TGFβ-Activated USP27X Deubiquitinase Regulates Cell Migration and Chemoresistance via Stabilization of Snail1

The epithelial-to-mesenchymal transition (EMT) is a reversible and highly conserved biological process originally described during embryonic development that shifts epithelial cells toward a mesenchymal phenotype (1). EMT is observed at the invasive tumor front and correlates with enhanced invasiveness and poor clinical prognosis (2). Phenotypically, tumor cells that have undergone an EMT become spindle-shaped and motile and acquire new biological characteristics, such as a higher resistance to chemotherapy (3, 4).

EMT is defined by the downregulation of the epithelial junctional protein E-cadherin, which is preceded by the expression of Snail1 repressor that directly binds to the E-cadherin gene (CDH1) promoter inhibiting its transcription (3). Snail1 also plays an active role in the activation of mesenchymal genes, such as Fibronectin (3). Moreover, Snail1 expression enhances resistance to apoptosis and the survival of tumor cells in conditions of stress (3). For all these reasons, Snail1 is considered a key protein controlling EMT.

In accordance with its functions, Snail1 expression in human tumors is associated with tumor progression, enhanced invasiveness, and poor clinical prognosis (1). In pancreatic cancer, carcinoma cells utilize an EMT program for metastatic dissemination (5). In breast tumors, expression of Snail1 and other mesenchymal markers associates with poor clinical parameters and to the particularly aggressive triple-negative breast cancer (TNBC) subtype with a high rate of recurrence and distant metastasis (6). This suggests that Snail1 is a possible target for therapeutic intervention.

Snail1 is a very unstable protein, tightly regulated by ubiquitination by several E3-ubiquitin ligases that catalyze the transfer of ubiquitin (Ub) from the E2-conjugating enzyme to a lysine residue in the substrate protein and the posterior elongation of the Ub chains (7). Snail1 is targeted to the proteasome by several Skp-Culin-Fbox (SCF) E3 ligases (SCF-Fbxw1, SCF-Fbxl14, SCF-Fbxl5, SCF-FbxO11, and SCF-FbxO45) that limit Snail1 expression in epithelial cells (8, 9). Different reports have demonstrated the stabilization of Snail1 with the concomitant downregulation of Snail1 E3 ligases in response to hypoxia, several proapoptotic insults such as γ-irradiation or antineoplastic drugs (10–14).

Ubiquitination is a reversible process, and the action of E3 ligases is antagonized by deubiquitinating enzymes (DUB) that recognize and remove Ub molecules (15). The 98 DUBs described so far are classified as cysteine proteases (Ub-specific proteases (USP), Ub carboxy-terminal hydrolases, ovarian tumor-like proteases (OTU) or "Machado-Joseph disease" proteases (MJD) and zinc metalloenzymes (the JAMM/MPN class). The role of DUBs in cancer is stressed by the low stability exhibited by many oncoproteins and the deregulation of their DUBs observed in tumors (16).

Despite the high number of E3 described targeting Snail1, only one DUB, DUB3 (USP17L2), has been recently shown to deubiquitinate Snail1 (17, 18). In an effort to identify new DUBs acting on Snail1, we have performed an siRNA screening on Snail1 expression in the MDA-MB-231 mammary cancer cell line and identified USP27X as a DUB for Snail1.

Authenticated cell lines were obtained from the European Collection of Authenticated Cell Cultures (ECACC) or the ATCC and supplied by the Cancer Cell Line Repository from IMIM. All cell lines were used for no more than 20 passages and routinely tested for Mycoplasma contamination by PCR (19). The generation of murine embryonic fibroblasts (MEF) and cancer-associated fibroblasts (CAF) from MMTV-PyMT tumors has been previously reported (20). Cells were maintained in DMEM (GIBCO-BRL) with 10% heat-inactivated FBS (GIBCO-BRL) at 37°C in 5% CO₂. Where indicated, cells were treated with 5 ng/mL of TGFβ1 (PeproTech), 10 to 100 μmol/L of cisplatin (Calbiochem), 20 μmol/L of the DUB inhibitor PR619 (LifeSensors), 20 μg/mL of cycloheximide (Sigma-Aldrich), or 1 μg/mL of doxycycline (Sigma-Aldrich). The antibodies used in this work were mouse monoclonal antibodies against Snail1 (21) and Myc (clone 9E10; a gift from Dr. Gabriel Gil, IMIM, Barcelona); rabbit anti-fibronectin (Dako); goat anti-lamin B (C-20), goat anti-Snail2 (Slug; both from Santa Cruz); rat anti-HA tag (Roche); rabbit anti-HA tag, rabbit anti-flag, rabbit anti-Myc, mouse anti-tubulin (all from Sigma-Aldrich); rabbit anti-GFP, rabbit anti-H3 (Abcam); goat anti-GST (GE Healthcare); rabbit anti-pSmad2, rabbit anti-Snail1 (Cell Signaling) and rabbit anti-USP27X (human and murine; ref. 22).

