Ubiquitylome analysis identifies dysregulation of effector substrates in SPOP-mutant prostate cancer

See allHide authors and affiliations

Science  03 Oct 2014:
Vol. 346, Issue 6205, pp. 85-89
DOI: 10.1126/science.1250255

Mutant protein in tumors hits the DEK

Cancer genome sequencing projects have uncovered a multitude of mutations in human tumors. Understanding whether and how these mutations contribute to tumor development and progression could ultimately lead to new therapies. Theurillat et al. studied the protein product of a gene that is recurrently mutated in prostate cancer. Normally this protein helps attach a biochemical tag to cellular proteins that marks them for degradation. The new work shows that the tumor-associated mutant protein loses this tagging ability, which results in the stabilization of a handful of cellular proteins that would otherwise be degraded. One of the most intriguing of these proteins was DEK, which helps prostate cancer cells invade into surrounding tissue.

Science, this issue p. 85


Cancer genome characterization has revealed driver mutations in genes that govern ubiquitylation; however, the mechanisms by which these alterations promote tumorigenesis remain incompletely characterized. Here, we analyzed changes in the ubiquitin landscape induced by prostate cancer–associated mutations of SPOP, an E3 ubiquitin ligase substrate-binding protein. SPOP mutants impaired ubiquitylation of a subset of proteins in a dominant-negative fashion. Of these, DEK and TRIM24 emerged as effector substrates consistently up-regulated by SPOP mutants. We highlight DEK as a SPOP substrate that exhibited decreases in ubiquitylation and proteasomal degradation resulting from heteromeric complexes of wild-type and mutant SPOP protein. DEK stabilization promoted prostate epithelial cell invasion, which implicated DEK as an oncogenic effector. More generally, these results provide a framework to decipher tumorigenic mechanisms linked to dysregulated ubiquitylation.

Genome sequencing studies have revealed unanticipated roles for the ubiquitylation machinery in cancer. For example, the cullin-RING ubiquitin ligase adaptor protein speckle-type POZ protein (SPOP) is mutated in 8 to 14% of prostate and endometrial cancers (14). In prostate cancer, SPOP mutations are confined to specific amino acid residues within the substrate-binding cleft, which mediates substrate interaction and ubiquitin transfer (5). This exquisite localization suggests that SPOP mutations have undergone positive selection during tumorigenesis by altering binding and ubiquitylation of distinct effector substrates in a protumorigenic manner. However, the mechanisms and substrates underlying this phenomenon remain incompletely understood.

In principle, mutant SPOP could enhance ubiquitylation of SPOP substrates (gain-of-function effect) or ubiquitylate new substrates (neomorphic effect). Alternatively, SPOP mutants could dimerize with their wild-type counterparts (e.g., through the BTB and BACK domains), and so repress wild-type SPOP function (dominant-negative effect). In support of the latter model, SPOP mutations typically occur in a heterozygous state with a retained wild-type allele and are able to dysregulate known substrates [e.g., nuclear receptor coactivator 3 (NCOA3)] in a dominant-negative manner (1, 6).

Characterizing the ubiquitin landscape (or “ubiquitylome”) that results from cancer genomic alterations affecting ubiquitin ligase components may provide new insights into tumorigenesis. To interrogate changes in ubiquitylation conferred by prostate cancer SPOP mutations, we stably overexpressed vector control (C); wild-type SPOP (SPOP-WT); or a mutated variant (SPOP-F133L or SPOP-Y87N, in which Phe133 is replaced by Leu or Tyr87 is replaced by Asn (SPOP-MT)], in immortalized prostate epithelial cells expressing endogenous SPOP (fig. S1, A and B) (7). In each case, we characterized the resulting ubiquitylome by measuring glycine-glycine remnants of ubiquitylated lysines (K-ε-GG) after trypsin digestion and stable isotope labeling of amino acids in cell culture (SILAC)–based mass spectrometry (Fig. 1A). To account for ubiquitylation-related changes in protein expression, all K-ε-GG values were normalized to protein abundance (Fig. 1A) (database D1 in supplementary materials).

