DUB1 Suppresses Hippo Signaling Via Modulating TAZ Protein in Gastric Cancer

The Hippo pathway functions as a tumor-suppressor pathway in human cancers, while the dys-function of Hippo pathway is frequently observed in malignancies. Although the YAP/TAZ activity is tightly controlled by the phosphorylation cascade of MST-LATS-YAP/TAZ axis, it is still unclear why YAP/TAZ protein is activated in human cancers, even Hippo pathway is still active. Besides phosphorylation, recent studies implicate that several post-translational modications also play critical roles in modulating TAZ function, including ubiquitination. inhibitor MG132, the stabilization effect of DUB1 on TAZ did not further increase TAZ protein levels. HEK293 cells were transfected with 0.5 ug Flag or Flag –DUB1 plasmids. After 24 h, cells were treated with 10 mM MG132/vehicle for 6 h. Cell lysates were prepared for Western blot analysis. The results are representative for three independent experiments.

sequences were: UUC UCC GAA CGU GUC ACG UTT; ACG UGA CAC GUU CGG AGA ATT. The RNA iMAX reagent (13778150, invitrogen) was used for siRNA transfection. For lenti-virus based DUB1 silencing, shDUB1 was cloned into the vector pLKO.1, which was co-transfected with pMD2.G and psPAX2 envelop plasmids into HEK293 cells. The DUB1 shRNA lenti-virus was harvested after 48 hours. The gastric cancer cells were incubated with 2 ml antibiotic-free medium containing 200 ul lentivirus.

Luciferase reporter assays
For TEAD luciferase activity assays, the MGC803 and AGS with siDUB1 or siControl cells transfected with the TEAD luciferase reporter vector for 24 h. Cells were then harvested for assays. Luciferase reporter assays were performed using the dual luciferase assay kit (Promega). The pRL-null vector expressing renilla luciferase (Promega) was used as an internal control to normalize the transfection e ciency.

Wound healing and Transwell assays
For the wound-healing assay, the MGC803 and AGS with siDUB1 or siControl cells were seeded in a 6-well plate until con uent and then wounded with a sterile tip. The cells pictures were acquired at the indicated time points after scratching. The distances between the two edges of the scratched wound were measured using Image J software. The trans-well system (8 μm pore size, Corning) was employed for cell migration and invasion assays. For invasion assays, the upper chambers were coated with matrigel (BD Biocoat, USA). After 24 h, the gastric cancer cells that had migrated through to the bottom of the insert membrane were xed, stained with crystal violet and counted under ×20 objective lens. The experiments were in three repeats.
Cycloheximide assay MGC803 cells were transfected with siRNA or siControl for 24 hours. After that, the Cycloheximide was added into culture medium with the nal concentration of 100 μmol/L. Cell lysis were collected at 0, 3, 6 and 9 hours after the treatment of cycloheximide. For HEK293 cells, the cells were transfected with 2ug Flag-DUB1 or Flag vector. After 24 hours, cells were treated with cycloheximide with nal concentration of 100 μmol/L. Cell lysis were collected at 0, 3, 6 and 9 hours after the treatment of cycloheximide.

Immuno uorescence (IF) staining
The MGC803 Cells on the coverslips were xed with 4% paraformaldehyde and incubated with the primary antibody against DUB1 (Sigma, A300-940A), TAZ (CST, 71192) at 4 °C overnight. After that, the cells were washing with PBS. Then the cells were then incubated together with uorescence-conjugated secondary antibody (Invitrogen, Carlsbad, CA). Finally, the cells were subsequently counterstained with DAPI (Life Technology). The staining Images were captured through the confocal laser-scanning microscope (Leica TCS SP8 STED). The uorescence-integrated density was measured by ImageJ software.