pLex-Snail-F-Luc-Puro (lentiviral) and pMSCV-pLuc-Hygro (retroviral) constructs were kindly provided by Dr. Kang lab and were previously described (23) and used to generate viral stocks. For the doxycycline inducible experiments, we used the lentiviral pF-CMV-TetR-PGK-Hygro; pF-CMV-USP27X-Puro-3xFLAG and pF-CMV-USP27XC/A-Puro-3XFLAG vectors, a gift from Dr. A. Weber (24). For in vivo imaging studies, we used the retroviral pLHCX-LUC plasmid, a gift from Dr. C. Fillat. Retroviral pBABE-Snail1-HA was previously described (12). Retroviral pBABE-Puro-hUSP27X-C-V5-FL-WT was obtained after BamHI/SalI subcloning from the corresponding pBABE-Zeo vector (22). In general, lentivirus and retroviruses were produced by transient transfection of HEK293T or 293T Phoenix gag-pol cells, respectively, as previously described (12, 13) and in Supplementary Materials and Methods. To generate MDA-MB-231–expressing Snail-F-Luc and R-Luc, cells were first transduced with the lentivirus for Snail1-F-Luc and selected with puromycin (1 μg/mL) and subsequently infected with the Renilla-Luciferase (R-Luc) retrovirus and selected with hygromycin B (200 μg/mL). To generate MDA-MB-231 expressing USP27X induced by doxycycline, cells were first infected with the TetR lentivirus, selected with hygromycin B, and subsequently transduced with the lentiviral expression system following selection with puromycin, as previously described (24). For the Snail1 rescue experiments, MDA-MB-231 USP27X KO cells were infected with the Snail1-HA retrovirus as described previously (12).

The siRNA screening was performed according to the manufacturer's instructions (Dharmacon). Specific experimental conditions may be found in Supplementary Materials and Methods.

For stable USP27X and CEZ2 gene silencing, the next MISSION shRNA plasmids (Sigma-Aldrich) were used to produce lentiviral particles: USP27X (#1: TRCN0000336909; #2: TRCN0000336851; #3: TRCN0000336912); OTUD7A/CEZANNE2 (#1: TRCN0000425854; #2: TRCN0000050160; #3: TRCN0000423073). After transduction, stable cell lines expressing the shRNA were isolated by puromycin (2.5 μg/mL) selection. For DUBs or Snail1 depletion, cells were transfected using DharmaFECT I (Dharmacon) with the specific synthetic siRNAs for: SNAI1; USP27X (human and murine); OTUD7A; VCPIP1; MPND; YOD1; USP15; TRABID; USPL1 (all from Dharmacon) during 48 and 72 hours for RNA and protein analysis, respectively. ON-TARGETplus Nontargeting siRNAs #2 and #3 (Dharmacon) were used as controls.

For in vivo xenografting experiments, 5 × 10⁵ cells were diluted in cold 100 μL 1:1 PBS-Matrigel (growth factor reduced; #354230, Corning), and subcutaneously inoculated in each flank of 6-week-old NOD/SCID mice. After 30 days, mice were euthanized and tumors were photographed and measured. Volumes (V) were determined by using the formula V = (width)² × length/2. Tail-vein injection experiments were performed with 1 × 10⁶ cells/100 μL and immediately analyzed for in vivo bioluminescent determination by using the IVIS 50 Imaging System (Xenogen Imaging Technologies) and the Living Image software 2.60.2. Bioluminescent quantification was performed after intraperitoneal inoculation of 100 μL of luciferin (PerkinElmer) solution (25 mg/mL). All experiments with mice were previously approved by the PRBB Committee for Ethical Animal Research (CEEA AGH-16-0034-P2 and AGH-16-0034-P3) following the EU Directive 2010/63/EU.

For USP27X knockout (KO) generation, the CRISPR/Cas9 system for human (MDA-MB-231) and murine (NMuMG and MEFs) cells was transfected with the human (sc-407628) and murine (sc-424918) USP27X CRISPR/Cas9 Knockout plasmids (Santa Cruz), respectively, both consisting of a pool of three guide RNA (gRNA) plasmids and codifying for GFP. Transfected cells were selected with the BD Influx Cell Sorter (Becton and Dickinson) and grown as single clones during 2 to 3 weeks. Positive KO clones were identified by Western blot.