Fig. 1 K-ε-GG (ubiquitin)–profiling and validation for DEK in immortalized prostate epithelial cells (LHMAR) overexpressing vector control, SPOP-WT and SPOP-F133L/Y87N mutants (MT).

(A) Schematic showing the design of the proteomics experiments. Isogenic cell line expressing either vector control, SPOP-WT or SPOP-MT were isotopically labeled (see fig. S1, A and B). K-ε-GG ratios were normalized to protein expression changes, assessed in parallel. (B) Scatter plot of protein normalized K-ε-GG level changes between SPOP-MT and SPOP-WT (depicted as log2 ratio of MT/WT) shows coordinate down-regulation of distinct K-ε-GG peptides across replicates (see also figs. S1, C and D, and S2, A to G). (C) To visualize K-ε-GG sites that undergo inverse changes at the protein level (e.g., more than twofold), normalized K-ε-GG data shown in (B) was inverted and multiplied by corresponding total protein expression changes (see also fig. S3, C and D, for expression changes). (D) DEK protein expression across isogenic LHMAR cells assessed by immunoblotting with and without proteasome inhibition by MG132. DMSO, diemthyl sulfoxide. Error bars represent means ± SD, n = 4 (n.s., not significant; ∗∗P < 0.01, Student's t test). (E) Degradation of Flag-DEK over time. Flag-DEK was induced by doxycycline (Dox) and protein decay measured by immunoblotting after doxycycline withdrawal in inducible LHMAR cells (P value, Friedman test). See also fig. S3E for DEK mRNA expression.

We reasoned that putative K-ε-GG peptide substrates might show distinct patterns of protein abundance depending on whether mutant SPOP caused a dominant-negative, loss-of-function, gain-of-function, or neomorphic effect. In a dominant-negative context, substrate peptides should exhibit a [SPOP-WT > control > SPOP-MT] pattern of abundance, whereas a pure loss-of-function effect might yield a [SPOP-WT > control = SPOP-MT] pattern. In contrast, a gain-of-function effect may result in a [SPOP-MT > SPOP-WT > control] pattern, and a neomorphic effect should cause a [SPOP-MT > SPOP-WT = control] pattern.

Only 12 of 7181 K-ε-GG peptides detected exhibited more than twofold changes in peptide abundance in the SPOP-mutant contexts (z score >3 or <–3) compared with control or SPOP-WT (Fig. 1B; fig. S1, C and D; table S1; and database D1). Differentially regulated (more than twofold) K-ε-GG peptides generally followed a dominant-negative pattern, including those corresponding to WIZ, SCAF1, GLYR1, DEK, and TRIM24 proteins (Fig. 1B and figs. S1, C and D, and S2, A to E). K-ε-GG peptides from two proteins (corresponding to G3BP1 and CAPRIN1) showed a loss-of-function pattern (figs. S1, C and D, and S2, F and G), and two K-ε-GG peptides from SPOP itself showed up-regulation in the setting of both SPOP-WT and SPOP-MT overexpression, suggestive of autoubiquitylation (fig. S1, C and D). Although we observed no evidence for gain-of-function or neomorphic effects, we cannot exclude the existence of such mechanisms.

We next determined whether SPOP-binding motifs were enriched in the differentially regulated proteins. Indeed, of the seven proteins described above (excluding SPOP itself), three (43%) contained a previously characterized SPOP-binding sequence (DEK, SCAF1, and CAPRIN1) (5). In contrast, the frequency of SPOP-binding sites was significantly lower (~3%) across the ubiquitylome as a whole (P < 0.001) (table S2). We identified one K-ε-GG peptide corresponding to DAXX, a known SPOP substrate (8). As expected, this peptide was enriched—and the corresponding total protein was down-regulated—after overexpression of wild-type SPOP (fig. S3A). Expression of the transcriptional coactivator NCOA3, another SPOP substrate, followed a dominant-negative pattern, as reported previously (6, 9) (fig. S3B).

