BRCA1-associated protein 1 serves as a tumor suppressor in hepatocellular carcinoma by deubiquitinating and stabilizing PTEN

Abstract

BRCA1-associated protein 1 (BAP1) or its mutants have been known to play critical regulatory roles in tumor biology, yet their role in hepatocellular carcinoma (HCC) remains largely unclear. In this study, we detected the mutations of all the exons of BAP1 in 105 HCC patients using Sanger sequencing, and found eight somatic mutations in 6 (5.71%) patients. We also found that the mRNA and protein levels of BAP1 were markedly downregulated in HCC versus the adjacent non-tumor tissues. Wild-type BAP1 but not mutant BAP1 significantly inhibited HCC cell proliferation, invasion, epithelial-mesenchymal transition (EMT) in vitro, and tumor progression and metastasis in vivo. Mechanistically, BAP1 complexed with PTEN and stabilized PTEN via deubiquitination and, furthermore, negatively regulated HCC cell EMT by deactivating the AKT/GSK-3β/Snail pathway. However, those tumor-inhibitory effects of BAP1 were abolished by inactivating mutations. Clinically, low BAP1 expression was positively correlated to aggressive tumor phenotypes, which also independently associated with poorer recurrence-free survival and overall survival after curative hepatectomy. Conclusively, our results indicate that BAP1, significantly downregulated, somatically mutated and negatively regulating EMT in HCC, serves as a tumor suppressor of HCC by deubiquitinating and stabilizing PTEN.

Introduction

The incidence and mortality of hepatocellular carcinoma (HCC) has kept increasing in recent years, making HCC as one of the most lethal cancer worldwide [1,2]. Currently, liver transplantation or hepatectomy is the main standard curative therapy for patients with HCC, but the long-term prognosis after these treatments remains dissatisfactory because of frequent recurrence and metastasis [3,4]. Still worse, the development of molecular targeted therapies or chemotherapies is obstructed by the poor understanding of HCC pathogenesis. Therefore, exploring the molecular mechanisms underlying HCC progression and metastasis is critical for identifying new effective therapeutic targets.

BRCA1-associated protein 1 (BAP1) belongs to the deubiquitinating enzyme (DUB) superfamily, which is homologous to other ubiquitin carboxyl-terminal hydrolase (UCH) containing a N-terminal catalytic domain [5]. BAP1 locates at chromosome 3p21, which is a hotspot that frequently mutates in human cancer [5,6]. Loss of BAP1 expression or inactivating genetic variants in the BAP1 gene locus have been found in various malignancies [5,7-16]. The germline or somatic mutations in BAP1 gene often lead to reduced expression or loss of protein function, which then accelerates tumor initiation, progression, and metastasis [9,14,16]. Importantly, recurrent somatic mutations of BAP1 have been found in HCC, and some of which located on the deubiquitinase region, which is vital to its deubiquitinating activity [17,18]. Therefore, it is possible that BAP1 deficiency may play a crucial role in HCC progression and metastasis.

In the present study, we identified that BAP1 harbored somatic mutations in HCC. BAP1 was markedly downregulated in HCC, and its deficiency was correlated to aggressive tumor phenotypes and poor postoperative prognosis. Further functional experiments indicated that BAP1 suppressed HCC cell proliferation, invasion, epithelial-mesenchymal transition (EMT) in vitro and tumor progression and metastasis in vivo, by forming a complex with phosphatase and tensin homolog (PTEN) and stabilizing PTEN via deubiquitination. We further demonstrated that those tumor-inhibitory effects of BAP1 were abolished by inactivating mutations. Therefore, BAP1 serves as a tumor suppressor of HCC, which provides a potential candidate treatment target in HCC.

Materials and methods

HCC samples and follow-up

Human liver tissues were collected from HCC patients that underwent curative hepatectomy at the Liver Cancer Institute, Zhongshan Hospital, Fudan University (Shanghai, China). The study was granted approval from the research ethics committee of Zhongshan Hospital, and each patient in this study provided informed consent.

Paired snap-frozen tumor and adjacent non-tumor samples from 105 patients with HCC treated between March and July 2011 were consecutively collected for Sanger sequencing. Paired snap-frozen tumor and adjacent non-tumor samples from 66 patients with HCC treated between July and September 2011 were used for analyzing the expression of BAP1 messenger RNA (mRNA), while 12 pairs of which were randomly selected for evaluating the expression of BAP1 protein. Snap-frozen HCC samples from 12 patients without or with early tumor recurrence were used for analyzing the mRNA and protein expression of BAP1 to assess the correlation with early recurrence. Paired paraffin-embedded HCC tissues and adjacent non-tumor tissues from 396 consecutive patients with HCC treated with curative hepatectomy between 2000 and 2002 were retrieved for constructing tissue microarray (TMA) and immunohistochemistry (IHC) analysis. Histopathological diagnosis was performed according to World Health Organization criteria. We classified tumor grade using the classification proposed by Edmondson and Steiner [19], and used the Chinese HCC staging system for tumor stage determination [2]. Clinical information collection and postoperative surveillance were conducted as previously described [20].