Other methods are described in the Supplementary Information.

The fusion protein Snail1-Firefly-Luciferase (Snail-F-Luc) has been used to identify E3-ligases controlling Snail1 stability (23). Analysis of F-Luc activity (and Renilla-Luc, R-Luc, as control) allows a reliable measure of the stability of the Snail-F-Luc fusion protein as demonstrated after treatment with the proteasome inhibitor MG132 or with the protein synthesis inhibitor cycloheximide (CHX; Supplementary Fig. S1A and S1B). The rapid and concomitant regulation of Snail-F-Luc protein and activity by these compounds demonstrate the low stability of Snail1. Interestingly, Snail1-F-Luc levels were sensitive to DUB inhibition by PR619, a cell-permeable and nonselective DUB inhibitor (Supplementary Fig. S1B and S1C; ref. 25). This PR619 effect suggests the participation of one or more DUBs in the control of Snail1 stability.

We generated MDA-MB-231 cells stably expressing Snail-F-Luc and R-Luc and screened MDA-MB-231-Snail1-F-Luc/R-Luc cells with a library of siRNAs targeting all known DUBs (Human ON-TARGETplus; Dharmacon). Four different siRNAs were used for each target and F-Luc/R-Luc ratio was determined after 72 hours of transfection. Five different control siRNAs (CTL#1 to #4 and a pool mix) were used as nontargeting controls to define the basal F-Luc/R-Luc reference value (Fig. 1A, red dotted line). A SNAI1 siRNA was also included to determine the minimal luciferase activity (Fig. 1A, green squares). MG132 was used as a positive control of stabilization (Fig. 1A, yellow dots). In contrast, PR619 strongly decreased F-Luc activity (Fig. 1A, blue crosses). Following the experimental workflow summarized in Supplementary Fig. S1D, two independent rounds of screening were performed to calculate the F-Luc/R-Luc ratio. Inhibition of 17 DUBs caused a 2-fold decrease of the F-Luc/R-Luc ratio, among them, the previously described Snail1 DUB, DUB3 (Fig. 1A). To narrow down the list, we arbitrarily increased the threshold to 2.5 (Fig. 1A, black dotted line); seven DUBs were selected that decreased at least 2.5-fold the F-Luc/R-Luc ratio and further analyzed to discard false positives (Supplementary Fig. S1D). Four DUBs were validated analyzing the effect of the selected DUBs siRNAs on endogenous Snail1 protein: MPND, VCPIP1, USP27X, and OTUD7A/CEZANNE2 (Supplementary Fig. S2A). A quantitative RT-PCR (qRT-PCR) analysis revealed that the efficiency of DUB depletion by the specific siRNAs was higher than 80% and off-side effects on other DUBs were minimal (Supplementary Fig. S2B). SNAI1 mRNA levels were not affected by USP27X, MPND, or VCPIP1 siRNAs; only OTUD7A/CEZANNE2 (CEZ2) siRNA decreased it (Supplementary Fig. S2B).

We tested the capability of the selected DUBs to bind Snail1 after cotransfection and immunoprecipitation; only USP27X clearly interacted with Snail1 (Fig. 1B, right). Interestingly, in the presence of MG132, a faint interaction was also observed with Cezanne2 and MPND (Supplementary Fig. S2C). Ectopic Snail1 was stabilized after cotransfection of USP27X or CEZ2 to a much higher extent than by MPND or VCPIP1 (Fig. 1B, left). The interaction between USP27X and Snail1 was also detected with endogenous proteins in MDA-MB-231 cells (Fig. 1C). Transfection of USP27X or CEZ2 siRNAs to these cells downregulated endogenous Snail1 protein (Fig. 1D). A final confirmation was obtained after stable infection of three independent shRNAs against USP27X (Fig. 1E) or CEZ2 (Supplementary Fig. S2D); both of them also decreased the Snail1-downstream target Fibronectin. Therefore, because USP27X consistently interacts with Snail1 and its inhibition decreases Snail1 protein but not mRNA, we chose it as a bona fide Snail1 DUB.

We studied the expression of USP27X and Snail1 in a panel of mammary and pancreatic cancer cell lines, two cell systems where Snail1 associates with malignancy (6, 26). In general, both proteins showed a significant correlation in pancreas and breast cancer cells, although some cells exhibit high Snail1 even with low USP27X (Supplementary Fig. S2E). Inhibition of USP27X in the highly expressing pancreatic RWP-1 and PANC-1 cell lines, or in breast BT-474 cells caused a strong decrease in Snail1 (Fig. 1F), reproducing the results obtained in MDA-MB-231 (Fig. 1D and E). Inversely, ectopic expression of USP27X in SKBR3 cells upregulates the levels of Snail1 and the downstream targets Fibronectin and N-Cadherin (Fig. 1G).

