HDAC2 Regulates Site-Specific Acetylation of MDM2 and Its Ubiquitination Signaling in Tumor Suppression.

Histone deacetylases (HDACs) are promising targets for cancer therapy, although their individual actions remain incompletely understood. Here, we identify a role for HDAC2 in the regulation of MDM2 acetylation at previously uncharacterized lysines. Upon inactivation of HDAC2, this acetylation creates a structural signal in the lysine-rich domain of MDM2 to prevent the recognition and degradation of its downstream substrate, MCL-1 ubiquitin ligase E3 (MULE). This mechanism further reveals a therapeutic connection between the MULE ubiquitin ligase function and tumor suppression. Specifically, we show that HDAC inhibitor treatment promotes the accumulation of MULE, which diminishes the t(X; 18) translocation-associated synovial sarcomagenesis by directly targeting the fusion product SS18-SSX for degradation. These results uncover a new HDAC2-dependent pathway that integrates reversible acetylation signaling to the anticancer ubiquitin response.

Introduction

Synovial sarcoma is an incurable malignant soft tissue tumor, which primarily affects children and young adults. Like many other sarcomas, this disease is histologically composed of mesenchymal cells. However, synovial sarcoma displays variable degrees of epithelial differentiation and contains a unique chromosomal translocation t(X; 18), which most commonly fuses the SS18 gene with SSX1 or SSX2 (Nielsen et al., 2015). Depletion of SS18-SSX by small interfering RNAs (siRNAs) causes apoptotic cell death of human synovial sarcoma cells (Peng et al., 2008, Cai et al., 2011, Carmody Soni et al., 2014). Conversely, overexpression of SS18-SSX in noncancerous rat fibroblast cells shows transforming activity in a xenograft model (Nagai et al., 2001). Notably, mice conditionally expressing the SS18-SSX fusion gene in certain cell lineages develop tumors that are pathologically indistinguishable from and molecularly consistent with synovial sarcoma in humans (Haldar et al., 2007), thus confirming the critical role for SS18-SSX in the pathogenesis of synovial sarcoma.

Fundamental progress has been made in understanding how the SS18-SSX fusion protein promotes tumorigenesis, which indeed involves multiple parallel mechanisms, such as epigenetic remodeling (Su et al., 2012, Kadoch and Crabtree, 2013, Banito et al., 2018, McBride et al., 2018), cellular adhesion (Eid et al., 2000), mesenchymal-to-epithelial transition (Saito et al., 2006, Barrott et al., 2015), protein translocation (Pretto et al., 2006), and microRNA regulation (Hisaoka et al., 2011, Minami et al., 2014). Such complexities in SS18-SSX action make the development of targeted therapies for synovial sarcoma extremely challenging. Despite the lack of effective treatment options, several lines of evidence have shown that human synovial sarcoma cells are highly sensitive to histone deacetylase (HDAC) inhibitors in cell cultures and in a cell-line-based xenograft model (Ito et al., 2005, Laporte et al., 2017a). One well-supported explanation for the action of HDAC inhibitors is histone acetylation through which key tumor suppressor genes become epigenetically reactivated (Lubieniecka et al., 2008, Su et al., 2010, Su et al., 2012, Laporte et al., 2017b). In the present study, we propose an additional, transcription-independent mechanism whereby HDAC inhibition facilitates proteasomal degradation of the SS18-SSX fusion protein. This action relies mostly on a novel combination of HDAC2 and MDM2 activities in concert with the MULE E3 ligase function. Our findings connect HDAC2 activity to oncogenic protein stabilization via a series of post-translational events, which constitute an acetylation-dependent ubiquitin pathway that may serve as a common therapeutic target in human cancers.

