Selective Autophagy of BES1 Mediated by DSK2 Balances Plant Growth and Survival

SUMMARY

Plants encounter a variety of stresses and must fine-tune their growth and stress-response programs to best suit their environment. BES1 functions as a master regulator in the brassinosteroid (BR) pathway that promotes plant growth. Here, we show that BES1 interacts with the ubiquitin receptor protein DSK2 and is targeted to the autophagy pathway during stress via the interaction of DSK2 with ATG8, a ubiquitin-like protein directing autophagosome formation and cargo recruitment. Additionally, DSK2 is phosphorylated by the GSK3-like kinase BIN2, a negative regulator in the BR pathway. BIN2 phosphorylation of DSK2 flanking its ATG8 interacting motifs (AIMs) promotes DSK2-ATG8 interaction, thereby targeting BES1 for degradation. Accordingly, loss-of-function dsk2 mutants accumulate BES1, have altered global gene expression profiles, and have compromised stress responses. Our results thus reveal that plants coordinate growth and stress responses by integrating BR and autophagy pathways and identify the molecular basis of this crosstalk.

In Brief

Plants must carefully balance growth and survival. Nolan et al. show that brassinosteroid-regulated and growth-promoting transcription factor BES1 is degraded during drought and starvation stress via DSK2, a phosphorylation-regulated selective autophagy receptor, thus revealing a mechanism that allows plants to shut down growth during unfavorable conditions.

INTRODUCTION

Organisms encounter constantly changing environments and must respond appropriately to optimize their fitness and ensure survival. Growth and stress response programs generally antagonize one another, and as such need to be balanced (Claeys and Inze, 2013; Lopez-Maury et al., 2008). This need is exacerbated in sessile organisms such as plants that cannot easily relocate to escape adverse environmental conditions. Plants are therefore an excellent system for the study of coordination of growth and stress responses, and research in this area has important implications in optimizing plant growth under adverse environments (Skirycz et al., 2011).

Brassinosteroids (BRs) are one major family of growth-promoting plant hormones (Li and Chory, 1997). BRs are perceived by a receptor kinase, BRASSINOSTERIOID INSENSITIVE 1 (BRI1), and many other signaling components, to regulate the BRI1-EMS SUPRESSOR 1 (BES1) and BRASSINAZOLE-RESISTANT 1 (BZR1) family of transcription factors (Clouse, 2011). In the absence of BRs, a GSK3-like kinase BIN2 (Li and Nam, 2002) phosphorylates and inhibits BES1/BZR1 function through multiple mechanisms (Li and Jin, 2007). In the presence of BR, BIN2 kinase activity is inhibited, thus leading to the accumulation of dephosphorylated BES1/BZR1 in the nucleus to regulate target gene expression (Belkhadir and Jaillais, 2015; Guo et al., 2013). Recently, BR signaling has also been linked with stress responses, in part through BIN2 activity (Youn and Kim, 2015; Hao et al., 2013), but many molecular details are still unclear.

Post-translational regulation adds another level of complexity to BR signaling. BES1 and BZR1 can be degraded by the proteasome (Wang et al., 2013; He et al., 2002), and gain-of-function bes1-D or bzr1-D mutants exhibit stabilized BES1 or BZR1, respectively (Wang et al., 2002; Yin et al., 2002). BES1 is targeted for ubiquitin-mediated degradation by the Skp-CULLIN-F-box (SCF) E3 ubiquitin ligase MORE AXILLARY GROWTH LOCUS 2 (MAX2) during strigolactone-mediated control of shoot branching (Wang et al., 2013), and BZR1 is degraded in a COP1-dependent manner in response to darkness (Kim et al., 2014). These results demonstrated that regulated proteolysis of BES1/BZR1 plays important roles in diverse plant responses, and the key downstream components required for such processes remain to be fully defined.

Typically, ubiquitin-mediated protein degradation occurs through proteasome or autophagy pathways (Floyd et al., 2012; Kraft et al., 2010). Autophagy functions in the degradation and recycling of macromolecules and cytoplasmic organelles, often in response to stress conditions (Yang and Bassham, 2015). There is also recent evidence for selective autophagy in plants, whereby specific proteins or organelles are recognized by receptor proteins and degraded, although many details remain to be elucidated (Michaeli et al., 2016). A subset of these receptors contain a ubiquitin-binding domain and an ATG8-interacting motif (AIM), allowing them to recruit ubiquitinated cargo to ATG8-labeled autophagosomes (Floyd et al., 2012). Two such receptors in Arabidopsis are the NEXT TO BRCA1 GENE 1 (NBR1) homolog, which mediates degradation of ubiquitinated protein aggregates (Zhou et al., 2013; Svenning et al., 2011), and REGULATORY PARTICLE NON-ATPASE 10 (RPN10), which can target ubiquitinated proteasomes for autophagic degradation (Marshall et al., 2015).

DOMINANT SUPPRESSOR OF KAR 2 (DSK2) is a ubiquitin-binding receptor protein with known connections to protein degradation pathways in yeast, animals, and plants (Lee and Brown, 2012; Lin et al., 2011; Funakoshi et al., 2002). In Arabidopsis, two DSK2 paralogs exist as a result of tandem duplication (DSK2A and DSK2B), with 87% amino acid identity (Farmer et al., 2010). Both DSK2 proteins contain an N-terminal ubiquitin-like (UBL) domain that mediates their interaction with the proteasome and a C-terminal ubiquitin-associated (UBA) domain that can bind both K48 and K63 polyubiquitin chains (Lin et al., 2011). Interestingly, studies of the human DSK2 homologs (Ubiquilins) revealed that they can function in autophagy as LC3-interacting partners (Lee et al., 2013).

In this study, we found links between the regulation and activity of DSK2 to BR signaling, which leads to altered plant growth under drought and fixed-carbon starvation conditions. Specifically, we show that BES1 is targeted for autophagy-mediated degradation by direct interaction with DSK2 following abiotic stress. In addition, the interaction between DSK2 and ATG8 is regulated by BIN2 phosphorylation of DSK2. Loss-of-function dsk2 mutants have increased BES1 protein levels, altered global gene expression profiles, and compromised survival during stresses. Our results thus provide a previously unknown mechanism by which plants coordinate growth and stress responses by targeting a central growth regulator to the selective autophagy pathway via a phosphoregulated receptor protein.

RESULTS

BES1 Is Degraded through Autophagy and Proteasome Pathways

Although BES1 is known to be degraded in a ubiquitin-dependent manner, the role of autophagy in this process has not been extensively examined. To test the possibility that BES1 is regulated by autophagy in addition to the proteasome, we first investigated BES1 protein levels after treatment with concanamycin A (ConA), the cysteine protease inhibitor E64d, or MG132. ConA and E64d cause accumulation of proteins degraded by autophagy whereas MG132 blocks proteasomal degradation (Kisselev et al., 2012; Inoue et al., 2006; Droese et al., 1993). Remarkably, BES1 accumulated in response to inhibitors of both pathways (Figure 1A) while another transcription factor, RD26, accumulated after MG132 treatment, but not following treatment with ConA or E64d (Figure S1A). This indicates that BES1 can be degraded by both the autophagy and proteasome pathways, similar to HIF2α in hypoxia response (Liu et al., 2014) and β-catenin in the Wnt signaling pathway (Petherick et al., 2013). To further verify that BES1 is degraded via autophagy, we examined BES1 accumulation in the autophagy-deficient mutants atg5-1 and atg7-2 (Chung et al., 2010). Consistently, we found that BES1 protein levels accumulated in these autophagy-deficient mutants during mock treatments and that the application of autophagy inhibitors had no effect in these mutant backgrounds (Figure 1B). Quantification of BES1 protein levels showed that atg5-1 and atg7-2 plants had higher BES1 levels after treatment with MG132 compared with mock-treated controls (Figure S1B). Further, coapplication of MG132 with ConA in wild-type (WT) plants led to slightly increased levels of BES1 compared with the application of either inhibitor alone (Figure S1C, lanes 2, 3, and 5). These results indicate that BES1 can be degraded by both proteasome and autophagy pathways, but that the two pathways do not function completely redundantly, which is consistent with previous reports regarding proteins targeted by both proteasome- and autophagy-mediated degradation (Liu et al., 2014).

Figure 1. BES1 Is Degraded by Proteasome and Autophagy Pathways and Interacts with Ubiquitin Receptor Protein DSK2.

(A) Response of 10-day-old WT seedlings to proteasome and autophagy inhibitors. Seedlings treated for 6 hr in half-strength Murashige and Skoog medium with DMSO, 50 μM MG132, 20 μM E64d, or 1 μM ConA were analyzed by western blotting with anti-BES1 antibodies. HERK1 and TUBULIN served as loading controls.

(B) Response of BES1 to inhibitors as described above in 4-week-old WT, atg5-1, or atg7-2 leaf tissue.

(C) Ubiquitination of BES1. BES1-GFP was expressed in N. benthamiana and treated with mock solvent or inhibitors for 16 hr. Immunoprecipitated BES1-GFP was analyzed by western blotting with GFP or ubiquitin antibodies.

(D) Confocal microscopy images of Arabidopsis protoplasts expressing BES1-YFP under −sucrose conditions treated with DMSO or 1 μM ConA for 12 hr. Arrows indicate BES1-YFP puncta. Scale bar represents 5 μm.

(E) Confocal microscopy of BES1P:BES1-GFP transgenic plants grown for 5 days in light followed by 2 days of sucrose starvation. DMSO or 1 μM ConA was applied 16 hr prior to microscopy. Arrows indicate BES1-GFP puncta. Scale bar represents 20 μm.

(F) Quantification of BES1P:BES1-GFP puncta. Data represent mean ± SEM, n = 8; ***p < 0.001 (t test).

(G) GST pull-down showing interactions of GST-DSK2A and GST-DSK2B with BES1-MBP. BES1-MBP was detected using anti-MBP antibody. Loading indicates amounts of GST proteins used in the pull-down reactions.

(H) Interaction of BES1 with DSK2 in BiFC assays in N. benthamiana. Bright-field (BF) images are shown. Arrows indicate BES1-DSK2 puncta.

