Guo-Ming Shi, Ai-Wu Ke, Jian Zhou, and Xiao-Ying Wang performed biochemical and cell biology experiments. Yang Xu, Zhen-Bin Ding, Ranjan Prasad Devbhandari, Shuang-Jian Qiu, and Ying-Hong Shi developed the spontaneous metastasis assay. Xiao-Yong Huang, Zhi Dai, Xin-Rong Yang, and Guo-Huan Yang contributed to the immunohistochemistry experiments. Jia Fan provided team leadership and project management. All authors discussed the results and commented on the article.
These authors contributed equally to this work.
Liver Cancer Institute, Zhongshan Hospital and Shanghai Medical School, Fudan University, Shanghai, People's Republic of China
Cancer Center, Institutes of Biomedical Sciences, Fudan University, Shanghai, People's Republic of China
Potential conflict of interest: Nothing to report.
Tetraspanin CD151 is involved in several pathological activities associated with tumor progression, including neoangiogenesis. However, the role and molecular mechanism of CD151 in the neoangiogenesis of hepatocellular carcinoma (HCC) remain enigmatic. We found that the level of expression of matrix metalloproteinase 9 (MMP9) was positively associated with CD151 expression in HCC cells. We developed a zone-by-zone blockade and demonstrated that overexpression of CD151 in HCC cells facilitated MMP9 expression through a phosphatidylinositol-3-kinase (PI3K)/protein kinase B (Akt)/glycogen synthase kinase 3β (GSK-3β)/Snail signaling pathway. In contrast, down-regulation of CD151 expression impaired the ability of HCC cells to form microvessels in vitro and reduced their in vivo metastatic potential. In a clinical setting, a significant correlation of the expression of CD151 with MMP9 expression and with microvessel density (MVD) was revealed by Pearson correlation analysis of HCC patients. The postoperative 3-, 5-, and 7-year overall survival rates of HCC patients with CD151high/MMP9high/MVDhigh were significantly lower than those of the CD151low/MMP9low/MVDlow group or groups in which only one or two of CD151, MMP9, and MVD were highly expressed. Cumulative recurrence rates were also highest in HCC patients with CD151high/MMP9high/MVDhigh in comparison with the other groups. Multivariate Cox proportional hazards analysis showed that the concomitant overexpression of CD151, MMP9, and MVD was an independent marker for predicting poor prognosis of HCC. Conclusion: Overexpression of CD151 up-regulated the expression of MMP9 through the PI3K/Akt/GSK-3β/Snail pathway. CD151-dependent neoangiogenesis appeared to promote the progression of HCC, and this suggests that CD151 may be useful as a high-priority therapeutic target for antiangiogenesis in HCC. HEPATOLOGY 2010
Hepatocellular carcinoma (HCC) is a highly vascular tumor characterized by neoangiogenesis, which contributes to the high rate of metastasis and dismal prognosis.1 Assessment of the microvessel density (MVD) by immunohistochemical staining for specific endothelial cell markers, such as CD34, has been shown to provide prognostic information independent of conventional pathological parameters in HCC patients.2 Repression of neoangiogenesis has become a promising approach for HCC therapy.1, 3 Recently, an increasing number of studies have shown that tumor cells have an important role in tumor angiogenesis.1 However, full details of the molecular mechanism underlying tumor-associated neoangiogenesis in HCC remain to be elucidated.