Clone formation assays
The MGC803 and AGS were seeded in six-well plates overnight and treated with 50nM DUB1 siRNA or 50nM siControl. After twenty-four hours, the gastric cancer cells were washed with PBS, trypsinized and plated at low density (5000 cell/well in six-well plate). The cells were cultured for 10 days and the medium was refreshed every two or three days. The colonies were stained with crystal violet. The number of the clones in a given area was counted for each condition.

Co-IP assays
The co-immunoprecipitation assays were performed as previous described. The MGC803 total cell lysls were collected and pre-cleared with rabbit IgG for 2 h and subsequently immunoprecipitated with DUB1 antibody (Sigma, A300-940A) over night, while rabbit IgG (Santa Cruz) was used as the negative control.
The bounded protein was analyzed by Anti-TAZ (CST, 71192). For the overexpression experiment, HEK293 cells were transfected with 5ug Flag-DUB1 (Full length or deletion domains) and Myc-TAZ (Full length or deletion domains) in 10 cm dish. Cell lysates were pre-cleared with IgG and subsequently incubate with anti-Flag-M2 (A8592, Sigma, 1:1000) antibody, while mouse IgG was used as the negative control. The bound proteins were analyzed by western blotting.
In vivo ubiquitination assays For in vivo ubiquitination assays, cells were transfected with vectors, including expressing Myc-TAZ, Flag-DUB1 and HA-Ub for 24 h respectively. Cells were then treated with MG132 (10 μM) for 6 h, and the levels of Myc-TAZ ubiquitination was determined by IP with anti-HA antibody (2013819001, Roche,1:1000) followed by western-blot assays with an anti-Myc (9E10, Santa Cruz, 1:1000).

In vivo tumorigenesis assay
For in vivo tumorigenic experiment, MGC803 cells were infected with shControl virus or shDUB1 virus.
After 48 hours of infection, cells were treated with 1ug/ml puromycin for 3 days. MGC803 cells (2×10 6 ) were injected into the right dorsal ank of 4-week-old female BALB/c nude mice. Tumor formation in nude mice was monitored over a 4-week period. The tumor volume was calculated by the formula: tumor volume = 0.5 × length × width 2 . The mice were sacri ced ve weeks after injection. The mice were sacri ced and the tumors were weighted and photographed. The experiments were performed under the protocols approved by ethnic committee of Xinxiang Medical University.
Cell proliferation assay MGC803 and AGS cells were transfected with siDUB1 or siControl in 24-well plates. Twenty-four hours after transfection, the cell number was countered and 4000 cells were seeded into 96-well plates. The relative cell viability was measured at the indicated time points. Cell numbers determined determined using CCK8 cell proliferation reagent by measuring the absorbance at 450 nm. Cell proliferation was further analyzed by EdU incorporation and ow cytometry. Gastric cancer cells were determined by using the 5-ethynyl-20-deoxyuridine (EdU) assay kit (Ribobio, Guangzhou, China). For quanti cation analysis of the images, each data point represents the positive uorescence area calculated from a minimum of ve randomly chosen elds from three individual experiments. EdU incorporation FACS assay was performed according to the manufacturer's instructions. The experiments were performed in triplicate. For cell cycle analysis, MGC823 cells were transfected with 50 nM siDUB1 or siControl. After 24 hours, the cells were xed with 70% ethanol and stained with propidium iodide. Twenty-four hours post transfection, the cells were stained with propidium iodide and annexin V. A BD LSR FACS was used to measure the uorescence intensity.
Tissue microarray (TMA) and immunohistochemistry (IHC) One hundred para n-embedded gastric cancer samples were acquired from Shanhai OTUDO Biotech Company (http://www.superchip.com.cn). All the gastric tumor samples were examined with pathological specialists. The pathological grade plus lymph node metastasis status of each sample were acquired from Shanhai OTUDO Biotech Company. The usage of the samples was approved from the Shanhai OTUDO Biotech Company with written informed consent from all the patients. Speci c antibodies against TAZ (CST, 71192) and DUB1 (Sigma, HPA12082) were used to detect the staining density in human samples. The scores were calculated on the intensity and percentage of positive tumor cells in the whole tissue, which were evaluated according to the Fromowitz Standard. The staining intensity was graded as: no staining, 0; weakly positive, 1; moderately positive, 2 and strongly positive, 3. The percentage for positive cells was into four grades: 0-25% staining, 1; 26-50% staining, 2; 51-75% staining, 3 and 76-100% staining, 4. The staining 1-2 was regarded as low expression, while the stain 3-4 was regarded as high expression. All staining were assessed at 200X magni cations and at least three elds from each core were counted.