(I) Time course showing movement of BES1-DSK2 BiFC signals. Time of image acquisition is shown in seconds.

(J) Colocalization of BES1-DSK2 BiFC signals (YFP channel) with Cerulean-ATG8e (CFP channel).

Scale bar for all BiFC experiments represents 10 μm. See also Figure S1.

Since ubiquitination often triggers degradation through the proteasome and autophagy pathways, we analyzed BES1 ubiquitination. Immunoprecipitated BES1-GFP showed extensive high molecular weight laddering after MG132 or E64d treatment, reminiscent of ubiquitination (Figure 1C). These high molecular weight forms of BES1 cross-reacted with anti-ubiquitin antibody, indicating that ubiquitinated BES1 accumulates in response to proteasome and autophagy inhibitors. Furthermore, treatments of protoplasts expressing BES1-YFP with ConA resulted in accumulation of BES1-YFP in puncta within the vacuole, consistent with BES1 degradation through autophagy (Figure 1D). Similar results were obtained using BES1P:BES1-GFP transgenic lines in which BES1 expression was driven by its native promoter (Figures 1E and 1F), supporting the idea that BES1 puncta observed in protoplasts were not the result of artificially high levels of BES1 expression. Taken together, these results demonstrate that BES1 can be degraded by autophagy.

BES1 Interacts with the Ubiquitin Receptor Protein DSK2

To further explore the mechanisms of BES1 degradation, we performed yeast-two-hybrid screening and identified one of the BES1 interactors as DSK2A, a ubiquitin-binding receptor. We hypothesized that DSK2 might be involved in targeting BES1 for degradation. We first confirmed that BES1 interacts with both DSK2A and DSK2B in vitro using glutathione S-transferase (GST) pull-down and pairwise yeast-two-hybrid assays (Figures 1G and S1D) and also in planta with bimolecular fluorescence complementation (BiFC) and coimmunoprecipitation (coIP) (Figures 1H, S1E, and S1F). Coexpression of BES1-YFPN with DSK2A-YFPC resulted in strong YFP fluorescence signals (Figure 1H, left panels), some of which were present in mobile puncta (Figure 1I). Fluorescence signals were not observed in negative controls (Figure 1H, middle panels and Figure S1E, lower panels) and were greatly diminished when the ubiquitin-binding UBA domain of DSK2 was deleted (DSK2ΔUBA) (Figure 1H, right panels), indicating that efficient binding of BES1 to DSK2 in vivo may be promoted by ubiquitination of BES1. Consistent with this idea, coIP using DSK2A-GFP transgenic lines treated with the autophagy inhibitor ConA demonstrated that DSK2A-GFP immunoprecipitated with anti-GFP antibodies pulled down high molecular weight forms of BES1 (Figure S1F) that likely represent ubiquitinated BES1. We also tested the effect of BES1 phosphorylation on the BES1-DSK2 interaction using BES1 phosphorylated in vitro with BIN2 kinase. DSK2 interacted with both phosphorylated and unphosphorylated BES1, indicating that phosphorylation does not markedly influence the interaction between DSK2 and BES1 in vitro (Figure S1G).

To examine whether BES1-DSK2 BiFC puncta were of autophagic origin, we performed colocalization experiments using the autophagosome marker Cerulean-ATG8e (Liu et al., 2012). Reconstituted YFP signal resulting from coexpression of BES1-YFPN with DSK2A-YFPC colocalized extensively with the Cerulean-ATG8e (Figure 1J). DSK2 could also interact with itself (Figures S1E and S1H), in common with several autophagy receptors that often dimerize or multimerize to recruit their cargo for autophagic degradation (Ciuffa et al., 2015; Floyd et al., 2012). YFP signals resulting from interaction of DSK2A-YFPN with DSK2A-YFPC also colocalized with Cerulean-ATG8e whereas controls expressing DSK2A-YFPN with DSK2A-YFPC or Cerulean-ATG8e alone did not result in any colocalization signal (Figure S1H). DSK2 therefore interacts with BES1 and the DSK2-BES1 complex can localize to autophagosomes, suggesting that DSK2 recruits BES1 to the autophagy pathway.

DSK2 Acts as a Receptor for BES1 Degradation

To test the possibility that DSK2 functions as an autophagy receptor mediating BES1 degradation, we examined colocalization of BES1-YFP with Cerulean-ATG8e in Arabidopsis protoplasts under starvation and osmotic stress conditions in which autophagy is induced (Liu et al., 2009; Doelling et al., 2002; Hanaoka et al., 2002). Strong colocalization of BES1-YFP and Cerulean-ATG8e in autophagic bodies occurred upon sucrose starvation (Figure 2A, top panels) or osmotic stress with mannitol treatment (Figure 2A, bottom panels), but not in unstressed controls (Figure 2A, middle panels). The colocalization was not observed in single transformations of BES1-YFP or Cerulean-ATG8e (Figure S2A), indicating that colocalization signals were not an artifact of crosstalk between YFP and CFP channels. Both BES1-YFP and Cerulean-ATG8e puncta were absent in autophagy-deficient atg7-2 mutants (Figure 2B). On the other hand, DSK2 RNAi protoplasts in which both DSK2A and DSK2B were knocked down (Figure S2B) displayed normal ATG8e puncta, but failed to recruit BES1 to ATG8-labeled autophagosomes (Figure 2C), suggesting that DSK2 is required to target BES1 to autophagy but is not required for proper function of the core autophagy machinery. These observations were supported by quantification of protoplasts with visible BES1 autophagy, which was reduced in DSK2 RNAi as compared with WT under starvation (−suc) or osmotic (mannitol) stress (Figures 2D and S2C). In contrast, quantification of protoplasts displaying ATG8e autophagy did not reveal significant differences between WT and DSK2 RNAi (Figure S2D); confirming that bulk autophagy is not impaired in DSK2 RNAi. Furthermore, accumulation of BES1 was observed in DSK2 RNAi lines during mock inhibitor treatments, and DSK2 RNAi had reduced response to application of E64d or MG132 (Figure 3A), consistent with a role for DSK2 in degrading BES1.

Figure 2. DSK2 Recruits BES1 to ATG8-Labeled Autophagosomes during Stress.

(A) Representative images showing colocalization of BES1-YFP and Cerulean-ATG8 in WT Arabidopsis protoplasts. Protoplasts treated with control (+Suc), starvation (−Suc), or mannitol stress conditions were incubated with 1 μM ConA for 12 hr and imaged by confocal microscopy. CFP, YFP, or merged (CFP + YFP) fluorescence channels are shown along with bright-field (BF) images.

(B) Colocalization of BES1-YFP and Cerulean-ATG8 in atg7-2 protoplasts.

(C) Colocalization of BES1-YFP and Cerulean-ATG8 in DSK2 RNAi protoplasts.

(D) Quantification of protoplasts with BES1 autophagy. Protoplasts expressing BES1-YFP were treated with indicated treatments as described in (A). BES1 autophagy was defined by the presence of ≥3 autophagosomes per protoplast. Data represent mean of three biological replicates ± SEM, n ≥ 50. *p < 0.05 and ***p < 0.001, t test.

Scale bars indicate 10 μm. See also Figure S2.

Figure 3. BES1 Degradation by DSK2 and Autophagy Affects BR-Regulated Plant Growth Responses.

(A) BES1 accumulation in 4-week-old Arabidopsis plants following 6-hr inhibitor treatments. HERK1 was used as a loading control.

(B) BRZ treatments under stress conditions. Five-day-old plants were transferred to medium with indicated combinations of BRZ with or without sucrose or 350 mM mannitol to induce stress. Plants were incubated in darkness for 3 days followed by imaging and hypocotyl measurements. Scale bars represent 5 mm.

(C) Quantification of BRZ sensitivity (hypocotyl length BRZ 1,000 nM/hypocotyl length BRZ 0) from (B). **p < 0.01 and ***p < 0.001, t test.

(D) Response to BRZ under non-stress conditions. Data represent mean ± SEM. ***p < 0.001, t test.

See also Figure S3.

To investigate the effects of stress-induced BES1 degradation on BR-mediated plant growth responses, we measured hypocotyl lengths of WT, DSK2 RNAi, atg7-2, and bes1-D treated with sucrose starvation or mannitol-induced osmotic stress. Under stress conditions, DSK2 RNAi and atg7-2 displayed decreased sensitivity to the BR biosynthesis inhibitor brassinazole (BRZ) (Asami et al., 2000) (Figures 3B and 3C). DSK2 RNAi seedlings showed a mild BRZ-resistant phenotype under normal conditions, whereas atg7-2 was not significantly different than WT, and bes1-D was resistant to BRZ under all of the tested conditions (Figures 3D, S3A, and S3B). The BR response phenotype was consistent across two DSK2 RNAi lines (Figure S3C) that have been previously well characterized (Lin et al., 2011) and showed depleted DSK2 levels as monitored by immunodetection with anti-DSK2 antibodies (Figure S1D). These results indicate that DSK2 functions as an autophagy receptor during stress conditions to reduce BR-mediated plant growth and imply that DSK2 may also operate in an autophagy-independent manner in non-stress conditions, which is consistent with the known role of DSK2 in other protein degradation pathways (Farmer et al., 2010).

DSK2 Is Phosphorylated by BIN2 and Serves as a Phosphoregulated Autophagy Receptor

Since autophagy receptor proteins typically interact with ATG8, we tested the interaction of DSK2 with several representative ATG8 family members (Marshall et al., 2015; Yoshimoto et al., 2004). DSK2 physically interacted with all ATG8 members tested in GST pull-down assays, including ATG8e (Figure 4A). To further characterize the role of DSK2 as a possible autophagy receptor, we examined the DSK2 protein sequence for predicted AIMs using iLIR prediction software (Kalvari et al., 2014). AIMs are typified by the consensus sequence W/F/Y-X-X-L/I/V, which is often adjacent to acidic or phosphorylated residues (Kalvari et al., 2014). DSK2A has two regions with high-scoring AIMs, each of which is surrounded by multiple BIN2 consensus phosphorylation sites (S/T-X-X-X-S/T) (Figure 4B) (Zhao et al., 2002). This observation led us to hypothesize that DSK2 might be phosphorylated proximal to its AIM sequences by BIN2, thereby promoting physical interaction between DSK2 and ATG8 (Farre et al., 2013; Zhu et al., 2013; Wild et al., 2011). Further examination of the DSK2A (hereafter referred to as DSK2) protein sequence revealed 22 potential BIN2 phosphorylation sites (Figure 4B). We first performed phosphatase treatments of immunoprecipitated DSK2-GFP protein, which caused DSK2 to shift from a higher- to lower-migrating form. Thus, DSK2 can be phosphorylated in vivo (Figure 4C).

