Journal of Proteomics

Proteomics analysis reveals the role of ubiquitin specific protease (USP47) in Epithelial to Mesenchymal Transition (EMT) induced by TGFβ2 in breast cells
Virgínia Campos Silvestrinia,b, Carolina Hassibe Thoméa,b, Daniele Albuquerquea,b, Camila de Souza Palmaa,b, Germano Aguiar Ferreiraa,b, Guilherme Pauperio Lanfredia, Ana Paula Massona, Lara Elis Alberici Delsina,b, Fernanda Ursoli Ferreirab,
Felipe Canto de Souzab, Lyris Martins Franco de Godoyc, Adriano Aquinod, Emanuel Carrilhod,
Rodrigo Alexandre Panepuccib, Dimas Tadeu Covasb, Vitor Marcel Façaa,b,⁎
a Department of Biochemistry and Immunology, FMRP – University of São Paulo, Av. Bandeirantes, 3900, 14049-900 Ribeirão Preto, SP, Brazil b Center for Cell Based Therapy, Hemotherapy Center of Ribeirão Preto, Rua Tenente Catão Roxo, 2501, 14051-140 Ribeirão Preto, SP, Brazil c Instituto Carlos Chagas, Fiocruz Paraná, 81350-010 Curitiba, PR, Brazil
d Instituto de Química de São Carlos, Universidade de São Paulo, Av. Trabalhador São-carlense, 400, 13566-590 São Carlos, SP, Brazil


Epithelial to mesenchymal transition
Cancer Metastasis
Proteomics analysis TGFβ2 and USP47


Epithelial to Mesenchymal Transition (EMT) is a normal cellular process that is also triggered during cancer progression and metastasis. EMT induces cellular and microenviromental changes, resulting in loss of epithelial features and acquisition of mesenchymal phenotypes. The growth factor TGFβ and the transcription factor SNAIL1 (SNAIL) have been described as inducers of EMT. Here, we carried out an EMT model with non-tu- morigenic cell line MCF-10A induced with the TGFβ2 specific isoform of TGF protein family. The model was validated by molecular, morphological and functional experiments and showed correlation with the up-reg- ulation of SNAIL. In order to identify additional regulators of EMT in this non-tumorigenic model, we explored quantitative proteomics, which revealed the Ubiquitin carboXyl-terminal hydrolase 47 (USP47) as one of the top up-regulated proteins. USP47 has a known role in cell growth and genome integrity, but not previously corre- lated to EMT. After validating USP47 alterations using MRM and antibody-based assays, we demonstrated that the chemical inhibition of USP47 with the inhibitor P5091 reduced expression of EMT markers and reverted morphological changes in MCF-10A cells undergoing EMT. These results support the involvement of USP47 in our EMT model as well as potential applications of deubiquitinases as therapeutic targets for cancer progression management.
Biological significance: Metastasis is responsible for most cancer-associated mortality. Additionally, metastasis
requires special attention, as the cellular transformations make treatment at this stage very difficult or occa- sionally impossible. Early steps in cancer metastasis involve the ability to detach from the solid tumor mass and invade the surrounding stromal tissues through cohesive migration, or a mesenchymal or amoeboid invasion. One of the first steps for metastatic cascade is denominated epithelial to mesenchymal transition (EMT), which can be triggered by different factors. Here, our efforts were directed to better understand this process and identify new pathways that contributes for acquisition of EMT, mainly focused on post translational modifica- tions related to ubiquitin proteasome system. Our model of EMT induction by TGFβ2 mimics early stage of metastatic cancer in epithelial breast cells and a proteomic study carried out for such model demonstrates that the deubiquitinase enzyme USP47 acts in SNAIL stabilization, one of the most important transcription factors for EMT phenotype acquisition and consequent metastasis. In addition, the inhibiton of USP47 with P5091, reverted the EMT phenotype. Together the knowledge of such processes of cancer progression and regulation can help in designing new strategies for combined therapies for control of cancer in early stages.
Corresponding author at: Department of Biochemistry and Immunology, FMRP – University of São Paulo, Av Bandeirantes, 3900, 14049-900 Ribeirão Preto, SP, Brazil.
E-mail address: [email protected] (V.M. Faça).


Received 3 June 2019; Received in revised form 4 February 2020; Accepted 10 March 2020

1. Introduction
Metastasis is responsible for as much as 90% of cancer-associated mortality [1]. As an early step in cancer metastasis, tumor cells need to gain the ability to disseminate from the solid tumor mass and invade the surrounding stromal tissues through cohesive migration, or a mesench- ymal or amoeboid invasion [2]. Epithelial mesenchymal transition (EMT) is one of the processes that induce cell migration, allowing polarized epithelial cells to undergo multiple biochemical changes that culminate in a mesenchymal phenotype, which includes increased migratory capacity, invasiveness, resistance to apoptosis, and production of extracellular ma- triX (ECM) components. This process can be stimulated by extracellular cytokines, such as TGFβ, EGF, FGF, or by intracellular signals, such as oncogenic Ras or NFkB signaling, or transcription factor families such as SNAIL1 (SNAIL), SNAIL2 (SLUG) and ZEB1[3–5].
The TGFβ signaling pathway mediates cell proliferation, apoptosis, differentiation and ECM production, among others [6]. High levels of TGFβ have been found in plasma of breast cancer patients and in in- vasive human breast cancer tissues, correlating with the presence of metastasis [7]. TGFβ plays contrasting roles in cancer, acting as a tumor suppressor during the first stages of tumorigenesis and as a tumor promoter during advanced stages of progression [8–13]. Moreover, the TGFβ family of proteins is composed of three different isoforms (TGFβ1, 2 and 3), which have different roles in normal and diseased processes [14].
In all cells, the modulation of processes and changes triggered by signaling cascades also depends on mechanisms finely controlled by post-translational modifications related to the ubiquitin-proteasome protein degradation system [15]. Recent basic and translational studies have revealed a connection between malfunction of the ubiquitin-pro- teasome system with both tumor formation and tumor metastasis [16]. During these processes, target proteins are post-translationally modified through ubiquitination by specific enzymes participating in the fine control of the pathways in which they are involved. Specifically during EMT, the ubiquitin proteasome system modulates important proteins for the acquisition of mesenchymal phenotype that is essential for metastasis dissemination [17,18].
The inverse of ubiquitination, a process known as deubiquitination, is also carried out by specialized enzymes denominated deubiquitinases (DUBs). This process counteracts the ubiquitination cascade, including inhibiting E2 ubiquitin conjugating enzymes and E3 ligases [19]. Pro- teasome-related DUBs help to prevent degradation of proteins con- jugated with an ubiquitin chain. DUBs can remove or edit ubiquitin chains to alter the signals mediated by non-degradation ubiquitin. Many DUBs are altered in the tumor, including USP47, a deubiquitinase enzyme involved in many processes, such growth control and cell sur- vival [20]. Several studies have revealed a connection between mal- function of the ubiquitin-proteasome system with both tumor formation and tumor metastasis [20–22].
To increase the understanding of the complex molecular mechan- isms of EMT at the protein level, we analyzed global proteome changes during EMT induction promoted by TGFβ2 in the non-tumorigenic epithelial cell line MCF-10A. We found a significant up-regulation of USP47, suggesting the involvement of ubiquitination/deubiquitination- related proteins in TGFβ2-induced EMT. In addition, we have observed that inhibition of USP47 with P5091 inhibitor leads to modulation of SNAIL and inhibited morphological changes promoted by EMT. Therefore, our results support the participation of specific ubiquitin- proteasome post-translational modifications in EMT, which can provide opportunities for effective clinical management of cancer progression.
2. Experimental procedures
2.1. Cell culture
MCF-10A cells, from a non-tumorigenic epithelial cell line, were