RNA sequencing and data analysis
The global gene expression analysis (siControl and siDUB1) was based on RNA sequencing platform from BGI (Beijing Genomic Institute). The RNA sequence data are deposited in the Gene Expression Omnibus (GEO) database (Assessing number: GSE143947). Analysis was performed for differentially expressed genes (P < 0.01 and fold change > 2) by Ingenuity Pathway Analysis (IPA). For gene set enrichment analysis of RNA-seq data, gene sets of Conserved Hippo Signature were used and downloaded from Molecular Signatures Database v7.4, GSEA was implemented using the GSEA 4.1.0 software, with default parameters. Volcano plot was generated using 'ggplot2' package in R (threshold P<0.05 and fold change>1.5).
Analysis of TCGA data and Progression-free survival data analysis Gene expression data for 385 TCGA gastric cancer patients were downloaded from the webpage (http://gepia.cancer-pku.cn/index.html). The expression of DUB1 mRNA level between normal gastric tissue and different gastric cancer stages were generated from GEPIA online software. The progressionfree survival (PFS) survival data of TAZ and DUB1 were generated from KMPLOT online analysis database (https://kmplot.com). The gene affy ID was 202133_at for TAZ and 227093_at for DUB1. The PFS survival data of DUB1 and TAZ in gastric cancer patients were generated from KMPLOT database.

Statistical analysis
No speci c statistical tests were used to predetermine the sample size. Statistical analysis was performed using GraphPad Prism 7 software or SPSS version 23.0. Data are expressed as the mean ± s.e.m. values. Differences between two independent groups were evaluated with Student's t-test. The Kaplan−Meier method with the log-rank test was applied for survival analysis. Differences were considered to be statistically signi cant when P < 0.05 (*P < 0.01; **P < 0.001).

Results
DUB1 is elevated in human gastric cancer and correlates with Hippo signaling activity in whole genomic scale Since this studies aimed to identify novel de-ubiquitinating enzymes in regulation Hippo signaling, we carried out a DUB (Deubiquitinases) siRNA screening via the DUB siRNA library (Dharmacon Company, Cat: G104705). Since HEK293 cells were widely used for Hippo signaling study and could be transfected with high e ciency, we utilized HEK293 cells for initial screening. Since CTGF was regarded as one of the most classical target genes, we used CTGF as the endpoint to indicate Hippo signaling activity (Fig. 1A). The siRNA screening coupled with QPCR revealed several con rmed DUBs, such as USP9X and USP7, which were proved to regulate Hippo pathway in previous publications (Supplementary Table 1). However, we discovered an unreported deubiquitinase-DUB1 could also be an important regulator in Hippo pathway (Fig. 1B). We further analyzed the DUB1 mRNA level in gastric cancer patients. The TCGA data showed that DUB1 expression was signi cantly elevated in gastric cancer samples compared with normal gastric tissue (Fig. 1C). The immunohistochemistry data showed that DUB1 was signi cantly higher in gastric cancer samples, which expression was correlated with lymph node metastasis and later clinical stage (P<0.001; P<0.001, P<0.001 respectively; Fig. 1D -1E). Interestingly, the protein level of DUB1 positively correlated with TAZ expression in gastric cancer samples (P<0.001, Fig. 1F). The prognostic data analysis showed that both DUB1 and TAZ correlated with poor survival in gastric cancer patients ( Fig. 1G-1H). In order to approach the function of DUB1 in gastric cancer in an unbiased way, we deplete DUB1 in MGC803 cells for the whole genomic expression analysis. The KEGG pathway analysis indicated that DUB1 depletion could affect several cancer-related pathways, including Hippo signaling and PI3K/AKT pathway (Fig. 1I). The GSEA analysis implicated that DUB1 depletion signi cantly inhibited Hippo signaling activity (Fig. 1J). The volcano plot analysis showed DUB1 silencing dramatically inhibited Hippo classical target gene expression, including CTGF, CYR61 and ANKRD1 (Fig. 1K).