Figure 4. DSK2 Interacts with ATG8 and Is Phosphorylated by BIN2 Kinase around AIMs.

(A) Interaction of DSK2 with ATG8 (letters indicate ATG8 family members) in GST pull-down assays. DSK2A-MBP was detected with anti-MBP antibodies. Loading indicates amounts of GST proteins used in pull-down reactions.

(B) Schematic showing DSK2A protein sequence. Text colors indicate the following: red, predicted ATG8 interacting motifs (AIMs); white, AIM mutations present in DSK2 mAIM constructs; yellow, putative BIN2 phosphorylation sites (also marked with an asterisk in the schematic).

(C) Phosphatase treatment of DSK2-GFP expressed in N. benthamiana, immunoprecipitated with anti-GFP antibodies and treated with (+) or without (−) calf intestinal alkaline phosphatase (CIP) followed by Phos-tag SDS-PAGE and western blotting with anti-GFP antibodies.

(D) Interaction of BIN2 (BIN2-YFPN) with DSK2 (DSK2-YFPC) in N. benthamiana BiFC experiments. Arrows indicate signal from DSK2-BIN2 interaction.

(E) In vitro phosphorylation of DSK2 by BIN2. MBP or DSK2-MBP proteins (Full, aa 1–538; C1, aa 90–538; C2, aa 403–538) were phosphorylated by BIN2 kinase. Arrows indicate phosphorylated DSK2A or BIN2 autophosphorylation.

(F) Modulation of DSK2 phosphorylation by BIN2 in vivo as monitored by Phos-tag SDS-PAGE and western blotting with anti-DSK2 antibody. Arrows denote gel shifts indicating phosphorylated or unphosphorylated DSK2.

(G) Schematics of DSK2 proteins showing mutated BIN2 phosphorylation sites and AIM mutations. Predicted BIN2 phosphorylation sites are marked with an asterisk. When mutated, residues are represented as A (Ser or Thr to Ala) or D (Ser or Thr to Asp). AIM domains are indicated, with red color representing mutant AIMs (mAIM).

(H) Yeast-two-hybrid interaction of DSK2 with ATG8e as detected by β-galactosidase (LacZ) activity.

(I) BiFC assays of DSK2 (DSK2-YFPN), phosphomimic forms of DSK2 (DSK2^D-YFPN), or DSK2 AIM mutant (DSK2^mAIM-YFPN) with ATG8e (ATG8e-YFPC) in N. benthamiana. Arrows indicate signal from DSK2-ATG8 interactions.

(J) Quantification of DSK2-ATG8e BiFC signal from (I). Fluorescence was measured using ImageJ and normalized to controls expressing DSK2-YFPN with YFPC or ATG8e-YFPC with YFPN. Data represent mean of four independent images ± SEM. Different letters indicate statistically significant differences at p < 0.05 (t test).

(K) Coimmunoprecipitation in N. benthamiana showing that Cerulean-ATG8e immunoprecipitated with GFP-Trap interacts more strongly with DSK2^D−MYC than DSK2-MYC.

(L) Phenotype of two independent T2 transgenic lines overexpressing phosphomimic DSK2 plants (DSK2^D-MYC).

(M) BES1 protein levels in DSK2^D-MYC lines. Protein extracts from indicated lines were analyzed by western blotting. DSK2^D-MYC was detected with anti-MYC antibodies, while BES1 was detected with anti-BES1. A non-specific band from anti-MYC was used as a loading control.

All DSK2 constructs presented in this figure are derived from the DSK2A gene. See also Figure S4.

Next, we found that DSK2 interacts with and is phosphorylated by BIN2, consistent with the predicted BIN2 phosphorylation sites. Specifically, we observed physical association between DSK2 and BIN2 in BiFC assays (Figure 4D), yeast-two-hybrid assays (Figure S4A), and in vitro pull-down assays (Figure S4B). BIN2 efficiently phosphorylated DSK2 in in vitro kinase assays (Figure 4E). Phosphorylation was unaffected by deletion of the N-terminal UBL domain of DSK2 (DSK2 C1, amino acids [aa] 90–538), but was blocked in a truncated version of DSK2 (DSK2 C2, aa 402–538) that removed 18 of the 22 predicted BIN2 sites (Figures 4B and 4E).

We confirmed that BIN2 phosphorylates DSK2 in vivo by western blotting using anti-DSK2 antibodies coupled with Phostag SDS-PAGE, which causes slower migration of phosphorylated proteins (Kinoshita et al., 2006). In this assay, BIN2 loss-of-function bin2-3 bil1 bil2 triple mutants (bin2-T) (Yan et al., 2009) displayed decreased DSK2 phosphorylation (Figure 4F). Conversely, gain-of-function (bin2-1D) mutants (Li and Nam, 2002) showed slower migration of DSK2 due to increased phosphorylation (Figure 4F).

To test the effect of BIN2 phosphorylation on DSK2 function, we generated a series of mutants, changing BIN2 sites to aspartic acid (D) or alanine (A) to mimic or abolish phosphorylation, respectively. We found that at least a portion of these sites can be phosphorylated in vivo since coexpression of DSK2-MYC with BIN2 in Nicotiana benthamiana led to a higher-migrating form of DSK2, but no shift was observed when BIN2 was expressed with DSK2^A, a construct in which all putative BIN2 phosphorylation sites were mutated to alanine (Figure S4C).

Given the precedence for phosphorylation in increasing autophagy receptor interactions with ATG8 (Wild et al., 2011), we selected ATG8e as a representative ATG8 family member for the investigation of the effect of phosphorylation on interaction of DSK2 with ATG8 (Figures 4G and 4H). Using yeast-two-hybrid assays we found that WT DSK2 interacted with ATG8e, and a phosphomimic version of DSK2 (DSK2^D) had increased interaction with ATG8e (Figure 4H, columns 2 and 3). In contrast, loss-of-function (DSK2^A) mutants showed decreased interaction (Figure 4H, column 4). The phosphomimic mutation of the putative phosphorylation sites proximal to the AIM domains (DSK2^D12) was sufficient to increase the interaction of DSK2 with ATG8e (Figure 4H, column 7). Moreover, mutation of the AIM sequences of DSK2 abolished the interaction with ATG8e (Figure 4H, columns 5 and 6). These DSK2 mutants showed a similar trend when tested in plants using BiFC. Specifically, DSK2^D showed increased interaction with ATG8e compared with DSK2 while mutation of DSK2 AIMs in DSK2^mAIM also abolished the interaction in planta (Figures 4I and 4J). Furthermore, coIP assays demonstrated that immunoprecipitated Cerulean-ATG8e interacted more strongly with phosphomimic DSK2^D-MYC than with DSK2-MYC (Figure 4K), confirming that phosphorylation of DSK2 promotes interaction with ATG8 in vivo.

We next confirmed site-specific phosphorylation of DSK2 by BIN2 using peptide mass spectrometry. Several phosphorylated peptides were detected from DSK2-MBP phosphorylated in vitro with BIN2-MBP (Tables S1 and S2), whereas no phosphorylation was detected in mock-treated DSK2-MBP. These phosphorylation sites include at least four phosphorylated residues near AIM1 (four sites between aa 5 and 41) and two residues around AIM2 (S240 and S244), suggesting that BIN2 can phosphorylate DSK2 proximal to its AIM domains. To investigate the function of DSK2 phosphorylation on plant growth and BES1 stability, we generated transgenic plants overexpressing phosphomimic DSK2 (DSK2^D). These lines exhibited reduced growth compared with WT plants and had decreased BES1 protein levels (Figures 4L and 4M), consistent with the role of phosphorylated DSK2 in promoting BES1 degradation. In contrast, transgenic lines expressing inactive DSK2^A or DSK2^mAIM variants did not cause dramatic changes in plant growth or BES1 protein levels under the conditions tested (Figures S4D and S4E). Taken together, these results demonstrate that DSK2 is a phosphoregulated ATG8-interacting receptor protein mediating BES1 degradation.

DSK2 Is Involved in BES1 Degradation during Drought Stress

Autophagy is induced by numerous stimuli, including nutrient stress and osmotic or drought stress (Liu et al., 2009). Thus, we hypothesized that autophagic degradation of BES1 may occur under such stress conditions. We confirmed that autophagy-defective atg7-2 plants were susceptible to drought (Zhou et al., 2013) (Figure 5A). We next tested the role of BR responses in drought. Constitutive activation of BR responses in bes1-D led to increased susceptibility to drought, while bri1-301, a loss-of-function mutant in the BR pathway, was resistant to drought treatments (Figure 5A). Based on these findings, we examined how BES1 levels may be regulated during drought conditions. To simulate drought conditions in a controlled and reproducible manner, we subjected plants to dehydration treatments, which have been widely used to examine plant drought responses (Sakuma et al., 2006; Urao et al., 1993). Following 4 hr of dehydration, BES1 levels were substantially reduced in WT plants, while BES1 accumulated in atg7-2 and DSK2 RNAi under control conditions and minimal reduction was observed in response to dehydration (Figure 5B). BES1 levels followed a similar trend in plants treated with sublethal drought in soil (Figure 5C), showing accumulation in DSK2 RNAi and atg7-2 backgrounds. Accumulation of BES1 in the mutant backgrounds was not due to increased transcription, as BES1 expression levels were not markedly increased in DSK2 RNAi or atg7-2 compared with WT during control or stress conditions (Figures S5A and S5B). Moreover, experiments in which translation was inhibited via cycloheximide (CHX) treatment showed that BES1 stability was decreased to a greater extent in drought versus control treated WT plants, especially at later time points of CHX treatment (Figures 5D [top panels] and S5C), whereas a similar reduction in BES1 was not observed after drought and CHX treatment in DSK2 RNAi or atg7-2 plants (Figure 5D, middle and bottom panels). These results demonstrate that BES1 is regulated at the post-translational level during drought stress.

