CDK8 phosphorylation of DELLA limits Mediator recruitment in gibberellin signaling

Abstract

Gibberellin (GA) promotes plant growth primarily by triggering degradation of DELLA transcription regulators, yet how DELLA activity is fine-tuned dynamically by phosphorylation independently of proteolysis remains poorly understood. Here we show that the CDK8 kinase module of the Mediator complex attenuates activity of the Arabidopsis DELLA protein REPRESSOR OF ga1-3 (RGA) by modulating coactivator recruitment. Using TurboID-based proximity labeling, biochemical and genetic analyses, we identify CDK8 as an in-planta kinase that phosphorylates RGA at Ser170 within its disordered PolyS/T region. This phosphorylation does not affect RGA stability, localization or interactions with transcription factors or histone H2A, but selectively weakens RGA association with the Mediator subunit MED15, thereby reducing DELLA-dependent transcription activation. Consistently, cdk8 mutants show impaired GA-responses and delayed developmental phase transitions that are partially rescued by loss of DELLA function. Our findings uncover a phosphorylation-dependent mechanism by which the Mediator kinase module fine-tunes hormone-responsive transcription through selective control of DELLA–coactivator interactions.

INTRODUCTION

The phytohormone gibberellin (GA) is a central regulator of plant development, controlling seed germination, vegetative growth, floral transition, and reproductive development (1). GA perception is mediated by the receptor GIBBERELLIN INSENSITIVE1 (GID1), which triggers the ubiquitin–proteasome-dependent degradation of DELLA proteins, the principal repressors of GA signaling (2, 3). Dominant mutations in DELLA genes underlie the semi-dwarf, high-yielding wheat varieties that fueled the Green Revolution of the 1960s, highlighting the key role of DELLA proteins in regulating plant growth (4, 5). First identified in Arabidopsis thaliana as repressor of GA signaling (6, 7), DELLAs are evolutionarily conserved across land plants (8) and function as integrative hubs that coordinate hormonal and environmental signals to balance growth, development, and stress responses (5, 9).

DELLA proteins are plant-specific GRAS family of transcription regulators, characterized by a conserved C-terminal GRAS domain (6, 7, 10) and a unique N-terminal DELLA domain required for GA-dependent proteolysis (11, 12). Upon GA binding, GID1 forms a ternary GA–GID1–DELLA complex that recruits the SCF^SLY1/GID2 E3 ubiquitin ligase, leading to DELLA polyubiquitination and rapid degradation, and thereby enabling transcriptional reprogramming of GA-responsive genes (2, 13–19). While this proteolytic pathway is a major determinant of DELLA abundance, accumulating evidence indicates that DELLA activity can also be modulated independently of protein degradation (5, 18).

At the molecular level, DELLA proteins regulate gene expression primarily through protein–protein interactions rather than direct DNA binding. Through their GRAS domain, they associate with a wide array of transcription factors (TFs) and chromatin remodelers to reprogram transcriptional outputs (9, 20–23). Recent work has shown that DELLA function requires the formation of TF–DELLA–histone H2A complexes at target loci, enabling both transcriptional repression and activation depending on the interacting partners (24). Notably, DELLA-mediated transcriptional activation depends on recruitment of the Mediator coactivator complex via interaction with the tail subunit MED15, a mechanism conserved across land plants (25). How DELLA interactions with Mediator and other cofactors are dynamically regulated remains largely unknown.

Structurally, the GRAS domain comprises five conserved subdomains—Leucine Heptad Repeat 1 (LHR1), VHIID, Leucine Heptad Repeat 2 (LHR2), PFYRE, and SAW—three of which are named after conserved sequence motifs (10). Functional and structural studies have defined distinct roles to these subdomains: LHR1 mediates interactions with TFs to recruit DELLA proteins to target promoters, whereas the PFYRE subdomain stabilizes chromatin-associated complexes through interaction with histone H2A (9, 24). In contrast, the VHIID, LHR2 and SAW subdomains interact with F-box proteins, linking these regions to DELLA turnover via the ubiquitin–proteasome pathway (18, 26–28). Together, these findings highlight the modular architecture of DELLA proteins and suggest multiple layers of regulatory control.

In addition to GA-induced degradation via polyubiquitination, DELLA activity is regulated by multiple post-translational modifications (PTMs), including Small Ubiquitin-Like Modifier-conjugation (SUMOylation), glycosylation, and phosphorylation (29–32). Among these, phosphorylation has been proposed as an important regulatory layer, although its functional role remains unclear (5, 29). Early studies suggested that phosphorylation of the rice DELLA SLR1 promotes interaction with the F-box protein GID2 and thereby facilitates proteasomal degradation (14, 33, 34), but later work found no effect of phosphorylation on GID2 binding, arguing against a direct role in DELLA turnover (35). Consistent with this view, pharmacological inhibition of phosphatases or genetic suppression of the protein phosphatase TOPP4 attenuated GA-induced DELLA degradation, suggesting that dephosphorylation rather than phosphorylation may favor proteolysis (36–39). Several kinases, including casein kinase I EARLIER FLOWERING1 (EL1) and a GSK3/SHAGGY-like kinase, have been reported to phosphorylate DELLA proteins and enhance their stability, although these activities have only been demonstrated in vitro (40, 41). More recently, mass spectrometry analyses of Arabidopsis DELLA, RGA (for REPRESSOR OF ga1-3), have identified multiple phosphorylation sites within the intrinsically disordered PolyS and PolyS/T regions flanking the DELLA domain, and mutational studies showed that phosphorylation in these regions modulates DELLA activity without affecting protein stability by promoting association with histone H2A at target loci (42). However, the identity of the protein kinase(s) responsible for DELLA phosphorylation in planta, and the mechanistic consequences of specific phosphorylation events, remain largely unknown (5).