acquired from ATCC (ATCC® CRL-10317™) and were maintained in MEBM (Cat. CC3150, MEGM bullet kit, Lonza) supplemented with 13 mg/mL BPE, 0.5 mg/mL hydrocortisone, 10 μg/mL hEGF, 5 mg/mL bovine insulin, 100 ng/mL cholera toXin and penicillin and strepto- mycin in 100 U/mL (GIBCO, Carlsbad, CA, USA). The cell line was kept at 37 °C with 5% CO2.
2.2. EMT induction
For EMT induction, MCF-10A cells (1 × 106 cells) were seeded in 100 mm X 20 mm culture plates with 6 mL of culture medium and monitored for 48 hours until reaching a confluence of approXimately 60%. After this time, 10 ng/mL of TGFβ2 (302-B2, R&D Systems) was added and the culture was maintained for additional 72 hours. The follow-up of EMT induction was performed by monitoring morpholo- gical changes while under the microscope.
2.3. Wound healing
MCF-10A cells treated with TGFβ2 and control cells were grown to 80% confluence in a 6-well plate, and received a scratch wound made using a 1 mL pipette tip. The cells were washed three times with PBS and fed with complete media. Pictures were taken at 0, 24, 48 and 72 hours to monitor the recovery of the scratch wound.
2.4. Stable isotope labeling by amino acids in cell culture
The MCF-10A cell line was cultured for 20 days (9 replications) in SILAC Advanced D-MEM/F12 Flex Media medium containing natural (light) (12C6) or isotopically heavy (13C6) lysine (SILACTM advanced DMEM/F12-Flex, Life Technologies, Inc.). The medium was supple- mented with 5% inactivated dialyzed FBS (SILAC kit). The follow-up of heavy lysine isotope incorporation into the cell line was performed by monitoring the relative amount of light and heavy peptides present for the β-actin protein in the cell extracts obtained with each new strain expansion in the SILAC medium. At each time point aliquots were withdrawn with approXimately 1 X 106 cells (corresponding to 50 μg) of total proteins, which were lysed and the protein extract was subjected to enzymatic digestion by trypsin. The total extract in the form of peptides was then analyzed by MRM (multiple reaction monitoring), as will be described in detail below, in order to monitor the relative amount of light and heavy peptides from β-actin protein present in the extracts. The MCF-10A strain grown in SILAC medium containing the heavy lysine isotope was used as a reference for comparison of the TGFβ2-induced proteomic alterations in the MCF-10A line grown in its ideal culture medium.
2.5. PCR analysis
The cDNA was synthesized using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems), according to the manufacturer’s instructions. The fragments generated by reverse transcription were analyzed by 1% agarose gel electrophoresis and then submitted to real- time PCR analysis. Quantitative analysis of CDH1, CDH2 and FN gene expression was performed using the TaqMan methodology (Applied Biosystems) using the GAPDH gene as calibrator. The gene expression of the selected factors was performed for the control and TGFβ2 treated cells. Real-time PCR reactions were performed on the 7500 equipment (Applied Biosystems) according to the manufacturer’s instructions.
2.6. Subcellular fractionation
Nuclear, cytoplasmic and membrane fractions were obtained by using a successive detergent extraction method as described by Thomé et al. [23]. Culture plates (15 cm) containing MCF-10A (1 X 107) cells were washed twice with PBS and lysed with 100 μL of hypotonic buffer

(caption on next page)

Fig 1. Characterization of EMT induction by TGFβ2 in MCF-10 cells. A) Morphological changes in MCF-10A cells induced by TGFβ2 10 ng/mL detected by contrast phase; B) PCR Analysis by fibronectin, CDH1 and CDH2. C) Fluorescence microscopy demonstrating the intensity of SNAIL when MCF-10A cells are treated with TGFβ2 10 ng/mL and compared with control. D) Immunostaining analysis of cellular SNAIL during EMT induction; Statistical significance was assessed by one-way analysis of variance **p < 0.05; E) Wound healing assay shows that TGFβ2 induction also increased cell migration in MCF-10A cells compared with control. Pictures were taken at 0, 24, 48 and 72 hours by contrast phase microscopy; F) Images demonstrating the observed result for wound healing assay.

M (50 mM HEPES, pH 7.4, 10 mM NaCl, 5 mM MgCl2, 0.1 mM EDTA,
1 mM Na3VO4, 1 mM NaF, and 1 mM Na4P2O7•10H2O) containing 5% v/v protease inhibitor miXture (P8340; Sigma-Aldrich). Samples were vortexed for 20s, passed through a 25-gauge needle (20 times), and homogenized (D-130 tissue homogenizer, Biosystems, São José dos Pinhais, PR, Brazil) for 1 min. The nuclear pellet was obtained by
centrifugation at 500× g for 20 min at 4 °C. Supernatant was cen- trifuged at 16,000× g for 15 min at 4 °C, collected, and designated as
the cytoplasmic fraction. Protein extraction of the nuclear pellet was performed with lysis solution containing 8 M urea, 2% CHAPS, and protease inhibitor miXture. Three sonication cycles were carried out for 5 min each in an ultrasonic bath (Unique, São Paulo, SP, Brazil) with
cooled water. Samples were centrifuged at 20,000× g for 30 min at
4 °C, and supernatant was collected and stored at −80 °C. Prior to fractionation, the cell samples either controlled or treated with TGFβ2 and homogenized by passage through the syringe were miXed to the ratio 1:1 in terms of total amount of protein with the homogenate of MCF-10A cells grown in the heavy SILAC medium (described in the sections above). Protein quantitation of the cytoplasmic and nuclear extracts was performed through Bradford method (Quick Start Bradford Kit, cat #500-0205, Bio-Rad, Hercules, CA), following the manufac- turer's instructions [24].