Figure 5. Degradation of BES1 Mediated by DSK2 and Autophagy during Drought Stress.

(A) Plant phenotypes for indicated lines after drought recovery assays.

(B) BES1 protein levels during dehydration. Plants were subjected to control or 4-hr dehydration conditions and BES1 levels analyzed by western blotting. HERK1 was used as a loading control.

(C) BES1 protein levels during drought in soil. Plants were treated with control (well-watered) or sublethal drought conditions, and protein was extracted from plant leaf tissue for analysis by western blotting as described in (B).

(D) BES1 protein levels following cycloheximide (CHX) treatment during control or drought treatments. Indicated genotypes were grown for 4 weeks in control or sublethal drought conditions followed by treatment with 500 μM CHX for the indicated time points. Ponceau S staining was used as loading control.

(E) Quantification of percent survival following drought recovery. Plants producing new leaves after the 7-day recovery period were scored as survivors. Data represent mean survival of three biological replicates of 12–16 plants ± SEM. **p < 0.01 (t test).

(F) Detached leaf water loss assays. Water loss represents proportion of total weight lost compared with initial weight. Data represent mean ± SEM from two to three biological replicates. Differences between WT and mutants are statistically significant at p < 0.05 for all time points shown except 30 min (t test).

(G) Comparison of dehydration-regulated and DSK2 RNAi dehydration DE genes in whole-transcriptome RNA-seq.

(H) Top 20 significantly enriched GO terms in DSK2 RNAi dehydration DE genes as ranked by false discovery rate (FDR).

(I) Clustering of 2,522 genes differentially expressed in DSK2 RNAi dehydration versus WT dehydration. Color legend indicates normalized gene expression values.

(J) Comparison of DSK2 RNAi dehydration DE genes with BES1/BZR1 target genes from genome-wide chromatin immunoprecipitation datasets.

See also Figures S5–S7.

Since DSK2 RNAi plants have elevated BES1 levels under drought conditions, we expected them to behave similarly to bes1-D. In support of this hypothesis, DSK2 RNAi plants also had drastically reduced survival in drought assays (Figures 5A, 5E, and S6A), and lost water more quickly in detached leaf water loss assays (Figure 5F). Meanwhile, DSK2 expression remained decreased in DSK2 RNAi compared with WT under stress conditions (Figures S5D and S5E).

Given the strong phenotype of DSK2 RNAi under drought conditions, we next examined changes in the transcriptome during control and dehydration conditions using RNA sequencing (RNA-seq). In dehydrated WT plants, 554 genes were differentially expressed (DE) compared with mock-treated controls (dehydration DE genes; Table S3). These dehydration-responsive transcripts exhibit a high concordance with previously published dehydration datasets and an enrichment of drought-related gene ontology (GO) terms (Figures S6B and S6C). We observed few genes that were DE in DSK2 RNAi plants compared with WT under control conditions (Table S3), but strikingly, 2,522 genes were DE in DSK2 RNAi compared with WT under dehydration stress (referred to as DSK2 RNAi dehydration DE genes). The DE genes in dehydrated DSK2 RNAi lines opposed dehydration-regulated genes (Figure 5G) and were enriched for drought- and/or dehydration-related GO terms (Figure 5H). Furthermore, clustering analysis of DSK2 RNAi dehydration DE genes (Figures 5I and S6D) revealed many genes with increased expression during dehydration in WT that showed lower expression patterns in DSK2 RNAi and minimal induction by the treatment (group I), while others were repressed in WT by dehydration but had higher expression in DSK2 RNAi (group II). These transcriptional changes following dehydration are consistent with the drought-hypersensitive phenotype of the DSK2 RNAi mutant.

Many previously published drought-regulated genes (Maruyama et al., 2009) were present in these datasets and showed altered expression in DSK2 RNAi lines (Figure S6E). There is also a high degree of overlap between BES1/BZR1 target genes and DSK2-regulated genes (Figure 5J and Table S3). Comparison of BR- and dehydration-regulated genes indicated that drought and BR pathways antagonize each other (Figures S6F–S6H). These results demonstrate that perturbation of BES1 levels in DSK2 RNAi plants coincides with the dramatic changes in drought phenotypes and drought-related gene expression.

We further confirmed the role of BES1 in drought stress resistance by testing the drought phenotype of previously described BES1-RNAi plants (Yin et al., 2005). BES1 RNAi had increased resistance to drought compared with WT (Figures S7A and S7B) and decreased BES1 protein levels (Figure S7C). We then generated DSK2 RNAi BES1 RNAi double transgenic plants in which both BES1 and DSK2 were reduced (Figure S7D) to determine the extent to which BES1 accumulation was responsible for the increased drought sensitivity in DSK2 RNAi. Drought survival assays indicated that decreased BES1 in DSK2 RNAi BES1 RNAi plants could restore the drought survival phenotype of DSK2 RNAi to near WT levels (Figures S7E and S7F). These observations indicate that BES1 accumulation in DSK2 RNAi is a major contributor to the impaired survival of DSK2 RNAi plants during drought stress.

We also tested the effect of modulating BIN2 activity on BES1 protein levels and plant survival during drought stress. We found that loss-of-function bin2-T mutants had higher BES1 protein levels compared with WT after dehydration treatment, whereas BES1 levels were reduced in gain-of-function bin2-1D mutants during dehydration (Figure S7G). Furthermore, bin2-1D plants were resistant to drought stress compared with WT. However, bin2-T mutants did not show obviously decreased drought survival under the conditions tested (Figures S7H and S7I), likely due to the large array of substrates targeted by BIN2 kinase (Youn and Kim, 2015). Taken together, these results support a role for BIN2 in modulating BES1 protein levels and survival under stress conditions.

DSK2 Is Involved in BES1 Degradation during Fixed-Carbon Starvation

In addition to drought, autophagy is strongly induced under nutrient-limiting conditions, including fixed-carbon starvation, and autophagy-deficient mutants have reduced survival under carbon or nitrogen starvation (Thompson et al., 2005; Contento et al., 2004). Recently, it has been shown that BRs are required for sugar-promoted hypocotyl elongation in the dark and that BZR1 is both transcriptionally and post-translationally regulated by sucrose, which suggests that BES1/BZR1 protein levels may be controlled under low-energy conditions (Zhang et al., 2015). We confirmed the sensitivity of atg7-2 to starvation stress in our assays and found that constitutive BR response in bes1-D led to enhanced sensitivity to fixed-carbon starvation (Figures 6A and 6B). We next examined BES1 protein levels after 5 days of dark treatment. Fixed-carbon starvation reduced the level of BES1 in WT plants, whereas BES1 was reduced to a lesser extent in atg7-2 and DSK2 RNAi (Figures 6C and S8A). Regulation of BES1 under these conditions was at least partially due to reduced energy availability rather than dark conditions, since addition of sucrose during starvation treatments increased BES1 levels (Figure S8B). Furthermore, we tested the fixed-carbon starvation survival phenotypes of DSK2 RNAi lines. In line with the BES1 protein accumulation, DSK2 RNAi plants had markedly reduced recovery compared with WT in a fixed-carbon starvation time course with seedlings grown on medium without sucrose (Figures 6B and S8C) or after 8 days of dark treatment on soil-grown plants (Figure 6A). Additionally we tested the role of BIN2 in starvation stress by examining the phenotypes of bin2 mutants in starvation survival assays. Similar to our observations in drought stress, gain-of-function bin2-1D mutants were resistant to starvation; however, bin2-T mutants were also more resistant than WT to starvation, raising the possibility that additional BIN2 homologs or substrates may be regulated during starvation in bin2-T mutants (Figure S8D).

Figure 6. Degradation of BES1 under Fixed-Carbon Starvation via DSK2 and Autophagy.

(A) Representative images of plant survival following 8-day fixed-carbon starvation treatment. Percentage survival is indicated from two biological repeats of 7–12 plants ± SD. Differences between WT and mutants are statistically significant, p < 0.05 (t test).

(B) Time course of fixed-carbon starvation survival. Plants were grown on half-strength Linsmaier and Skoog plates without sucrose for 1 week and transferred to darkness for indicated time points followed by 1 week of recovery under light. Data represent mean ± SEM for three to four biological replicates of 20 plants each. Differences between WT and mutants are statistically significant, p < 0.05 (t test).

(C) Western blot analysis of BES1 protein levels following 5-day dark treatment. A general transcription factor, IWS1, served as a loading control.

(D) Comparison of starvation-regulated and DSK2 RNAi starvation DE genes in whole-transcriptome RNA-seq.

(E) Clustering analysis of 3,077 genes that are starvation upregulated and downregulated in DSK2 RNAi starvation or starvation downregulated and upregulated in DSK2 RNAi starvation. Color legend indicates normalized gene expression values.

(F) Top 20 significantly enriched GO terms in DSK2 RNAi starvation DE genes as ranked by false discovery rate (FDR).

See also Figure S8.

We also performed RNA-seq experiments under starvation conditions and found 6,737 genes that were DE in DSK2 RNAi lines compared with WT (designated DSK2 RNAi starvation DE genes). Examination of these genes showed that starvation-responsive gene expression was attenuated in DSK2 RNAi plants as illustrated by the opposing pattern of overlap observed between DSK2 RNAi starvation DE genes and genes regulated by starvation in WT. A total of 1,677 out of 4,070 DSK2 RNAi starvation upregulated genes and 1,400 of 2,667 DSK2 RNAi starvation downregulated genes overlapped with genes downregulated or upregulated in WT starvation, respectively (Figures 6D and S8E). These differential changes were also evident from clustering analysis (Figure 6E). The DE genes in starved DSK2 RNAi lines were enriched for GO terms related to plant growth and stress responses (Figure 6F). Further comparisons revealed that many DE genes in starved DSK2 RNAi lines are BES1/BZR1 targets (24.7%, 1,664/6,737 genes) (Figure S8F). Moreover, we observed a large degree of overlap among DSK2 RNAi starvation DE genes and those DE in atg7-2 and bes1-D during starvation (Figures S8G and S8H), suggesting functional overlap among DSK2, ATG7, and BES1. Furthermore, 41% of BR-induced genes (1,601/3,898) and 33.6% of BR-repressed genes (1,274/3,789) were regulated in opposite directions (down- and upregulated, respectively) during starvation (Figures S8I–S8K), indicating a mostly inverse relationship between BR function and starvation response. Taken together, these results demonstrate that BES1 and BRs play a negative role in survival during fixed-carbon limitation and that BES1 is degraded by DSK2 and autophagy pathways under these conditions.