DUB1 facilitates Hippo/TAZ axis in gastric cancer cells
We further utilized two independent siRNAs for DUB1 to avoid off-target effects. The western blotting assays showed that DUB1 depletion could not affect YAP protein level in AGS and MGC803 cells (Data not shown). However, DUB1 depletion could dramatically decrease TAZ protein level in AGS and MGC803 cells ( Fig. 2A-2B). We examined the Hippo target gene expression in gastric cancer cells. We found that DUB1 depletion could inhibit Hippo target gene expression, including CTGF and CYR61, in AGS and MGC803 cells (Fig. 2C-2D). We further evaluated the effect of DUB1 in Hippo signaling activity. The luciferase reporter assay showed that DUB1 silencing inhibited TEAD response element activity in AGS and MGC803 cells ( Fig. 2E-2F). Consistently, the transient DUB1 overexpression in HEK293 cells showed increased TAZ protein level, Hippo target gene expression (CTGF and CYR61) and the increased luciferase activity of TEAD response elements ( Fig. 2G-2I).

DUB1 depletion inhibits gastric cancer progression in vivo and in vitro.
In order to investigate the impact of DUB1 in gastric cancer phenotype, we depleted DUB1 in MGC803 and AGS cells. The CCK8 assay indicated that DUB1 depletion signi cantly inhibited gastric cancer cell growth ( Fig. 3A-3B). Besides, The EdU incorporation assays showed that DUB1 silencing could dramatically decrease the EdU positive cells in MGC803 and AGS cells ( Fig. 3C-3D). The cell cycle analysis by PI staining and FACS analysis showed that DUB1 depletion signi cantly caused the increased proportion of G1 phase cells ( Fig. 3E-3F). Besides, the trans-well assay implicated that DUB1 was necessary for gastric cancer invasion capacity in AGS and MGC803 cells (Fig. 3G-3H). The woundhealing assay showed that DUB1 depletion could decrease the cancer cell migration speed in AGS and MGC803 cells (Fig. 3I-3J). In the in vivo tumor growth assay, stable silencing of DUB1 in MG803 cells could inhibit tumor growth in the xenograft mice model (Fig. 3K-3M).

DUB1 controls gastric cancer progression via Hippo/TAZ axis
In order to investigate the logic link between the gastric cancer phenotype and Hippo/TAZ signaling in DUB1 function, we carried out several rescue experiments. DUB1 depletion could decrease TAZ protein level, which could be rescued by TAZ over-expression in MGC803 and AGS cells (Fig. 4A-4B). The QPCR assay showed that DUB1 silencing could decrease Hippo target gene expression, which could be rescued by TAZ overexpression in MGC803 and AGS cells (Fig. 4C-4D). The luciferase reporter assay showed that DUB1 silencing inhibited TEAD response element activity in AGS and MGC803 cells, which could be rescued by TAZ overexpression (Fig. 4E-4F). The CCK8 assay showed that the inhibited cell growth by DUB1 depletion could be rescued by further TAZ overexpression in MGC803 and AGS cells (Fig. 4G-4H).
The EdU staining assay showed that the TAZ over expression could rescue the EdU positive cells, which was decreased by DUB1 depletion (Fig. 4I-4J). The cell cycle analysis showed that G1 arrest caused by DUB1 depletion could at least partially rescued by further TAZ overexpression (Fig. 4K-4L). The clone formation assays showed that the decreased clone numbers caused by DUB1 depletion could be rescued by further TAZ overexpression in MGC803 and AGS cells (Fig. 4M). In gure 4N-4O, DUB1 knocking-down in gastric cancer cells could inhibit cell invasion, which effect could be at least partial rescued by further TAZ overexpression. The wound-healing assay showed that DUB1 depletion could inhibit cell migration speed, which effect could be further rescued by TAZ overexpression in MGC803 and AGS cells (Fig. 4P-4Q). These data indicated that DUB1 modulated Hippo signaling through TAZ protein in gastric cancer.