2.7. In-gel digestion and LC-MS/MS analysis
The cytoplasmic and nuclear fractions (30 μg) were dissolved in 30 μL of Laemmli sample buffer [25] containing dithiothreitol (1 mg/mg of total protein), boiled for 5 min to reduce disulfide bonds, and alkylated with iodoacetamide (5 mg/mg of total protein). Samples were then separated using a 10% SDS-PAGE (Bio-Rad,). Gel lanes corresponding to each fraction were cut into 6 pieces of ~1 cm each, washed, and digested with trypsin as described previously [23]. Tryptic peptides were successively extracted with aqueous solutions containing 0.1% formic acid, 50% acetonitrile, and then 70% acetonitrile and then dried by SpeedVac (Thermo Scientific, Marietta, OH). Peptide miXtures were dissolved in 10 μL of 5% acetonitrile, 0.1% formic acid and desalted using ZipTip columns (Supelco, Sigma-Aldrich) according to the manufacturer's instructions. Samples were eluted in 52.5% acetonitrile/water in 0.1% formic acid, dried again, and resuspended in 15 μL of 5% acetonitrile, 0.1% formic acid for LC-MS analysis. Samples

considered for protein quantification. The list of proteins was generated with a Protein Prophet cutoff value of 0.90, representing an overall protein false discovery rate of approXimately 1%. Proteins were quantitated as described previously, using the Q3 algorithm (1.22a release) to measure SILAC peak intensities [30,31]. The mass spectrometry proteomics data were deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD013193 [32].

2.8. Multiple reaction monitoring (MRM) analysis
Aliquots containing 400 μg of proteins from cytosol-enriched frac- tion were concentrated in Microcon YM-50 Millipore (cat. no. 42415) according to the manufacturer’s instructions. After quantification by Bradford assay, 100 μg of proteins were diluted in Urea 8 M, Tris-HCl
0.1 M, pH 8.5. Protein cysteine residues were reduced with 100 μg dithiotreitol at 37 °C for 30 min and then alkylated with 300 μg io- doacetamide for 30 min. Protein solution was then diluted with 740 μL Tris/HCl 0.2 M, pH 8.0 to reduce urea concentration and digested with 4 μg of trypsin at 37 °C overnight. Solid phase extraction was performed in Oasis HLB Cartridges (Waters). Detection of proteotypic peptides was programmed in a scheduled method using 4 transitions per peptide and monitoring windows of 2 min. Method development and data analysis were conducted using Skyline software V. 19.1. Each sample was ana- lyzed in triplicate (10 μL injections) in a LC-MS/MS Xevo TQs system
(Waters). Chromatographic separation was performed in a UPLC (I- class, Waters) using a C18 column (1.8 μm particle size, 100 Å pore size, 1 mm X 150 mm, Waters) in a linear gradient of 5 to 30% of acetonitrile over 25 min and at 100 uL/min in a water:acetonitrile:formic acid solvent system [33].

2.9. MTT assay
MTT assay was performed to evaluate the cellular viability of MCF- 10A cells treated with P5091 (SML0770, Sigma-Aldrich). MCF-10A cells (4 X 103 cells) were plated in 96-well plates (Corning) and after
48 hours (cell adhesion time) the inhibitor was added at different
concentrations (0, 0.1, 0.5, 1, 5 and 10 μM), and after 72 hours 50 μL of MTT ready-made solution was added ((3-(4,5-dimethylthiazol-2-yl)-5-

were analyzed by LC-MS/MS using an LTQ-Orbitrap Velos (Thermo-Fin-

(3-carboXymethoXyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, Sigma-

nigan) mass spectrometer. Chromatography was performed using nLC ThermoScientific equipment in-house packed 75-μm inner diameter (New
Objective) × 25-cm-long C18 column packed with Magic C18 resin, 5 μm particle size, 300 Å pore size, at 600 nL/min, with 90-min linear gradients from 5 to 35 % acetonitrile in 0.1% formic acid. The spectra were acquired in a data dependent mode in the range of m/z from 400 to 1800 and the 10 most abundant ions of +2 or +3 charge of each MS spectrum were selected for CID (collision-induced dissociation) and for MS/MS analysis. Peptides and proteins were identified with the Computational Proteomics Analysis Systems (CPAS) [26] using the Comet search algorithm (v.2019.1) [27], Peptide Prophet [28] and Protein Prophet [29] algorithms for statistical validation of data and protein grouping. MS data were searched with using
human proteome database (Uniprot, August 2019, 74,449 entries). Search parameters for semi-tryptic peptides included up to two missed cleavages, mass tolerance of 0.5 Da for fragment ions, fiXed cysteine modification with

Aldrich, Cat 11465015001). After 2 hours of incubation, the absorbance was read at 450 and 650 nm in a plate reader (Versamax Tunable Microplate Reader, Molecular Devices). The MTT assay was carried out in triplicate.

2.10. USP47 inhibition
MCF-10A (1 X 106) cells were seeded on 100 mm X 20 mm culture plates (Corning) and monitored for 48 hours until reaching approXi- mately 60% of confluence. Then, cells were incubated with 0.5 and
1 μM of P5091 with or without 10 ng/mL TGFβ2 for 72 hours of treatment. After this time, the culture plates were scraped together with hypotonic buffer M (50 mM HEPES, pH 7.4, 10 mM NaCl, 5 mM MgCl2,
0.1 mM EDTA, 1 mM Na3VO4, 1 mM NaF, and 1 mM Na4P2O7·10H2O)

carbamidomethylation (+57.02146), variable methionine oXidation

and extraction buffer (8 M urea and 0.1 M Tris, pH 8.5) containing 5%

(+15.99491), and variable lysine modification (+6.020129) to account for both heavy and light SILAC labels. Only peptides with a Peptide Prophet score above 0.90 and precursor ions with delta mass less than 20 ppm were

v/v protease inhibitor miXture (P8340; Sigma-Aldrich), for fractiona- tion and total extract, respectively.

Fig 2. Subcellular fractionation enriches cellular compartments and increases protein identification by mass spectrometry. A) Table showing the peptides and proteins identified by global proteomics analysis in each compartment and B) The involvement of proteins in each compartment and their communication, where 2492 proteins are found in both nucleus and cytoplasm; C) Gene ontology classification of proteins enriched in each subcellular compartment. Subcellular compartment annotations were enriched in expected fractions and proteins representing specific cellular components were uncovered.