The USP27X gene is predicted to encode for a 49.6 kDa protein (Uniprot: A6NNY8; refs. 24, 27). This USP27X form interacts with Snail1 and increases the levels of ectopic Snail1 as shown in Fig. 1B. However, the recent generation of specific antibodies against USP27X has demonstrated that the most abundant form of this protein presents a molecular weight of 72 kDa (22). This form is originated through translation using an upstream CTG codon and results in an additional N-terminal extension of 198 amino acids (22). This 72 kDa form was the most abundant detected in MDA-MB-231 cells and was specifically depleted by siUSP27X and shUSP27X (Fig. 1D and E). Overexpression of this 72 KDa form also upregulated Snail1 as shown in Fig. 1G. Pulldown experiments indicate that both the full-length USP27X [FL-USP27X; amino acids (aa) 1-636] and the short form (S-USP27X; aa 198–636) bind to GST-Snail1 (Supplementary Fig. S3A and S3B), suggesting that the N-terminal part of the 72 kDa protein is not required for Snail1 binding and stability. Other data demonstrate that the central region of USP27X (438–561) is involved in the interaction with Snail1 (Supplementary Fig. S3A and S3B).

USP27X–Snail1 association was verified by reverse experiments using recombinant GST-USP27X, which demonstrate that Snail1 and USP27X physically interact and that the binding region is mainly located at the Snail1 C-terminal domain, involving Zn fingers 2 and 3 (Supplementary Fig. S3C and S3D, summarized in S3E). Finally, the direct interaction was also verified using purified recombinant Snail1 and USP27X proteins (Supplementary Fig. S3F). Similar assays were also carried out with Snail2 that did not interact with USP27X.

USP27X shares a high degree of homology with USP22 (82%) and USP51 (70%), and these three proteins constitute a phylogenetically related subfamily (28). Like USP27X, overexpression of USP51, and USP22 to a lower extent upregulated ectopic Snail1 expression (Supplementary Fig. S4A, left). Surprisingly, coimmunoprecipitation experiments indicated that Snail1 interacted with USP27X and USP22 but not with USP51 (Supplementary Fig. S4A, right). Accordingly, GST-Snail1 pulled-down USP22 but not USP51 or any of the other DUBs tested (OTUD7A/CEZANNE2, MPND, VCPIP1; Supplementary Fig. S4B). All these results indicate that USP27X and the close homologue USP22 specifically interact and stabilize Snail1.

Next, we studied the capability of USP27X to stabilize Snail1 in MDA-MB-231 cells. We used the USP27X form starting from the predicted ATG either wild-type (WT) or an inactive mutant where Cys87, present in the catalytic center, is replaced by an Ala (USP27X C/A; refs. 24, 27). USP27X WT but not the inactive mutant greatly enhanced Snail1 endogenous expression (about 6-fold) without increasing SNAI1 mRNA (Fig. 2A and B). Snail2 was not significantly modified. Analysis by immunofluorescence revealed that both USP27X and Snail1 colocalized in the nucleus, although USP27X was also observed in the cytoplasm (Fig. 2C). Similar results were obtained transfecting the 72-kDa USP27X long form, which also caused a strong stabilization of ectopic Snail1 in all the subcellular compartments analyzed (cytosol, nucleoplasm, and chromatin; Supplementary Fig. S5A). Inversely, a USP27X siRNA caused a strong decrease in Snail1 in its main subcellular localization, the nucleus (Supplementary Fig. S5B and S5C). Curiously, the two forms of endogenous USP27X present a different localization being the short form more abundant in the cytosol, whereas the long form in the nucleus (Supplementary Fig. S5B and S5C). To determine more precisely whether Snail1 is stabilized in the nucleus or the cytosol, we evaluated the action of USP27X on Snail1-SD, a Snail1 mutant highly unstable due to its preferential localization in the cytoplasm and on Snail1-SA or Snail1-LA, mutants that are mainly nuclear (29). As shown in Supplementary Fig. S5D and S5E, the three mutants interacted with USP27X; however, USP27X strongly stabilized Snail1-SD, partially Snail1-LA, and minimally affected Snail1-SA. The lower effect on the nuclear versions of Snail1 suggests that USP27X stabilization is more evident with cytosolic Snail1, which is the one prone to be degraded. However, we cannot rule out the fact that USP27X might target both nuclear and cytosolic Snail1, as suggested by its localization in both compartments.

