascular patterning in the leaf.

Although the application of chemical genetics to plant research is gaining interest (1–3), unraveling the mode of action of the small molecules remains a major challenge. The cellular thermal shift assay (CETSA) is a label-free method that can assess target engagement directly in live cells (4), but its application to plant cells remains limited (5, 6). The technique is based on the biophysical principle that a ligand can induce changes in the thermal stability of the target protein, allowing the generation of so-called protein melting curves (7, 8). Similar to the classical thermal shift assay with purified proteins, small molecule binding typically leads to protein stabilization and an increase in melting temperature (Tm). Coupling CETSA with multiplexed quantitative mass spectrometry (MS) enables the monitoring of an entire proteome for changes in protein thermostability in the presence of a small molecule (9, 10). Consequently, proteins interacting with this molecule can be identified without previous knowledge of the pathways or molecular mechanisms involved (11). As the thermal stability of a protein may also be affected by posttranslational modifications or by binding to other proteins, cofactors, or metabolites, CETSA MS carried out on intact cells, in which active signaling takes place, allows the identification of effector proteins downstream of the direct target (9).

The small molecule bikinin (12) is an inhibitor specifically targeting the Arabidopsis thaliana Shaggy/glycogen synthase kinase 3 (GSK3)-like kinases (AtSKs), including the key negative brassinosteroid (BR) signaling regulator BR-INSENSITIVE2 (BIN2)/AtSK21 that phosphorylates and inactivates two main transcription factors, BRASSINAZOLE RESISTANT1 (BZR1) and BRI1-EMS-SUPPRESSOR1 (BES1)/BZR2 (13, 14). The Arabidopsis genome encodes 10 AtSKs, which belong to four groups (15); evidence exists for at least six, AtSK11, AtSK12, AtSK13, AtSK22, AtSK23, and AtSK32, that negatively regulate the BR signaling just as BIN2/AtSK21 (16–18). Like the mammalian GSK3s (19), AtSKs also phosphorylate numerous substrates and control many developmental and physiological processes in plants, such as root, stomatal, and flower development; xylem differentiation; responses to light; and different abiotic and biotic stresses (15). BIN2/AtSK21 also mediates the cross-talk between BR and other plant hormones, including auxin (15).

Despite the reported interdependency and cooperation (20–25), the molecular mechanisms of the signaling cross-talk between BRs and auxin are still not well understood. Although auxin does not affect the phosphorylation state of the BZR1 and BES1/BZR2 transcription factors (14), BIN2/AtSK21 interacts and phosphorylates the AUXIN RESPONSE FACTOR2 (ARF2) (21). This phosphorylation results in a loss in the DNA binding and repressor activity of ARF2 and facilitates auxin responses (21). BIN2/AtSK21 also phosphorylates and activates ARF7 and ARF19 to promote lateral root development through an increase in auxin response (22). Moreover, BRs have been shown to control posttranscriptionally the endocytic sorting of PIN-FORMED 2 (PIN2) (23) and stimulate the nuclear abundance and signaling of auxin via repressing the accumulation of PIN-LIKES (PILS) proteins at the endoplasmic reticulum (24).

Here, by adapting the CETSA MS to Arabidopsis intact cells and combining it with phosphoproteomics, we discovered that the auxin efflux carrier PIN1 is a substrate of the AtSKs. We found that phosphorylation mediated by the AtSKs is required for PIN1 polarity and for leaf venation. In summary, we demonstrate that CETSA MS is a powerful method for the identification of small molecule targets as well as for the discovery of new protein–protein interactions in plant cells.

Results

CETSA Monitoring of Small Molecule–Protein Interactions in Intact Arabidopsis Cells.