Here, using TurboID-based proximity labeling (43) together with biochemical and genetic approaches, we identify the CDK8 kinase module (CKM) of the Mediator complex as an in-planta regulator of DELLA phosphorylation and uncover a broad set of DELLA-associated transcription regulators. We show that CDK8 phosphorylates the Arabidopsis DELLA protein RGA at Ser170 within its intrinsically disordered PolyS/T region, attenuating RGA transactivation activity by weakening its interaction with MED15. These findings uncover a mechanism by which CKM fine-tunes hormone-responsive transcription through phosphorylation-dependent control of DELLA–coactivator interactions.

RESULTS

Proximity labeling reveals the RGA proximitome

To identify kinase(s) that associate with the Arabidopsis DELLA protein RGA in planta, we performed TurboID (TbID) proximity labeling (43–45) to capture target proteins such as protein kinases that may only transiently interact with RGA in planta. We generated transgenic Arabidopsis expressing P_RGA:His-FLAG-RGA-TbID in ga1-13 della pentuple (ga1 dP) background (RGA-TbID line). The ga1-13 null allele causes a GA-deficient dwarf phenotype, whereas ga1-13 dP plants are tall due to the removal of all five DELLA genes. The RGA-TbID fusion protein in ga1 dP restored the ga1 dwarf phenotype and underwent GA-induced degradation similar to FLAG-RGA, indicating normal functionality (SI Appendix, Fig. S1A–1D). As a control for nonspecific labeling, we generated a P_RGA:His-FLAG-NLS-GFP-TbID line in the ga1 dQ (quadruple della with WT RGA) background (GFP-TbID line) with comparable expression levels as RGA-TbID (SI Appendix, Fig. S1A–1D). Seedlings were treated with 100 μM biotin for 3 h (Fig. S1E), and biotinylated proteins were purified with streptavidin beads and identified by liquid chromatography-tandem mass spectrometry (LC-MS/MS) with label-free quantification (LFQ). Proteins significantly enriched in RGA-TbID relative to GFP-TbID were identified from three biological replicates using statistical thresholds of FDR < 0.01 and S0 = 0.8 (Fig. 1A and Dataset S1). Among the 199 RGA-proximal proteins, Gene Ontology (GO) analysis revealed strong enrichment for nuclear factors involved in transcriptional regulation, including transcription factors and regulators, Mediator complex components, chromatin remodelers, and histone-modifying enzymes (Fig. 1B, Datasets S1–S2). The specificity of the labeling is validated by the detection of many known RGA-interacting transcription factors/regulators (3, 22), including BZR1, IDDs, EIN3, TCPs, ARFs, ARR1, bHLHs and JAZ9, as well as the Mediator tail subunit MED15 (25) and the chromatin remodelers BRAHMA (BRM) (46) and SWI3C (47) (Fig. 1C and Dataset S1).

Figure 1. RGA proximitome identified by TbID-labeling.

(A) Volcano plot of RGA proximal proteins, identified by label-free quantification (LFQ) MS analysis of biotinylated proteins purified from RGA-TbID and GFP-TbID samples. Proteins significantly enriched in RGA-TbID are labeled in red, while those enriched in GFP-TbID are in black. (B) Enriched molecular functions (MFs) of RGA proximal proteins by GO term analysis. (C) STRING network analysis (81) of RGA proximal proteins involved in transcription regulation. Known RGA interactors (22) are labeled in blue and new RGA proximal proteins are in yellow. (D-F) Co-IP assays showing direct interactions between RGA and three proximal proteins CDK8, SEU and LUG. FLAG-RGA (FR) was expressed alone or together with Myc-CDK8, Myc-SEU, or -YFP-LUG, separately, in N. benthamiana. Myc-GUS or YFP-GUS was included as a negative control. IP was performed using anti-Myc agarose (D-E) or anti-GFP agarose (F). Immunoblots containing input and IP eluates were probed with anti-FLAG, anti-Myc or anti-GFP antibodies. Representative images of 2 biological repeats are shown.

Notably, the catalytic subunit of the Mediator kinase module, CDK8, was strongly enriched together with the module components MED12 and MED13 (48, 49), identifying the CDK8 kinase module (CKM) as a candidate regulator of RGA (Fig. 1C and Dataset S1). Additional Mediator subunits, MED14 and MED16, were also identified along with MED15. The RGA proximitome further include transcriptional cofactors such as SEUSS (SEU) and SEU-LIKEs (SLKs) (50, 51) and Groucho/Tup1-type corepressors LEUNIG (LUG)/LUG HOMOLOG (LUH) and TOPLESS (TPL)/TPL-RELATEDs (TPRs) (52), and components of histone-acetyltransferase complexes, including HAC12 in the CBF family, SAGA complex subunits (SCS1, TAF12B, and PHL/SPT20), and NuA4 complex components TRA1A and TRA1B (53, 54) (Fig. 1C and Dataset S1). To validate these candidates, CDK8, SEU, and LUG were tested by co-immunoprecipitation (IP) assays and each showed direct interaction with FLAG-RGA (FR) when co-expressed in Nicotiana benthamiana (Fig. 1D–1F).