2.11. Western blotting
MCF-10A cells were washed in cold PBS and lysed with 8 M urea, and 0.1 M Tris buffer containing protease inhibitor miXture (Sigma- Aldrich). Three sonication cycles were carried out for 5 min each in an
ultrasonic bath (Unique, São Paulo, SP, Brazil) with cooled water. Lysates were centrifuged at 20,000× g for 30 min at 4 °C, and the supernatants were designated as total cell lysates. The protein con- centration was determined by the Bradford assay (cat #500-0205, Bio- Rad), separated by SDS-PAGE and transferred to polyvinylidene fluoride membranes (GE Lifesciences, Pittsburgh, PA, USA). Membranes were blocked with 5% non-fat dry milk in 0.05% Tween/TBS and in- cubated with the specific antibodies. Rabbit anti-Vimentin (#5741, Cell Signaling), rabbit anti-E-cadherin (#3195, Cell Signaling), rabbit anti- β-catenin (#8480, Cell Signaling), rabbit anti-SNAIL (#3879, cell sig- naling), mouse anti-USP47 (WH0055031M1, Sigma-Aldrich) or rabbit
anti-GAPDH (#5174, Cell Signaling) were all used in the dilution of 1:1000. The protein-antibody complex was detected using the ECL Western blotting detection reagents (GE Lifesciences), and signals were detected using a CCD camera (Image Quant LAS 4000 mini, Uppsala, Sweden).
2.12. Immunostaining
Cells (MCF-10A) were plated on a 96-well plate (3603, Costar) (2 X 103/well) in complete growth medium and incubated at 37°C in a humidified atmosphere of 5% CO2 for 48 hours. After medium removal and treatment with TGFβ2 (10 ng/mL) and USP47 inhibitor (P5091) (0.5 and 1 μM) for over 72 hours on the same incubation conditions, cells were washed twice with PBS (100 μL/well) and fiXed/permeabi- lized with ice-cold 2% formaldehyde in methanol solution (50 μL/well) at -20 °C for 20 min. Then, cells were washed twice with PBS (100 μL/ well), quenching solution glycine (0.1 M) (50 μL/well) was added at room temperature for 15 min and blocked with 1% FBS (Gibco) in PBS

(50 μL/well) for 30 min. Subsequently, cells were incubated overnight at 4 °C with mouse anti-USP47 (WH0055031M1, Sigma-Aldrich) (1:100) and rabbit anti-Snail-Slug (1:250) (ab180714, Abcam). After incubation, cells were washed twice with PBS (100 μL/well) and in-
cubated with secondary antibodies (1:300) of goat anti-mouse IgG (DyLight®594) (3551, Thermo Scientific) and donkey anti-rabbit IgG (488 DyLight®) (ab96919, Abcam). The nuclei and cytoplasm staining was also carried out using Hoechst (33342, Life Technologies) (1:1000) and CellMask Blue (H32720, Life Technologies,) (1:1000), respectively, for 45 min at room temperature, while protected from light. Next, the wells were washed three times with PBS and maintained with 100 μL of PBS. Images were acquired with an automated ImageXpress micro XLS High Content Screening system (Molecular Devices, USA), using a 10X objective and the excitation/emission filter cubes for DAPI (Hoechst/ CellMask Blue), FITC (DyLight®488) and Texas Red (DyLight®594).

2.13. Image analysis
Image analysis was carried out using MetaXpress software (Molecular Devices). Briefly, images that had been marked with fluor- escent dyes were analyzed within their integrated intensity after nuclei, cytoplasm and cell segmentation were performed. Several morphology features, as well as intensity-related measurements (from SNAIL/SLUG and USP47 staining), were obtained from the nuclei and cytoplasm of each cell.

2.14. USP47 TCGA and protein atlas expression analysis
We used the Cancer Genome Atlas (TCGA) and The Human Protein Atlas database [34] to investigate the expression of USP47 in relation to breast cancer.

Table 1
Regulated proteins after TFGβ2-treatment. The 40 proteins more up and downregulated are shown. A) Cytoplasm. B) Nucleus.
A) Cytoplasm B) Nucleus
GENE Spectral Counts Ratio (TGF/CTR) GENE Spectral Counts Ratio (TGF/CTR) Up regulated PYGM 4 6.40 Up regulated DBN1 6 12.65
PDLIM5 4 5.89 SERPINA3 5 8.22
TF 7 5.48 ANKAR 25 7.16
MYH14 7 4.89 TUBB6 11 5.77
EEF1A1 8 4.54 FN1 46 4.96
ROCK2 6 4.52 MARS 6 4.39
NDRG1 10 4.47 SUPT6H 4 3.98
FUS 4 4.44 S100A16 17 3.74
GALK2 5 4.30 RSU1 6 3.71
SBDS 5 4.22 MRTO4 5 3.71
PFKL 8 3.81 PEA15 9 3.69
AP2B1 5 3.54 KRT6C 9 3.67
TGFBI 12 3.49 ERO1A 12 3.34
DCTD 6 3.40 SAR1A 5 3.12
ELAC2 4 3.38 UPF1 5 3.08
USP47 6 3.17 LAMC2 17 3.03
ANKAR 15 3.09 CMPK1 16 2.99
ANPEP 103 3.08 LAMA3 28 2.95
CLYBL 4 2.96 COPA 16 2.76
VAT1 16 2.85 XRN2 6 2.75
Down regulated MAT2A 13 0.45 Down regulated ALDH1B1 10 0.48
ACAA2 15 0.44 DEK 17 0.48
DCTN2 5 0.44 RPS15 6 0.48
VDAC1 8 0.43 UQCRFS1P1 6 0.47
GFPT1 13 0.43 MT1X 4 0.47
STMN1 12 0.43 PCNA 10 0.47
ALOX15B 5 0.43 HSD17B12 5 0.47
RPL26L1 7 0.39 CBX5 15 0.46
RPL21 13 0.39 AK1 4 0.43
SNRPD1 9 0.39 DDX23 6 0.43
ALB 13 0.38 EEF1D 16 0.39
UBA3 9 0.37 SF3A3 7 0.38
PRRC1 9 0.37 CSTF3 6 0.35
RTCB 15 0.34 COPG1 7 0.31
CA2 12 0.31 KRT14 9 0.30
MRPL27 4 0.30 EIF2AK2 4 0.18
PDAP1 10 0.28 DNM2 4 0.13
CD81 4 0.22 DDX39B 11 0.12
KRT6A 6 0.13 RPL35A 5 0.11
PLIN3 5 0.08 PHGDH 8 0.07

3. Results
3.1. Molecular, morphological and functional evaluation of EMT induction by TGFβ2 in MCF-10A cells
Here we developed a new model for EMT induction of MCF-10A mammary epithelial cells promoted by TGFβ2. To ensure that changes in MCF-10A cells promoted by TGFβ2 treatment are correspondent to EMT, we performed a detailed evaluation of the induction process using morphological, molecular and functional methods. First, we evaluated the morphological changes using microscopy and after 72 hours of treatment we observed loss of adhesion and acquisition of a fusiform phenotype, characteristic of cells undergoing EMT (Fig 1A). Then, using qPCR analysis, we detected increased expression of CDH2 and Fi- bronectin, two major mesenchymal markers (Fig 1B). Also, the protein expression of the transcription factor SNAIL was found to have in- creased after treatment with TGFΒ2 using immunostaining analysis (Fig. 1C/D). Finally, we demonstrate increased migratory capacity of cells induced to EMT as evaluated by the wound-healing assay (Fig. 1E/ F). Altogether, these results support the EMT-like phenotype in MCF- 10A induced by treatment with TGFβ2 at 10 ng/mL for 72 hours.