Recurrence-free survival (RFS) was the time from surgery to tumor recurrence. Overall survival (OS) was the duration from surgery to death. Data for patients alive were censored at the last follow-up.

Other materials and methods

Details of TMA construction, IHC analysis, Sanger sequencing, protein structural/array analysis, sequence alignment analysis, cell lines and transfection, polymerase chain reaction (PCR), qRT-PCR, immunoprecipitation (IP), deubiquitination assay, western blot, immunofluorescence, functional in vitro and in vivo assays, and statistical analysis are described in the Supplementary Materials and Methods. The sequences of mutant BAP1 variants, target sequences of short hairpin (sh) RNA, antibodies, and primer sequences used for Sanger sequencing are respectively listed in Tables S1, S2, S3, S4.

Results

Prevalence screening of somatic BAP1 mutations in HCC

All the exons of BAP1 in 105 HCC patients were sequenced by Sanger sequencing, and eight somatic mutations were identified in six patients (6/105, 5.71%). Among the six HCCs containing somatic mutations in BAP1, two of which harbored two mutations each, and the others only had one single mutation each. The eight somatic mutations of BAP1 were located in exons 4, 5, 7, 8, 9, 10, and 11 of BAP1, and five of which were located in the UCH domain (Figure 1A; Table 1). According to Polyphen-2 analysis [21], four missense mutations located in the UCH domain (Phe50Ser, Glu125Val, Arg150Gly, and Tyr189Cys) were forecasted to adversely affect the function of BAP1 protein, whereas the other two missense mutations (Leu294Arg and Gln322His) were benign (Table 1). Furthermore, multiple species conservation analysis showed that the four residues mutated in the UCH domain (Phe50Ser, Glu125Val, Arg150Gly, and Tyr189Cys) affected the highly conserved regions of BAP1 (Figure 1B).

Figure 1.

Somatic mutations of BAP1 in HCC and multiple sequence alignments of BAP1 paralogs in different vertebrate species. A. Trace images of BAP1 mutations identified by Sanger sequencing in HCC. The sequenced regions of tumor tissues were compared with those of adjacent non-tumor tissues to confirm the somatic nature of these mutations. The red arrows indicated the mutation sites. B. Multiple sequence alignments of BAP1 paralogs in different vertebrate species, including human, bovine, rat, mouse, chicken, clawed frog, and zebrafish. The mutation sites in BAP1 were marked with red arrows.

Table 1.

BAP1 mutations in 105 HCC samples

Exon No.	Nucleotide change (CDS)	Amino acid sequence change	Amino acid change	Mutation type	Polyphen-2 prediction (Score; Sensitivity; Specificity)	Case No.
4	c.T149C	TTC>TCC	p.Phe50Ser	Missense	PRD (1.000; 0.00; 1.00)	1
5	c.A374T	GAG>GTG	p.Glu125Val	Missense	PRD (1.000; 0.00; 1.00)	1
7	c.C448G	CGC>GGC	p.Arg150Gly	Missense	PRD (0.991; 0.71; 0.97)	1
7	c.A566G	TAC>TGC	p.Tyr189Cys	Missense	PRD (1.000; 0.00; 1.00)	1
8	c.C622A	CGG>AGG	p.Arg208Arg	Coding silent	-	1
9	c.A730C	AGG>CGG	p.Arg244Arg	Coding silent	-	1
10	c.T881G	CTG>CGG	p.Leu294Arg	Missense	Benign (0.148; 0.92; 0.86)	1
11	c.A966C	CAA>CAC	p.Gln322His	Missense	Benign (0.021; 0.95; 0.80)	1

Abbreviations: CDS, Coding sequence; PRD, probably damaging.

Structural and functional implications of BAP1 mutations

The mature BAP1 protein contains 729 amino acid residues, and all the somatic mutations of BAP1 identified in our study are located in the first half of BAP1 (Figure 2A and 2B). According to the protein structure model, five mutations (Glu125Val, Arg150Gly, Tyr189Cys, Leu294Arg, and Gln322His) are located on the surface of BAP1 protein, which may play an important role in protein-protein interactions (Figure 2C). Based on the structural analysis (Figure 2D), we identified that four mutations (Phe50Ser, Glu125Val, Arg150Gly, and Gln322His) dramatically change the physicochemical characteristics of BAP1 such as hydrophobic residues to polar or acid residues etc. and the other two mutations (Tyr189Cys and Leu294Arg) also play important roles on the surface of BAP1. Combined with the results of the Polyphen-2 analysis and multiple species conservation analysis, we speculate that the four missense mutations located in the UCH domain (Phe50Ser, Glu125Val, Arg150Gly, and Tyr189Cys) have implied structural and functional significance, which may inactivate BAP1 by impacting its deubiquitinase activity, structure stability, or substrate binding ability.