Previously, we had used the Western blot–based CETSA for small molecule target validation in cell lysates of Arabidopsis seedlings (6). To extend the use of CETSA to intact plant cells, we adapted the available protocol (10) to Arabidopsis cell suspension cultures (). First, we tested whether the plant cell wall complicated the protein isolation by evaluating the efficiency of the freeze–thaw lysis method applied to mammalian cells (10). After washing and resuspension in protein extraction buffer, 100-µL aliquots of cells were freeze–thawed multiple times and the protein concentration of the supernatant was measured with the Bradford protein assay. The protein concentration in the lysate plateaued at the seventh freeze–thaw cycle (). Moreover, considering that plants grow over a wide temperature range, we assessed whether in intact Arabidopsis cells, proteins follow a melting profile similar to that in lysates when heated (5, 26). To this end, we heated 100-µL aliquots of Arabidopsis cells in protein extraction buffer, heated to different temperatures (from 25 °C to 80 °C) for 2 min, and analyzed the lysates by sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE) after seven freeze–thaw cycles and centrifugation. As expected, the Arabidopsis proteins unfolded and precipitated at high temperature ().

Subsequently, as a proof of concept, we aimed to apply CETSA to cells treated with bikinin, an inhibitor of several AtSKs in Arabidopsis (12). Application of bikinin at concentrations of 30 to 50 µM to Arabidopsis seedlings induces BR responses that can be measured by changes in the phosphorylation status of the transcription factor BES1/BZR2 (14). To check whether bikinin is effective in Arabidopsis cell suspension cultures, we treated cells with 50 µM bikinin for 30 min. As expected, bikinin induced the dephosphorylation of BES1 in the cell cultures similarly to the most active BR, brassinolide (BL) (). Taken together, bikinin can induce BR responses in Arabidopsis cell suspension cultures.

Next, we investigated the effect of bikinin on the thermal stability of its direct targets, the ten AtSKs (12), by means of the adjusted CETSA protocol () and by using Western blots for detection. In brief, after 30 min of treatment with bikinin or dimethyl sulfoxide (DMSO), the Arabidopsis cells were washed, resuspended in protein extraction buffer containing either bikinin or DMSO, and aliquoted into PCR tubes. Then, the aliquots were heated to 12 distinct temperatures (30, 35, 40, 43, 46, 49, 52, 55, 58, 61, 65, and 70 °C) for 2 min. Afterward, the heated cells were lysed through seven freeze–thaw cycles followed by Western blot–based protein detection. As specific antibodies for all AtSKs are not available, Arabidopsis cell suspension cultures overexpressing the hemagglutinin (HA)-tagged AtSKs were utilized to generate protein-melting curves and to assess the bikinin-induced Tm shifts. First, we examined whether bikinin induced Tm shifts for AtSK12 and AtSK13 at a 50-µM concentration, but, surprisingly, found that it stabilized AtSK13 with a Tm shift of 5.94 °C (), but not AtSK12 (). By contrast, the thermal stability of the ATP synthase ß (ATPβ), used as a control, was not affected by bikinin. We next performed an isothermal dose–response fingerprinting (ITDRFCETSA) for AtSK12 at 45 °C, a temperature selected between the Tm of AtSK12 and Tm of AtSK13 under control (DMSO) conditions and determined the bikinin half-maximum effective concentration (EC50) to be around 70 µM (). To ensure saturation and achieve sufficiently sized Tm shifts for all AtSKs, we used 250 µM bikinin (Fig. 1). Staining of the cells with the cell viability tracer, fluorescein diacetate (FDA), excluded the potential cytotoxic effect of bikinin when used at high concentrations (). Of the 10 putative bikinin targets (12), AtSK11, AtSK12, AtSK13, BIN2/AtSK21, AtSK22, and AtSK41 showed Tm shifts, whereas the thermal stability of AtSK23, AtSK31, AtSK32, and AtSK42 was not affected by the small molecule at 250 µM (Fig. 1) as well as the thermal denaturation of the ATPβ control (). Collectively, these results showed that bikinin stabilized most of its targets, indicating that the CETSA protocol was applicable to intact Arabidopsis cells.

Fig. 1.