Tetraspanins, also known as the transmembrane 4 superfamily, are a family of proteins characterized by four highly conserved transmembrane domains. These proteins are thought to be involved in the regulation of a broad range of cellular functions, including fertilization, platelet aggregation, mobility, differentiation, and tumor metastasis.4 An unusual biochemical property of tetraspanins is that they form complexes by interacting with other tetraspanins and/or with a variety of transmembrane proteins, such as integrins and growth factor receptors, which are required for their function.4
CD151, one of the most important of the tetraspanins, has been extensively studied, especially in connection with the progression and prognosis of malignant tumors, including breast cancer, colon cancer, prostate cancer, and HCC.5-8 Initial evidence for the involvement of CD151 in metastasis came from a study that showed specific in vivo inhibition of metastasis in a human epidermoid carcinoma by an unknown antibody. Since then, reduction of CD151 expression in primary melanocytes by small interfering RNA (siRNA) has been shown to result in the loss of motility, whereas it has little effect on the steady-state levels of integrins. These alterations can also be reversed if CD151 is re-expressed.9 Recent work continues to clarify the role of CD151 in metastasis.5 For example, CD151 now appears to probably function by linking laminin-binding integrins to signaling molecules, growth factor receptors, and other tetraspanins, and this results in the increased motility and invasiveness of cancer cells.5 This process most likely contributes to the activated pathways mediated by small guanosine triphosphatases, which increase GTP binding to cell division control protein 42 homologue (Cdc42) and Rac, important organizers of the cell cytoskeleton.10
Previously, we demonstrated that CD151 was positively associated with both in vivo and in vitro invasiveness of HCC. We also found that CD151 was a novel marker for predicting the prognosis of HCC.6 In addition, overexpression of CD151 has been reported to activate the phosphatidylinositol-3-kinase (PI3K)/protein kinase B (Akt) signal and promote neovascularization in ischemia animal models.11 Moreover, mouse lung endothelial cells from CD151 null mice displayed a marked reduction in angiogenesis-related endothelial events, including migration, spreading, invasion, Matrigel contraction, tube and cable formation, and spheroid sprouting.12 Interestingly, immunohistological analysis of xenografts showed that neoangiogenesis observed at the subcutaneous border of CD151(+) tumors was less pronounced or absent in CD151(−) xenografts, and this contributed to a new role for CD151 as a regulator of communication between tumor cells and the endothelium.13 These data strongly suggest that CD151 orchestrates the tumor association of angiogenesis in HCC. However, the precise molecular mechanisms are still poorly defined, and the combined value of CD151 and neoangiogenesis in predicting the prognosis of HCC patients needs to be further evaluated.
Multiple lines of evidence have shown a link between CD151 and matrix metalloproteinases (MMPs), a family of multidomain, zinc-containing neutral endopeptidases that can degrade extracellular matrix components and thus promote the formation of a favorable microenvironment for tumor growth.14, 15 Recently, MMPs such as MMP9/gelatinase B, MMP2/gelatinase A, MMP3/stromelysin 1, and MMP7/matrylysin have emerged as master regulators of angiogenesis and tumor progression.15, 16 Of the MMPs, MMP9 is of particular interest because it seems to act as a switch for tumor angiogenesis.15, 16 CD151 appears to facilitate pericellular activation of MMPs by associating with proMMPs. The signal, initiated by CD151 homophilic interactions, prompts c-Jun binding to activator protein 1 sites in the MMP9 gene promoter and enhances MMP9 expression in MelJuSo cells.17 Reduced expression of MMP9 in a CD151-knockdown carcinoma cell line provides direct evidence to support the notion that CD151 is involved in MMP9 expression.17 In our previous study, HCC cell lines with CD151high were found to show higher MMP9 expression,6 and this made a profound impression on us. Given the special function of CD151 in cancer progression, further investigating the role and mechanism of CD151 in the expression of MMP9 and tumor neoangiogenesis in HCC is significant.
Here we investigated the expression and secretion of MMP9 in more depth by modifying the expression levels of CD151. We then explored the molecular mechanism underlying modulation of CD151 in MMP9 expression in HCCLM3 cells through the zone-by-zone blockade of the PI3K/Akt/glycogen synthase kinase 3β (GSK-3β)/Snail signal. We further explored the role of CD151 in tumor-associated neoangiogenesis and metastasis in vitro and in vivo. Finally, we evaluated the combined expression of CD151, MMP9, and MVD as a prognostic marker in HCC patients.
A highly metastatic human HCC cell line (HCCLM3), low-metastatic human HCC cell lines (MHCC97-L, PLC/PRF/5, Hep3B, and HepG2; American Type Culture Collection),6, 18, 19 and human umbilical vein endothelial cells (HUVECs; American Type Culture Collection) were used in this study. Male, athymic BALB/c nude mice (8 weeks old; Shanghai Institute of Material Medicine, Chinese Academy of Science, Shanghai, China) were raised under specific pathogen-free conditions. Animal care and experimental protocols were in accordance with the guidelines established by the Shanghai Medical Experimental Animal Care Commission.
Patients and Follow-Up.