Figure 2.

Schematic representation of the somatic BAP1 mutations in HCC. A. Distribution of the somatic BAP1 mutations in HCC. B. The full-length three-dimensional model of BAP1 protein and BAP1 missense mutations. The wild and mutated residues were shown in red and yellow spheres, respectively. C. Surface view of BAP1 protein and BAP1 missense mutations. The mutation residues were shown in red. D. A detailed view of the mutation residues F50S, E125V, and the other 4 residues (R150G, Y189C, L294R, and Q322H), respectively. The wild and mutated residues were shown in red and yellow sticks, respectively.

Clinical implications of BAP1 mutations in HCC

The clinical implications of somatic BAP1 mutations were investigated in the 105 HCC patients. We found that those somatic BAP1 mutations were more inclined to happen in HCC patients who had high level of AFP (5/58, 8.6%), tumor size >5 cm (4/42, 9.5%), multiple tumors (2/14, 14.3%), poor tumor differentiation (3/33, 9.1%), presence of vascular invasion (5/35, 14.3%), and presence of tumor recurrence (4/56, 7.1%). Additionally, the correlation analysis showed that these mutations were significantly correlated to vascular invasion (P = 0.015), but not with other clinicopathological characteristics (P>0.05) (Table S5).

BAP1 was downregulated in human HCC

Given that BAP1 is somatically mutated in HCC, we then evaluated the expression of BAP1 in HCC, to further explore the functional role of BAP1 in HCC. We found that the mRNA expression of BAP1 was significantly reduced in 74.2% (49/66) of HCC, versus the adjacent non-tumor tissues (P = 0.032) (Figure 3A). Western blot assays showed similar results (Figure 3B). Next, IHC staining of BAP1 was performed in TMAs containing a cohort of 396 HCC patients. Low BAP1 expression (scored as negative or weak) was observed in 46.2% of tumor tissues (183/396) but only 27.2% of corresponding adjacent non-tumor liver tissues (108/396) (Figure 3C). Further, we examined the mRNA expression of BAP1 in the 105 HCC tissues that had been sequenced by Sanger sequencing to investigate the pertinence of mutation and expression of BAP1 in HCC. The results revealed that the mRNA expression level was comparable in HCC tissues without or with BAP1 mutations (P = 0.386), indicating that the mRNA expression of BAP1 was independent of its mutation status (Figure S1). Moreover, to assess the correlation of BAP1 expression with early tumor recurrence, we further analyzed the mRNA and protein expression of BAP1 in 12 HCC samples and found that the expression of BAP1 in patients with tumor recurrence were much lower than in those without tumor recurrence (Figure 3D).

Figure 3.

BAP1 was downregulated in HCC and correlated to poor postoperative survival. A. The expression level of BAP1 mRNA in paired tumor (T) and adjacent non-tumor samples (P) from 66 patients with HCC. B. The expression level of BAP1 protein in paired tumor (T) and adjacent non-tumor samples (P) from 12 patients with HCC. The protein level of BAP1 was expressed relative to GAPDH in the densitometry analysis. C. Representative IHC images of BAP1 protein in HCC tumor and matched adjacent non-tumor samples. The statistics of the score of BAP1 staining intensity in TMA was shown in the bar graph. Scale bars = 50 µm. D. The expression levels of BAP1 mRNA and protein in HCC tissues without and with tumor recurrence after curative hepatectomy. E. Kaplan-Meier curves for OS and RFS in HCC patients based on BAP1 expression. Bar graphs described quantification of three independent results. ***P<0.001.

Clinical significance of BAP1 protein expression in HCC

In order to explore the clinical relevance of BAP1 in HCC, we dichotomized the 396 patients with HCC as BAP1^high group (scored strong or moderate, n = 213) or BAP1^low group (scored weak or negative, n = 183). By analyzing the relationship of BAP1 expression and clinicopathological characteristics, we identified that low BAP1 expression was positively correlated to aggressive tumor phenotypes, such as greater tumor number (P = 0.044), larger tumor size (P = 0.007), appearance of vascular invasion (P = 0.012), and higher Chinese HCC stage (P = 0.002) (Table 2). Notably, the RFS and OS of patients with low BAP1 expression were significantly poorer compared with patients with high BAP1 expression, respectively (P<0.001 and P<0.001) (Figure 3E). The median RFS and OS were 55.0 months and 43.0 months for patients in BAP1^low group as compared with 118.5 months and 110.0 months for patients in BAP1^high group, respectively. Importantly, multivariate Cox analysis, which incorporated all significant variables identified by univariate analysis, revealed that low BAP1 expression was an independent predictor for RFS and OS (Table 3).