We also generated MDA-MB-231 cells stably expressing USP27X WT or the C/A mutant under the control of a doxycycline (Doxy) inducible promoter (24). Upon induction of USP27X WT, Snail1 endogenous levels increased gradually until 7-fold; in contrast, the USP27X C/A mutant did not upregulate Snail1 (Fig. 2D). Snail2 expression was only slightly affected by USP27X induction (Supplementary Fig. S6A). Snail1 relative half-life was increased 3-fold by USP27X WT compared with cells transfected with the empty vector or with the USP27X C/A-inactive mutant (Fig. 2E). The lack of Snail1 stabilization with the mutant USP27X is not due to an impaired interaction with Snail1 because both USP27X forms bound similarly (Supplementary Fig. S6B). Ectopic expression of the USP27X increased Snail2 protein stability (Supplementary Fig. S6C) but the effect was much less pronounced than that on Snail1 (Supplementary Fig. S6D).

To determine if Snail1 stabilization is associated with USP27X deubiquitinating activity, we coexpressed Snail1 with ubiquitin and the WT or C/A USP27X forms. After purification of ubiquitinated proteins, Snail1 was detected as a mix of mono- and polyubiquitin forms of high molecular weight (Fig. 2F). Polyubiquitinated Snail1 was strongly decreased upon ectopic expression of WT USP27X (Fig. 2F, left); however, monoubiquitinated Snail1 was relatively resistant to deubiquitination (Fig. 2F). USP27X but not the C/A mutant counteracted the ubiquitination and also the degradation triggered by the β-TrCP1 and FBXL14 ubiquitin ligases (Fig. 2F and G). All these results indicate that USP27X deubiquitinates and stabilizes Snail1.

Because DUB3 has been recently found as a DUB for Snail1/2 (17, 18), we compared this DUB with USP27X. According to our screening data (Fig. 1A), in MDA-MB-231 cells depletion of USP27X was more effective than that of DUB3 in decreasing Snail1 (Supplementary Fig. S6E), whereas USP27X ectopic expression was also more potent in promoting Snail1 stabilization (Supplementary Fig. S6F, compare with Fig. 2E). Both DUBs similarly interacted with Snail1 (Supplementary Fig. S6G).

We generated MDA-MB-231 cells depleted of USP27X using the CRISPR/Cas9 technology (Fig. 3A). Two independent USP27X knockout (KO) cell clones (named #1 and #2) showed undetectable USP27X (Fig. 3A); in these cells, the levels of Snail1 protein were remarkably downregulated with respect to the control (Fig. 3A) without significant changes in SNAI1 mRNA (Supplementary Fig. S7A). Other independent clones (KO #3 and KO #4) caused a similar Snail1 protein downregulation (Supplementary Fig. S7B). USP27X KO cells showed a lower growth rate than controls (Fig. 3B), in accordance with the previously reported role of USP27X regulating proliferation (22). The clonogenic capability (Fig. 3C; Supplementary Fig. S7C) was also severely decreased in USP27X KO cells. Elimination of USP27X also retarded the growth of MDA-MB-231 cells when subcutaneously xenografted in NOD/SCID mice (Fig. 3D; Supplementary Fig. S7D). USP27X KO tumors do not express Snail1 and very low levels of fibronectin compared with WT tumors (Supplementary Fig. S7E). Snail1 was infected in USP27X KO cells to obtain levels similar to those detected in control cells (see below); however, Snai1l overexpression did not rescue cell growth (Fig. 3B), colony formation (Fig. 3C), or in vivo tumor growth (Fig. 3D), suggesting that USP27X regulates tumor growth by acting on other substrates distinct of Snail1.

We also analyzed in MDA-MB-231 USP27X KO cells other characteristics dependent on Snail1 expression. USP27X KO cells or cells transfected with a USP27X siRNA exhibited impaired cell migration or cell invasion through Matrigel or Collagen I–coated Transwells (Fig. 4A and B); this inhibition was similar to that observed with an SNAI1 siRNA (Fig. 4A). Importantly, Snail1 overexpression rescued cell invasion in USP27X KO cells (Fig. 4B). The metastatic potential was also assessed after intravenous inoculation of 1 × 10⁶ cells of WT, USP27X KO, and USP27X KO-Snail1 cells expressing luciferase into NOD/SCID mice. In vivo bioluminescent imaging showed that WT cells consistently originated lung metastasis after 21 days (when control mice were euthanized to avoid animal suffering); in contrast, USP27X KO cells showed only residual luciferase activity in the lung till the end of the experiment (day 40; Fig. 4C). Snail1 ectopic expression in USP27X KO cells partially rescued metastasis formation after 3 weeks, being luciferase activity highly increased after 5 weeks, although USP27X KO-Snail1 cells never reached the levels of WT (Fig. 4C and D). Ex vivo quantification of luminescence gave similar results (Fig. 4E). Postmortem histologic analysis of lungs revealed a high number of metastatic foci generated by WT cells and USP27X KO-Snail1 cells, which were absent for USP27X KO cells (Fig. 4F). These results clearly showed that USP27X was critical for extravasation and/or colonization of the lung by breast tumor cells and that this effect was significantly dependent on Snail1 expression.