Bikinin stabilized a subset of the AtSKs. Thermal denaturation curves for 10 HA-tagged AtSKs stably overexpressed in Arabidopsis cell suspension cultures in the presence of 250 μM bikinin (BIK) or 0.1% (vol/vol) DMSO. The relative band intensities from the Western blot analysis were calculated based on the lowest temperature (30 °C). Melting temperatures (Tm) are indicated. Individual data points are plotted for three biological replicates. For the blots these graphs are based on, see .

CETSA MS of Bikinin in Intact Arabidopsis Cells.

Several recent studies in mammalian cells reported the use of proteome-wide CETSA MS for obtaining a comprehensive view on small molecule–protein interactions through determination of individual temperature shifts (8–10, 27). Therefore, we extended the CETSA protocol as described above () to the Arabidopsis proteome () by using 50 µM bikinin, the BR response-inducing concentration in cell cultures. Briefly, following the heating at 25, 30, 35, 40, 45, 50, 55, 60, 70, and 80 °C and freeze–thaw cycles, samples were analyzed with a nanoscale liquid chromatography coupled to tandem mass spectrometry (nano LC-MS/MS), whereafter they were digested with trypsin and labeled with 10-plex tandem mass tag (TMT10).

In total, 6,000 proteins were identified, of which the melting profiles were defined for 4,225 proteins in samples treated with both bikinin and DMSO (Dataset S1). Approximately 96% of the identified proteins had a melting temperature within the range of 35 °C to 60 °C (Fig. 2 and Dataset S1), of which only 61 proteins displayed a significant change in thermal stability (27 were stabilized and 34 were destabilized) in the presence of 50 µM bikinin (absolute value of Tm shift ≥ 2 °C, analysis of variance [ANOVA]-based F test P < 0.01) (Fig. 2 and Dataset S1). However, because none of the AtSKs were identified by the CETSA MS, we examined the protein expression of AtSKs in Arabidopsis cell suspension cultures. Although qRT-PCR revealed expression of all AtSKs in the cell cultures (), only AtSK11, AtSK21, AtSK31, and AtSK41, albeit at a low intensity, were detected with shotgun proteomics ( and Dataset S2). Given that all AtSKs were efficiently extracted using the CETSA protocol when overexpressed (Fig. 1), probably the low number of peptides identifying these proteins in cell cultures obstructed the creation of their melting curves (Dataset S2).

Fig. 2.

CETSA MS in the presence of bikinin. (A) Melting temperatures of the 4,225 proteins identified in the CETSA MS in the presence of 50 µM BIK or 0.1% (vol/vol) DMSO. (B) Distribution of the Tm shifts of all proteins. Proteins in red highlight significant changes (Tm shift > 2 °C, ANOVA-based F test P < 0.01). (C) Thermal denaturation curves generated for WIN2 in the presence of 50 μM BIK or 0.1% (vol/vol) DMSO subtracted from the CETSA MS in Arabidopsis cell suspension cultures. The Tm is indicated. Individual data points are plotted for n = 2 technical replicates. (D) Coimmunoprecipitation of WIN2 with all AtSKs when coexpressed in tobacco. The p35S::WIN2-HA construct was transiently coexpressed with either p35S::AtSKs-GFP or p35S::GFP (as a negative control). Proteins were extracted (input) and immunoprecipitated (IP) by means of GFP beads. AtSKs-GFP, GFP, and WIN2-HA were detected with anti-GFP and anti-HA antibody, respectively. IB, immunoblot.