Results

HDAC Inhibitor Treatment Reduces SS18-SSX Levels through the Ubiquitin System

To assess the in vivo efficacy of HDAC inhibition in synovial sarcoma, we generated transgenic mice expressing human SS18-SSX2 fusion oncogene within the myogenic factor 5 (Myf5) lineage (Haldar et al., 2007). Treatment with the HDAC inhibitor FK228 on a weekly basis significantly reduced growth of mouse synovial sarcomas (Figures S1A and S1B), associated with remarkable cytoreductive activity (Figures S1C–S1H). In addition to the histological observations, we noticed that SS18-SSX2 protein abundance was substantially decreased in FK228-treated tumors (Figures 1A and 1B). Considering the fusion oncogene dependency in synovial sarcoma, we decided to examine the molecular mechanism of SS18-SSX downregulation upon HDAC inhibition. To this end, we first developed a CRISPR/Cas9-based genome editing approach for FLAG epitope tagging of endogenous SS18-SSX2 fusion oncoprotein in patient-derived SYO-1 cells (Figures S1I–S1K). Anti-FLAG western blots revealed that SS18-SSX levels remained constant through the early time points of FK228 treatment, but fell drastically after overnight stimulation (Figure S1L). Similar results were obtained in cells treated with other structurally different HDAC inhibitors, such as SB939 and PCI-24781 (Figure 1C). We also tested this in SS18-SSX1-associated synovial sarcoma cells (Yamato-SS) and found that treatment with the HDAC inhibitors FK228 and SB939 led to a marked reduction of SS18-SSX protein levels, coupled with impaired tumor cell growth (Figures 1D and 1E). Importantly, the mRNA levels of SS18-SSX remained unchanged (Figure S1M), whereas its protein stability was significantly reduced (Figures 1F and 1G). This effect was efficiently blocked by the proteasome inhibitor MG-132 (Figure 1C), and restoration of SS18-SSX levels correlated positively with increased conjugation of poly-ubiquitin chains (Figure 1H). Our findings indicate the existence of as-yet-undefined ubiquitin ligases that target SS18-SSX for proteasomal degradation.

Figure 1

HDAC Inhibition Downregulates SS18-SSX Fusion Oncoprotein

(A) Representative immunofluorescence analysis of human SS18-SSX2 expression in tumor sections prepared from SSM2 mice treated with vehicle (−) or FK228 (3 mg/kg). Human SSX2 antibody was used to detect the fusion oncoprotein; DAPI was used for nuclear staining. Scale bar, 25 μm.

(B) Western blot analysis of human SS18-SSX2 expression in vehicle or FK228-treated SSM2 mouse tumors. Actin was used as a loading control.

(C) Western blot analysis of the lysate from CRISPR/Cas9-modified SYO-1 cells treated with vehicle (−) or HDAC inhibitors (FK228, SB939, and PCI-24781), in the presence and absence of MG-132. FLAG/tubulin ratios were normalized to vehicle and are shown in the top panel.

(D) Cell viability assay showing the sensitivity of Yamato-SS cells to FK228 and SB939 at different concentrations. Results represent mean ± SD of three independent experiments.

(E) Western blot analysis of SS18-SSX1 protein levels in DMSO-, FK228-, (100 nM), and SB939- (1 μM) treated Yamato-SS cell lysate. Tubulin was used as a loading control.

(F) Western blot analysis of SS18-SSX (anti-FLAG) protein abundance in CRISPR/Cas9-modified SYO-1 cells upon 12-h treatment of DMSO or FK228, followed by exposure to cycloheximide (CHX, 100 μg/mL) for the indicated time. Actin serves as a loading control, and HDAC1 serves as a negative control, which stays constant regardless of FK228 treatment.

(G) FLAG/actin ratios were normalized to the 0-h time point. Data represent mean ± SD of three independent experiments.

(H) Ubiquitination analysis of anti-FLAG immunoprecipitates from CRISPR-modified SYO-1 cells treated with DMSO, FK228, or SB939, in the presence of MG-132. Cell lysate (input) was applied to western blot analysis showing equal amounts of ubiquitin protein under all conditions, and mouse IgG was used as a negative control.