Table 2.

Correlation between BAP1 expression and clinicopathological characteristics in 396 HCC patients

Clinicopathological indexes		BAP1
		Low	High	P
Age (year)	≤50	89	111	0.490
	>50	94	102
Sex	Female	28	22	0.138
	Male	155	191
HBsAg	Negative	32	39	0.831
	Positive	151	174
HCV	Negative	179	207	0.758*
	Positive	4	6
AFP (ng/ml)	≤20	68	73	0.550
	>20	115	140
ALT (U/l)	≤75	167	192	0.704
	>75	16	21
Liver cirrhosis	No	35	39	0.836
	Yes	148	174
Tumor size (cm)	≤5	93	137	0.007
	>5	90	76
Tumor number	Single	138	178	0.044
	Multiple	45	35
Satellite	No	162	198	0.126
	Yes	21	15
Tumor encapsulation	Complete	101	131	0.204
	None	82	82
Vascular invasion	No	127	171	0.012
	Yes	56	42
Tumor differentiation	I-II	121	139	0.857
	III-IV	62	74
Chinese HCC Stage	I	104	153	0.002
	II+IIIa	79	60

Abbreviations: AFP, alpha-fetoprotein; ALT, alanine aminotransferase.

^*Fisher’s exact tests, and x² tests for all the other analysis.

Table 3.

Univariate and multivariate analyses of prognostic factors in 396 HCC patients

Variables	OS		RFS
	HR (95% CI)	P	HR (95% CI)	P
Univariate analysis
Age, year (>50 versus ≤50)	0.930 (0.719-1.204)	0.584	0.980 (0.735-1.306)	0.890
Sex (male versus female)	1.205 (0.796-1.825)	0.379	1.074 (0.688-1.675)	0.755
HBsAg (positive versus negative)	1.417 (0.989-2.031)	0.058	1.106 (0.763-1.605)	0.594
AFP, ng/mL (>20 versus ≤20)	1.606 (1.211-2.130)	0.001	1.149 (0.853-1.549)	0.361
ALT, U/L (>75 versus ≤75)	0.919 (0.581-1.454)	0.719	0.776 (0.457-1.316)	0.347
Liver cirrhosis (yes versus no)	1.662 (1.148-2.406)	0.007	2.041 (1.326-3.141)	0.001
Tumor size, cm (>5 versus ≤5)	1.460 (1.127-1.892)	0.004	0.920 (0.682-1.241)	0.584
Tumor number (multiple versus single)	2.115 (1.578-2.836)	0.000	1.969 (1.399-2.769)	0.000
Satellite (yes versus no)	1.651 (1.090-2.503)	0.018	1.210 (0.724-2.021)	0.467
Tumor encapsulation (none versus complete)	1.447 (1.116-1.875)	0.005	1.146 (0.854-1.536)	0.363
Vascular invasion (yes versus no)	1.792 (1.351-2.377)	0.000	1.618 (1.171-2.236)	0.004
Tumor differentiation (III-IV versus I-II)	1.086 (0.829-1.424)	0.550	0.915 (0.672-1.248)	0.576
Chinese HCC stage (II+IIIa versus I)	2.075 (1.597-2.696)	0.000	1.861 (1.385-2.502)	0.000
BAP1 (low versus high)	1.847 (1.424-2.396)	0.000	1.679 (1.257-2.242)	0.000
Multivariate analysis
AFP, ng/mL (>20 versus ≤20)	1.561 (1.173-2.076)	0.002	NA	NA
Liver cirrhosis (yes versus no)	1.751 (1.203-2.550)	0.003	1.971 (1.281-3.033)	0.002
Tumor size, cm (>5 versus ≤5)	1.401 (1.073-1.829)	0.013	NA	NA
Chinese HCC stage (II+IIIa versus I)	1.726 (1.320-2.257)	0.000	1.741 (1.293-2.343)	0.000
BAP1 (low versus high)	1.745 (1.340-2.272)	0.000	1.573 (1.177-2.103)	0.002

Cox proportional hazards regression model. Variables for multivariate analyses were adopted for their prognostic significance by univariate analysis (P<0.05), and these variables were assessed for prognostic significance by univariate analysis with forward stepwise selection (forward, likelihood ratio). Abbreviations: 95% CI, 95% confidence interval; AFP, alpha-fetoprotein; ALT, Alanine aminotransferase; HR, Hazard Ratio; NA, not applicable.