As changes in thermal stability had also been observed for downstream effectors of the direct small molecule target, possibly as a result of altered posttranslational modifications or interactions with other proteins (4, 9), we examined whether any of the 61 proteins (P < 0.01) (Dataset S2) functioned together with the AtSKs and found that only the MITOGEN-ACTIVATED PROTEIN KINASE3 (MPK3), which acts downstream of the known BIN2/AtSK21 interactor YODA (YDA) (28), exhibited a Tm shift (−2.96 °C, P < 0.01). Furthermore, we checked whether any of the 61 proteins with significant Tm shifts were putative AtSK interactors () according to the STRING database (29). Based on this analysis, the HOPW1-1-INTERACTING2 (WIN2, AT4G31750) (Tm shift = 7.80 °C, P < 0.01) (Fig. 2) was selected as a possible AtSK-interacting protein. Interaction between HA-tagged WIN2 and each of the 10 GFP-tagged AtSKs was observed by coimmunoprecipitation experiments carried out using tobacco (Nicotiana tabacum) cells transiently overexpressing the proteins (Fig. 2). In summary, in cell suspension cultures bikinin affected the thermal stability of the Arabidopsis proteome and induced Tm shifts in several proteins that might be putative AtSK-interacting proteins or AtSK downstream effectors.

The Phosphoproteome of Bikinin.

As, besides MPK3, the proteins with altered thermal stability identified by the CETSA MS (Dataset S1) were neither direct bikinin targets nor known downstream AtSK effectors, we carried out a phosphoproteomics analysis on bikinin-treated Arabidopsis cell suspension cultures. We hypothesized that the putative AtSK-interacting proteins or AtSK downstream effectors might modify their phosphorylation state upon inhibition of their kinase activity. Arabidopsis cell suspension cultures were treated with 50 µM bikinin or DMSO for 30 min, under the same CETSA MS. The phosphoproteomics analysis revealed that, in total, 9,351 phosphopeptides were mapped to 1,751 proteins (Dataset S3). Bikinin treatment significantly down-regulated (P < 0.05) the phosphorylation intensities of 972 phosphosites that belong to 665 proteins (Fig. 3 and and Dataset S3 ), of which six proteins were known AtSK-interacting proteins, including, BZR1 (30), ARF2 (21), YDA (28), GLUCOSE-6-PHOSPHATE DEHYDROGENASE6 (G6PD6) (31), TETRATRICOPETIDE-REPEAT THIOREDOXIN-LIKE3 (TTL3) (32), and OCTOPUS (33). In addition, the phosphorylation intensities of 101 phosphosites belonging to 84 proteins were significantly up-regulated (P < 0.05) (Fig. 3 and and Dataset S3 ). Of note, both down-regulated and up-regulated phosphosites were identified in 35 proteins (Fig. 3 and Dataset S3 ). The gene ontology (GO) enrichment analysis (34) revealed that most of the enriched terms for the identified phosphorylation-regulated proteins were related to mRNA splicing, metabolic process, and transport ( and Dataset S3). However, of the bikinin-regulated phosphoproteins, only six (phosphorylation down-regulated) proteins showed significant Tm shifts in the CETSA MS (Dataset S3), including the auxin efflux carrier PIN1 (Tm shift = 3.83 °C) (Fig. 3 ). Taken together, only a few proteins identified in the bikinin phosphoproteome displayed bikinin-induced changes in their thermal stability.

Fig. 3.

Identification of PIN1. (A) Venn diagram comparing the CETSA MS and the BIK phosphoproteome. Only six proteins (phosphorylation down-regulated) showed significant Tm shifts in CETSA MS. (B) Tm shifts distribution of the proteins of which the phosphorylation intensities were down-regulated in the presence of 50 µM BIK. (C) Volcano plot representation of the quantitative phosphoproteomics analysis in the presence of 50 µM BIK and 0.1% (vol/vol) DMSO. Diamonds represent phosphopeptides quantified in five biological replicates, each phosphopeptide log2 (fold change) is the average logarithmic ratio of the phosphopeptide abundance in cell suspensions treated with BIK vs. DMSO plotted against the log10 (P value) determined by the Student’s t test. Phosphopeptides of PIN1 are indicated. (D) The thermal denaturation curves generated for PIN1 in the presence of 50 μM BIK or 0.1% (vol/vol) DMSO subtracted from the CETSA MS in Arabidopsis cell suspension cultures. The Tm is indicated. Individual data points are plotted for n = 2 technical replicates.