SINAT Family E3 Ubiquitin Ligases Are Involved in BES1 Degradation during Starvation

In addition to DSK2, we also recovered a RING E3 ubiquitin ligase, SINAT2, as a BES1-interacting partner via yeast-two-hybrid screening. Several SINAT family members directly interact with and ubiquitinate BES1 (Yang et al., 2017 [this issue of Developmental Cell]). To test the possible role of SINAT2 in DSK2-mediated degradation of BES1, we first examined the physical interaction of SINAT2 with DSK2. Strikingly, SINAT2 strongly interacted with DSK2 in BiFC assays (Figure 7A), likely in the nucleus and also in puncta. GST pull-down assays demonstrated that the DSK2-SINAT2 interaction is direct (Figure 7B). These data indicate that SINAT2 and DSK2 form a complex to carry out BES1 degradation and suggest that SINAT family E3 ligases might be involved in targeting BES1 for degradation during stress. SINAT E3 ligases have been previously implicated in stress responses (Bao et al., 2014), and we found that three out of six SINAT family members were transcriptionally induced by starvation (Figure 7C). Furthermore, BES1 protein accumulated following starvation treatment in SINAT RNAi lines compared with WT (Figures 7D and S9A) and SINAT RNAi plants exhibited a starvation hypersensitive phenotype, showing dramatically reduced survival in fixed-carbon starvation assays (Figure 7E). Accordingly, BES1 ubiquitination was reduced in SINAT RNAi plants compared with WT under starvation conditions when autophagy-mediated degradation was blocked with E64d (Figure S9B). These findings reveal that SINAT E3 ubiquitin ligases are involved in targeting BES1 for degradation during starvation (Figure 7F).

Figure 7. SINAT E3 Ligases Are Involved in BES1 Degradation during Starvation.

(A) Interaction of DSK2 and SINAT2 in BiFC assays in N. benthamiana. Bright-field (BF) images are shown. Scale bars indicate 10 μm.

(B) GST pull-down assays showing interaction of SINAT2 with DSK2. SINAT2-MBP was detected with anti-MBP antibodies. Loading indicates amounts of GST proteins used.

(C) Expression levels of SINAT family E3 ligases from starvation RNA-seq experiments. Data represent mean ± SEM. *p < 0.05 and **p < 0.01 (t test).

(D) Western blot analysis of BES1 protein levels following 3-day starvation treatment. IWS1 served as a loading control.

(E) Representative images of plant survival following 8 days of fixed-carbon starvation treatment. Numbers indicate recovered plants after a 1-week recovery period. The experiment was repeated twice with similar results.

(F) A Model for BES1 protein degradation through selective autophagy. Under stress conditions, BES1 is ubiquitinated and targeted for degradation by SINAT E3 ubiquitin ligases, partly due to the induction of SINAT genes by stresses. Ubiquitin receptor DSK2 interacts with BES1 and recruits BES1 to autophagy through DSK2-ATG8 interactions. When activated under stress conditions, BIN2 phosphorylates DSK2 flanking DSK2’s AIM domains and potentiates DSK2-ATG8 interactions, promoting BES1 degradation through selective autophagy.

See also Figure S9.

DISCUSSION

Organisms develop strategies to coordinate growth and stress responses. Multiple mechanisms have been reported that allow for inhibition of growth when stress is encountered, including global reprogramming of gene expression (Harb et al., 2010), RNA processing or sequestration (Weber et al., 2008), and translation inhibition (Ren et al., 2011). Our studies established a major mechanism for the coordination of plant growth and stress responses. We demonstrated that targeting of a central growth regulator, BES1, to autophagy-mediated degradation by ubiquitin receptor DSK2 plays an important role in slowing down plant growth under dehydration and fixed-carbon starvation (Figure 7F). We showed that GSK kinase BIN2, which is repressed by growth hormone BRs and activated by stresses (Youn and Kim, 2015; Dal Santo et al., 2012; Zhang et al., 2009; Charrier et al., 2002), controls BES1 autophagy by modulating DSK2-ATG8 interaction. Our studies also revealed a function for SINATs, which encode E3 ubiquitin ligases that target the active unphosphorylated form of BES1 to influence BR-regulated growth in a light-dependent manner (Yang et al., 2017 [this issue of Developmental Cell]). Our results indicate that SINATs are induced by starvation stress and are involved in targeting BES1 for degradation under starvation conditions. Thus, both BIN2 and SINATs are activated by stress to potentiate BES1 degradation.

Our results revealed crosstalk between autophagy and plant steroid hormone signaling pathways. Both autophagy (Liu et al., 2009; Doelling et al., 2002; Hanaoka et al., 2002) and BRs (Zhang et al., 2015; Hao et al., 2013) have been linked to drought and starvation, but the role of BRs in controlling plant survival and mechanisms regulating BES1 during these stresses is not completely understood. Recently, it was reported that TARGET OF RAPAMYCIN (TOR) functions to activate BR signaling under energy-replete conditions by stabilizing BZR1 and that tor rnai-induced BZR1 degradation was attenuated by treatment with 3-methyladenine, an inhibitor of autophagy (Zhang et al., 2016). These results suggest that BZR1 may be degraded through autophagy, but genetic and cell biological evidence for autophagy-mediated degradation of BES1/BZR1 family transcription factors as well as the E3 ligase and autophagy receptor that mediate this process remained to be established. The BES1-DSK2-ATG8 interaction demonstrated in this study provides a molecular mechanism connecting autophagy and BR pathways, which allows plants to slow down growth under stress conditions. Our studies revealed that BES1 negatively regulates plant survival under drought and starvation and that DSK2 mediates degradation of BES1 under these conditions (Figures 5 and 6). Accumulation of BES1 in DSK2 RNAi or atg7-2 mutants coincided with decreased plant survival under both drought and fixed-carbon starvation. In contrast, DSK2 RNAi and atg7-2 mutants were resistant to the inhibition of BR biosynthesis during stress (Figure 3), suggesting that BES1 accumulation in these mutants promotes plant growth during stress. Furthermore, expression of a constitutive active DSK2 (DSK2^D, mimicking BIN2 phosphorylation) led to reduced BES1 protein levels and decreased plant growth (Figure 4). Taken together, these genetic and physiological studies indicate that DSK2-mediated BES1 degradation leads to reduced plant growth under stress conditions and is required for optimal plant responses to stresses.

This study established DSK2 as a selective autophagy receptor. DSK2 has ties to both proteasome and autophagy pathways. The human homologs of DSK2, called Ubiquilins, have been implicated in protein degradation pathways including autophagy as LC3 (ortholog of ATG8) interactors (Lee et al., 2013), but their specific functions in autophagy have not been fully defined. In Arabidopsis, DSK2 was shown to participate in ubiquitinated cargo delivery to the proteasome, along with other receptors (Lin et al., 2011; Farmer et al., 2010; Vierstra, 2009). Our results demonstrate that DSK2 targets BES1 through selective autophagy. DSK2 is required for recruitment of BES1 to autophagy, but ATG8-labeled autophagosomes are not affected in DSK2 RNAi lines (Figure 2), suggesting that DSK2 functions in selective degradation of BES1 but does not regulate bulk autophagy. Although selective autophagy has been shown to function in plant systems, the full repertoire of autophagy receptors and their specific cargoes is only beginning to be fully appreciated (Michaeli et al., 2016). Multiple lines of evidence demonstrated that BES1-DSK2-ATG8 interactions function to target BES1 to selective autophagy during stress (Figures 1 and 2). Considered together, our results expand the role of DSK2 in ubiquitin-mediated proteolysis by showing DSK2’s role in selective autophagy and by providing a molecular basis for DSK2 in targeting specific proteins such as BES1 for degradation.

Another significant finding of this study is that DSK2-ATG8 interaction is modulated by BIN2 kinase, a negative regulator of the BR pathway. BIN2 phosphorylates BES1/BZR1 as well as numerous other substrates involved in diverse aspects of plant growth, development, and stress responses (Youn and Kim, 2015). DSK2 is phosphorylated by BIN2 flanking its AIM domains (Figures 4 and S4), which is a typical pattern of autophagy receptors that are regulated by phosphorylation (Farre et al., 2013; Zhu et al., 2013; Wild et al., 2011). Mutational analysis indicated that DSK2 phosphomimic forms had increased interaction with ATG8, suggesting that BIN2 induces BES1 degradation by phosphorylating DSK2 and enhancing its interaction with the autophagy protein ATG8. These results provide a new layer of regulation of BES1 by BIN2 kinase.

The role of DSK2 in the regulation of BES1 during stress responses is further corroborated by global gene expression studies. Our RNA-seq analyses showed that BES1 accumulation in DSK2 RNAi was associated with altered expression of a large number of genes under dehydration and starvation conditions. In contrast, few genes were affected in DSK2 RNAi plants in unstressed control conditions, indicating that DSK2 functions primarily during stresses. BES1 and BZR1 bind to the promoters of about 6,600 target genes as determined by genome-wide chromatin immunoprecipitation studies (Yu et al., 2011; Sun et al., 2010). Strikingly, a large proportion of the misregulated genes under both stress conditions were BES1/BZR1 targets (Figures 5 and S9). These results suggest that BES1 accumulation under stress conditions in DSK2 RNAi is associated with changes in BES1/BZR1 target gene expression. Moreover, the expression of stress-regulated genes in dehydration and starvation generally opposed that of BR-regulated genes, indicating that BR responses are reduced during stress through degradation of BES1 by DSK2 and autophagy.