Together, these results indicate that RGA associates with transcriptional regulatory machinery comprising transcription factors, Mediator components, transcription cofactors, chromatin remodelers, and histone acetyltransferase complexes. Among these candidates, the CDK8 kinase module was of particular interest because it contains a catalytic kinase, suggesting a potential mechanism for DELLA phosphorylation.

Phosphorylation of RGA by CDK8 kinase module

To examine whether CDK8 phosphorylates RGA in planta, we compared phosphorylation levels of endogenous RGA in WT vs cdk8 mutants using Phos-tag SDS-PAGE because phospho-RGA (pRGA) could not be separated clearly from the unphosphorylated form by standard SDS-PAGE (42). Phos-tag gel analysis showed that pRGA levels were reduced in both cdk8-1 and cdk8-2 mutants compared with WT, and RGA phosphorylation was restored in the 35S:Myc-CDK8 cdk8-1 transgenic line (Fig. 2A). To facilitate monitoring of RGA phosphorylation, we generated transgenic Arabidopsis (Col-0 ecotype) carrying P_RGA:His-FLAG-RGA^GKG in ga1-13 or ga1-13 cdk8-1 backgrounds. His-FLAG-RGA^GKG was previously shown to be functional in planta (55); RGA^GKG contains an additional trypsin cleavage site created by inserting a Lys (K) residue within the PolyS/T region, enabling MS detection of this region (55). Similar to the endogenous RGA, FLAG-RGA^GKG showed reduced phosphorylation in the cdk8 mutant compared with WT (Fig. 2B).

Figure 2. RGA Phosphorylation by CDK8 kinase module.

(A) cdk8 mutations reduced phosphorylation of endogenous RGA, while 35S:Myc-CDK8 rescued this defect. pRGA, phosphorylated RGA. Protein blots contained total proteins extracted from seedlings that were grown in medium containing 1 μM paclobutrazol (PAC, a GA biosynthesis inhibitor). (B) Phosphorylation of His-FLAG-RGA^GKG (FR) was reduced by cdk8. pFR, phosphorylated His-FLAG-RGA. (C) Co-expression of both Myc-CDK8 and -CYCC increased His-FLAG-RGA (FR) phosphorylation in N. benthamiana. FR was expressed alone (−) or co-expressed with Myc-CDK8 or both Myc-CDK8 and -CYCC. (D) Co-IP assays showing CDK8-RGA interaction in Arabidopsis. IP was performed using anti-Myc conjugated beads and protein extracts from transgenic lines in the ga1 dP background carrying P_RGA:His-FLAG-RGA (FR) alone or together with P_Ubq10:Myc-CDK8. Immunoblots were detected with anti-Myc and anti-FLAG antibodies, separately. In (A-C), the ratios of phosphorylated RGA/unphosphorylated RGA are shown below the Phos-tag gel blots. (A-D) Representative images of two biological repeats are shown.

CKM consists of CDK8, Cyclin C (CYCC), MED12 and MED13 (48, 49). We found that co-expression of both CDK8 and CYCC with FLAG-RGA (FR) in N. benthamiana dramatically enhanced pFR levels, whereas co-expression of CDK8 alone did not (Fig. 2C). Addition of the remaining two CKM subunits, MED12 and MED13, in the co-expression assay did not further increase pFR levels (SI Appendix, Fig. S2), indicating that these two CKM subunits are not limiting in this heterologous system. As described above, Myc-CDK8 interacted with FLAG-RGA when they were co-expressed in N. benthamiana (Fig. 1D). Co-IP assays further show that the VHIID subdomain in RGA is required for CDK8 interaction (SI Appendix, Fig. S3). To detect their interaction in Arabidopsis, co-IP assays were performed using transgenic Arabidopsis in the ga1 dP background carrying P_RGA:His-FLAG-RGA alone or together with P_Ubq10:Myc-CDK8 (Fig. 2D). His-FLAG-RGA was co-immunoprecipitated with Myc-CDK8. Taken together, these results support that RGA is phosphorylated by CDK8 of the CKM. Using the transient expression system in N. benthamiana, we further showed that CDK8 also enhanced phosphorylation of three other Arabidopsis DELLA proteins, GAI, RGL1 and RGL2, although only weakly for RGL2 (Fig. S4).

CDK8 promotes GA signaling by inhibiting RGA function

The role of CDK8 in regulating RGA activity and GA response in planta was analyzed by hypocotyl elongation assay. Importantly, the cdk8 mutants displayed reduced GA responses compared with WT (Fig. 3A–3B), suggesting that CDK8 promotes GA signaling by inhibiting RGA function. CDK8, but not the kinase-dead mutant (CDK8^D176A) (56), rescued the cdk8 hypocotyl response to GA (Fig. 3A–3B, SI Appendix, Fig. S5A–S5C), indicating that CDK8 kinase activity is essential for CDK8-induced GA response.

Figure 3. CDK8 reduced RGA function to promote GA signaling.