Specimens taken from areas next to the margins of tumors were collected from 327 consecutive patients with HCC who underwent curative resection between 1997 and 2000 at the Liver Cancer Institute of Fudan University (Shanghai, China). The histopathological diagnosis was based on the World Health Organization criteria.20 The histological grade of tumor differentiation was determined according to the classification proposed by Edmondson and Steiner.21 Liver function was assessed by the Child-Pugh scoring system. Clinical tumor typing was performed according to the sixth edition of the tumor node metastasis (TNM) classification system of the Union Internationale Contre le Cancer. Ethical approval was obtained from the research ethics committee of Zhongshan Hospital, and written, informed consent was obtained from each patient. Follow-up data were summarized at the end of March 2007 with a median follow-up of 62 months (range = 4-121 months). The follow-up procedures were described in detail in our earlier study.6
Transfection and Clone Selection.
The pGCSIL-GFP-shRNA-CD151, Akt, and MMP9 lentiviral vectors (pGCSIL is a lentiviral vector) were purchased from Shanghai GeneChem Co., and the target shRNA sequences are listed in the supporting information. A pGCSIL-GFP-lentiviral vector was used as a control. The siRNA duplexes of Snail and GSK-3β were synthesized by Shanghai GeneChem, and the target siRNA sequences are included in the Supporting Information. The pcDNA3-CD151 plasmids were kindly provided by Hansoo Lee (Kangwon National University, Korea). The pcDNA3-GSK-3βS9A plasmids22 were a gift from Professor Yi-Zheng Wang (Institute of Neuron Sciences, Chinese Academy of Science). pcDNA3 plasmids were used as controls. The lentiviral vector and plasmid were transfected into cells as described elsewhere.6 Transfection of the siRNAs was performed with Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. Stably transfectant clones were validated by quantitative real-time polymerase chain reaction (qRT-PCR) and immunoblotting for the level of target gene expression (as shown in Supporting Information Fig. 1).
qRT-PCR, Reverse-Transcription Polymerase Chain Reaction (RT-PCR), and Immunoblotting.
HCC cell lines were used to investigate the expression of messenger RNA (mRNA) via qRT-PCR as described in a previous study.6 Sixty blindly chosen HCC samples were used to investigate the expression of mRNA via RT-PCR as described in a previous study.12 Primers for RT-PCR are listed in the Supporting Information. Five HCC cell lines, modified HCC cell lines, 60 HCC samples blindly chosen from the same cohort, and subcutaneous xenografts were used for immunoblotting.6 Antibodies used in this study are listed in the Supporting Information. All experiments were performed in triplicate.
Gelatin Zymography and Enzyme-Linked Immunosorbent Assay (ELISA).
The supernatant from HCCLM3, MHCC97-L, PLC/PRF/5, shRNA-CD151-HCCLM3, shRNA-MMP9-HCCLM3, HCCLM3-mock (HCCLM3-pGCSIL-GFP), and Hep3B cultured in a serum-free medium and from HCCLM3 treated with laminin 5 (Abcam, United Kingdom) was collected. The type IV collagenase activity and the concentrations of MMP9 in a conditioned medium were determined by gelatin zymography6 with a human MMP9 ELISA kit (R&D Systems); the concentrations of vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) were determined with human VEGF and bFGF ELISA kits (R&D Systems), respectively, essentially as described previously.23 All experiments were performed in triplicate.
Matrigel Assay, Aortic Ring Assay, and Mouse Cornea Micropocket Angiogenesis Assay.
The supernatant was collected from HCCLM3, shRNA-CD151-HCCLM3, shRNA-MMP9-HCCLM3, HCCLM3-mock, and Hep3B cells cultured in a serum-free medium. Matrigel angiogenesis and aortic ring assays were performed as described previously.12 Neoangiogenesis in HCCLM3-derived, shRNA-CD151-HCCLM3–derived, shRNA-MMP9-HCCLM3–derived, HCCLM3-mock–derived, and Hep3B-derived tumor tissues was assayed with the mouse cornea micropocket angiogenesis model as described previously.24 The vessel area was calculated as follows:
All experiments were performed in triplicate.
Spontaneous Metastasis Assay and Immunohistochemical Analysis.