Snail1 is upregulated by chemotherapeutic agents and increases resistance to cell death triggered by these compounds (30). Western blot analysis showed a clear increase of Snail1 in WT cells upon addition of cisplatin (Fig. 5A), consistent with an SNAI1 mRNA upregulation (Fig. 5B). USP27X remained constant. In contrast, Snail1 protein was not upregulated by cisplatin in USP27X KO cells (Fig. 5A), despite SNAI1 mRNA was raised similarly to WT cells (Fig. 5B). These results suggest that Snail1 stabilization after treatment with this drug requires the participation of USP27X. We also determined the effect of USP27X expression on the viability of MDA-MB-231 cells in the presence of cisplatin. Exposure to the drug was limited to 48 hours, a time where the differences in cell growth between WT and USP27X KO cells were not significant yet (see Fig. 3B). USP27X depletion strongly affected cisplatin sensitivity because KO cells showed a much lower IC₅₀ (9 μmol/L) than WT cells (30 μmol/L; Fig. 5C). Importantly, Snail1 reexpression in USP27X KO cells rescued cell survival to a similar extent than that observed in WT cells (Fig. 5C). According with this, the quantification of apoptotic cells (early and late apoptosis) after cisplatin treatment showed a higher number in USP27X KO cells than in WT cells (Fig. 5D). Snail1 reexpression in USP27X KO cells significantly decreased the number of apoptotic cells after cisplatin addition (Fig. 5D). Overall, these results show that USP27X depletion enhances sensitivity to cisplatin and apoptosis in a Snail1-dependent fashion.

The cytokine TGFβ induces an EMT (1) in the normal murine mammary NMuMG cells initiated by the rapid increase in Snail1 expression (31, 32). Interestingly, the expression of USP27X was also strongly upregulated by TGFβ both at mRNA and protein levels (Fig. 6A and B). Snail1 protein levels were not stimulated by TGFβ in USP27X KO cells (Fig. 6A). Accordingly, the Snail1 downstream target fibronectin was upregulated at 24 hours of treatment in WT but not in USP27X KO cells (Fig. 6A). The inability of TGFβ to promote an EMT in USP27X KO cells was not due to a general inhibition of TGFβ signaling because phospho-Smad2 was increased in both WT and KO cells (Fig. 6A). In accordance with the induction of Snail1 and fibronectin by TGFβ, this cytokine significantly boosted migration and invasion through Collagen I (4-fold and 6-fold, respectively, in TGFβ treated vs. control cells); this stimulation was completely abrogated in USP27X KO cells (Fig. 6C). The absence of USP27X also affected other genes modulated by TGFβ downstream of Snail1: Cdh1 RNA was much more rapidly repressed in WT cells upon TGFβ treatment than in USP27X KO cells, whereas the mesenchymal marker Zeb1 showed a much higher rise in WT cells than in USP27X KO cells (Fig. 6D). Snai1 mRNA increased in a time-dependent fashion in both WT and KO cells (Fig. 6D), reinforcing the idea that KO cells respond to TGFβ similar to WT cells. In fact, the upregulation of Snai1 mRNA by TGFβ was even higher in USP27X KO than in WT cells, suggesting the existence of a Snail1 self-regulatory inhibition (33). Confocal immunofluorescence analysis also showed that USP27X, Snail1, and fibronectin were upregulated by TGFβ in control but not in USP27X KO cells (Fig. 6E). Both USP27X and Snail1 were preferentially nuclear in TGFβ-treated cells, although they also showed a partial cytoplasmic localization (Fig. 6E). Finally, ectopic expression of USP27X, either the 50 kDa or 72 kDa forms in NMuMG cells recapitulated the EMT program triggered by TGFβ because it increased Snail1, fibronectin, and N-cadherin expression (Fig. 6F and G).

The TGFβ/Snail1 axis is also required for fibroblast activation (20). Similarly to that observed in epithelial cells, TGFβ also increases USP27X and Snail1 in control MEFs and in mCAFs obtained from PyMT-MMTV breast tumors (Figs. 7A and B) but not in MEFs KO for USP27X or mCAFs after silencing of USP72X by siRNA. In both cell systems, USP27X depletion also affected the upregulation of fibronectin (Fig. 7A and B) and cell invasion (Fig. 7C and D) by TGFβ. Altogether, these results indicate that Snail1 protein stabilization by USP27X is a key event to develop a full EMT and fibroblast activation in response to TGFβ.