(A-B) The cdk8 mutants displayed a reduced GA response. Seedlings were grown in medium containing 1 μM PAC and varying concentrations of GA₄ for 9d. (A) Bar = 3 mm. (B) Average hypocotyl lengths. Means ± SE (n =11). (C-E) The cdk8 mutants showed delayed phase transitions in leaf development and flowering time, and rga-29 gai-t6 partially rescued these phenotypes. (C) Photo showing representative 31d-old plants grown under LD. Bar = 3 cm. (D) Average rosette leaf numbers after bolting. Juvenile leaves are rosette leaves that lack abaxial trichomes. The values plotted are the means ± SE (n =20-22). (E) Days to flower are shown in the boxplot (n=20-22). Center lines and box edges are medians and the lower/upper quartiles, respectively. Whiskers extend to the lowest and highest data points within 1.5× interquartile range (IQR) below and above the lower and upper quartiles, respectively. (D-E) Different letters represent significant differences: p < 0.01 (D) or p < 0.05 (E), two-tailed Tukey’s HSD test. Experiments were repeated two times with similar results.

At the adult stage, both cdk8 mutants exhibited delayed vegetative and reproductive phase transitions compared with WT (Fig. 3C–3E). To examine the genetic interaction between CDK8 and two DELLAs (RGA and GAI), the cdk8-1 rga-29 gai-t6 triple mutant was generated, in which the rga-29 and gai-t6 null alleles are caused by T-DNA and transposon Ds insertions in their respective coding regions (7, 57). We found that rga-29 gai-t6 partially rescued hypocotyl growth, juvenile-to-adult leaf transition and flowering time of cdk8-1 (Fig. 3C–3E, SI Appendix, Fig. S5D–S5E), consistent with the notion that CDK8 promotes GA signaling by reducing RGA and GAI activity.

CDK8-mediated phosphorylation of S170 reduced RGA activity

To identify CDK8-mediated phosphosite(s) in RGA, we performed liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS) using affinity-purified His-FLAG-RGA^GKG protein from transgenic Arabidopsis Col-0 ecotype plants carrying P_RGA:His-FLAG-RGA^GKG in ga1-13 and ga1-13 cdk8-1 backgrounds, separately. Relative peptide abundance was estimated by semi-quantitative analysis of ion currents detected in MS1 survey scans. All RGA phosphorylation sites detected were located within the PolyS and PolyS/T regions flanking the DELLA domain: LSN-peptide [Pep1, LSNHGTSSSSSSISK(DK), 5% phosphorylation] and LKS-peptide [Pep2, (LK)SCSSPDSMVTSTSTGTQIGK, 32.9%] (Fig. 4A, 4B, SI Appendix Fig. S6, Table S1). These phosphosites largely match those identified in our previous MS analysis of FLAG-RGA^GKG purified from ga1-3 in the Ler ecotype background (42). However, Pep1 phosphorylation was higher in Ler than in Col-0 ecotype (30.5% vs. 5.0%). Additionally, only S170 phosphorylation within Pep2 was detected in Col-0, whereas multiple phosphosites were identified in the Ler background.

Figure 4. Identification of CDK8-mediated phosphosite in RGA.

(A) RGA phosphorylation sites identified by MS/MS analysis. The schematic shows the RGA protein structure, with disordered regions depicted as black lines and the DELLA domain shaded pink. The phosphorylated LSN-peptide (Pep1) and LKS-peptide (Pep2) are in bold letters. Green S/T residues denote confirmed phosphorylation sites. Underlined residues in Pep1 indicate regions containing additional modifications that could not be mapped to specific residues. The underlined K in parenthesis indicates the additional Lys residue in RGA^GKG, which creates an extra trypsin cleavage site. Figure modified from Huang et al. (42) (B) MS1 extracted ion chromatograms (XICs) showing reduced Ser170 phosphorylation in RGA Pep2 in cdk8. Relative XICs of Pep2 from Arabidopsis ga1-13 (top) and ga1-13 cdk8-1 (bottom) are shown. Red “pS” denotes phosphorylated serine; lowercase “c” denotes alkylated cysteine. (C) Amino acid substitutions in rga^S170A, rga^S170D and rga^m2A are highlighted in red (S/T-to-A) or blue (S-to-D). (D) S170A abolished CDK8/CYCC-mediated RGA phosphorylation in N. benthamiana. (E) Co-IP assays showing that S170A did not affect Myc-CDK8 binding in N. benthamiana. The amount of FLAG-RGA pulled down by Myc-CDK8 was set as 1.0. –, not detectable. Myc-GUS served as a negative control. (F) Phosphorylation pattern of P_RGA:His-FLAG-RGA/-rga^S170A/rga^S170D ga1 dP transgenic lines. Protein blots were probed with anti-FLAG antibody or anti-tubulin (TUB; loading control). Representative results of 2 biological repeats are shown. In (D) and (F), the ratios of pFR/unphosphorylated FR are indicated below the Phos-tag gel blots.

Importantly, the cdk8-1 mutation reduced relative phosphorylation abundance of S170 within Pep2 by 2-fold (32.9% in ga1 vs 16.0% in ga1 cdk8) (Fig. 4B), while phosphorylation within Pep1 was unchanged (Table S1). To test whether S170 in RGA is the primary CDK8-dependent phosphosite, we expressed wild-type RGA or the mutant rga^S170A, either alone or together with CDK8 and CYCC, in N. benthamiana. S170A is expected to block phosphorylation at S170. Phos-tag gel analysis showed that the S170A mutation abolished CDK8-mediated RGA phosphorylation (Fig. 4C, 4D), although co-IP assays indicated that this mutation did not affect CDK8-RGA interaction (Fig. 4E). These results indicate that S170 is the major CDK8 phosphosite in RGA.