BAP1 deficiency promoted HCC progression and metastasis

We examined the expression of BAP1 in seven HCC cell lines. MHCC97L, MHCC97H and HCCLM3 cell lines showed low BAP1 expression, while relatively high BAP1 expression was observed in SMMC7721, Huh7, HepG2, and Hep3B cell lines (Figure 4A). To investigate the function role of BAP1 in HCC, wild-type (WT) BAP1 and shBAP1 were transfected into HCCLM3 and HepG2 cell lines respectively, both of which were validated to be WT for BAP1. The transfection efficiency of these lentiviral vectors was validated using qRT-PCR, western blot, and immunofluorescence (Figures 4A, S2).

Figure 4.

BAP1 deficiency enhanced HCC progression and metastasis in vitro. A. The expression level of BAP1 mRNA and protein in indicated HCC cell lines. B. Effects of overexpressing and downregulating BAP1 on HCC cell proliferation by using CCK8 assay. C. Effects of overexpressing and downregulating BAP1 on HCC cell migration and invasion by using scratch assay and Matrigel invasion assay, respectively. D. Effects of overexpression of WT BAP1 and mutant BAP1 on cell proliferation using CCK8 assay. E and F. Effects of overexpression of WT BAP1 and mutant BAP1 on cell migration and invasion using scratch assay and Matrigel invasion assay, respectively. Scale bars = 200 µm. Bar graphs described quantification of three independent results. *P<0.05, **P<0.01, ***P<0.001.

In the CCK8 proliferation, scratch and Matrigel invasion assays, overexpressing BAP1 in HCCLM3 cells markedly inhibited cell viability, proliferation, migration and invasion, whereas downregulating BAP1 in HepG2 cells yielded an opposing effect (Figure 4B and 4C). Next, to validate the biological significance of BAP1 mutations, we generated plasmids encoding WT BAP1 or mutant BAP1 (Phe50Ser, Glu125Val, Arg150Gly, and Tyr189Cys), and then transfected them into HCCLM3 cells. As expected, overexpressing WT BAP1 in HCCLM3 cells notably inhibited cell proliferation. However, overexpressing mutant BAP1 in HCCLM3 cells did not yield a significant inhibiting effect. Moreover, HCCLM3 cells overexpressing mutant BAP1 displayed markedly increased proliferation, migration and invasion capability, versus HCCLM3 cells overexpressing WT BAP1 (Figure 4D-F). Evidently, those results indicated that BAP1 was a tumor suppressor of HCC and inactivated by its somatic mutations in HCC.

In the in vivo study, human HCC orthotopic transplantation model was established. The tumor size of xenografts originated from HepG2-shBAP1 and HCCLM3-Mock cells were 1.81 ± 0.27 cm³ and 2.33 ± 0.43 cm³, which were notably larger than that originated from HepG2-Mock and HCCLM3-BAP1 cells (0.81 ± 0.46 cm³ and 1.18 ± 0.52 cm³, P = 0.003 and P = 0.005, respectively) (Figure 5A). Pulmonary metastasis of HCC occurred in 60% (3/5) of HepG2-shBAP1 mice and 100% (5/5) of HCCLM3-Mock mice, and the rates were significantly high than the rates of pulmonary metastasis in HepG2-Mock and HCCLM3-BAP1 mice (0% [0/5] and 20% [1/5], P = 0.038 and P = 0.010, respectively) (Figure 5B). Node number of pulmonary metastasis of each grade was huger in HepG2-shBAP1 and HCCLM3-Mock mice, as compared with HepG2-Mock and HCCLM3-BAP1 mice, respectively (Figure 5C).

Figure 5.

BAP1 deficiency enhanced HCC progression and metastasis in vivo. A. Representative xenograft tumors originated from indicated HCC cell lines (Left panel). Tumor volume of the tumors was gauged at 6 weeks (Right panel). B. Incidence of lung metastasis among the different animal groups. C. Representative IHC images of hematoxylin-eosin staining of lung metastatic nodules in different animal groups (Left panel). The grades of lung metastatic nodules in each animal group were shown (Right panel). Scale bars = 200 µm. Date depicted the mean ± SD (n = 5) and were represented quantification of three independent results. **P<0.01.

BAP1 complexed with PTEN and stabilized PTEN via deubiquitination

The biological function role of DUBs in cancer primarily depends on their substrate proteins, which are always important factors in tumor initiation and progression [22]. Therefore, an online protein-protein interaction repository (BioGRID database) were searched, and we identified PTEN was one of the BAP1 binding proteins (Figure 6). PTEN is an important tumor suppressor in cancers, and its ubiquitination status is of great importance to its activity and expression [23]. Therefore, we sought to determine whether BAP1 can complex with PTEN and stabilize PTEN through deubiquitination.

Figure 6.

BAP1 interacting proteins by searching BioGRID database. Brown node represents the query gene (BAP1); Blue node represents associated gene from same organism (Homo sapiens); Yellow node represents associated gene from different organism; Yellow line represents association with physical evidence; Greater node size represents increased connectivity and thicker edge sizes represent increased evidence supporting the association.