3.2. Proteomic studies of TGFβ2-treatment in MCF-10A cells
A proteomic study was carried out with the purpose of revealing

new proteins regulated during EMT in the MCF-10A epithelial non-tu- morigenic breast cell line. For high-throughput global proteomic ana- lysis we relied on an adaptation of the conventional SILAC strategy since MCF-10A cells did not respond to TGFβ2 treatment in SILAC media. To this end, we used cells grown in heavy SILAC medium as a spiking reference for cells grown in the ideal MEBM medium and in- duced or not to EMT. Also, in order to increase the depth of our pro- teomic analysis, cell fractionation was performed, and subcellular fractions enriched in nucleus and cytoplasm were obtained. We iden- tified a total of 3619 proteins (Fig. 2A) with less than 1% false discovery rate or 2967 proteins carrying two or more peptide observations. Comparing the number of proteins detected in each fraction, an en- richment of 31% of proteome coverage can be attributed to cell frac- tionation (Fig. 2B). Moreover, based on cellular location gene ontology annotation, the fractionation of EMT-induced MCF-10A cells enriched cytoplasmic and nuclear fractions accordingly while also slightly changing the distribution of mitochondrial proteins, accumulated in cytoplasmic fraction (Fig. 2C). The table with all the proteins identified in each independent fraction is presented in the Supplementary Table 1. The subset of proteins altered during EMT was selected based on the following criteria: 1) Protein identified and quantified in both the control experiment (non-treated cells miXed with heavy-labeled re- ference cells) and TGFβ2-treatment experiment (treated cells miXed with heavy-labeled reference cells) 2) Regulation greater than 2-fold; and 3) observation of 2 or more quantification events. Among all, 80

Fig 3. Network analysis of proteins differentially expressed in EMT induction. A) STRING-DB analysis indicates that proteins differentially expressed with confidence of 0.4. The network indicates the main biological processes that regulate the 80 proteins differently expressed: In red are marked proteins linked to cellular component organization and green metabolic process; B) During the process of EMT, USP47 has been rarely reported, without correlation with the process marker proteins in the literature.

Fig 4. Levels of USP47 RNA expression in breast cancer cell line and patients. A) Analysis of The Human Protein Atlas of USP47 expression measured in FPKM (Fragments Per Kilo base Million) in several tissues of cancer. Patients with breast cancer appear to be the mostly affected with mutated USP47. B) Analysis of The Human Protein Atlas of USP47 expression in four breast cancer cell lines in TPM (Transcripts per million). Breast cancer tissues presented a higher medium expression compared to
others (https://www.proteinatlas.org/). C) Summary of clinical outcomes. Probability of Overall Survival (OS) in Breast Cancer patients according to USP47 ex- pression (TCGA and protein atlas). The analyzes were divided into quartiles, Q1 and Q2, where Q1 show the expression of USP47 higher than 9.1 and Q2 lower than
9.1. Probability of Overall survival (OS) in patients with Breast Cancer (n = 1176) according to USP47 expression, C) patients in stage I (180), (D) Patients in stage II (n = 609) and (E) patients in stage III (n = 243) (TCGA and https://www.proteinatlas.org/).

regulated proteins were selected and are presented in Table 1 A and B. For the selected regulated proteins we performed a protein network analysis of these proteins using STRING DB [35] to identify major in- teractions and sets of coordinated cellular processes altered with TGFβ2-treatment (Fig. 3A). The main biological processes involved during EMT are linked to cellular component organization and meta- bolic process. These processes are known to be involved in EMT, cor- roborating the acquisition of the mesenchymal phenotype and corre-
spond to an in vitro tumor progression event.
Table 1A shows the 20 most regulated proteins with TGβ2- treat- ment in cytoplasm. Several proteins such as ROCK2 (Rho-associated protein kinase 2 [36]), NDRG1 (Protein NDRG1[37]), FUS (RNA- binding protein FUS [38]) and TGFβI (Transforming growth factor- beta-induced protein [39]) have already been previously related to EMT, and therefore not interesting to pursue. On the other hand, those observations support the validity of our EMT model. Among the up- permost regulated proteins, we highlight USP47, which participates in cellular control related to the ubiquitin proteasome system, being in- creased 3-fold with TGFβ2-treatment (Table 1A). This important en- zyme for cell homeostasis has not been previously described to parti- cipate in EMT process as evidenced in the protein network analysis (Fig 3B). USP47 represents then a potential new effector of EMT in breast cancer cells. Thus, we focused the next steps of our study to validate the correlation of USP47 deubiquitinase function in EMT, using a potent small molecule inhibitor (P5091) commercially available.

3.3. USP47 data mining in cancer and EMT
In order to find additional external evidences of USP47 participation in cancer development or EMT, we searched The Human Protein Atlas
[40] and The Cancer Genome Atlas Program (TCGA) for alterations in terms of expression in several cell lines of breast cancer as well as in patient samples. The Protein Atlas database (Fig. 4B) demonstrated a variation in USP47 expression depending on the cell line. In addition, USP47 is expressed in several types of cancer in which breast cancer appears to present a higher mean expression when compared to other types of tumors (Fig. 4A).
Moreover, considering the correlation between USP47 expression and patient survival from TCGA, the overall survival seems to be af- fected by lower USP47 expression at early stages of the disease. Furthermore, this difference is diminished as days advance, which could be speculated as a potential role of USP47 in early stage devel- opment of metastasis (Fig. 4C/D/E). Besides this, USP47 expression appears to be determinant depending on the stage when the breast cancer is diagnosed in these patients. Finally, higher expression of USP47 is correlated with metastasis of breast cancer and thus a lower survival probability.

3.4. Inhibition of USP47 suggests reversion of EMT process
MRM analysis and immunohistochemistry staining in a high-content

Fig 5. Increase of USP47 by immunohistochemistry staining in a high-content screening system in MCF- 10A during EMT induced. A) Analysis by im- munohistochemistry staining for USP47 during
EMT induction by TGFβ2-treatment. B) USP47 fluorescence intensity quantitation. Statistical significance was assessed by one-way analysis of variance **p < 0.05.
screening system were used to demonstrate higher levels of USP47 in MCF-10A cells treated with TGFβ2 (Fig. 5 and Fig 6B/C/D). Similar results were observed by western blotting (Fig. 6E), which confirm our proteomics findings. In order to evaluate if USP47 participates in EMT, we used the small molecule P5091 that specifically inhibits its deubi- quitinase activity. After determining that the concentrations of 0.5 and 1 μM of P5091 do not interfere significantly in cell viability (Fig. 6A) or cellular morphology, we treated cells with the inhibitor using the conditions for EMT induction with TGFβ2. Our MRM analysis did not show large variation in USP47 expression with P5091 inhibition. However, evident changes in protein expression of EMT markers by inhibition of USP47 can be observed in (Fig 6E), such as down- regulation of CDH1, CTNNB1 and SNAIL, are concentration-dependent and support the participation of USP47 in EMT.
To further test whether inhibition of USP47 plays a role in major pathways that regulate SNAIL, we tracked EMT in a high content screening platform. Fig. 7C and D indicate that USP47 inhibition has effect on levels of nuclear and cytoplasmic SNAIL in EMT process (P5091-treatment combined TGFβ2) during 72 hours of treatment, al- ready using P5091 in 0.5 and 1 μM, again suggesting that there is cross talk between USP47 activity and SNAIL levels levels in both con- centrations. Moreover, during EMT induction an increase of cellular area and nuclear size was observed, which is reverted with P5091- treatment (Fig. 7E–F). This finding supports USP47 as a regulator of EMT driven by TGFβ2 and SNAIL transcription factors and supports further studies of P5091 in therapeutics applications and USP47 as a target for controlling the EMT, and eventually, the metastatic process.