DUB1 associates with TAZ and modulates TAZ stability in gastric cancer cells
We further investigate the localization of DUB1 and TAZ in gastric cancer cells. The immuno-staining shows that both DUB1 and TAZ are mainly localized in the nuclear (Fig. 5A). This data was further con rmed by nuclear cytoplasm separation assay (Fig. 5B). The endogenous immuno-precipitation showed that DUB1 could interact with TAZ in MGC803 cells (Fig. 5C) (Fig. 5D-5E). We made these deletion constructs and further investigated the associated domain by IP assay. The data showed that WW domain was required for TAZ to interact with DUB1, while the CTD domain of DUB1 was responsible domain for interaction with TAZ ( Fig. 5F-5G). DUB1 depletion could decrease TAZ protein level in MGC803 cells, which effects could rescued by MG132 treatments (Fig. 5H). In consistent, transfection of DUB1 in HEK293 cells showed that DUB1 could increase TAZ protein level, which effect could be minimized with the presence of the proteasome inhibitor MG132 (Fig. 5I). These data indicate that DUB1 could modulate TAZ protein level through proteasome degradation system. We further tested the protein stability in HEK293 system, which showed that DUB1 overexpression could enhance TAZ stability (Fig. 5J). This was further con rmed in MGC803 cells via endogenous TAZ depletion (Fig. 5K).

DUB1 stabilizes TAZ via inhibiting TAZ K48-linked poly-ubiquitination.
Since DUB1 is one Deubiquitinating enzyme, we further investigated the role of DUB1 in TAZ polyubiquitination. The ubiquitination-based immuno-precipitation in HEK293 cells showed that DUB1 overexpression could inhibit TAZ overall poly-ubiquitination (Fig. 6A). Since K48-linked ubiquitination is the most common degradation manner, we examined the DUB1 effect on K48-linked ubiquitination of TAZ. In HEK293 cells, we observed that DUB1 overexpression could inhibit K48-linked ubiquitination of TAZ (Fig. 6B). In the MGC803 cells, we observed that depletion of endogenous DUB1 could enhanced the overall poly-ubiquitination and K48-linked poly-ubiquitination (Fig. 6C-6D). We further investigated the functional domain for DUB1 to exert its function on TAZ ubiquitination. In the ubiquitination-based immuno-precipitation assay, we observed that the USP domain (1-420 AA) of DUB1 was necessary for its deubiquitination function on TAZ protein (Fig. 6E-6F).