AtSKs Phosphorylate PIN1 and Regulate Its Polarity.

We hypothesized that PIN1 is a direct substrate of the AtSKs, because the PIN1 protein had been identified in the CETSA MS (Fig. 3 and Dataset S1) and the phosphorylation intensities of five residues in the hydrophilic loop (HL) (Ser271, Ser282, Thr286, Ser290, and Ser337) () were reduced in the presence of 50 µM bikinin (Fig. 3 and Dataset S3). To verify this observation, we carried out a protein kinase assay by incubating the polyhistidine (HIS)-tagged HL of PIN1 with HIS-small ubiquitin-like modifier (HIS-SUMO)-tagged AtSK proteins in an in vitro phosphorylation reaction. The results showed that eight AtSK proteins phosphorylated the His-PIN1-HL (). Remarkably, four of the five PIN1 residues, of which the phosphorylation was altered by bikinin, had previously been identified as targets of the serine/threonine protein kinases PINOID/AGCVIII kinases (PID/WAGs) (35), D6 Protein Kinases (D6PKs) (36), and MPKs (MPK3, MPK4, and MPK6) (37, 38). Moreover, these residues were conserved () and essential for the polar localization of the PIN proteins with long HLs (39).

Given the prominent role of phosphorylation in regulating polarity, we next assessed whether treatments with bikinin and BL affected the polar localization of PIN1 and PIN2 in the root meristem by means of immunolocalization. Wild-type Arabidopsis plants were treated in liquid medium with 50 µM bikinin, 10 nM BL, and 0.1% (vol/vol) DMSO for 12 h. As anticipated, PIN1 displayed a more apolar localization in the presence of bikinin and BL (Fig. 4 ) than with DMSO, whereas the PIN2 polarity remained unaffected (). To provide genetic evidence for the AtSK function in PIN1 polarity regulation, we analyzed the PIN1 localization in the root tips of the AtSK quadruple (atsk13RNAi bin2 bil1 bil2, designated atsk) and sextuple (atsk11RNAi atsk12RNAi atsk13RNAi bin2-3 bil1 bil2, designated atsk) mutants (40). PIN1 had a more apolar localization in the root tips of the atsk mutant than the wild-type (Fig. 4 ). Because of the seedling lethality of the atsk mutant (40), the plant phenotypes were analyzed in the T1 generation. The atsk mutant also displayed an aberrant PIN1 polar localization (Fig. 4 ). Given that PIN1-mediated polar auxin transport regulates the foliar vascular patterning (41), we examined whether AtSKs were involved in the regulation of leaf venation. Five-day-old wild-type plants were treated in liquid medium with 50 µM bikinin, 10 nM BL, and 0.1% (vol/vol) DMSO for 2 d. As expected, plants treated with bikinin and BL had an abnormal leaf vascular patterning, when compared to the mock control (). Consistent with BR and bikinin treatments, the atsk and atsk mutants exhibited an aberrant vascular patterning, with missing loops and disconnected upper loops (Fig. 4 ). Subsequently, we examined the PIN1 localization in the leaf veins of atsk and atsk mutants. As anticipated, atsk and atsk mutants displayed impaired PIN1 localizations, including missing PIN1 in some cells and more PIN1 localizations parallel to the vein axis, when compared to the wild-type control (Fig. 4 ).

Fig. 4.