4. Discussion
EMT induces complex cellular and microenviromental changes that, in cancer, culminate in enhanced invasive and migratory capabilities of

tumor cells. In this study, we developed an EMT in vitro model using the non-tumorigenic epithelial cell line from mammary glands MCF-10A, induced by the specific isoform of the tumor growth factor TGFβ2. Since MCF-10A cells are non-tumorigenic, we consider this model as a potential recapitulation of early stages of tumor development stimu- lated by TGFβ. Of note, the tests with TGFβ1 did not promote pro- nounced alterations compatible with EMT as pronounced as TGFβ2 in MCF-10A cells, even though other models had indicated some success [41].
TGFβ superfamily contains several proteins such as activins, BMP (bone morphogenetic protein) and GDNF (glial cell derived neuro- trophic factor), which share the intracellular pathway of SMADs (mo- thers against decapentaplegic homologs) [12]. These proteins have several functions, which range from cellular communication in multi- cellular organisms to regulation of growth, proliferation, adhesion and apoptosis of several cell types. More specifically, the TGFβ subfamily of proteins has five known isoforms, but only three have been identified in mammals, TGFβ1, TGFβ2 and TGFβ3 [42], which share about 70% of sequence similarity but are coded in distinct regions of human chro- mosomes, 19q13, 1q41 and 14q24, respectively [12], supporting the observation of different expression patterns and, therefore, functional roles. The action of TGFβ is mediated through TGFβ receptors, types 1 (TβR-1), 2 (TβR-2), and 3 (TβR-3). This mechanism of activation is especially important for TGF-β2, which only interacts with TβR-2 when bound to TβR-3, propagating signals that act on several key proteins of the EMT process [12]. The specific isoform used in this study, TGFβ2, has been described for specific roles in epithelial–stromal cross-talk and matriX remodeling, in enhancing tumor proliferation and decreasing immune surveillance during tumor development, all characteristics compatible with EMT induction [43].
In our model, it was evident that there is increased expression of some important mesenchymal markers, such as FN and CDH2 and a

Fig 6. Inhibition of USP47 with P5091-treatment and its correlation with EMT. MCF-10A cells treated with the indicated concentrations of P5091 and TFGβ2-treatment for 72 hours. A) Cell viability assay showed changes with 0.5 and 1 μM already capable of leading to functional changes in EMT markers. B–C–D) MCF-10A cells treated with TFGβ2 and P5091 were evaluated using targeted multiple reaction monitoring (MRM) approach. Proteotypic peptides used in the analysis are shown on top of each graph. ACT4 protein was used as endogenous control. Statistical significance was assessed by one-way analysis of variance **p < 0.05, ***p < 0.001.
E) Western blotting validation of protein extract with antibody specific to EMT biomarker, mainly showing the decrease of SNAIL with P5091-treatment. The analyses were carried out with 50 μg of total protein extract. GAPDH was used as loading control.

decrease in CDH1, and more importantly, the transcription factor SNAIL [1,3,33]. The increase in CDH2 observed in carcinomas is di- rectly linked to the regulation of cell migration, invasion and resistance to apoptosis [44]. The CDH1 and CDH2 proteins share many structural and functional characteristics, for example, both establish calcium de- pendent cell-cell adhesion with their extracellular domains and are connected with the Catenins in their intracellular domains. In breast cancer, CDH2 is related to an aggressive character being overexpressed in invasive ductal carcinomas and in tumors with metastatic propensity [45]. Additionally, after treatment with TGFβ2, the cells lost their cubic form, and acquired a fusiform character with more cell motility, char- acteristic of mesenchymal cells. With all these results of morphological, functional and molecular markers, we can conclude that the MCF-10A cells were induced to the process of EMT by the treatment for 72 hours with TGFβ2.
After characterizing the developed EMT model for MCF-10A non- tumorigenic cell line, we sought to explore subcellular proteomics to reveal regulators of EMT. The subcellular fractionation reduced sample complexity and increased the coverage of the proteome. As previously

shown, after cellular and protein fractionation, 400,000 scans were obtained that yielded the confident identification of 2967 proteins. Of this total, 14% were detected only in the cytoplasmic fraction, while 18% were unique to the nuclear fraction. In quantitative terms, we obtained a large number of quantified proteins even though we used the special criteria to observe the protein in both control vs. heavy-labeled reference MCF-10A and TGFβ2-treated vs. heavy-labeled reference MCF-10A. In fact, the duplication of data acquisition intrinsically im- proved the data quality and reliability of the quantitative analysis. From the 80 proteins regulated in the study, proteins representing the cellular processes of cellular component organization and metabolic process were regulated. Particularly, cell cycle changes have been re- cently explored by our group, supporting a cell cycle arrest during EMT reprogramming [33,46]. Specifically in ovarian cancer cells, the EGF- induced EMT process led to alteration in cell cycle proliferation and arrest demonstrated by a remarkable increase in p21 protein levels indicating accumulation of cells in G1 phase and consequent blockage of S phase [33].
Of special attention among the top-regulated proteins in our results,

Fig 7. EMT induction in MCF-10A cells by TFGβ2-treatment and inhibition with P5091. A) Integrated intensity of cytoplasm USP47 during EMT and its inhibition with P5091 0.5 and 1 μM; B) nuclear USP47; C) Cytoplasmic SNAIL; D) nuclear SNAIL; E) All mean area and E) Nuclear area. Statistical significance was assessed by one- way analysis of variance ***p < 0.01 and ** p < 0.05.