Discussion
In our study, we discovered a novel deubiquitinase family member DUB1 through DUB library siRNA screening. As an important endogenous modulator for Hippo/TAZ axis in gastric cancer progression, DUB1 was elevated in human gastric cancer and related to poor survival for gastric cancer patients. The IHC analysis showed that DUB1 expression correlated with metastasis and TAZ expression, while DUB1 depletion signi cantly inhibited Hippo signature genes expression in whole genomic scale. DUB1 inhibition caused cell growth inhibition, decreased cell cycle progression and decreased cell invasion capacity. The molecular biology assays showed that DUB1 could associate with TAZ protein and stabilize TAZ protein via inhibited TAZ K48-linked poly-ubiquitination in gastric cancer (Fig. 6G). Based on these, we can propose inhibition of DUB1 expression or pharmaceutical targeting DUB1 function could be a promising strategy to inhibit the Hippo/TAZ-driven gastric cancers.
The conserved Hippo signaling maintains tissue hemostasis and organ size control [26]. Several studies showed that the dys-regulation of Hippo pathway was also exist in gastric cancer [27][28][29][30]. For example, the Hippo pathway effector YAP/TAZ was signi cantly elevated in human gastric cancer and related to poor overall survival, while depletion of YAP/TAZ or pharmaceutically targeting YAP/TAZ-TEAD interaction could inhibit gastric cancer growth in vitro and in vivo [31]. Based on these facts, we can propose that the dys-function of Hippo signaling could be one of the driver events for gastric cancer. As the "Oncogenic addiction" pathway, targeting Hippo signaling effector YAP/TAZ could a plausible strategy for gastric cancer. On the other hand, as an auto-inhibitory pathway, the control of Hippo The Hippo signaling effector TAZ plays important roles in controlling Hippo signaling activity, which contains three domains [33]. The TEAD interaction domain is responsible for its association with several transcriptional factors, while the WW domain mainly interacts with several TAZ modulators [34]. Due to the extensive interactive face between TAZ and several transcriptional factors, directly targeting TEAD-TAZ interaction becomes technically challenging. Based on the fact that TAZ protein is highly active, while the inhibitory factors remains functional in gastric cancer. Our research team starts to target Hippo effectors via modulating protein stability. Through a DUB siRNA library, which contained more than 100 deubiquitinases, we discovered DUB1 as a novel effector for TAZ stability and gastric cancer progression.
Such ndings identi ed novel endogenous modulators for Hippo signaling, but also promising targets to Hippo/TAZ in gastric cancer treatments.
Our previous studies characterized several E3 ubiquitin ligases in modulator Hippo signaling effectors and cancer progression, such as RNF187 and ZNF213 [20,24]. However, the protein stability is controlled via a balance between ubiquitinations and deubiquitinations. The deubiquitinases, which remove the ubiquitin chains from the target proteins, mainly function to stabilize target proteins [35]. Currently, there are approximately more than 100 DUBs identi ed in human genomes, while the ubiquitin speci c peptidases is the largest group [36]. DUB1 (also known as USP36) gene was rstly discovered from Hela cells in 2004, which subsequently reported to be function as one deubiquitinase through the USP domain [37]. DUB1 was found to modulate RNA helicases stability for RNA transcription in cancer[38].
Besides, DUB1 was also reported to stabilize Myc and facilitate myc-driven cancer progression, while depletion of DUB1 could cause growth inhibition and autophagy through P62/SQSTM1 pathway [39]. Couple with our conclusion that DUB1 stabilizing TAZ in cancer progression, we can propose that targeting DUB1 could block multiple oncogenic pathway and synergize in cancer therapeutics.

Conclusions
We have discovered DUB1 as an oncogene for gastric cancer both in clinical samples and experimental studies. We demonstrated that DUB1 was elevated in gastric carcinoma and related to poor survival. DUB1 associated with TAZ protein, inhibited TAZ poly-ubiquitination and proteasome-dependent degradation in gastric cancer cells. Our studies revealed a novel function of DUB1 in Hippo signaling in multiple layers. As a novel modulator for Hippo signaling, modulation of DUB1 activity or gene expression level could be an appealing strategy to treat gastric cancer.       G: The hypothetical model for DUB1 regulating Hippo/TAZ axis in gastric cancer. DUB1 protein associated with TAZ and enhanced TAZ protein stability via inhibiting TAZ K48-linked poly-ubiquitination, which facilitated the activation of Hippo/TAZ axis and gastric cancer progression.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.