To examine BAP1 complexing with PTEN, co-IP assays were performed in HepG2 cells. The results confirmed that endogenous BAP1 formed a complex with endogenous PTEN in HepG2 cells (Figure 7A). Several ubiquitin ligases have been reported to ubiquitinate PTEN protein, such as WWP2, CHIP, and TRIM27, etc. [24-26]. In order to investigate whether BAP1 could block the degradation of PTEN induced by ubiquitin ligases, PTEN, WWP2, or BAP1 were transfected into HCCLM3 cells and PTEN protein level was detected in indicated cells. Interestingly, WWP2 overexpression downregulated PTEN protein level, while BAP1 overexpression antagonized the degradation of PTEN mediated by WWP2 (Figure 7B).

Figure 7.

BAP1 complexed with PTEN and stabilized PTEN via deubiquitination. A. Endogenous BAP1 or PTEN was immunoprecipitated from HepG2 cells with BAP1 or PTEN antibodies, respectively. IgG was used for negative control. Western blot was used for testing the endogenous BAP1 or PTEN. B. HCCLM3 cells were transfected with BAP1, PTEN, or E3 ligase WWP2, and the indicated cell lysates were analyzed by western blot. C. HCCLM3 cells were transfected with PTEN, Ub, WT BAP1, or mutant BAP1. HepG2 cells were transfected with shBAP1 and Ub. All the cells were treated by MG132 (10 mM, 6 h) before collection, immunoprecipitated by PTEN antibody and immunoblotted with Ub and PTEN antibodies. D. HCCLM3 cells were transfected with WT BAP1 or mutant BAP1, treated by cycloheximide (CHX) for indicated times before collection, and then immunoblotted with BAP1, PTEN and GAPDH antibodies (Upper panel). HepG2 cells were transfected with shBAP1, treated by cycloheximide (CHX) before collection for indicated times, and then immunoblotted with BAP1, PTEN and GAPDH antibodies (Lower panel). The protein level of PTEN was expressed relative to GAPDH in the densitometry analysis. E. Immunostaining images of BAP1 and PTEN in representative HCC patients using Serial HCC sections. Patient 1 had high expression of BAP1 and PTEN, whereas patient 2 had low expression of BAP1 and PTEN. Scale bars = 50 µm. F. Prognostic values of BAP1 and PTEN in HCC by Kaplan-Meier survival analysis. I, BAP1^high and PTEN^high; III, BAP1^low and PTEN^low; and II, others. Bar graphs described quantification of three independent results. *P<0.05, **P<0.01, ***P<0.001.

To explore whether BAP1 can deubiquitinate PTEN, we transfected PTEN, Ub, WT BAP1 or inactivating mutant BAP1 (Phe50Ser, Glu125Val, Arg150Gly, and Tyr189Cys) into HCCLM3 cells. The degradation of cell protein was blocked by treating indicated cells using MG132 (a proteasome inhibitor). Then, we tested polyubiquitinated PTEN proteins by western blot using anti-Ub antibody. The results showed that WT BAP1, instead of inactivating mutant BAP1 significantly decreased PTEN protein polyubiquitination (Figure 7C). By contrast, knockdown of endogenous BAP1 significantly increased endogenous PTEN protein polyubiquitination in HepG2 cells (Figure 7C). To further investigate whether BAP1 could stabilize PTEN against degradation, we overexpressed WT BAP1 or inactivating mutant BAP1 (Phe50Ser, Glu125Val, Arg150Gly, and Tyr189Cys) into HCCLM3 cells. The protein translation was blocked by treating indicated cells using cycloheximide (CHX, a protein synthesis inhibitor). Evidently, overexpression of WT BAP1, but not inactivating mutant BAP1, prominently increased the stability of endogenous PTEN protein (Figure 7D).

Conversely, knockdown of endogenous BAP1 markedly decreased the stability of endogenous PTEN protein in HepG2 cells (Figure 7D).

Moreover, to validate the correlation of BAP1 and PTEN expression in HCC, we further performed IHC staining of BAP1 and PTEN in TMAs containing a cohort of 396 HCC patients. Likewise, IHC results demonstrated that the expression of BAP1 and PTEN were positively correlated with each other in HCC (Figure 7E; Table S6). To assess the combined effects of BAP1 and PTEN on the postoperative survival of HCC, the 396 patients with HCC were categorized into three groups as follows based on the BAP1 and PTEN expression: HCC patients with both high BAP1 and PTEN expression were classified as group I; HCC patients with both low BAP1 and PTEN were classified as group III; the other HCC patients were classified as group II. Strikingly, HCC patients in group I had better OS and RFS than HCC patients in group II or group III (Figure 7F).