To investigate how S170 phosphorylation affects RGA function in planta, we generated transgenic Arabidopsis in the ga1 dP background carrying either P_RGA:His-FLAG-rga^S170A or - rga^S170D. While the S170A substitution abolishes phosphorylation at this site, S170D serves as a phosphomimetic substitution. Three independent homozygous lines for each construct, with expression levels comparable to the previously reported P_RGA:His-FLAG-RGA lines (24), were selected for phenotypic analysis (Fig. S7A). Phos-tag gel analysis showed that the S170A mutation reduced overall RGA phosphorylation to ~16% of the wild-type level (Fig. 4F). As expected, His-FLAG-RGA restored the dwarf phenotype in ga1 dP plants. Notably, the phosphomimetic S170D substitution slightly decreased RGA activity in repressing stem growth and floral induction (Fig. 5A, 5B and SI Appendix Fig. S7B–S7D), supporting the idea that CDK8-mediated phosphorylation of S170 attenuates RGA activity. In contrast, S170A did not further reduce growth compared with wild-type RGA, possibly because the control RGA lines already display an extremely dwarf phenotype. Consistent with these observations, hypocotyl elongation assays showed that S170D enhanced the GA response in hypocotyl growth (Fig. 5C and 5D). Together, these results indicate that CDK8 promotes GA signaling by down-regulating RGA activity through phosphorylation of S170.

Figure 5. Phosphomimetic S170D reduced RGA function in planta.

(A)-(B) Phenotypes of P_RGA:His-FLAG-RGA/rga transgenic lines in the ga1 dP background. FR, line #16; S170A, line #9; S170D, line #1. (A) Representative 93d-old plants grown under LD. Bar = 5 cm. (B) Boxplot showing plant heights (n=10). (C-D) His-FLAG-rga^S170D ga1 dP displayed enhanced GA responses. Seedlings were grown for 9d on medium containing varying concentrations of GA₄. Bar = 3 mm. (D) Average hypocotyl lengths are means ± SE (n =12). (E-F) RT-qPCR analyses showing cdk8 increased expression of RGA-induced genes (E) but not RGA-repressed genes (F). Expression was normalized to PP2A. Means ± SE of three biological replicates are shown. Different letters indicate significant differences (p < 0.01, two-tailed Student’s t-test) in (E); no significant differences were detected in (F) (p > 0.05). (G-H) Dual LUC assays in N. benthamiana. Upper panels show relative fLUC activity. C, empty effector as a negative control. Lower panels show effector proteins detected by anti-FLAG and anti-Myc antibodies. PS, Ponceau staining. (G) CDK8/CYCC reduced RGA-induced SCL3 transcription. (H) S170A enhanced, whereas S170D reduced, RGA activity. Boxplots in (B) and (G-H) are presented as described in the legend to Fig. 3E. Different letters indicate significant differences (p < 0.01, two-tailed Tukey’s HSD test). Assays in (A-D) and (G-H) were repeated twice with similar results.

Interestingly, this inhibitory effect of CDK8 on RGA activity contrasts with our prior finding that substituting eight S/T-to-A residues in RGA Pep2 (the m2A mutant, Fig. 4C) reduced RGA function, suggesting that phosphorylation within Pep2 can enhance RGA activity (42). We therefore propose that CDK8 and another, as yet unidentified, protein kinase may regulate RGA differently by targeting distinct phosphorylation sites. While CDK8-mediated phosphorylation of S170 reduces RGA activity, phosphorylation of other S/T residues within Pep2 may enhance its function.

CDK8 does not affect RGA stability or its interaction with PIF3, BZR1 or H2A

To understand how CDK8-mediated phosphorylation reduces RGA function, we examined whether cdk8 affects RGA subcellular localization or stability in response to GA. The cdk8 mutation did not alter GA-induced RGA degradation (Fig. S8A) or its nuclear localization (Fig. S8B). RT-qPCR analysis showed that cdk8 mutation increased expression of RGA-induced genes (Fig. 5E) but did not affect RGA-repressed genes (Fig. 5F), suggesting that CDK8 mainly suppresses the transactivation activity of RGA. To further test this idea, we performed dual luciferase (LUC) assays using the transient expression system in N. benthamiana. The P_SCL3:firefly LUC (fLUC) was used as the reporter because SCL3 is a direct DELLA target whose transcription is activated by DELLA (58, 59). 35S:Renilla LUC (rLUC) served as an internal control to normalize transformation efficiency. Co-expression of CDK8 and CYCC reduced RGA-induced P_SCL3:fLUC expression (Fig. 5G), indicating that CDK8 phosphorylation suppresses RGA activity. Consistently, the S170A mutation enhanced RGA activity, whereas the phosphomimetic S170D reduced it (Fig. 5H), supporting that CDK8-mediated phosphorylation of S170 attenuates RGA function.

Because formation of TF-RGA-H2A complexes at target chromatin is essential for RGA function (24), we next examined whether CDK8 affects RGA interactions with TFs or H2A. In vitro pulldown assays were performed using recombinant GST-PIF3 and GST-BZR1 together with protein extracts from transgenic Arabidopsis expressing FLAG-RGA^GKG in either the WT or cdk8-1 background. GST-PIF3 and GST-BZR1 pulled down similar amounts of His-FLAG-RGA from both backgrounds (Fig. S9A). Co-IP assay using FLAG-RGA^GKG transgenic lines also showed that the cdk8 mutation did not affect the RGA–H2A interaction in planta (Fig. S9B). Similarly, the S170A and S170D substitutions did not alter RGA binding to H2A. In contrast, the previously described m2A mutation (eight S/T-to-A substitutions in Pep2) markedly weakened the RGA–H2A interaction (Fig. S9C), consistent with our earlier findings (42).