In metazoans, regulation of β-catenin, a positive regulator of cell proliferation in the Wnt pathway, provides a remarkable example of how growth and stress pathways can crosstalk through targeted protein degradation. β-Catenin reduces autophagy and inhibits the expression of the autophagy receptor p62 under normal conditions, but is targeted for degradation through interaction with the autophagy protein LC3 under autophagy-inducing conditions (Petherick et al., 2013). Similarly, our results showed that autophagy-mediated degradation of BES1 under stress conditions affects BR-regulated growth and stress responses; however, whether the BR pathway also influences autophagy remains an open question. Given the similarities in signaling mechanisms between the Wnt and BR signaling pathways (Yin et al., 2002), future studies should determine whether there is a reciprocal regulation of autophagy by BRs, which could provide additional means for plants to balance growth and stress responses.

STAR★METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-BES1	(Yu et al., 2011)	N/A
Rabbit polyclonal anti-HERK1	(Guo et al., 2009)	N/A
Rabbit polyclonal anti-DSK2	(Lin et al., 2011)	N/A
Rabbit polyclonal anti-GFP	This study	N/A
Rabbit polyclonal anti-c-MYC	Sigma-Aldrich	Cat#C3956; RRID: AB_439680
Rabbit polyclonal anti-IWS1	This study	N/A
Chicken polyclonal anti-Ubiquitin	(Pratelli et al., 2012)	N/A
Mouse monoclonal anti-MBP	New England Biolabs	Cat#E8032S; RRID: AB_1559732
Mouse monoclonal anti-Tubulin	Sigma-Aldrich	Cat#T8203; RRID: AB_1841230
Mouse monoclonal anti-c-MYC	Cell Signaling Technology	Cat#2278S; RRID: AB_10693332
Chemicals, Peptides, and Recombinant Proteins
MG132	Sigma-Aldrich	Cat#M7449
Concanamycin A	Sigma-Aldrich	Cat#C9705
E64d	Sigma-Aldrich	Cat#E8940
Mannitol	Sigma-Aldrich	Cat#M4125
Cycloheximide	Sigma-Aldrich	Cat#C4859
Brassinazole	Tadao Asami (Asami et al., 2000)	N/A
TUBE2 Agarose	LifeSensors	Cat#UM402
Phos-tag	Wako	Cat#AAL-107
Complete protease inhibitor cocktail tablets	Roche	Cat#11836170001
Trizol	Thermo Fisher	Cat#15596018
RNeasy Kit	Qiagen	Cat# 74904
Glu-C	Thermo Fisher	Cat#90054
Trypsin	Roche	Cat# 03708969001
Linsmaier and Skoog	Caisson Laboratories	Cat#LSP03-1LT
Deposited Data
Raw mass spectra	MassIVE repository	MassIVE ID: MSV000079641
Raw data files for DNA sequencing	NCBI Gene Expression Omnibus	GEO: GSE93420
Experimental Models: Organisms/Strains
E. coli BL21 Gold (DE3)	Thermo Fisher	Cat#50-125-348
Agrobacterium tumefaciens (strain GV3101)		N/A
Arabidopsis thaliana: WT Col-0		N/A
Arabidopsis thaliana: atg5-1	(Thompson et al., 2005)	N/A
Arabidopsis thaliana: atg7-2	(Chung et al., 2010)	N/A
Arabidopsis thaliana: bri1-301	(Li and Nam, 2002)	N/A
Arabidopsis thaliana: bes1-D	(Vilarrasa-Blasi et al., 2015; Yin et al., 2002)	N/A
Arabidopsis thaliana: bin2-1D	(Li et al., 2001)	N/A
Arabidopsis thaliana: bin2-3 bil1 bil2	(Yan et al., 2009)	N/A
Arabidopsis thaliana: BES1P-BES1-GFP	(Yin et al., 2002)	N/A
Arabidopsis thaliana: DSK2-RNAi 2-6	(Lin et al., 2011)	N/A
Arabidopsis thaliana: DSK2-RNAi 1-4	(Lin et al., 2011)	N/A
Arabidopsis thaliana: SINAT RNAi #5	(Yang et al., 2017)	N/A
Arabidopsis thaliana: SINAT RNAi #14	(Yang et al., 2017)	N/A
Arabidopsis thaliana: DSK2AD-MYC-1	This study	N/A
Arabidopsis thaliana: DSK2AD-MYC-2	This study	N/A
Arabidopsis thaliana: DSK2AA-MYC-1	This study	N/A
Arabidopsis thaliana: DSK2AA-MYC-2	This study	N/A
Arabidopsis thaliana: DSK2AmAIM-MYC-1	This study	N/A
Arabidopsis thaliana: DSK2AmAIM-MYC-2	This study	N/A
Arabidopsis thaliana: DSK2A-GFP	This study	N/A
Recombinant DNA
pET42a GST	Novagen	Cat#70561
pET42a GST-DSK2A	This study	N/A
pET42a GST-DSK2B	This study	N/A
pET42a GST-ATG8a	This study	N/A
pET42a GST-ATG8e	This study	N/A
pET42a GST-ATG8f	This study	N/A
pET42a GST-ATG8i	This study	N/A
pET42a GST-BIN2	(Yin et al., 2002)	N/A
pET28a HIS-DSK2A	This study	N/A
pETMALc-H MBP	Pryor and Leiting, 1997	N/A
pETMBP-H	This study	N/A
pETMALc-H MBP-BES1	(Yin et al., 2002)	N/A
pETMALc-H MBP-SINAT2	This study	N/A
pETMALc-H MBP-BIN2	(Yin et al., 2002)	N/A
pETMBP-H MBP-DSK2A	This study	N/A
pETMBP-H MBP-DSK2A C1 (AA 90-538)	This study	N/A
pETMBP-H MBP-DSK2A C2 (AA 403-538)	This study	N/A
pGBKT7	TAKARA	Cat#630443
pGADT7	TAKARA	Cat#630442
pGADT7 BES1	(Yin et al., 2002)	N/A
pGADT7 BIN2	(Yin et al., 2002)	N/A
pGADT7 DSK2A	This study	N/A
pGADT7 ATG8e	This study	N/A
pGBKT7 DSK2A	This study	N/A
pGBKT7 DSK2AD	This study	N/A
pGBKT7 DSK2AA	This study	N/A
pGBKT7 DSK2AmAIM	This study	N/A
pGBKT7 DSK2AA mAIM	This study	N/A
pGBKT7 DSK2AD12	This study	N/A
pGBKT7 DSK2AA12	This study	N/A
pYY46 35S:BES1-GFP	(Yin et al., 2002)	N/A
pYY46 35S:DSK2-GFP	This study	N/A
pXY136 35S:BES1-YFP	This study	N/A
pLZ0102 35S:BIN2-1D-YFP	This study	N/A
pPZP211-BES1P:BES1-GFP	(Yin et al., 2002)	N/A
pPZP211-35S:Cerulean-ATG8e	This study	N/A
pAN578 35S:Cerulean-ATG8e	(Liu et al., 2012)	N/A
pYY361 BRI1P: DSK2AD-MYC	This study	N/A
pYY361 BRI1P: DSK2AA-MYC	This study	N/A
pYY361 BRI1P: DSK2AmAIM-MYC	This study	N/A
pXY103 35S:YFPN	(Yu et al., 2008)	N/A
pXY104 35S:YFPC	(Yu et al., 2008)	N/A
pXY103 35S:BES1-YFPN	(Wang et al., 2014)	N/A
pXY103 35S:DSK2A-YFPN	This study	N/A
pXY103 35S: DSK2AD -YFPN	This study	N/A
pXY103 35S: DSK2AmAIM -YFPN	This study	N/A
pXY103 35S:BIN2-YFPN	This study	N/A
pXY104 35S:DSK2A-YFPC	This study	N/A
pXY104 35S:DSK2A DUBA-YFPC	This study	N/A
pXY104 35S:DSK2B-YFPC	This study	N/A
pXY104 35S:ATG8e-YFPC	This study	N/A
pXY104 35S:SINAT2-YFPC	This study	N/A
Oligonucleotides
Full list of DSK2A variants is presented in Table S4
Full list of primers is presented in Table S5
Software and Algorithms
R statistical package	www.rproject.org	ver 3.3.0; SCR_001905
JMP Pro	www.jmp.com	ver 12

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents may be directed to and will be fulfilled by the Lead Contact, Yanhai Yin (yin@iastate.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Plant Materials and Growth Conditions

Arabidopsis thaliana accession Columbia (Col-0) was used along with previously described mutants: atg5-1 (Thompson et al., 2005), atg7-2 (Chung et al., 2010), bri1-301 (Li and Nam, 2002), bes1-D (Vilarrasa-Blasi et al., 2015; Yin et al., 2002), bin2-1D (Li et al., 2001), bin2-3 bil1 bil2 (Yan et al., 2009). Well characterized DSK2 RNAi lines (Lin et al., 2011) were used in this study. The majority of experiments in this manuscript were conducted with DSK2 RNAi 2-6 (referred to as DSK2 RNAi) and another independent line (DSK2 RNAi 1-4) was used to verify the phenotype. DSK2 RNAi BES1 RNAi plants were generated by crossing DSK2 RNAi with BES1 RNAi and F1 progeny used to examine BES1 and DSK2 protein levels as well as drought phenotypes. Plants were grown on 0.5X Linsmaier and Skoog (LS; Caisson Laboratories) plates or in soil under long day (16 h light/8 h dark) conditions at 22°C unless otherwise specified.

METHOD DETAILS

Inhibitor Treatments

Treatment with proteasome or autophagy inhibitors was performed by soaking plants in 0.5X LS liquid medium containing DMSO, 50 μM MG132, 20 μM E64d or 1 μM ConA. Plants were vacuum infiltrated for 5 minutes and then kept under light for 6 hours before sample collection. Similarly, cycloheximide treatments were performed by soaking plants in 0.5X LS medium containing 500 μM cycloheximide for the indicated time points. For BES1 ubiquitination, BES1-YFP was expressed in N. benthamiana. 24 hours post-inoculation, infiltration medium (10 mM MgCl₂, 10 mM MES, pH 5.7) containing DMSO, 50 μM MG132, or 20 μM E64d was in-filtrated into the lower side of the leaves. Samples were collected 16 hours after addition of inhibitors.