MULE Ubiquitin Ligase Binds to SS18-SSX and Promotes Its Degradation

Next, we performed anti-FLAG immunoprecipitation to purify SS18-SSX-interacting proteins. Mass spectrometric analysis of the peptides uniquely enriched after HDAC inhibitor stimulation led to identification of MCL-1 ubiquitin ligase E3 (MULE), a HECT-type enzyme that plays a central role in the regulation of cell proliferation and tumorigenesis (Shmueli and Oren, 2005, Kao et al., 2018) (Figures 2A and 2B). Its interaction with the fusion oncoprotein was readily detected in both proximity ligation and immunoprecipitation assays (Figures 2C–2E). This binding event seemed to involve two fundamental domains in MULE, WWE and UBM, which co-occupied the repression domain (SSXRD) at the C-terminal end of SS18-SSX (Figures 3A–3D). Interestingly, there exists an alternatively spliced product of the SSX2 gene (dos Santos et al., 2000), containing a different C-terminal region (Figure S2A). We found that this variant failed to interact with MULE in SYO-1 cells, regardless of HDAC inhibitor stimulation (Figures S2B–S2D). Given the fact that MULE neither binds wild-type SS18 nor does its deletion affect SS18 protein expression (Figures S2E and S2F), we reasoned that the C-terminal SSXRD may play an indispensable role for SS18-SSX recognition by MULE E3 ligase.

Figure 2

MULE Binds to SS18-SSX upon HDAC Inhibitor Treatment

(A) Mass spectrometric analysis of anti-FLAG immunoprecipitates from CRISPR/Cas9-modified SYO-1 cells treated with DMSO or FK228. MULE was identified under the treatment of FK228; anti-FLAG western blot was performed to confirm the pull-down efficiency.

(B) List of the MULE peptides detected from anti-FLAG immunoprecipitates in FK228- (but not DMSO-) treated CRISPR/Cas9-modified SYO-1 cells.

(C) Proximity ligation assay for endogenous SS18-SSX (anti-FLAG) and MULE in CRISPR/Cas9-modified SYO-1 cells treated with DMSO, FK228, or SB939. DAPI was used to stain the nuclei; scale bar, 15 μm.

(D and E) MULE immunoprecipitation (IP) analysis of its interaction with the fusion oncoprotein (anti-FLAG) in DMSO-, FK228 (D)-, and SB939 (E)-treated CRISPR cells. The lysate (input) and rabbit IgG serve as positive and negative controls, respectively.

Figure 3

Molecular Mechanism of MULE-Mediated SS18-SSX Degradation

(A) Schematic of hemagglutinin (HA)-tagged MULE domains. ARLD, armadillo repeat-like domain; WWE, tryptophans-/glutamate-rich domain; BH3, BCL-2 homology 3 domain; UBM, ubiquitin-binding domain; PIP, PCNA-interacting protein domain; HECT, homologous to the E6-AP carboxyl terminus domain. BR indicates the SS18-SSX-binding region highlighted in blue.

(B) Anti-HA immunoprecipitation (IP) analysis of HEK293 cells expressing GFP-tagged SS18-SSX2, together with HA-tagged MULE constructs (or empty vector), under FK228 and MG-132 treatment. Cell lysate (input) serves as a positive control; *mouse IgG heavy chain.

(C) Schematic of GFP-tagged SS18-SSX2 construct. SNH, SYT N-terminal homolog domain; QPGY, glutamine-/proline-/glycine-/tyrosine-rich domain; SSXRD, SSX repression domain.

(D) Anti-FLAG immunoprecipitation analysis of HEK293 cells expressing full-length FLAG-MULE, together with GFP-SS18-SSX2 and its truncated mutants, under treatment with FK228 and MG-132. Cell lysate (input) was used as a positive control.

(E) Ubiquitination analysis of GFP-SS18-SSX2 and its mutants in FK228-treated HEK293 cells, in the presence of MG-132. Cell lysate (input) was applied to western blot analysis showing equal amounts of ubiquitin protein under all conditions.

(F) Model proposed for MULE-mediated SS18-SSX ubiquitination.