CDK8 attenuates RGA transactivation by weakening RGA-MED15 interaction

The coactivator MED15 interacts with RGA to promote transcription of RGA-induced genes (25, 60). Because MED15 and several other Mediator subunits were identified in the RGA proximitome (Fig. 1C), we tested whether CDK8/CYCC affects RGA–MED15 interaction. Yeast three-hybrid (Y3H) assays were performed using Gal4 DNA-binding domain (DB)-MED15, Gal4 transactivation domain (AD)-RGA in the absence or presence of Myc-CDK8/Myc-CYCC (Fig. 6A and Fig. S10). Myc-CDK8^D176A served as a kinase-dead control. Interaction strength was assessed by growth on His⁻ media containing increasing concentrations of the competitive HIS3 inhibitor 3-aminotriazole (3-AT). Co-expression of CDK8 and CYCC weakened MED15-RGA interaction, as indicated by yeast growth only in the absence of 3-AT rather than at 10 mM. In contrast, CDK8^D176A had no effect, indicating that CDK8 kinase activity is required.

Figure 6. CKM weakened RGA-MED15 interaction in Y3H and pulldown assays.

(A) Y3H assay showing CDK8/CYCC reduced RGA-MED15 interaction. Full-length MED15 and RGA were fused to the Gal4 DB and AD domains, respectively. CDK8, CDK8^D176A and CYCC were expressed from pH3-Myc vector. Immunoblot analyses showed comparable RGA and CDK8 protein levels across samples (Fig. S10). Interaction strength was assessed by yeast growth on –His media containing increasing concentrations of 3-AT. – indicates empty vector, or no growth at 0 mM 3-AT. Max, growth at the highest 3-AT concentration tested. (B) In vitro pulldown assay showing stronger binding of FLAG-RGA^GKG from the cdk8-1 background to MBP-MED15^N than that from WT CDK8 background. MBP and MBP-MED15^N immobilized on amylose beads were used separately to pull down FR^GKG from extracts of P_RGA:His-FLAG-RGA^GKG transgenic Arabidopsis in ga1 or ga1 cdk8-1 background. Immunoblots were detected with an anti-FLAG antibody. Comparable amounts of MBP and MBP-MED15^N were confirmed by Ponceau S-staining (Fig. S13B). (A-B) Results were reproducible in two biological replicates.

A limitation of these Y3H assays was that full-length (FL) MED15 protein was not detectable by immunoblotting and was also unstable when expressed in N. benthamiana. To address this, we mapped RGA-interacting region of MED15 using co-IP and Y2H assays. Three MED15 fragments were detectable and capable of binding RGA, but the N-terminal region (amino acids 1–450, MED15^N), which contains the KIX domain, showed the strongest interaction (Fig. S11 and S12). The KIX domain is necessary and sufficient for RGA binding, although weaker than MED15^N (Fig. S12).

The inhibitory effect of CDK8 on MED15-RGA interaction was further verified by co-IP and in vitro pulldown assays using MED15^N. For the co-IP assay, FLAG-RGA was transiently expressed alone or together with YFP-MED15^N in N. benthamiana leaves, with or without Myc-CDK8 and Myc-CYCC. YFP-GUS and Myc-CDK8^D176A were included as negative controls. FLAG-RGA co-immunoprecipitated more strongly with YFP-MED15^N in the absence of CDK8/CYCC than when CDK8/CYCC were co-expressed (Fig. 7A and Fig. S13A). Consistently, CDK8^D176A did not weaken RGA-MED15^N interaction. In vitro pulldown assays using recombinant MBP-MED15^N and protein extracts from Arabidopsis carrying P_RGA:His-FLAG-RGA^GKG in either ga1-13 or ga1-13 cdk8-1 backgrounds further showed that the cdk8-1 mutation increased MED15^N–RGA binding (Fig. 6B and Fig. S13B).

Figure 7. CKM reduced RGA-MED15 interaction.

(A) Co-IP assays showed that CDK8/CYCC reduced RGA-MED15^N interaction, and that S170D and S170A substitutions reduced and enhanced MED15^N binding, respectively, in N. benthamiana. FLAG-RGA/rga was expressed alone or together with YFP-MED15^N in the presence or absence of Myc-CDK8/-CYCC. YFP-GUS and Myc-CDK8^D176A served as negative controls. Phos-tag gel confirmed increased phosphorylated FLAG-RGA by CDK8/CYCC in the input sample (Fig. S13A). YFP-MED15^N or -GUS were immunoprecipitated using anti-GFP agarose and immunoblots were probed with αFLAG, αMyc or αGFP. Relative amounts of co-immunoprecipitated FLAG-RGA/rga are shown below the blot. –, not detectable. Representative results of 3 biological repeats are shown. (B) Model for CKM-mediated attenuation of DELLA function. DELLA proteins are recruited to target chromatin through interactions with TFs, forming TF–DELLA–H2A complexes that regulate the transcription of GA-responsive genes. CDK8 kinase module (CKM) promotes GA responses by phosphorylating RGA at Ser170 within the disordered PolyS/T region. This modification weakens RGA-MED15 interaction, thereby limiting recruitment of the core Mediator complex (cMED) and the RNA polymerase II preinitiation complex (PIC) to RGA-induced, GA-repressed promoters.