PIN1 polarity modulation by the AtSKs. (A) Immunolocalization of PIN1 in root tips after 12 h of 50 µM BIK, 10 nM BL, or 0.1% (vol/vol) DMSO treatments. Zoom-ins of areas marked with dashed boxes are shown below each main image. (Scale bars, 20 µm.) (B) Quantitative evaluation of A showing mean of PIN1 lateral-to-basal signal intensity ratio in endodermal cells. n > 180 cells corresponds to a minimum of 15 roots per treatment from three independent experiments. (C) Immunolocalization of PIN1 in root tips of Wassilewskija (Ws), AtSK quadruple (atsk), and sextuple (atsk) mutants. Zoom-ins of areas marked with dashed boxes are shown below each main image. (Scale bars, 20 µm.) (D) Quantification of C showing mean of PIN1 lateral-to-basal signal intensity ratio in endodermal cells. n > 95 cells corresponds to a minimum of 15 roots per treatment from two independent experiments. (E) Representative images of venation patterning defects in cotyledons of Ws, atsk, and atsk mutants. (Scale bars, 1 mm.) (F) Quantification of venation defects in atsk and atsk mutants (n > 60 of each genotype from three independent experiments). (G) Immunolocalization of PIN1 in the leaf veins of Ws, atsk, and atsk mutants. Red arrows indicate the defects in PIN1 localization, including missing PIN1, and more PIN1 localizations parallel to the vein axis. (H) Quantitative evaluation of G showing defects in PIN1 localization; n > 45 leaves from five independent experiments. (I) Subcellular localization of PIN1-GFP and PIN1-GFP5A in root tip cells. Zoom-ins of areas marked with dashed boxes are shown below each main image. Arrowheads indicate the lateral membrane signal in the endodermal cells. (J) Quantitative evaluation of I showing mean of PIN1-GFP and PIN1-GFP5A lateral-to-basal signal intensity ratio in endodermal cells. n > 60 cells corresponds to a minimum of 15 roots per genotype from three independent experiments. (K) Representative images of venation patterning defects in cotyledons of PIN1pro::PIN1-GFP and two transgenic lines of PIN1pro::PIN1. (Scale bars, 1 mm.) (L) Quantification of venation defects in PIN1pro::PIN1-GFP and two PIN1pro::PIN1 transgenic lines (n > 60 of each genotype from three independent experiments). (B, D, and J) Scatter dot plots showing all the individual points with means and SEs. One-way ANOVA with Tukey’s post hoc test compared to DMSO or PIN2-GFP, respectively. ∗∗∗P < 0001.

To test whether the identified five phosphorylation sites were relevant for the AtSK-mediated regulation of PIN1, we generated phospho-inactive His-PIN1-HL5A by substituting the Thr and Ser residues with Ala. The phosphorylation of His-PIN1-HL5A by BIN2/AtSK in vitro was reduced but not abolished (), suggesting that more sites were phosphorylated by BIN2/AtSK21. Next, we introduced the phospho-inactive PIN15A-GFP (PIN1pro::PIN1) into wild-type Arabidopsis plants. Noteworthy, 6 of the 26 GFP-positive transgenic plants had naked inflorescence stems in the first generation (), reminiscent of the pin1 null mutant (42). Then, we examined the localization of PIN15A-GFP in the root tip cells of wild type–resembling transgenic plants. When expressed at the same levels as the native PIN1-GFP (pPIN1::PIN1-GFP) (43) (), PIN15A-GFP exhibited a more apolar localization in two independent transgenic lines (Fig. 4 ) that, in a similar way to the atsk and atsk mutants, had an abnormal vascular patterning in the cotyledons (Fig. 4 ). In summary, these observations show that AtSK-mediated phosphorylation of PIN1 is important for polar localization and leaf venation.

Discussion

Given the challenges in small molecule target identification in plants (44), experimental strategies, especially label-free methods, are required to facilitate chemical genetics studies. CETSA is such a method that has been proven useful for the detection of direct drug targets and downstream effects of drug-induced perturbations in several cellular systems and tissues (9, 27, 45, 46). In plants, CETSA has been successfully applied for small molecule target validation in cell lysates of Arabidopsis (6) and, recently, the thermal profiles of more than 2,000 proteins in Arabidopsis lysates have been reported (5, 26). Here, we explored the potential of CETSA to identify the protein targets of the AtSK kinase inhibitor bikinin (12) in intact Arabidopsis cells. In Western blot–based CETSA, bikinin stabilized six of the 10 AtSKs. As expected, bikinin had no effect on either AtSK31 or AtSK42, of which the kinase activity was not inhibited in vitro (12), but surprisingly, on AtSK23 and AtSK32 as well, of which the in vitro inhibition of the kinase activity had previously been reported (12). One reason might be that the binding affinity of bikinin to AtSK23 and AtSK32 is lower than that to other AtSKs, thus requiring higher bikinin concentrations to induce the noticeable Tm shifts. Moreover, in a previous study, ∼30% of the target proteins of the promiscuous kinase inhibitor staurosporine did not show Tm shifts (9), suggesting that some protein kinases are not responsive in CETSA.