BAP1 deficiency led to activation of AKT signaling and induction of EMT

AKT signaling is an important downstream target of PTEN, and AKT/GSK-3β/Snail signaling has been reported to induce EMT in cancer [23,27,28]. Therefore, we further detected the cellular morphology, and the expression of molecular markers of AKT/GSK-3β/Snail signaling and EMT in indicated HCC cells. Overexpression of WT BAP1 in HCCLM3 cells induced a typical normal epithelial phenotype (cobblestone-like appearance), Downregulation of BAP1 in HepG2 cells led to an obviously morphological change to a fibroblastic morphology (spindle-like appearance) (Figure S3A). Furthermore, activating AKT/GSK-3β/Snail signaling, downregulation of E-cadherin, as well as upregulation of vimentin, N-cadherin, and snail (a EMT regulator), were observed in HCCLM3-Mock cells and HepG2-shBAP1 cells, as compared with HCCLM3-WT BAP1 cells and HepG2-Mock cells, respectively (Figure 8A). Similarly, activating AKT/GSK-3β/Snail signaling and the typical EMT phenotype were also observed in HCCLM3 cells overexpressing inactivating mutant BAP1, as compared with HCCLM3-BAP1 cells (Figures 8B, S3B). Together, these results demonstrated that BAP1 deficiency activated AKT/GSK-3β/Snail signaling and induced EMT in HCC.

Figure 8.

BAP1 deficiency induced hyperactivity of AKT signaling through PTEN, which is necessary for EMT and invasive capacity of HCC cells. A. Protein levels of the AKT/GSK-3β/Snail signaling and the EMT markers were compared in indicated HCC cells. B. Protein levels of the AKT/GSK-3β/Snail signaling and the key EMT markers were compared in indicated cells. C. HCCLM3-BAP1 cells were transfected with shPTEN, or treated with AKT inhibitor (MK2206) after the transfection of shPTEN. Protein levels of the AKT/GSK-3β/Snail signaling and the key EMT markers were compared in indicated cells (Left panel). The proliferation, migratory, and invasive capacities of indicated HCC cells were examined by functional assays in vitro (Right panel). D. HepG2-shBAP1 cells were transfected with PTEN. Protein levels of the AKT/GSK-3β/Snail signaling and the key EMT markers were compared in indicated cells (Left panel). The proliferation, migratory, and invasive capacities of indicated HCC cells were examined by functional assays in vitro (Right panel). Bar graphs described quantification of three independent results. **P<0.01, ***P<0.001.

PTEN protein level is vital for BAP1-regulated EMT and cell invasion

To elucidate the role of PTEN in EMT and cell invasion regulated by BAP1 in HCC, we transfected shPTEN into HCCLM3-BAP1 cells to inhibit of PTEN expression. Interestingly, shPTEN transfection in HCCLM3-BAP1 cells induced evident cell morphology change from epithelial to fibroblastic phenotype, reversing the morphology change caused by BAP1 overexpression (Figure S4A). Consistent changes in key molecular markers of EMT and activating AKT/GSK-3β/Snail signaling were also observed in these cells, and the proliferation, migratory and invasive capacities of these cells were significantly enhanced (Figures 8C, S5A). However, after we treated these cells using AKT inhibitor MK2206, the effects of shPTEN in HCCLM3-BAP1 cells were abolished, thus indicating that activating AKT/GSK-3β/Snail signaling was necessary for HCC cell invasion and EMT (Figures 8C, S4A, S5A). To further validate these findings, we also transfected PTEN into HepG2-shBAP1 cells. As expected, PTEN overexpression in HepG2-shBAP1 cells obviously induced cell morphology change from fibroblastic to epithelial phenotype, reversing the morphology change caused by BAP1 downregulation (Figure S4B). Simultaneous changes in key molecular markers of EMT were consistently observed in these cells, which also showed inactivating AKT/GSK-3β/Snail signaling (Figure 8D). Moreover, PTEN overexpression markedly suppressed cell viability, proliferation, migration and invasion of HepG2-shBAP1 cells (Figures 8D, S5B). These findings supported the crucial role of PTEN in BAP1-regulated EMT and cell invasion.

Discussion

In our study, we systematically investigated the biological significance as well as the underlying mechanisms of BAP1 and its mutants in HCC progression and metastasis. The results demonstrated that BAP1 served as a tumor suppressor in HCC, which was evidently downregulated and somatically mutated. Moreover, we also revealed that BAP1 complexed with, deubiquitinated, and stabilized PTEN to inhibit EMT and cell invasion in HCC.