BRZ Response Assays

For BRZ response assays under normal (non-stressed) conditions, sterilized seeds were planted directly on 0.5X LS medium with 1% sucrose containing DMSO or indicated concentrations of BRZ (Asami et al., 2000). Plates were exposed to light for 6–8 hours and then kept in darkness for 7 days. Seedlings were then imaged and hypocotyls quantified using ImageJ software (http://imagej.nih.gov/ij/). For BRZ response under stress conditions, seedlings were grown under light for 5 days on 0.5X LS medium and then transferred to medium containing DMSO or 1 μM BRZ in combination with control (+ suc; containing 1% sucrose), starvation (without sucrose) or mannitol (350 mM mannitol) stresses. The seedlings were incubated in darkness for 3 days following transfer and then hypocotyl length was quantified as described above.

Drought and Dehydration Assays

Drought survival assays were performed by withholding water for 2–3 weeks to impose drought stress followed by rewatering. 7 days after rewatering plants were scored for survival as judged by the presence of new green leaves. Equal water and soil amounts were assured in drought assays by weighing the amount of dry soil for each pot and watering with equal amounts of water. Pots were randomized in trays to control for varying water loss due to position. Dehydration treatments were performed as previously described (Qin et al., 2008). Briefly, whole rosettes of 4-week-old plants were removed from pots and placed in empty petri dishes (dehydration) or in petri dishes containing moistened kimwipes (mock control) and sealed with parafilm for 4 hours. Detached leaf water loss assays were performed with 4–5 week old plants grown under short day conditions as previously described (Bao et al., 2014).

Fixed-Carbon Starvation Assays

Fixed-carbon starvation survival assays in seedlings were performed as described by (Chung et al., 2010). Seedlings were grown on 0.5X LS plates without sucrose in light for one week and then transferred to darkness for 8–12 days as indicated. After dark treatment, the seedlings were placed back in light conditions for one week. Plants with new growth were counted as surviving. For fixed-carbon starvation treatments of plants in soil, plants were grown for 4–5 weeks under short day conditions and then transferred to a dark chamber at 22°C. After 8 or 9 days dark incubation, plants were transferred back into light for one week recovery before being scored for survival.

Immunoprecipitation

The majority of immunoprecipitation experiments were carried out as follows: plant leaf tissue was ground to a fine powder and re-suspended in 2–3 volumes of protein extraction buffer. Extracts were then filtered through Miracloth (Calbiochem), and insoluble debris removed by centrifugation for 10 minutes at 5000 g, 4°C. The protein extracts were incubated for 2–4 hours with GFP-trap agarose beads (Chromotek) or with anti-GFP antibody followed by Protien-A agarose beads (Pierce). Beads were then washed 3–5 times with protein extraction buffer containing 0.1%–0.5% Triton X-100 and eluted by boiling in 2X SDS sample buffer. CIP treatments were performed at 37 degrees for one hour using Alkaline phosphatase (Roche) as described (Yin et al., 2002). Cerulean-ATG8 was Immunoprecipitated from the membrane fraction by grinding 5 g fresh N. benthamiana tissue (48 hours post-innoculation) in 20 mL cold extraction buffer (0.3 M Sucrose, 0.1 M Tris-HCl, pH7.5, 1 mM EDTA, 1 mM PMSF). The extract was then passed through Miracloth (Calbiochem) and centrifuged for 5 minutes at 1,000 g, 4°C to remove insoluble material. The supernatant was then centrifuged at 100,000 g for 30 minutes, 4°C. The membrane pellet was resuspended in Phosphate buffered saline (PBS) supplemented with 1% Triton X-100, 10 mM Beta-mercaptoethanol and Roche complete mini protease inhibitor cocktail by gentle rocking for 2 hours at 4°C. The extract was centrifuged again at 100,000 g for 30 minutes, 4°C and the supernatant was incubated with 20 μL GFP-Trap agarose beads (Chromotek) overnight. Beads were then washed 5 times with PBS containing 0.1% Triton X-100 and eluted by boiling in 2X SDS sample buffer.

TUBE Ubiquitination Assays

To detect BES1 ubiquitination, Tandem Ubiquitin Binding Entities (TUBEs; LifeSensors) were used to enrich total ubiquitinated proteins. Harvested tissue was ground to a fine powder in liquid nitrogen and extracted in protein extraction buffer (50 mM Tris-Cl, pH 7.4, 150 mM NaCl, 10% glycerol) supplemented with complete mini protease inhibitor cocktail (Roche) and deubiquitinase inhibitor (50 μM PR-619; LifeSensors). Extract was clarified by two rounds of centrifugation at 13,000 g for 10 minutes at 4LC and the supernatant was incubated with blocked agarose (control) or TUBE2 agarose beads for 3 hours with gentle rocking at 4°C. Beads were then washed 4 times with protein extraction buffer and eluted by boiling in 2X SDS sample buffer.

Western Blotting

Protein was extracted in protein extraction buffer (50 mM Tris-Cl, pH 7.4, 150 mM NaCl, 10% glycerol) supplemented with complete mini protease inhibitor cocktail (Roche) followed by quantification using Bradford assay or by fresh weight through directly adding 2–3 volumes of 2X SDS sample buffer to ground plant tissue. Western blotting was performed using standard laboratory techniques. Phos-tag SDS-PAGE was conducted by adding 1 mM MnCl₂ to protein samples and running SDS-PAGE gels containing 100 μM Phos-tag reagent (Wako) according to the manufacturer’s instructions. The following antibodies were used in this study in conjunction with appropriate secondary antibody-HRP conjugates: anti-BES1 (Yu et al., 2011), anti-DSK2 (Lin et al., 2011), anti-HERK1 (Guo et al., 2009), anti-GFP, anti-Tubulin (Sigma), anti-Ubiquitin (Pratelli et al., 2012), anti-MYC (Sigma or Cell Signaling Technology), anti-MBP (NEB).

RNA-Seq

RNA-seq experiments were performed using 4-week old plants under control, 4 hour dehydration, or 5 days dark treatment for starvation. 3 biological replicates were performed for control and dehydration conditions and 2 biological replicates for starvation. For each replicate, whole rosette tissue from 3–4 plants was pooled. RNA was then extracted using Trizol, followed by DNAse digestion and RNA cleanup using Qiagen RNeasy kit. Purified RNA was subject to quality control on an Agilent 2100 Bioanalyzer. Library preparation and RNA sequencing were performed by BGI Americas using an Illumina HiSeq 2000 with 50bp single-end reads and ~30 million reads per sample. Details of data processing and statistical analysis of RNA-seq data are provided in the Quantification and Statistical Analysis section.

Yeast-Two Hybrid

Yeast-two hybrid screening for BES1 interacting proteins using the Clontech matchmaker system was described previously (Yin et al., 2005). DSK2A was recovered from a screen using BES1-N domain as bait (amino acid residues 1-99) and SINAT2 was recovered using BES1-P domain (amino acid residues 100-150). For pairwise yeast-two hybrid assays, full length BES1, ATG8e, BIN2 and DSK2 were cloned into GBKT7 or GADT7 vectors and transformed into yeast strain Y187. Yeast cells were grown in medium lacking Trp and Leu and assayed for LacZ activity as described in the Yeast Protocols Handbook (Clontech) using X-gal (5- bromo-4-chloro-3-indolyl-b-D-galactopyranoside).

In Vitro Pull-down Assays

For GST-pull-down assays, GST or MBP fusions were generated by cloning into pET42a, pET-MALc-H or pET-MBP-H vectors. DSK2A-HIS fusion protein constructs were generated by cloning DSK2A into pET28a. Recombinant proteins were produced in E. coli strain BL21, and tested for interaction as described previously (Yin et al., 2002). GST or each GST tagged fusion were incubated with indicated MBP fusion proteins in 1 mL GST-pull-down buffer (50 mM Tris-HCl pH 7.5, 200 mM NaCl, 0.5% Trition X-100 and 0.5 mM Beta-mercaptoethanol, Roche complete mini protease inhibitor cocktail) at room temperature for 2 hours on a tube rotator. Following incubation, 20uL GST beads pre-blocked overnight with 1 mg/mL BSA and BL21 extract were added and the incubation was continued for an additional 30 minutes. GST beads were washed in GST-pull-down buffer 5–6 times and then eluted in 2X SDS sample buffer. For pull-down reactions exhibiting high background, an alternative pull-down buffer was substituted (25 mM HEPES-KOH [pH 8.0], 1 mM DTT, 50 mM KCL, 10% glycerol and 1% NP-40). GST-pull-down experiments were repeated 2–3 times with similar results. For pull-down of DSK2A-HIS, phosphorylated BES1-MBP protein was generated by incubating BES1-MBP with BIN2-GST in kinase buffer containing 10 mM ATP (rotated 5 hours at 37 degrees). Non-phosphorylated BES1 control was produced with an identical reaction lacking ATP. Pull-down assays were carried out as described above, except that blocked amylose resin (NEB) was used to capture the MBP proteins. DSK2A-HIS was detected with anti-DSK2 antibodies.

BiFC Assays

BiFC assays were conducted using constructs for the N- or C-terminus of YFP (Yu et al., 2008). BES1, DSK2, ATG8e, BIN2, and SINAT2 were cloned upstream of YFP fragments and transformed into Agrobacterium tumefaciens (strain GV3101). Agrobacterium cultures were grown overnight in LB medium containing 200 mM acetosyringone, washed with infiltration medium (10 mM MgCl₂, 10 mM MES, pH 5.7, 200 mM acetosyringone) and resuspended to an OD600 of 1.0. Agrobacterium carrying NYFP and CYFP constructs were mixed in equal ratios, and the Agrobacterium mixtures were infiltrated into the lower surface of N. benthamiana leaves. After 36–48 hours, YFP signal was detected using a Leica SP5 X MP confocal microscope equipped with an HCS PL APO CS 20.0×0.70 oil objective. YFP was excited with a 514-nm laser line and detected from 530 to 560 nm. Images were acquired with LAS AF software (Leica Microsystems) using identical settings for samples and controls. For BiFC colocalization studies, Agrobacterium carrying Cerulean-ATG8e was mixed along with NYFP and CYFP constructs at equal ratios and infiltrated into N. benthamiana as described above. Confocal microscopy was used to image YFP and CFP sequentially to reduce cross-excitation. YFP was excited with a 514 nm laser line and detected from 530–560 nm. CFP was excited at 405 nm and detected from 460–490 nm. BiFC experiments were repeated 2–3 times.