We continued to assess whether SS18-SSX is a substrate of MULE, and found that in vitro synthesized fusion oncoprotein could be poly-ubiquitinated by cell extracts prepared from wild-type but not Mule-knockout mouse embryonic fibroblasts (Figures 4A–4C). Consistently, depletion of MULE by short hairpin RNA in SYO-1 cells prevented endogenous SS18-SSX ubiquitination upon HDAC inhibitor stimulation (Figure 4D). Removal of MULE-binding SSXRD domain strikingly reduced SS18-SSX ubiquitination levels (Figure 3E), suggesting that MULE ubiquitinates the fusion oncoprotein largely through their physical interaction. It should be noted that the SYT N-terminal Homolog (SNH) motif of SS18-SSX also contributed to MULE-mediated ubiquitination (Figure 3E). Given that the SNH-deleted (ΔSNH) mutant retained binding to MULE (Figure 3D), we reasoned that the critical SS18-SSX ubiquitination sites may be located within the SNH region. Indeed, previous studies have identified a lysine residue (K13) within wild-type SS18 protein that is ubiquitinated in human cell lines (Mertins et al., 2013, Udeshi et al., 2013). However, among four lysines, only mutating K23 to arginine (K23R) effectively blocked SS18-SSX ubiquitination (Figure 3E). Thus the mechanism underlying SS18-SSX ubiquitination likely differs from that for wild-type SS18. Together, these results provide molecular insights into MULE E3 ligase activity toward the fusion oncoprotein (Figure 3F).

Figure 4

Knockdown of MULE Prevents SS18-SSX Downregulation and Promotes Resistance of Synovial Sarcoma Cells to HDAC Inhibition

(A) Anti-Myc western blot analysis of in vitro synthesized SS18-SSX2 protein. Empty vector (−) serves as a negative control.

(B) Wild-type (WT) and Mule-null mouse embryonic fibroblast (MEF) cells were subjected to western blots showing endogenous Mule protein levels. Tubulin was used as a loading control.

(C) Ubiquitin conjugation to Myc-tagged SS18-SSX2 protein after in vitro reactions supplemented with the indicated MEF cell extract. Reactions with no ATP serve as a negative control.

(D) SS18-SSX poly-ubiquitination in anti-FLAG immunoprecipitates prepared from control and MULE-knockdown CRISPR-modified SYO-1 cells treated with DMSO (−) or FK228 (30 nM), in the presence of the proteasome inhibitor MG-132. Input was applied to western blot analysis showing equal amounts of ubiquitin protein under all conditions.

(E) Western blot analysis of MULE and SS18-SSX (anti-FLAG) protein levels in CRISPR-modified SYO-1 cells stably expressing control or MULE short hairpin RNAs (shRNAs) upon treatment of DMSO and HDAC inhibitors (FK228 and SB939). Tubulin was used as a loading control.

(F) Viability assay of stable cell lines (used in E) in response to FK228 and SB939 treatment. Results represent mean ± SD of three independent experiments; *p < 0.01 by two-tailed Student's t test.

(G) Representative images of colony formation assay using DMSO-, FK228- and SB939-treated stable SYO-1 cell lines. Colony numbers under FK228 and SB939 treatment were normalized to vehicle (DMSO) and shown in the bottom panel. Data represent mean ± SD of three independent experiments; *p < 0.01 (two-tailed Student's t test).

To elucidate the biological significance of these findings, we examined the effect of the presence or absence of MULE on SS18-SSX protein expression upon HDAC inhibitor treatment. In SYO-1 cells, SS18-SSX levels fell after addition of FK228 and SB939. However, under the same conditions, the fusion oncoprotein avoided downregulation by MULE depletion (Figure 4E). More importantly, removal of MULE reduced the sensitivity of SYO-1 cells to HDAC inhibitors (Figures 4F and 4G). This resistance was neither due to the reported role of MULE in degrading histones (Liu et al., 2005) nor due to epigenetic changes in global histone acetylation upon HDAC inhibitor treatment (Figures S3A–S3C). This might result at least partly from impaired turnover of SS18-SSX, which retained the ability to promote cell proliferation and transformation via HDAC-independent mechanisms. In addition, among several known substrates of MULE, we found that HDAC inhibitor treatment downregulated CTCF and β-catenin protein expression, and this reduction was closely dependent on MULE (Figures S3D and S3E). CTCF is a multifunctional transcription factor that organizes chromatin architecture and controls genomic stability in various biological processes involved in tumorigenesis (Qi et al., 2012, Ghirlando and Felsenfeld, 2016, Song and Kim, 2017). β-Catenin acts as a central regulator of the WNT signaling pathway, which has been extensively studied in many human cancer types and directly linked to the progression of synovial sarcoma (Barham et al., 2013, Trautmann et al., 2014, Barrott et al., 2015, Barrott et al., 2018, Cironi et al., 2016, Sanchez-Vega et al., 2018). Collectively, our findings point to MULE-mediated ubiquitination signaling as a potential fundamental target for HDAC inhibitors in tumor suppression.