Because protein ubiquitylation is often a prerequisite for proteasomal degradation, we next sought to ascertain which differentially expressed K-ε-GG peptides showed an inverse correlation with total protein expression (database D1). Peptides corresponding to TRIM24, SCAF1, and GLYR1 fit this pattern (one peptide each) (Fig. 1C and fig. S3, C and D). TRIM24 has been shown to mediate ligand-dependent activation of the androgen receptor—a key prostate cancer driver—and to degrade TP53 through its E3 ligase activity (10, 11). This observation is of interest given that primary prostate cancers with SPOP mutations typically lack TP53 alterations (1).

However, the K-ε-GG peptides that exhibited the most profound down-regulation coupled with robust (e.g., more than twofold) up-regulation of the corresponding total protein mapped to DEK (Fig. 1C). DEK is a putative oncoprotein, first described as a fusion gene in an aggressive subset of acute myeloid leukemia (1214). This finding therefore raised the possibility that DEK might represent an SPOP substrate that becomes dysregulated by oncogenic prostate cancer SPOP mutations.

We next determined whether prostate cancer SPOP mutants affected DEK protein stability. Indeed, whereas forced expression of SPOP-WT in immortalized prostate epithelial cells reduced DEK protein levels, the SPOP-F133L and SPOP-Y87N mutants produced an accumulation of DEK (Fig. 1D). Treatment with the proteasome inhibitor MG132 also increased DEK, consistent with a possible role for the proteasome in DEK regulation. We also examined the effect of SPOP mutants on DEK protein stability, using immortalized prostate epithelial cells engineered to express a doxycycline-inducible Flag-tagged DEK. After doxycycline exposure, Flag-DEK decayed more rapidly in control cells than in cells containing SPOP mutants (Fig. 1E); there was no difference in DEK mRNA expression (fig. S3E). These results support the notion that prostate cancer SPOP mutants function in a dominant-negative fashion to impair ubiquitylation and proteasomal degradation of DEK.

Given that forced expression of SPOP reduced DEK protein levels, we sought to determine whether SPOP interacts with DEK directly to promote ubiquitylation. In support of this hypothesis, the primary DEK amino acid sequence contained a five–amino acid consensus SPOP-binding motif (fig. S3F) (5). To test this, we first overexpressed a Flag-tagged DEK protein harboring serine-to-alanine mutations at the motif (S287A/S288A) and assessed the ability of wild-type SPOP to repress DEK expression. As predicted, the S287A/S288A variant abolished the repressive effect of SPOP and produced elevated levels of DEK protein (Fig. 2A). To determine whether this motif mediated direct binding of SPOP to DEK, we performed immunoprecipitation experiments in cells expressing either wild-type DEK or the S287A/S288A variant. Whereas overexpressed SPOP protein was detectable after immunoprecipitation of Flag-DEK (Fig. 2B) and at the endogenous level (fig. S4A), the DEK(S287A/S288A) variant disrupted the DEK-SPOP interaction (Fig. 2B). Thus, the SPOP-binding motif within DEK appeared necessary for SPOP binding and destabilization.

Fig. 2 DEK is a SPOP substrate and SPOP mutants repress DEK ubiquitylation.