A transcriptomic analysis of USP27X in a public database showed a significant increase in patients with breast, pancreatic, and colorectal cancer (Supplementary Fig. S8). Because of the high expression of USP27X in breast cancer cell lines, we also analyzed the expression in a panel of 27 patient-derived xenografted (PDX) human breast tumors corresponding to HER2⁺ and triple-negative breast neoplasms (TNBC). PDXs recapitulate and preserve most of the traits from the patient sample at the transcriptomic, epigenomic, and histologic levels (34). Western blot (Fig. 7E) and densitometric analysis (Fig. 7F) showed a significant correlation between Snail1 and USP27X in TNBC but not in HER2⁺ tumors. These results reinforce the hypothesis that Snail1 requires USP27X to be efficiently expressed in TNBC cells.

In this article, we have identified USP27X as a highly active Snail1 DUB using an siRNA functional screening. This approach has been shown to be a valuable tool to characterize new kinases and ubiquitin ligases controlling Snail1 protein stability (12, 23, 35). Besides USP27X, three other DUBs were identified in our screening—MPND, VCPIP1, and OTUD7A/CEZANNE2—that stabilized Snail1, although they did not associate with this factor in coimmunoprecipitation experiments. However, MPND and Cezanne2 slightly interact with Snail1 after proteasome inhibition (Supplementary Fig. S2C), suggesting that accumulation of polyubiquitinated Snail1 might favor the association with these particular DUBs. They might also regulate Snail1 indirectly, as indicated above or even controlling the stability of USP27X. Curiously, our screening also detected other DUBs depletion of which enhances Snail1-F-Luc/R-Luc (Fig. 1A). Probably, these DUBs might remove other ubiquitination events that have been associated with Snail1 stability, such as Snail1 K63 polyubiquitination mediated by Pellino E3 ligase (36) or A20-dependent Snail1 multimonoubiquitination (37).

Snai1l ubiquitination occurs in the cytoplasm by the highly active ubiquitin ligases β-TrCP1 (Fbxw1) and Fbxl14, or in the nucleus, by Fbxl5 or Fbxo11 (9). Wherever Snail1 ubiquitination takes place, inhibition of nuclear export stabilizes Snail1, suggesting that degradation happens in the cytosol (9). USP27X contains a nuclear localization signal (22) and target nuclear substrates such as histone H2B or the transcription factor Hes1 (22, 27). Accordingly, USP27X might act on both nuclear and cytosolic Snail1; however, our data suggest that Snail1 deubiquitination likely occurs in the cytoplasm. This hypothesis is supported by experiments showing a strong stabilization of Snail1-SD mutant that is mostly localized in the cytosol and by immunofluorescence analyses showing USP27X presence in both the nucleus and cytosol. USP27X action in the cytosol has also been demonstrated by other authors (24). However, we cannot discard the possibility that nuclear deubiquitination also contributes to Snail1 stability by blocking nuclear export and favoring Snail1 nuclear retention. Regardless of whether Snail1 deubiquitination happens in the nucleus or cytosol, in both cases it contributes to Snail1 stability.

The USP27X homologous protein USP22 also interacts and stabilizes Snail1 (see Supplementary Fig. S4). However, inhibition of USP22 was not found to significantly decrease Snail1-F-Luc/R-Luc in the screening. Moreover, its endogenous levels cannot compensate the lack of USP27X in KO cells in the different cell lines analyzed. This suggests that USP27X has a predominant role in Snail1 deubiquitination in vivo, although we cannot discard that USP22 may substitute USP27X and participate in Snail1 stabilization in other cells or upon other stimuli, because many pieces of evidence support the functional relationship between USP22 and USP27X on other substrates. USP22 was first identified as an H2B deubiquitinase being part of the SAGA complex (38, 39). Depletion of USP22 favors USP27X (and also USP51) deubiquitination of H2B independently of SAGA (22). USP27X or USP22 depletion enhances sensitivity to cisplatin (Fig. 5) or to 5-fluoruacil (40), respectively, suggesting a functional relationship. Moreover, USP22 is associated with EMT in colorectal (41), lung (42), and pancreatic tumors (43). This action of USP22 might be related to its activity on nonhistone substrates such as c-Myc and Hes1 (27, 44, 45). Further work is needed to clarify the putative role of USP22 as a possible Snail1 DUB.