To examine directly the role of S170 phosphorylation in RGA-MED15 interaction, co-IP assays were performed in N. benthamiana expressing FLAG-RGA or rga^S170A or rga^S170D together with YFP-MED15^N, with or without CDK8/CYCC. The phosphomimetic S170D mutation reduced MED15 binding, whereas S170A increased binding compared with wild-type RGA (Fig. 7A and Fig. S13A). We also examined the m2A mutant (eight S/T-to-A substitutions in Pep2). In contrast to S170A, m2A weakened MED15 binding. These results suggest that phosphorylation of S170 decreases RGA activity by weakening MED15 interaction, whereas phosphorylation at other S/T residues within Pep2 enhances RGA function.

DISCUSSION

Based on our findings, we propose a working model in which the CDK8 kinase module (CKM) functions as a regulatory switch that modulates DELLA-dependent transcription and thereby fine-tunes GA responses (Fig. 7B). In this model, CKM attenuates the transactivation capacity of the DELLA protein RGA by limiting recruitment of the coactivator Mediator complex to DELLA-induced, GA-repressed promoters. Mediator is a conserved transcriptional hub that transmits regulatory inputs from transcription factors (TFs) to RNA polymerase II (Pol II). The core Mediator (cMED) complex contains head, middle, and tail modules that interact with Pol II and TFs, respectively. Association of the CDK8 kinase module—composed of CDK8, Cyclin C (CYCC), MED12, and MED13—adds a dynamic regulatory layer to Mediator function (48, 49). CDK8 substrates include Mediator subunits, Pol II and transcription factors. Although initially characterized as a transcriptional repressor that blocks Mediator–Pol II interaction, CKM is now recognized as a context-dependent regulator capable of either repressing or activating transcription through CDK8 kinase activity in response to developmental and stress signals (48, 61). In Arabidopsis, CDK8 has been implicated in ABA signaling during drought stress (62) and in salicylic acid–mediated defense responses (63, 64), but its connection to GA/DELLA signaling had not been established.

Previous studies showed that DELLAs are recruited to target chromatin through interactions with transcription factors (TFs), forming TF–DELLA–H2A complexes that regulate the transcription of GA-responsive genes (24). DELLA-mediated transcriptional activation further depends on recruitment of the Mediator coactivator complex via interaction with the tail subunit MED15 (25). Here we show that CDK8 phosphorylation of RGA at Ser170 weakens the interaction between RGA and MED15, thereby reducing Mediator recruitment and dampening transcriptional activation of RGA-induced genes. To our knowledge, this represents a distinct regulatory mechanism of CKM, which differs from those previously described for CDK8/CKM substrates, in which phosphorylation typically modulates the stability and/or intrinsic transcriptional activity of transcription factors (65, 66).

Our model is supported by multiple lines of evidence. TurboID proximity labeling identified CDK8 together with MED12 and MED13 in the RGA proximity proteome. Phos-tag gel and MS/MS analyses demonstrated CKM-mediated phosphorylation of RGA at Ser170 within the PolyS/T region in planta. Although the amino acid sequence spanning this phosphorylation site is not conserved among DELLA proteins (Fig. S14), CDK8 also phosphorylates three other Arabidopsis DELLA proteins (Fig. S4), suggesting that CKM-mediated phosphorylation is a conserved mechanism for regulating DELLA function. Functional assays further showed that CKM weakens RGA–MED15 association in both Y3H systems and co-IP assays. Consistent with this mechanism, cdk8 and rga gai mutants exhibit opposing phenotypes during vegetative and reproductive phase transitions—developmental processes that are promoted by GA signaling and repressed by DELLAs (67, 68). In addition, MED12 and MED13 have been reported to promote these developmental transitions (69–71). Together, these findings support a role for CKM in promoting GA signaling by limiting DELLA-mediated transcription activation through phosphorylation-dependent control of Mediator recruitment. It is intriguing that CDK8 does not alter expression of RGA-repressed, GA-induced genes, suggesting that Ser170 phosphorylation may not modulate recruitment of corepressors by RGA.

Our results further indicate that RGA activity is modulated by multiple phosphorylation sites within the Pep2 region. Ser170 lies within the intrinsically disordered PolyS/T region between the DELLA and GRAS domains. At first glance, the inhibitory effect of Ser170 phosphorylation appears inconsistent with our previous observation that Ala substitution of eight Ser/Thr residues within Pep2, including S170 (the m2A mutant, Fig. 4C), reduces RGA activity in planta (42), suggesting that phosphorylation within this region promotes RGA function. However, the underlying phenotypes reflect distinct molecular effects. The S170A mutation enhances RGA–MED15 interaction without affecting RGA binding to histone H2A, whereas the m2A mutation strongly disrupts the RGA–H2A interaction (42) (Fig. 7A, Fig. S9C). Although only S170 phosphorylation was detected in RGA Pep2 in the current study, we previously identified several additional potential phosphorylation sites within this peptide, including S167, S169, S173, and S179 (42). Together, these findings suggest that phosphorylation at different Pep2 residues selectively modulates RGA interactions with distinct cofactors. We therefore propose that RGA activity is dynamically tuned by multisite phosphorylation within Pep2, potentially mediated by different kinases with opposing effects. In this model, CDK8 phosphorylation of Ser170 attenuates RGA activity by limiting Mediator recruitment, whereas phosphorylation of other Pep2 residues enhances RGA function, at least in part by stabilizing interaction with H2A.