Coexpression of DSK2 and BIN2 in Tobacco

DSK2-MYC or DSK2 A-MYC (containing putative BIN2 phosphorylation sites mutated to alanine) was co-expressed with BIN2-D-YFP (bin2-1) (Li et al., 2001) or empty YFP vector in N. benthamiana as described for BiFC assays. Leaf tissue was collected for protein extraction 2 days post-infiltration.

Plasmid Constructs and Generation of Transgenic Plants

Plasmid constructs were generated using standard laboratory techniques via restriction enzyme digestion or Gateway technology (Invitrogen) and were confirmed by DNA sequencing. pET-MBP-H vector was generated by modifying pETMALc-H (Pryor and Leit-ing, 1997) via digestion with SacI and XhoI and replacement of the multiple cloning site with a redesigned sequence (GAGCTCCNGC GAATTCACGGGATCCCTGGGTACCCGCAAGCTTCGAGTCGACTACCTCGAG). gBlocks gene synthesis (IDT) was used to generate DSK2 mutants. An overview of mutated DSK2 sites and oligonucleotides used in this study is provided (Tables S4 and S5). DSK2-MYC constructs were generated by fusing DSK2 or mutated DSK2 variants with a C-terminal 2X MYC tag driven by the strong constitutive BRI1 promoter (Li et al., 2009; Li and Chory, 1997). Plasmid constructs were transferred into Agrobacterium tumefaciens (strain GV3101) and used to transform plants by the floral-dip method (Clough and Bent, 1998). Transgenic plants were screened on 0.5X LS plates supplemented with 50 mg/L kanamycin and further confirmed via western blotting using anti-MYC antibodies.

Protoplast Transient Expression Assays

Protoplasts were prepared from Arabidopsis rosette leaves grown under short day conditions and transformed using 30–50 μg of plasmid DNA via the PEG method (Wu et al., 2009; Yoo et al., 2007). Transformed protoplasts were incubated in darkness for 36–48 hours with control (+sucrose; .5% sucrose added), starvation (–sucrose; without sucrose) or mannitol (350 mM mannitol) treatments and concanamycin A (Sigma) or dimethyl sulfoxide (DMSO) were added 12 hours before visualization by confocal microscopy. Confocal microscopy was performed with a Leica (Leica Microsystems) SP5 X MP confocal microscope equipped with a resonance scanner and HPX PL APO CS 63.0×1.40 oil objective. For colocalization assays, YFP and CFP were imaged sequentially to avoid cross-detection between channels. Excitation and detection wavelengths were the same as those described for BiFC colocalization assays.

Confocal Microscopy of Arabidopsis Roots

For imaging of BES1-GFP signal, homozygous BES1P:BES1-GFP lines were grown for 5 days in light and then transferred to –sucrose plates for 2-day starvation treatments. DMSO or 1 μM conA was applied by transferring the plants to liquid 0.5X LS medium 16 hours prior to imaging. Confocal microscopy was performed with a Leica (Leica Microsystems) SP5 X MP confocal microscope equipped with a HCX PL APO CS 40.0×1.25 oil objective. GFP was excited with a 489 nm laser line and detected from 500–580 nm.

Phosphorylation Assays

For in vitro kinase assays, MBP or DSK2-MBP proteins were mixed with BIN2-HIS in 20 μL kinase buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 12 mM MgCl2 and 10μCi ³²P-γATP) as previously described (Yin et al., 2002). For Mass spectrometry analysis of phosphorylated proteins, DSK2-MBP was phosphorylated using BIN2-MBP in kinase buffer containing 20 mM ATP.

Protein Digestion and LC-MS/MS

Proteins were reduced with 5 mM TCEP in ammonium biocarbonate for 5 min at 94 °C. Proteins were then digested using either Glu-C (ThermoFisher) or trypsin (Roche) at 37 °C overnight and then alkylated with 12.5 mM iodacetamide for 15 min at 37 °C in the dark. Peptides were further digested using an additional aliquot of Glu-C or trypsin for 2 hrs. Samples were then acidified to a pH of ~3 with formic acid. Digested peptides were purified using Waters Oasis MCX cartridges and eluted using 45%IPA/500 mM NH₄HCO₃. Eluted peptides were dried using a speedvac (Thermo) and resuspended in 0.1% formic acid. Peptide amount was then quantified using the Pierce BCA Protein assay kit.

An Agilent 1260 quaternary HPLC was used to deliver a flow rate of ~600 nL min⁻¹ to a 3-phase capillary chromatography column through a splitter. The 3-phase capillary chromatography was assembled as follows. Using a Next Advance pressure cell a fused silica capillary column was packed with 5 mM Zorbax SB-C18 (Agilent) to form the first dimension reverse phase column (RP1). A 5 cm long strong cation exchange (SCX) column packed with 5 μm PolySulfoethyl (PolyLC) was connected to RP1 using a zero dead volume 1 μm filter (Upchurch, M548) attached to the exit of the RP1 column. A nanospray fused silica capillary was pulled to a sharp tip using a laser puller (Sutter P-2000) and packed with 2.5 μM C18 (Waters) to form RP2 and then connected to the SCX column. The 3 sections were joined and mounted on a custom electrospray source for on-line nested elutions. A new set of columns was used for every sample. Peptides were loaded onto RP1 using the Next Advance pressure cell. Peptides were eluted from RP1 unto the SCX column using a 0 to 80% acetonitrile gradient over 60 min. Peptides were then fractionated using the SCX column using a series of 9 salt gradients (0, 30, 50, 60, 70, 80, 90, 100 and 1000 mM ammonium acetate), followed by high resolution reverse phase separation using an acetonitrile gradient of 0–80% for 150 min.

Spectra were acquired on a Thermo Scientific Q-Exactive high-resolution quadrupole Orbitrap mass spectrometer. Data dependent acquisition was obtained using Xcalibur 3.0.63 software in positive ion mode with a spray voltage of 2.00 kV and a capillary temperature of 275 °C. MS1 spectra were measured at a resolution of 70,000, an automatic gain control (AGC) of 3e6 with a maximum ion time of 100 ms and a mass range of 400–2000 m/z. Up to 15 MS2 were triggered at a resolution of 17,500, an AGC of 1e5 with a maximum ion time of 50 ms and a normalized collision energy of 28. MS1 that triggered MS2 scans were dynamically excluded for 15 s.

The raw data were extracted and searched using Spectrum Mill v4.01 (Agilent Technologies). MS/MS spectra with a sequence tag length of 1 or less were considered to be poor spectra and were discarded. The remaining MS/MS spectra were searched against the Arabidopsis TAIR10 proteome. The enzyme parameter was limited to tryptic peptides with a maximum mis-cleavage of 2. Carbami-domethylation was set as a fixed modification while Ox-Met, and phosphorylation on Serine, Threonine, or Tyrosine were defined as variable modifications. A maximum of 6 phosphorylation events per peptide was used. A 1:1 concatenated forward-reverse database was constructed to calculate the false discovery rate (FDR). The tryptic peptides in the reverse database were compared to the forward database, and were shuffled if they matched to any tryptic peptides from the forward database. Cutoff scores were dynamically assigned to each dataset to obtain a peptide false discovery rate (FDR) of 0.1%. Phosphorylation sites were localized to a particular amino acid within a phosphopeptide using the variable modification localization (VML) score in Agilent’s Spectrum Mill software (Chalkley and Clauser, 2012).

QUANTIFICATION AND STATISTICAL ANALYSIS

RNA-seq Data Processing and Statistics

Raw reads were subject to quality control and trimming and aligned to the Arabidopsis TAIR10 reference genome using GSNAP (Wu and Nacu, 2010). Uniquely aligned reads were used to obtain read counts per gene. Only genes with an average read count greater than 1 across all samples were used for differential expression analysis. Normalization was conducted automatically by DESeq2, which corrects for biases introduced by differences in the total number of uniquely mapped reads between samples. Normalized read counts were used to calculate fold changes and test for differential expression. The R package DEseq2 (https://bioconductor.org/packages/release/bioc/html/DESeq2.html) was used to test the null hypothesis that expression of a given gene is not different between two genotypes or conditions being compared. This null hypothesis was tested using a model with a negative binomial distribution. P-values of all statistical tests were converted to adjusted p-values (q-values) (Benjamini and Hochberg, 1995). A false discovery rate of 10% (q-value) was used to account for multiple testing. Comparisons of gene lists were performed using Venny (http://bioinfogp.cnb.csic.es/tools/venny/index.html) and overrepresentation in overlapping gene lists tested using the Genesect tool in VirtualPlant (Katari et al., 2010). Reported overlaps were statistically significant at a p< 0.05 level. Clustering was performed using the ‘aheatmap’ function of the NMF package in R (https://cran.r-project.org/web/packages/NMF/index.html). Log2 reads per million mapped reads (RPM) values were used for clustering analysis and values were normalized for each gene by centering and scaling each row of the heatmap. GO analysis was performed using BINGO (Maere et al., 2005).

DATA AND SOFTWARE AVAILABILITY

RNA-seq data have been deposited into the Gene Expression Omnibus (GEO: GSE93420). Proteomics data relating to DSK2 phosphorylation by BIN2 have been deposited in the MassIVE repository (MassIVE: MSV000079641).

Highlights.

• BES1 is targeted to selective autophagy by ubiquitin receptor DSK2

• BIN2 phosphorylates DSK2, enhancing DSK2’s interaction with ATG8

• BES1 is ubiquitinated by SINAT2 during starvation stress

• Impaired BES1 degradation results in compromised drought and starvation survival