(A) Effects of stable transduced wild-type SPOP overexpression on protein levels of Flag-DEK and corresponding S287A/S288A mutant were assessed by immunoblotting in LHMAR cells. Error bars represent means ± SD, n = 4 (n.s., not significant; P < 0.05, ∗∗P < 0.01, Student's t test). (B) Flag-immunoprecipitation of Flag-DEK and corresponding S287A/S288A mutant in LHMAR cells with forced SPOP expression. (C) In vitro ubiquitylation of DEK. Flag-DEK was purified by Flag-specific antibody immunoprecipitation from LHMAR cells and incubated with the indicated components. Flag-DEK immunoprecipitates were analyzed by immunoblotting (see also fig. S4B). (D) In vivo ubiquitylation of DEK. Human embryonic kidney 293T cells were transfected with ubiquitin (Ub) tagged with eight histidines (8xHis-ubiquitin) and indicated constructs, with or without MG132, and lysed. Immunoblot of protein lysates and 8×His-ubiquitin pull down using nickel beads. (E) In vitro ubiquitylation as in (C) using indicated recombinant SPOP components in equimolar ratios. See also fig. S5A. (F) In vitro ubiquitylation as in (E) using indicated recombinant SPOP components. Dimerization-deficient mutants of SPOP-F133L: BTB domain (L186D/L190D/L193D/I217K = BTB-dd) and BACK domain (Y353E = BACK-dd). See also fig. S6.

Next, we tested whether SPOP could ubiquitylate DEK as part of a larger CUL3-RBX1 complex (8). In vitro, the addition of CUL3 and RBX1 caused wild-type DEK to migrate as high-molecular-weight species indicative of a multiple monoubiquitylation pattern (Fig. 2C and fig. S4B). In contrast, DEK(S287A/288A) failed to undergo efficient ubiquitylation (Fig. 2C). Addition of an ubiquitin moiety that is unable to form polyubiquitin chains because of lysine-to-arginine substitutions produced a similar pattern (K0) (fig. S4C), which supported the idea that SPOP may promote “multimonoubiquitylation” of DEK. This result aligns well with the observation that multiple ubiquitylated DEK lysine residues were detected in the initial proteome analysis (fig. S2E).

To determine whether DEK ubiquitylation targets this protein for proteasomal degradation, we cultured SPOP- and DEK-expressing cells in the presence or absence of the proteasome inhibitor MG132. Short-term (4- to 5-hour) MG132 treatment markedly increased ubiquitylated Flag-DEK, with a smaller effect on total protein expression (Fig. 2D and fig. S4D). Prolonged proteasomal inhibition (16 hours), or depletion of SPOP with two independent lentiviral short hairpin RNAs (shRNAs), also increased DEK protein levels (fig. S4E). Moreover, SPOP knockdown increased DEK stabilization after doxycycline was removed by using the inducible system described above, without affecting DEK mRNA levels (fig. S4, F and G). In aggregate, these data are consistent with a model in which multimonoubiquitylation of DEK promotes its proteasomal degradation, in line with recent studies showing that multiple monoubiquitylation can be sufficient for proteasomal targeting (15). The lower affinity of monoubiquitin moieties (as opposed to ubiquitin chains) for the proteasome may explain why DEK protein turnover is relatively slow (Fig. 1, D and E, and fig. S4F) (15).

We next asked whether prostate cancer SPOP mutations might disrupt the interaction between SPOP and DEK. Neither SPOP-F133L nor SPOP-Y87N could ubiquitylate DEK effectively in vitro (Fig. 2E). Moreover, both of these SPOP mutants suppressed ubiquitylation induced by SPOP-WT (Fig. 2E) and decreased basal DEK ubiquitylation in cell culture (fig. S5A). In prostate epithelial cells, overexpression of mutant SPOP resulted in increased total SPOP and DEK protein expression, as expected. However, both SPOP-F133L and SPOP-Y87N substantially reduced the interaction of wild-type SPOP with Flag-DEK in immunoprecipitation assays (fig. S5B). These data thus support the notion that prostate cancer SPOP mutants may impair DEK ubiquitylation in a dominant-negative manner.