The functional significance of DELLA phosphorylation has long been debated, with studies proposing roles in DELLA degradation, stabilization, or activity. Our results demonstrate that phosphorylation can regulate DELLA transcriptional activity independently of protein stability and that distinct phosphorylation events within intrinsically disordered regions (IDRs) can produce different regulatory outcomes. These findings are consistent with the emerging concept that IDRs serve as regulatory hubs enriched in clustered phosphorylation sites, enabling context-dependent control by tuning the specificity and affinity of protein-protein interactions rather than functioning as simple on-off switches (72–74).

Beyond the CDK8 kinase module highlighted here, our TbID proximity labeling analysis also uncovered a large set of previously unrecognized RGA-proximal proteins linked to transcriptional regulation. These findings markedly expand the known RGA interaction landscape and suggest that RGA engages a broad transcriptional regulatory network comprising transcription factors, transcriptional cofactors, chromatin remodelers, and histone acetyltransferase complexes. This expanded RGA proximitome reveals strong connections between RGA and core transcriptional regulatory machinery, providing a framework for mechanistic dissection of RGA-mediated transcriptional control.

Taken together, our work uncovers a mechanism linking the Mediator kinase module to GA signaling through phosphorylation-dependent control of DELLA activity. By modulating the interaction between RGA and Mediator, CKM provides a tunable layer of transcription regulation that connects kinase signaling with hormone-responsive gene expression. This mechanism highlights how phosphorylation within IDRs can dynamically shape transcriptional regulatory networks during plant development.

Methods

Plant materials, growth conditions and plasmid construction

All Arabidopsis lines used in this study are in the Col-0 ecotype background. Mutants, transgenic lines, plant growth conditions, GA response assays and adult phenotype analyses are described in SI Methods. Primers and plasmid constructs are listed in Tables S2 and S3, respectively, and addition construct information is described in SI Methods.

Statistical analyses

Statistical analyses for all quantitative data were conducted using Student’s t-test for comparisons between two samples, or one-way ANOVA followed by Tukey’s Honestly Significant Difference (HSD) mean separation test for comparisons among multiple samples, using SPSS Statistics 17.0 software.

Transient expression and dual luciferase assay in Nicotiana benthamiana

Transient expression of epitope-tagged proteins in N. benthamiana was performed as described (75). The N. benthamiana leaves were harvested 48 h after Agro-infiltration for protein extraction, followed by dual luciferase assays using dual-luciferase reporter assay system (Promega).

Reverse transcription (RT)-quantitative PCR (qPCR) and immunoblot analyses

Total Arabidopsis RNA extraction and RT-qPCR analysis were performed as described (42). Relative transcript levels were determined by normalizing with PP2A (AT1G13320) (76). Primers for the qPCR are listed in Table S2. Phos-tag gel and immunoblot analyses are described in SI Methods.

Yeast two-hybrid and yeast three-hybrid Assays

The yeast strain PJ69-4A and the Matchmaker Gold Yeast Two-Hybrid System (Clontech, Palo Alto, CA) were used. For Y3H assays, a third vector pH3-Myc-GW was used to express the third set of proteins (see SI Methods).

In vitro pulldown and co-IP assays

In vitro pulldown and co-IP assays were performed following the procedures published previously (24, 55) (see SI Methods).

TurboID-based proximity labeling and LC-MS/MS analyses

Identification of RGA proximity proteome by MS analysis was performed as described previously (77) with some modifications (see SI Methods). LFQ analysis was performed and LFQ intensities were imported to Perseus (version 2.1.3) where a Student’s t-test was executed to identify candidate proteins that are enriched by the RGA-TurboID in comparison to the GFP-TurboID control. The t-test settings were the following: ‘Permutation-based FDR’, ‘FDR=0.01’, S0=0.8, ‘Report q-value’, ‘Number of Randomizations=250’ and ‘-log10 p-value’.

Identification of PTM sites by liquid chromatography tandem MS (MS/MS) analyses

His-FLAG-RGA^GKG was purified from P_RGA:His-FLAG-RGA^GKG transgenic Arabidopsis lines in the ga1-13 and ga1-13 cdk8-1 backgrounds by the tandem affinity purification procedures described previously (78). Affinity-purified proteins were trypsin-digested, and peptides were analyzed by online LC-electrospray ionization (ESI) tandem MS [electron-transfer dissociation (ETD) and collisionally activated dissociation (CAD) MS/MS] using a Thermo^™ Orbitrap Fusion Lumos^™ Tribrid^™ mass spectrometer equipped with ETD (79) (See SI Methods).

Significant Statement.

DELLA proteins are central repressors of gibberellin signaling and genetic variants in DELLA genes were instrumental in the development of the semi-dwarf crops of the Green Revolution. DELLA proteins regulate gene expression through interactions with transcription factors, histone H2A, and the Mediator coactivator complex via its tail subunit MED15. However, how DELLA activity is modulated by phosphorylation has remained largely unsolved. We identify CDK8 as a physiologically relevant kinase for the Arabidopsis DELLA protein RGA and show that CDK8-mediated phosphorylation attenuates RGA activity by weakening its interaction with MED15. These findings uncover a mechanism linking the Mediator CDK8 kinase module to gibberellin signaling and provide new insight into how growth-regulatory networks can be dynamically fine-tuned in plants.

Data and Materials Availability

Materials generated in this study are available upon request. Material that can be shared will be released via a material transfer agreement. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (80) partner repository with the dataset identifier PXD077697 for RGA PTMs (Project DOI: 10.6019/PXD077697), and PXD078134 for RGA proximitome.