By combining CETSA with MS, we identified the thermal profiles of 4,225 proteins in the presence of bikinin in Arabidopsis cell suspension cultures, of which 61 (1.44%) showed significant Tm shifts (absolute value ≥ 2 °C, ANOVA-based F test P < 0.01). Although the CETSA MS assay did not detect any AtSKs, probably due to their low protein abundance in the cell cultures, MPK3, the substrate of the know BIN2/AtSK21 interactor YDA (28), was detected. In addition, the thus far unknown interaction between WIN2 (AT4G31750) and AtSKs was validated in vivo. WIN2 is implicated in the modulation of the defense responses induced by the Pseudomonas syringae effector protein HopW1-1 (47). AtSK11 has been shown to regulate the pattern-triggered immunity and the susceptibility to P. syringae, probably through G6PD6 phosphorylation (48). Our data suggest that AtSK11, as well as its homologs, might regulate the immune responses to P. syringae via direct interaction with WIN2.

To demonstrate the capability of CESTA MS to identify proteins that might bind the AtSKs directly or function in downstream pathways, we carried out a phosphoproteomics analysis in the presence of bikinin. Phosphorylation sites in 665 proteins were down-regulated upon bikinin treatment, comprising six known substrates of the AtSKs, hence confirming the quality of our data. Noteworthy, the phosphorylation intensities of 84 proteins were up-regulated upon bikinin treatment, indicating that these proteins might be indirect AtSK targets. The GO enrichment analysis for the bikinin-regulated phosphoproteins revealed that AtSKs might be involved in the regulation of RNA splicing and protein intracellular transport, like their mammalian homologs (49, 50). Moreover, of all proteins with differential phosphorylation, 76 proteins were also identified by proximity labeling with BIN2/AtSK21 as a bait (51) (Dataset S3). Thus, we speculate that some of these proteins might be substrates of the AtSKs. By comparing the CETSA MS and phosphoproteome datasets, we found that the thermal stability of most of these proteins was not affected, even though their phosphorylation intensities were altered by the bikinin treatment. A possible reason might be that either the phosphorylation cannot affect the protein melting behavior or that the bikinin-induced changes in the phosphorylation intensity are not sufficient to induce an important Tm shift for most of these proteins. However, for a few proteins, bikinin had an impact on both the phosphorylation and melting behaviors, indicating that the phosphorylation status might affect the thermal stability of some proteins. Although some studies in human cells reported that phosphorylation affected the protein thermal stability (52), others demonstrated that for most of the proteins, the melting behavior of phosphorylated and nonphosphorylated forms was concordant (53, 54).