Suppression of ubiquitylation by mutant SPOP may conceivably occur through heteromeric (e.g., WT-mutant) SPOP complexes. Along these lines, SPOP is known to undergo dimerization and multimerization through its BTB and BACK domains (5, 16, 17). In support of this model, Flag-tagged SPOP-F133L coimmunoprecipitated with hemagglutinin (HA)–tagged SPOP-WT (fig. S6A). To determine whether dimerization might mediate the suppressive function of SPOP-F133L, we introduced dimerization-deficient mutations involving the BTB (L186D/L190D/L193D/I217K; BTB-dd) or BACK (Y353E; BACK-dd) domains into SPOP-F133L (5, 17). Both dimerization-deficient mutants abolished formation of high-molecular-weight SPOP complexes and retained CUL3 binding, as evidenced by size-exclusion chromatography (fig. S6B). Similarly, HA-tagged versions of these SPOP mutants showed appropriate subcellular localization, and coimmunoprecipitated with MYC-tagged CUL3 (fig. S6, C and D). However, both the BTB and BACK domain mutations reversed the suppressive effect of SPOP-F133L mutant on DEK ubiquitylation in vitro (Fig. 2F) and in vivo (fig. S6E). In particular, the dimerization-deficient mutations impaired the interaction of wild-type SPOP with Flag-DEK, on the basis of immunoprecipitation experiments (fig. S6F). In aggregate, these data support the notion that prostate cancer SPOP mutations exert a dominant-negative effect on wild-type SPOP activity through formation of heteromeric complexes.

Next, we assessed whether DEK might contribute to oncogenic phenotypes induced by mutant SPOP in prostate cancer. Toward this end, both DEK and SPOP have previously been implicated in cell invasion (1, 18). We found that SPOP-F133L and SPOP-Y87N induced collagen invasion by prostate epithelial cells, whereas wild-type SPOP suppressed invasion (Fig. 3A). Similar effects were observed in the human model of prostatic cancer (LNCaP) cells (fig. S7A). To assess the contribution of DEK to this phenotype, we overexpressed DEK in cells expressing SPOP-WT or an empty vector control, and analyzed the resulting invasion phenotypes. In both settings, DEK overexpression promoted cellular invasion (Fig. 3B). Similarly, overexpression of Flag-DEK-S287A/S288A—the mutant form of DEK that cannot bind SPOP—also increased invasion, which implies that this phenotype is not simply a consequence of trapped SPOP-DEK complexes (fig. S7B). Conversely, shRNA knockdown of DEK decreased cellular invasion in cells overexpressing SPOP-87N (Fig. 3B) but the effect on control cells was minimal (fig. S7C). These results suggest that DEK may influence the invasive phenotype conferred by prostate cancer SPOP mutations, although a SPOP-independent role for DEK in cellular invasion cannot be ruled out.

Fig. 3 SPOP mutant-related changes in DEK expression contribute to invasion and sphere formation.

(A) Effects of different SPOP species on collagen invasion and DEK expression in LHMAR cells. FBS, fetal bovine serum. (B) Collagen invasion changes related to forced expression and depletion of DEK with two shRNAs in LHMAR cells. See also fig. S7. (C). Effects of different SPOP species and DEK on sphere formation in primary prostate epithelial cells (hPREC). (D) Effects of DEK depletion by three shRNAs on sphere formation. Error bars represent means ± SD, n = 3 in triplicate (n.s., not significant; ∗∗P < 0.01, Student's t test). See also fig. S7D.

DEK has also been implicated in the maintenance of stem cell–like properties, such as sphere formation under nonadherent culture conditions (14, 1820). Consistent with these observations, SPOP-Y87N produced elevated DEK levels and enhanced sphere formation in primary human prostate epithelial cells (hPREC) (Fig. 3C and fig. S7D). In contrast, overexpression of SPOP-WT suppressed sphere formation in this setting (Fig. 3C and fig. S7D). Overexpression of DEK in hPREC cells also augmented sphere formation, whereas shRNA depletion of DEK (with three independent shRNAs) in cells overexpressing SPOP-87N impaired sphere-forming capacity (Fig. 3D and fig. S7D). Taken together, these data suggest that SPOP-dependent DEK regulation may promote a stemlike phenotype.