USP47 which belongs to deubiquitinases family and counteracts the ubiquitin proteasome degradation system. The correlation of USP47 and classic EMT markers was not evident using the STRING.DB plat- form. Previous evidences indicated USP47 as a potential mediator of hypoXia-induced EMT [21]. Deubiquitinase enzymes [DUBs] are pro- teins that specifically remove mono- or poly-ubiquitin chains from target proteins and regulate multiple cellular signal transduction pathways [47]. DUBs are up-regulated by various environmental and endogenous stressors [16,19,48] Recently, association between dysre- gulated DUBs, such as USP4, −7, −11, −15, −19, −20, −22, −36,
−44, etc., and cancer has been suggested. However, the molecular basis for the possible role of USPs in cancer cell progression and me- tastasis remains poorly understood [49]. The DUB USP47 found here is involved in growth control and cell survival, alone or in combination with various chemotherapeutic agents. The silencing of USP47 in- creases the rate of apoptosis and the cytotoXic effects induced by che- motherapeutic agents in several strains, including osteosarcoma and breast cancer cells. USP47 is known to regulate DNA repair via deubi- quitination of mono-ubiquitinated DNA polymerase beta (POL-β), commonly mutated in many human tumors [22]. USP47 also increases Wnt signaling through deubiquitination of β-catenin, an important EMT biomarker, in A549 lung and PC3 prostate cancer cells [20]. Another

study of USP47 in gastric cancer cell lines showed that USP47 silencing or depletion leads to decreasing cell survival and contributes to anti- proliferative effects of cancer drugs [50].
Here we also expanded the relevance of USP47 from a clinical perspective with data mining, showing that USP47 is highly expressed in several types of cancer, especially breast cancer. Moreover, TCGA data supports the observation that the overall survival seems to be af- fected by lower USP47 expression at early stages of the disease and that USP47 expression appears to be a determinant in survival probability depending on the stage that the breast cancer has reached. Finally, higher expression of USP47 is correlated with metastasis of breast cancer and thus a worse survival of these patients.
To test more direct effects of USP47 in our EMT model, we inhibited this deubiquitinase with a specific inhibitor of USP47 and USP7 (P5091). The inhibition of USP47 in combination with TGFβ2 treat- ment, decreases levels of SNAIL in the nucleus and cytoplasm as de- monstrated by both high content screening and western blotting ana- lysis. A concomitant increase of CDH1 in expression and alterations of
cellular area and nuclear were observed with P5091-treatment. Our results are supported by another previous study of Choi et al, 2017, which showed expression of USP47 was elevated in three different human colorectal cancer cell lines [21]. The enhancement of USP47 in

colorectal cancer cells under hypoXic conditions induced the dis-

[11] L. Caja, et al., The transforming growth factor-beta (TGF-beta) mediates acquisition

assembly of CDH1 and promoted EMT through deubiquitination of SNAIL [21].
The alterations observed in SNAIL levels for different compartments can be explained by the fact that most biological processes involve changes in protein subcellular localization frequently associated with cellular dysfunction and disease, including cancer EMT [51]. Previous study demonstrated that this explanation may be due to the regulation by SoX9 to increase the expression USP47 which, in turn, stabilizes the SNAIL through deubiquitination [50]. SNAIL then translocates to nu- cleus where it induces the expression of proteins mediating EMT. In addition, knockdown of SNAIL has been shown to significantly inhibit tumor growth and metastasis by enhancing tumor infiltrating lympho- cytes in the cells, and by enhancing the growth of carcinomas in im- mune deficient mouse [52]. Together, these results suggest the in- volvement of USP47 in stabilizing SNAIL and therefore, regulating EMT.
In conclusion, the development of our model of EMT induction by TGFβ2 as well as the proteomic study of the model demonstrate that USP47 is an important deubiquitinase enzyme that acts in SNAIL sta- bilization. Together, these findings reflect the complex network of processes that occur during EMT, reinforcing the idea of cooperative and co-regulated processes. Detailed knowledge of such processes highlights the complexity of cancer progression and regulation as well as helps in designing new strategies for combined therapies targeting different and synergistic pathways to potentially control cancer pro- gression and increase patient life span.
Declaration fo Competing interest
This study was supported in part by grants from FAPESP (2011/ 09470-1, 2016/03809-3 and 2017/03960-6), CNPq, CAPES, Center for
Cell Based Therapy - CTC-CEPID (FAPESP 2013/08135-2) and CISBi- NAP (12.1.17598.1.3). V.M.F. received fellowship from CNPq (308561/ 2014-7).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jprot.2020.103734.

[1] C.L. Chaffer, R.A. Weinberg, A perspective on cancer cell metastasis, Science 331 (6024) (2011) 1559–1564.
[2] G.P. Gupta, J. Massague, Cancer metastasis: Building a framework, Cell 127 (4) (2006) 679–695.
[3] C.L. Chaffer, et al., EMT, cell plasticity and metastasis, Cancer Metastasis Rev. 35
(4) (2016) 645–654.
[4] M. Singh, et al., EMT: Mechanisms and therapeutic implications, Pharmacol. Ther. 182 (2018) 80–94.
[5] S. Lamouille, J. Xu, R. Derynck, Molecular mechanisms of epithelial-mesenchymal
transition, Nat. Rev. Mol. Cell Biol. 15 (3) (2014) 178–196.
[6] V.G. Keshamouni, et al., Differential protein expression profiling by iTRAQ-2DLC- MS/MS of lung cancer cells undergoing epithelial-mesenchymal transition reveals a
migratory/invasive phenotype, J. Proteome Res. 5 (5) (2006) 1143–1154.
[7] R. Kalluri, R.A. Weinberg, The basics of epithelial-mesenchymal transition, J. Clin. Invest. 119 (6) (2009) 1420–1428.
[8] J. FuXe, T. Vincent, A. Garcia, de Herreros, Transcriptional crosstalk between TGF- beta and stem cell pathways in tumor cell invasion: Role of EMT promoting Smad
complexes, Cell Cycle 9 (12) (2010) 2363–2374.
[9] Z.C. Yang, et al., Transforming growth factor-beta1 induces bronchial epithelial
cells to mesenchymal transition by activating the Snail pathway and promotes airway remodeling in asthma, Mol. Med. Rep. 8 (6) (2013) 1663–1668.
[10] H.J. Zhang, et al., Transforming growth factor-beta1 promotes lung adenocarci-
noma invasion and metastasis by epithelial-to-mesenchymal transition, Mol. Cell Biochem. 355 (1–2) (2011) 309–314.