HDAC Inhibition Stabilizes MULE by Acetylation and Dissociation of MDM2

We noted that MULE protein levels rose upon HDAC inhibitor treatment, without mRNA level changes (Figures S4A and S4B). Similar effects were also observed after proteasome inhibition (Figures 5A and S4C), indicating the unstable status of MULE protein in synovial sarcoma cells. Consistent with this notion, recent studies have reported MULE degradation by the oncogenic E3 ligase MDM2 in human cancers (Kurokawa et al., 2013, Canfield et al., 2016). We confirmed the endogenous interaction between MDM2 and MULE in both SYO-1 and Yamato-SS cells (Figures 5B and S4D). Knockdown of MDM2 by siRNAs led to increased levels of MULE, which inversely correlated with SS18-SSX downregulation and impaired cell viability (Figures 5C, 5D, S4E, and S4F). This effect seems independent of the p53 tumor suppressor, a key target of MDM2, because SYO-1 cells express wild-type p53, but Yamato-SS cells harbor a mutant p53 (R273C) (Vlenterie et al., 2016). Consistent with this view, depletion of endogenous p53 failed to rescue SYO-1 cells from MDM2 knockdown (Figures S4G–S4I). These results indicate a predominant role for MDM2 control of MULE (but not p53) signaling in synovial sarcoma cells. Notably, this action could be reversed by the addition of FK228 and SB939, which suppressed MDM2 binding to MULE (Figure 5E). As a result, HDAC inhibition led to a decrease in MULE ubiquitination and an increase in its protein stability (Figures 5F, 5G, S4J, and S4K). In addition, we found that the response of SYO-1 cells to FK228 and SB939 treatment, similar to MDM2 knockdown, remained unaffected after p53 deletion (Figures S4L–S4N), further supporting the notion that HDAC-inhibitor-induced anticancer action relies largely on MDM2 regulation of MULE, rather than p53, in synovial sarcoma.

Figure 5

MDM2 Functions as a Negative Regulator of MULE

(A) Anti-MULE western blot analysis of the lysate prepared from SYO-1 cells treated with or without MG-132. MULE-depleted cell extracts were used as a negative control; tubulin serves as a loading control.

(B) Western blot analysis of anti-MULE immunoprecipitates from MG-132-treated SYO-1 cells. The lysate (input) and rabbit IgG serve as positive and negative controls, respectively.

(C) Western blot analysis of the lysate from CRISPR-modified SYO-1 cells transfected with control (−) or MDM2 siRNAs. Tubulin was used as a loading control.

(D) Viability of SYO-1 cells in response to MDM2 deletion by two individual siRNAs. Values were normalized to control knockdown (−); data represent mean ± SD of three independent experiments.

(E) Immunoprecipitation analysis of the MULE-MDM2 interaction in SYO-1 cells treated with DMSO (−) or HDAC inhibitors (FK228 and SB939), in the presence of MG-132. Cell lysate (input) and rabbit IgG serve as positive and negative controls, respectively.

(F) Ubiquitination analysis of endogenous MULE protein in vehicle (−), FK228- and SB939-treated SYO-1 cells. Cell lysate (input) was used to show equal amounts of ubiquitin under all conditions.

(G) Western blot analysis of MULE protein abundance in DMSO- and FK228-treated SYO-1 cells, upon exposure to cycloheximide (CHX) for the indicated time. Tubulin serves as a loading control, and MULE/tubulin ratios were normalized to the 0-h time point.