We then investigated whether the effects of SPOP mutations on DEK regulation might be generalizable across a broad spectrum of cancer-associated SPOP mutations. Here, we tested SPOP mutants from both prostate and endometrial cancer for their effects on DEK protein levels, collagen invasion, and sphere formation. All prostate cancer–associated SPOP mutants examined [representing >80% of mutations identified in this tumor type (1)] conferred increased DEK protein expression, whereas two recurrent endometrial cancer–associated SPOP mutants (S80R and E50K) (2, 3) had no effect (Fig. 4A and figs. S8A, S9A, and S10A) in prostate cells. As expected, DEK expression correlated with sphere-forming capacity and cell invasion without affecting baseline cell proliferation or viability (figs. S8, B and C, and S9B).

Fig. 4 Up-regulation of DEK, TRIM24, and NCOA3 is a feature of prostate cancer SPOP mutations.

(A) Effects of different SPOP species on abundance of indicated proteins in unmodified primary prostate epithelial cells (hPREC) by immunoblotting. Protein changes relative to individual controls depicted as median in a heat map (n = 3). (B) Representative images of primary prostate cancer tissues stained for DEK, TRIM24, and NCOA3 by immunohistochemistry. Scale bar, 30 μm. (C) Corresponding expression analysis on 181/178 primary tumors stratified according to their SPOP mutations status (P values, Kendall beta-tau). See also figs. S8 to S10.

Finally, we considered whether other putative SPOP effector proteins—both known substrates and novel ones nominated by our ubiquitylome screen—might also undergo dysregulated expression in the setting of prostate cancer SPOP mutations (8, 9, 2123). Regarding the latter, we noted that TRIM24 represented another putative SPOP substrate from our screen with potential biological relevance to prostate cancer (Fig. 1, B and C, and fig. S3C) (10, 11). As expected several known or putative SPOP substrates were down-regulated in response to SPOP-WT overexpression (including DEK, TRIM24, NCOA3, DAXX, BRMS1, and GLI1 and 2) (Fig. 4A and fig. S10A). However, only DEK, TRIM24, and NCOA3 also became up-regulated upon overexpression of prostate cancer SPOP mutants in this setting (6). Moreover, nuclear expression of these proteins correlated significantly with SPOP mutations in 181/178 primary prostate cancers analyzed by immunohistochemistry (Fig. 4, B and C, and fig. S10B, table S3). These data are consistent with the ubiquitylome and experimental studies, which supports a model in which SPOP mutations may engage overlapping routes of transformation in prostate cancer by dysregulation of DEK, TRIM24, and NCOA3.

These data support a model wherein prostate cancer SPOP mutants impair ubiquitylation and degradation of a subset of SPOP substrates in a dominant-negative manner to promote tumorigenesis. Future studies of DEK dysregulation and its biological functions in normal and tumor tissue may therefore promote new biological insights relevant to many human cancers. More generally, large-scale profiling of the ubiquitin landscape linked to tumor genomic alterations in protein homeostasis genes may catalyze the identification of downstream effector mechanisms. Given the growing number of such genes identified by comprehensive cancer genome studies, this approach may provide an additional avenue for functional annotation of the cancer genome. In the future, these studies could uncover new biological and therapeutic avenues that exploit dysregulated protein homeostasis or ubiquitin-based regulation in cancer.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Tables S1 to S3

References (2430)


  1. Acknowledgments: We thank L. Gaffney for illustrations and A. Fitsche and M. Storz for excellent technical assistance. J.-P.P.T. is funded by a fellowship from the Swiss Science foundation (SSMBS PASMP3_145764). P.J.W. is funded by a grant (PhophoNet PPM), the Swiss initiative in systems biology. M.A.R. and Cornell University have filed a patent (WO 2012115789 A2) relating to the use of SPOP mutations as markers for human prostate cancer. L.A.G. is a paid consultant for the following pharmaceutical companies: Novartis Foundation Medicine, Boehringer Ingelheim, and Millennium/Takeda.
View Abstract

Navigate This Article