of a mesenchymal stem cell-like phenotype in human liver cells, J. Cell Physiol. 226 (5) (2011) 1214–1223.
[12] M. Tian, J.R. Neil, W.P. Schiemann, Transforming growth factor-beta and the hallmarks of cancer, Cell Signal 23 (6) (2011) 951–962.
[13] J.F. Santibanez, et al., Transforming growth factor-beta, matriX metalloproteinases,
and urokinase-type plasminogen activator interaction in the cancer epithelial to mesenchymal transition, Dev. Dyn. 247 (3) (2018) 382–395.
[14] M.K. Lichtman, M. Otero-Vinas, V. Falanga, Transforming growth factor beta (TGF-
beta) isoforms in wound healing and fibrosis, Wound Repair Regen. 24 (2) (2016) 215–222.
[15] I.A. Voutsadakis, The ubiquitin-proteasome system and signal transduction path-
ways regulating Epithelial Mesenchymal transition of cancer, J. Biomed. Sci. 19 (2012) 67.
[16] M. He, et al., Emerging role of DUBs in tumor metastasis and apoptosis: Therapeutic implication, Pharmacol. Ther. 177 (2017) 96–107.
[17] F. Zhang, M. Laiho, On and off: proteasome and TGF-beta signaling, EXp. Cell Res. 291 (2) (2003) 275–281.
[18] Q.P. Dou, J.A. Zonder, Overview of proteasome inhibitor-based anti-cancer thera-
pies: Perspective on bortezomib and second generation proteasome inhibitors versus future generation inhibitors of ubiquitin-proteasome system, Curr. Cancer
Drug Targets 14 (6) (2014) 517–536.
[19] J.M. Fraile, et al., Deubiquitinases in cancer: New functions and therapeutic op- tions, Oncogene 31 (19) (2012) 2373–2388.
[20] J. Shi, et al., Deubiquitinase USP47/UBP64E regulates beta-catenin ubiquitination and degradation and plays a positive role in wnt signaling, Mol. Cell Biol. 35 (19)
(2015) 3301–3311.
[21] B.J. Choi, et al., HypoXia induces epithelial-mesenchymal transition in colorectal
cancer cells through ubiquitin-specific protease 47-mediated stabilization of Snail: A potential role of SoX9, Sci. Rep. 7 (1) (2017) 15918.
[22] A. Peschiaroli, et al., The ubiquitin-specific protease USP47 is a novel beta-TRCP
interactor regulating cell survival, Oncogene 29 (9) (2010) 1384–1393.
[23] C.H. Thome, et al., Linker for activation of T-cell family member2 (LAT2) a lipid raft
adaptor protein for AKT signaling, is an early mediator of alkylphospholipid anti- leukemic activity, Mol. Cell Proteomics 11 (12) (2012) 1898–1912.
[24] J.B. Hammond, N.J. Kruger, The bradford method for protein quantitation, Methods
Mol. Biol. 3 (1988) 25–32.
[25] U.K. Laemmli, Cleavage of structural proteins during the assembly of the head of bacteriophage T4, Nature 227 (5259) (1970) 680–685.
[26] A. Rauch, et al., Computational Proteomics Analysis System (CPAS): an extensible,
open-source analytic system for evaluating and publishing proteomic data and high throughput biological experiments, J. Proteome Res. 5 (1) (2006) 112–121.
[27] J.K. Eng, T.A. Jahan, M.R. Hoopmann, Comet: An open-source MS/MS sequence database search tool, Proteomics 13 (1) (2013) 22–24.
[28] A.I. Nesvizhskii, et al., A statistical model for identifying proteins by tandem mass
spectrometry, Anal. Chem. 75 (17) (2003) 4646–4658.
[29] A. Keller, et al., Empirical statistical model to estimate the accuracy of peptide identifications made by MS/MS and database search, Anal. Chem. 74 (20) (2002) 5383–5392.
[30] V. Faca, et al., Quantitative analysis of acrylamide labeled serum proteins by LC-
MS/MS, J. Proteome Res. 5 (8) (2006) 2009–2018.
[31] V.M. Faca, et al., A mouse to human search for plasma proteome changes associated with pancreatic tumor development, PLoS Med 5 (6) (2008) e123.
[32] J.A. Vizcaino, et al., ProteomeXchange provides globally coordinated proteomics data submission and dissemination, Nat. Biotechnol. 32 (3) (2014) 223–226.
[33] M.L. Grassi, et al., Proteomic analysis of ovarian cancer cells during epithelial- mesenchymal transition (EMT) induced by epidermal growth factor (EGF) reveals
mechanisms of cell cycle control, J. Proteomics 151 (2017) 2–11.
[34] U. Kusebauch, et al., Human SRMAtlas: A resource of targeted assays to quantify the complete human proteome, Cell 166 (3) (2016) 766–778.
[35] D. Szklarczyk, et al., The STRING database in 2017: Quality-controlled protein- protein association networks, made broadly accessible, Nucleic Acids Res. 45 (D1)
(2017) D362–D368.
[36] J. Luo, Z. Lou, J. Zheng, Targeted regulation by ROCK2 on bladder carcinoma via Wnt signaling under hypoXia, Cancer Biomark. 24 (1) (2019) 109–116.
[37] A. Li, et al., Upregulation of NDRG1 predicts poor outcome and facilitates disease progression by influencing the EMT process in bladder cancer, Sci. Rep. 9 (1) (2019)
[38] Q. Wu, et al., DLX6-AS1 promotes cell proliferation, migration and EMT of gastric cancer through FUS-regulated MAP4K1, Cancer Biol. Ther. 21 (1) (2020) 17–25.
[39] J. Zou, et al., Secreted TGF-beta-induced protein promotes aggressive progression in
bladder cancer cells, Cancer Manag. Res. 11 (2019) 6995–7006.
[40] M. Uhlen, et al., A human protein atlas for normal and cancer tissues based on antibody proteomics, Mol. Cell Proteomics 4 (12) (2005) 1920–1932.
[41] J. Zhang, et al., TGF-beta-induced epithelial-to-mesenchymal transition proceeds through stepwise activation of multiple feedback loops, Sci. Signal 7 (345) (2014)
[42] T. Imamura, A. Hikita, Y. Inoue, The roles of TGF-beta signaling in carcinogenesis and breast cancer metastasis, Breast Cancer 19 (2) (2012) 118–124.
[43] M.Y. Hachim, et al., Differential expression of TGFbeta isoforms in breast cancer highlights different roles during breast cancer progression, Tumour Biol. 40 (1)
(2018) (p. 1010428317748254).
[44] A. Kourtidis, et al., A central role for cadherin signaling in cancer, EXp. Cell Res. 358 (1) (2017) 78–85.
[45] P.J. Marie, E. Hay, Cadherins and Wnt signalling: A functional link controlling bone formation, Bonekey Rep. 2 (2013) 330.

[46] S. Palma Cde, et al., Proteomic analysis of epithelial to mesenchymal transition (EMT) reveals cross-talk between SNAIL and HDAC1 proteins in breast cancer cells,
Mol. Cell Proteomics 15 (3) (2016) 906–917.
[47] F.E. Reyes-Turcu, K.H. Ventii, K.D. Wilkinson, Regulation and cellular roles of
ubiquitin-specific deubiquitinating enzymes, Annu. Rev. Biochem. 78 (2009) 363–397.
[48] C. Luise, et al., An atlas of altered expression of deubiquitinating enzymes in human cancer, PLoS One 6 (1) (2011) e15891.

[49] R. Pfoh, I.K. Lacdao, V. Saridakis, Deubiquitinases and the new therapeutic op- portunities offered to cancer, Endocr. Relat. Cancer 22 (1) (2015) T35–T54.
[50] L. Naghavi, et al., Deubiquitinylase USP47 promotes RelA phosphorylation and survival in gastric cancer cells, Biomedicines (2018) 6(2).
[51] E. Lundberg, G.H.H. Borner, Spatial proteomics: A powerful discovery tool for P5091 cell biology, Nat. Rev. Mol. Cell Biol. 20 (5) (2019) 285–302.
[52] V. Bergoglio, et al., Deregulated DNA polymerase beta induces chromosome in- stability and tumorigenesis, Cancer Res. 62 (12) (2002) 3511–3514.