Given that HDAC inhibition affects biological functions mostly through protein acetylation (Verdin and Ott, 2015), we investigated whether the same mechanism controls the MDM2-MULE interaction. Recently, it has been discovered that acetylation of the lysine-rich domain (KRD) can remove its positive charge and interrupt its association with the negatively charged acidic domain (Wang et al., 2016, Wang et al., 2017). Indeed, there does exist a lysine-rich stretch (amino acids 460–476) in the MDM2 protein, whereas MULE has an acidic domain between the residues 2425 and 2469. In support of these clues, we found structural evidence of MULE's acidic domain bound to the MDM2 KRD (Figures 6A, S5A, and S5B). This interaction likely involved five evolutionarily conserved lysine residues within KRD (Figures S5C–S5F), which were acetylated in endogenous and ectopically expressed MDM2 upon HDAC inhibitor treatment (Figures 6B, 6C, S5G, and S5H). According to these properties, we generated a mutant MDM2 (5KR) in which all KRD lysine sites were replaced by arginine (Figure 6D). This acetylation-deficient mutation retained the ability of MDM2 binding to MULE, but exhibited resistance to HDAC-inhibitor-stimulated dissociation (Figure 6E). Point mutation analysis identified two acetylated lysines, K469 and K470, responsible for the release of MDM2 from MULE (Figures 6F, 6G, S5I, and S5J). In line with this observation, K469 and K470 appeared to be the most effective sites of MDM2 acetylation after FK228 addition (Figure S5H). To further test this idea, we generated another MDM2 mutation in which the K469 and K470 residues were substituted to glutamine. This acetylation-mimicked mutation resulted in a dramatic loss of the MDM2-MULE interaction even without HDAC inhibitor treatment (Figures 6H and 6I). Intriguingly, Moshe Oren and colleagues reported that acetylation of the neighboring lysines K466 and K467 impairs MDM2's E3 ligase activity (Wang et al., 2004). However, a similar defect was not found when mimicking the acetylation of K469 and K470 (Figure S5K). Therefore different acetylation modifications in the KRD region might have distinct and cooperative roles in limiting MDM2 function.

Figure 6

MDM2 Acetylation at Specific Lysine Sites Negatively Regulates its Interaction with MULE

(A) Structural interface showing the interaction between MDM2 (lysine-rich domain; KRD) and MULE (acidic domain).

(B) Acetyl-lysine western blots showing increased signals in endogenous MDM2 immunoprecipitates prepared from HDAC inhibitor (FK228/SB939)-treated SYO-1 and Yamato-SS cells, compared to DMSO (–). Rabbit IgG was used as a negative control.

(C) MS/MS spectra of the MDM2 lysine-rich domain (KRD) peptides identified in anti-MDM2 immunoprecipitates prepared from SYO-1 cells under the treatment of FK228, but not DMSO.

(D) Schematic of HA-tagged MDM2 (WT), acetylation-deficient (KR) and -mimicked (KQ) mutations.

(E) Immunoprecipitation analysis of MULE association with ectopically expressed MDM2 (WT) and its 5KR mutant in HEK293 cells treated with DMSO (–) or FK228, in the presence of MG-132. Input serves as a loading control.

(F) Immunoprecipitation analysis of MULE association with ectopically expressed MDM2 (WT) and its mutants (KR1 and KR2) in HEK293 cells treated with DMSO (–) or FK228, in the presence of MG-132. Input serves as a loading control.

(G) MULE-bound MDM2 levels were quantified by ImageJ and normalized to the Input; values of FK228-treated samples were further normalized to DMSO. Error bars represent mean ± SD of three independent experiments.

(H) Immunoprecipitation assay showing Flag-MULE interaction with HA-MDM2 and its KQ mutants in MG-132 treated HEK293 cells. The lysate (Input) was used as a loading control. (I) MULE-bound/input ratios of KQ mutants were normalized to WT. Results represent mean ± SD of three independent experiments.