Initiation and progression of HCC is considered a complex biological process partially driven by genetic alterations that activate oncogenes or inactivate tumor suppressors [17]. In our previous study, we found that BAP1 harbored recurrent somatic mutations in HCC and was defined as a potential driver gene of HCC [18]. Here, our mutation prevalence screen by Sanger sequencing showed that BAP1 was mutated in 5.71% of HCC patients, especially in those with a higher level of AFP, larger tumor bulk, greater tumor number, poorly differentiated tumor, presence of vascular invasion or tumor recurrence. The mutation frequency of BAP1 in HCC identified in our study agrees with the previous finding that identified by whole-exome sequencing [17]. Meanwhile, our study showed that BAP1 mutations were mainly missense mutations and 62.5% of which were located in the UCH domain, which were also consistent with the typical BAP1 mutation pattern in cancers [29]. Structural and functional analysis indicated that the four missense mutations located in the highly conserved UCH domain region probably harbor implied functional significance, which may inactivate BAP1 by impacting its deubiquitinase activity, structure stability, or substrate binding ability. Our subsequent experimental results confirmed the inactivating effects of somatic mutations on protein BAP1 in HCC. However, whether other mechanisms, such as posttranscriptional or epigenetic modification, inactivate protein BAP1 in HCC requires further exploration.

Credible data identified that BAP1 was obviously reduced in HCC, and its downregulation positively correlated to aggressive tumor phenotypes, such as greater tumor nodules, larger tumor bulk, appearance of vascular invasion, and higher Chinese HCC stage, all of which are hallmarks of recurrence and poor prognosis of HCC [30]. Indeed, survival analysis confirmed that BAP1 was an independent predictor for postoperative survival, and its downregulation was associated with poor prognosis for HCC patients. Those data was in accordance with studies of other cancer types, in which tumor suppressor role of BAP1 was reported [5,8,9]. Furthermore, our functional assays revealed that WT BAP1 but not mutant BAP1 significantly inhibited HCC cell proliferation, invasion, EMT in vitro, and tumor progression and metastasis in vivo. Collectively, these results convinced us that BAP1 serves as a tumor suppressor of HCC and a potential treatment target for HCC.

The biological function of BAP1 in cancer always depends on its substrate proteins, and several lines of evidence in our study support BAP1 as a PTEN DUB. First, BAP1 did form a complex with PTEN. Second, BAP1 decreased PTEN polyubiquitination, and stabilized PTEN against degradation in a DUB activity-dependent manner. Third, BAP1 was downregulated and positively correlated to PTEN in human HCC; HCC patients with both high BAP1 and high PTEN expression showed the best prognosis. Finally, PTEN acts as a vital tumor suppressor in various malignancies, which is also validated in HCC [23,31]. BAP1 inhibited cell proliferation, invasion, and EMT partially via PTEN, because overexpressing PTEN in HCC cell partially reversed the BAP1-knockdown-induced cell proliferation, invasion, and EMT, while downregulating PTEN in HCC cell partially blocked the inhibitory effects. Notably, the four missense mutations of BAP1, located in the highly conserved UCH domain region, abolished the effect that deubiquitinated and stabilized PTEN, which may attribute to that these mutations of BAP1 impacted its deubiquitinase activity or its binding with PTEN. Furthermore, as PTEN harbored loss of one allele in a large fraction of human cancers including HCC [23], we speculate that BAP1 deficiency may also drive HCC progression and metastasis with heterozygous inactivation of PTEN, and we will explore this issue in our future research projects.

Furthermore, we demonstrated that BAP1 deficiency could induce HCC cell EMT through typical EMT patterns. Overexpressing PTEN suppressed the the typical EMT change in BAP1 deficient HCC cells, while knockdown of PTEN reinduced EMT in HCC cells overexpressing WT BAP1. Additionally, BAP1 has been reported to regulate various important cell signaling pathways, such as IP3R3 mediated Ca²⁺ signaling, ER stress gene-regulatory network, and the DNA damage response, etc. [5,32-34]. Here, we showed that BAP1 negatively regulated AKT/GSK-3β/Snail signaling by complexing with PTEN, and that BAP1 deficiency or PTEN downregulation induced activating AKT/GSK-3β/Snail signaling was necessary for HCC cell invasion and EMT. Hence, these observations suggested that activating EMT signaling events in HCC progression caused by BAP1 deficiency are vital for HCC cell survival.

In conclusion, this study identified that BAP1 served as a tumor suppressor of HCC, which was somatically mutated and markedly downregulated. BAP1 modulated tumor invasion and EMT by complexing with PTEN and stabilizing PTEN via deubiquitination, whereas these tumor-inhibitory effects of BAP1 were abolished by inactivating mutations. Importantly, deficiencies of BAP1 were significantly associated with aggressive tumor phenotypes, and poor postoperative prognostic outcomes. The biological features of BAP1 in HCC advocate the further investigation as a candidate treatment target for HCC.

Disclosure of conflict of interest

None.