We have observed a significant correlation between Snail1 and USP27X in PDX derived from TNBC tumors by Western blot. Unfortunately, no studies have so far analyzed the expression of USP27X in tumors due to the absence of good antibodies to detect this protein in paraffin-embedded samples. Specific studies directed to analyze USP27X in tumor samples are required to verify its relative contribution during cancer progression.

While this study was being performed, two groups reported that Snail1 is deubiquitinated by DUB3 (also known as USP17L2; refs. 17, 18). In our screening, DUB3 inhibition decreased Snail1-F-Luc/R-Luc by 2-fold, less than the threshold preestablished in our screening, and therefore this DUB was not initially selected. Anyway, we compared DUB3 and USP27X effects on Snail1: both DUBs stabilize and interact with Snail1, although USP27X seems to be more potent (Supplementary Fig. S6E–S6G). Moreover, USP27X and DUB3 present several differences. DUB3 is associated with the Snail1 N-terminal domain (18), whereas USP27X requires the C-terminal fragment. DUB3 also binds and stabilizes Snail2 (Slug) and Twist (46), whereas the action of USP27X on Snail2 is likely indirect because both proteins do not physically interact. Curiously, both DUB3 and USP27X proteases are also stimulated by cytokines; whereas USP27X is rapidly and potently stimulated by TGFβ, DUB3 is induced by IL6 (18), a cytokine also linked to EMT (47). It remains to be established whether Snail1 induction caused by IL6 or other cytokines in other cells require the expression of USP27X, DUB3, or a different DUB.

Snail1 transcriptional factor plays a key relevant action in several cellular traits related to tumor growth and progression because it promotes tumor invasion, resistance to classic antineoplastic agents, acquisition of cancer stem cell characteristics, and activation of cancer-activated fibroblasts (3, 20, 48). Therefore, Snail1 is an attractive target for chemotherapy. However, inhibitors of Snail1 function have not been characterized so far, a fact probably related to the lack of enzymatic activity of this protein. Snail1 is a very unstable protein, and its half-life is dependent on the coordinated action of several ubiquitin ligases and DUBs. Because developing E3s agonists is extremely challenging, an alternative and much more feasible possibility consists in generating DUB inhibitors that affect Snail1 expression and function. The characterization of USP27X as a very active Snail1 DUB provides new opportunities for the development of new molecules that specifically interfere with Snail1 expression and opens new lines for the treatment of cancer metastasis.

J. Arribas reports receiving other commercial research support from Hofman La Roche, Synthon, and Molecular Partners and is a consultant/advisory board member for Menarini Biotech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. García de Herreros, V.M. Díaz

Development of methodology: G. Lambies, C. Martínez-Guillamon, R. Olivera-Salguero, R. Peña, V.M. Díaz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Lambies, M. Miceli, C. Martínez-Guillamon, C.-P. Frías, I. Calderón, S.Y.R. Dent, J. Arribas

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Lambies, M. Miceli, C. Martínez-Guillamon, R. Olivera-Salguero, R. Peña, A. García de Herreros, V.M. Díaz

Writing, review, and/or revision of the manuscript: G. Lambies, S.Y.R. Dent, J. Arribas, A. García de Herreros, V.M. Díaz

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Peña, C.-P. Frías, I. Calderón

Study supervision: V.M. Díaz

Other (provided critical reagents including USP27X specific antibodies and expression vectors; discussed the results in the manuscript): B.S. Atanassov

We thank Drs. Yibin Kang (Princeton University, Princeton, New Jersey), Arnim Weber and Georg Häcker (University Medical Center Freiburg, Germany), Taeko Kobayashi (Institute for Virus Research, Kyoto University), Yadi Wu (The University of Kentucky, Lexington, Kentucky), and Cristina Fillat (IDIBAPS, Barcelona, Spain) for kindly sending plasmids and for advice. We really appreciate the technical support of Jordi Vergés and Cristina Bernardo. This study was funded by grants awarded by Ministerio de Economía y Competitividad (MINECO) and Fondo Europeo de Desarrollo Regional-FEDER to A. García de Herreros (SAF2013-48849-C2-1-R and SAF2016-76461-R) and to V.M. Díaz (SAF2013-48849-C2-2-R). Research at the A. García de Herreros lab is supported by funds from the Fundación Científica de la Asociación Española contra el Cáncer and from the Instituto de Salud Carlos III (PIE15/00008). Research at the B.S. Atanassov lab is supported by the Rosswell Park Cancer Institute and NCI grant P30CA016056. Research at the J. Arribas lab is supported by funds from the Breast Cancer Research Foundation (BCRF-17-008) and Instituto de Salud Carlos III (PI16/00253).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.