HDAC2 Regulates the MDM2-MULE Interaction and Maintains SS18-SSX Protein Stability

Finally, it is important to determine which of the 11 HDAC family members act upstream of MDM2 (Figure S6A). We focused on two highly homologous members, HDAC1 and HDAC2, because they are most abundantly expressed in synovial sarcoma cells (Figure 7A) (Pacheco and Nielsen, 2012), and because the compound FK228 mainly inhibits HDAC1/2 activity. Interestingly, MDM2 associated with both HDAC1 and HDAC2 (Figure S6B), and depletion of neither HDAC1 nor HDAC2, affected MDM2 protein abundance (Figures S6C and S6D). However, MDM2-MULE interaction was diminished in HDAC2-deficient cells (Figures 7B and S6E). Moreover, the amount of SS18-SSX protein (but not mRNA) was drastically reduced upon HDAC2 knockdown (Figures 7C, S6F, and S6H), similar to the results obtained in MDM2-knockdown cells. We did not observe any significant effects after depletion of either HDAC1 or its Class I homolog HDAC3 (Figures S6G–S6I). Therefore unlike HDAC1, which contributes to SS18-SSX-mediated gene regulation (Su et al., 2012, Cironi et al., 2016), HDAC2 performs a nonredundant role in safeguarding the fusion oncoprotein from ubiquitin-mediated degradation.

Figure 7

HDAC2 Contributes to MDM2-MULE Interaction and Governs SS18-SSX Protein Stability

(A) RNA sequencing analysis for the expression of HDAC family members in human synovial sarcoma cell lines, SYO-1 and Yamato-SS. FPKM, fragments per kilobase million.

(B) Immunoprecipitation analysis of the MDM2-MULE interaction in MG-132-treated SYO-1 cells depleted of HDAC1 or HDAC2. Cell lysate (input) and rabbit IgG serve as positive and negative controls, respectively.

(C) Western blot analysis of the lysate from nonspecific and HDAC2 knockdown CRISPR/Cas9-modified SYO-1 cells. Actin serves as a loading control.

(D) Proposed model for MDM2 control of MULE stability and activity through HDAC-inhibitor-induced lysine acetylation.

Discussion

The experiments presented here demonstrate a new mechanism for MDM2 inactivation, which is achieved through site-specific acetylation (Figure 7D). Repression of this acetylation is crucial to MDM2 interaction with its substrate MULE. In synovial sarcoma, the HDAC2 enzyme governs MDM2 substrate-binding activity; suppression of HDAC2 by RNA interference or small-molecule inhibitors allows dissociation and accumulation of the ubiquitin ligase MULE, leading to subsequent degradation of the SS18-SSX fusion oncoprotein, β-catenin, and CTCF. We provide mechanistic and functional evidence pointing toward the therapeutic implication for using HDAC inhibitors in synovial sarcoma treatment. Notably, this may not be limited to synovial sarcoma. For example, MDM2-mediated MULE downregulation has been reported to confer breast cancer resistance to the human epidermal growth factor receptor 2 inhibitor lapatinib (Kurokawa et al., 2013). It is tempting to speculate that use of HDAC inhibitors in combination therapy may resensitize some currently incurable cancers to conventional treatment.

Parallel with our current work, we recognize that MDM2 can also interact with MULE in an HDAC2-independent manner. This binding event requires the hydrophobic p53-binding pocket of MDM2, which in turn provides a targeting site for the small-molecule antagonist Nutlin-3a to prevent MULE accommodation and degradation (Kurokawa et al., 2013). In synovial sarcoma cells, however, Nutlin-3a treatment does not disrupt the MDM2-MULE complex, nor does it influence MULE protein expression (Figures S6J and S6K). These different results indicate that instead of a two-site binding model, there exist at least two levels of MDM2 control of MULE—one acting in the forward direction involving the intermolecular contact, whereas the second acting in the reverse direction involving lysine acetylation to unlock the binding interface. It will be important to further examine if any of the other post-translational modifications have a similar role as MDM2 in the regulation of MULE stability and activity. These efforts may also lead the way to a better understanding of how MDM2 integrates diverse input signals to execute its ubiquitination function with a high degree of specificity.

Limitations of the Study

We do not know if HDAC2 regulates MDM2 acetylation in any cancer forms other than synovial sarcoma. With our experiments, it is also not known if HDAC2 has a similar role in noncancerous context. Under different physiological conditions, HDAC2 and other related HDAC members may distinctly target the same lysine residues for control of MDM2 acetylation and function. Another limitation of our current study is the inability to answer whether HDAC inhibitors also control the substrate targeting of MULE E3 ligase through lysine acetylation. At this moment, the importance of post-translational modifications in the differential regulation of MULE downstream targets remains an open issue for further investigation.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.