Particularly, bikinin reduced the phosphorylation and induced the thermal stabilization of the auxin efflux carrier PIN1. The identified AtSK phosphorylation sites in the HL of PIN1 are conserved among the PIN proteins (), some of which are essential for their polar localization and the intercellular auxin transport (39). However, although some kinases shared phosphorylation sites in the PIN proteins, for instance PID/WAGs and D6PKs, their corresponding mutants or transgenic overexpression lines affected the PIN polarities differently (35, 36). Moreover, the auxin-regulated receptor CANALIZATION-RELATED AUXIN-REGULATED MALECTIN-TYPE RECEPTOR-LIKE KINASE (CAMEL) (55) and CALCIUM-DEPENDENT PROTEIN KINASE 29 (CPK29) (56) phosphorylate PIN1 and control its polarity via phosphorylation sites that are not shared with other kinases, including PID/WAGs, D6PKs, MPKs, and the AtSKs () (55, 56). Therefore, multiple parallel mechanisms for the maintenance of the PIN polarities and activities exist. Our data imply that BRs control the PIN1 phosphorylation and polarity via AtSKs. Two of the five phosphorylation sites identified in our phosphoproteomics study are also targeted by PID/WAGs and D6PKs, whereas another two phosphorylation sites are targeted by MPK3, MPK4, and MPK6, implying a functional redundancy of AtSKs with other kinases in the PIN phosphorylation. In addition, bikinin and BR treatments, which inhibit the activity of AtSKs directly and indirectly, respectively, impaired the PIN1 polarity, indicating that AtSKs control the polarity of PINs in combination with other kinases. Considering this overlap in the PIN1 phosphorylation sites, understanding of the precise environmental or developmental conditions and upstream pathways that coordinate these kinases will be needed to cooperatively control the PIN polarities. Furthermore, because the PIN phosphorylation induced by D6PKs and PID/WAGs control their activity (57), it will be necessary to examine whether bikinin and exogenous BRs disrupt the auxin transport activity of PIN1 and other PINs. Previous studies have shown that BRs affect PIN proteins through different mechanisms (23, 58), including transcriptional activation of PIN4 and PIN7 (58). Additionally, BRs and bikinin stabilized PIN2, but not PIN1, to interfere with the auxin distribution in gravistimulated roots (23), whereas BRs partially altered PIN2 polarity via the actin cytoskeleton (59). However, the actin cytoskeleton had been demonstrated not to be required for the polar localization of PIN2 (60). Interestingly, BRs did not affect the PIN2 polarity in our study, although four of the five identified phosphorylation sites are conserved in the HL of PIN2. Together, these data indicate that the mechanisms controlling the polarity and the stability of PINs by BRs differ.

PIN1 has been shown to be expressed in early leaf veins and margins and PIN1-mediated polar auxin transport to be important for the leaf venation pattern in Arabidopsis (40). Moreover, MPK6 could regulate the leaf venation pattern probably by regulating the PIN1 phosphorylation and polarity, although the main MPK6-targeted phosphorylation site Ser337, also targeted by AtSKs, was not involved (37). In addition, the auxin-regulated receptor CAMEL controls the cotyledon venation through modulation of the PIN1 phosphorylation and polarization (55). Our data showed that chemical activity inhibition or genetic knockout of the AtSKs resulted in impaired PIN1 polarity and defective cotyledon venation patterns, suggesting that BRs control leaf venation via AtSK-mediated phosphorylation of PIN1.

In summary, we adapted the CETSA MS method for intact Arabidopsis cells and demonstrated its ability to identify direct targets and downstream components of the small molecule target proteins. Such information is useful to understand the mode of action of the small molecules and the function of their target proteins. Moreover, we identified PIN1 as a substrate of AtSKs, uncovering an unknown BR mechanism and auxin cross-talk.

Materials and Methods

Plant Materials and Growth Condition.

A. thaliana (L.) Heynh. (accession Columbia-0 [Col-0]) was used, except for the atsk13RNAi bin2-3 bil1 bil2 quadruple and atsk11RNAi atsk12RNAi atsk13 bil2-3 bil1 bil2 sextuple mutants that were generated in Wassilewskija (Ws) background as previously described (40). The transgenic Arabidopsis plant expressing PIN1pro::PIN1-GFP had been described previously (43). The growth conditions are described in detail in .

The experimental procedures for CETSA, CETSA MS, phosphoproteomics, confocal microscopy, in vitro kinase assay, whole-mount in situ immunolocalization of PIN1, immunoprecipitation, qRT-PCR, FDA measurement, and statistical analysis are described in detail in . The data of CETSA MS and phosphoproteomics analysis are provided in Datasets S1–S3. The primers used in this study are listed in .