Samples of 8 × 106 HCCLM3, shRNA-CD151-HCCLM3, shRNA-MMP9-HCCLM3, HCCLM3-mock, and Hep3B cells were used for spontaneous metastasis assays, as described previously.25 Lung metastases of shRNA-CD151-HCCLM3, shRNA-MMP9-HCCLM3, and HCCLM3-mock were visualized with fluorescence stereomicroscopy (Leica Microsystems Imaging Solutions, Ltd., Cambridge, United Kingdom). Serial sections were made for every tissue block from the lung, and the total number of lung metastases was counted under the microscope as described previously.6 There were five animals in each group. Immunohistochemical analysis of subcutaneous xenografts was performed as described elsewhere.6 Antibodies used in this study are listed in the Supporting Information.
Construction of Tissue Microarrays and Immunohistochemical Double Staining.
The construction of tissue microarrays was described in detail in our earlier study.6 Immunohistochemical double staining was performed as described elsewhere.26 Mouse anti-human CD151 antibodies (1:100; 11G5a, Serotec, NK), rat anti-human MMP9 antibodies (1:50; Cell Signal Tec, United States), mouse anti-human CD34 antibodies (1:100; DakoCytomation, Denmark), and rabbit polyclonal VEGF antibodies (1:100; Neomarkers, Fremont, CA) were used to detect the expression of CD151, MMP9, MVD, and VEGF. The density of positive staining of CD151, MMP9, and VEGF was measured as described previously.6 The MVD was evaluated as described elsewhere.27
Statistical analysis was performed with SPSS12.0 software (SPSS, Chicago, IL). Values are expressed as means and standard deviations. The Student t test and one-way analysis of variance were used for comparisons between groups. Correlation analysis was performed. Overall survival (OS) and time to recurrence were defined as described previously.28 OS and the cumulative recurrence rates were calculated by the Kaplan-Meier method and the log-rank test. Cox's proportional hazards regression model was used to analyze the independent prognostic factors. P < 0.05 was set as the level of statistical significance.
CD151 Up-Regulated Expression and Secretion of MMP9 in HCC Cells.
In an earlier study,6 we demonstrated that the bioactivity of MMP9 in the supernatant from HCCLM3 cells with a high level of expression of CD151 was much stronger than that in shRNA-CD151-HCCLM3 cells and HepG2 cells with low CD151 expression by gelatin zymography. We now explore the relationship between the expression of CD151 and MMP9 in HCC cell lines (HCCLM3, MHCC97-L, PLC/PRF/5, Hep3B, and HepG2) with different metastatic potentials by qRT-PCR and immunoblotting. Highly metastatic HCCLM3 cells with CD151high expression showed the highest expression of MMP9 at both the mRNA and protein levels, whereas low-metastatic HCC cell lines (MHCC97-L, PLC/PRF/5, Hep3B, and HepG2) showed low levels of expression of CD151 and MMP9 (Fig. 1A,B). When expression of CD151 in HCCLM3 cells was down-regulated by transfection of lentiviral vectors with shRNA, the expression of MMP9 was correspondingly reduced in comparison with HCCLM3-mock cells (Fig. 1C,D and Supporting Information Fig. 1A,B). Then, the bioactivity and concentration of MMP9 in the supernatant were further assayed by gelatin zymography and ELISA, and the results were the same as those described previously (Fig. 1B,C). Previous publications have shown that CD151 may increase secretion of MMP2 and MMP9 by forming a complex with integrin α3β1 and facilitating the integrin signal.29 We further investigated the bioactivity and concentration of MMP9 and MMP2 in the supernatants from HCCLM3-mock and shRNA-CD151-HCCLM3 cells treated or not treated with laminin 5 by gelatin zymography. Gelatin zymography showed that the concentration of MMP9 in the supernatant in HCCLM3-mock cells was significantly increased when HCCLM3-mock cells were treated with laminin 5. However, laminin 5 had no influence on the secretion of MMP9 in shRNA-CD151-HCCLM3 cells. Similarly, no differences in the concentration of MMP2 were noted between HCCLM3-mock and shRNA-CD151-HCCLM3 cells treated or not treated with laminin 5 (Fig. 1D,E). These data indicate that CD151 is likely to form a complex with integrin and be involved in the modulation of MMP9 expression and secretion.
CD151 Promoted Secretion of MMP9 via the PI3K/Akt/GSK-3β/Snail Pathway in HCC Cells.
To scrutinize the molecular mechanism underlying overexpression of CD151 in the up-regulation of MMP9, we investigated the expression of focal adhesion kinase (FAK) and p38 and the phosphorylation level of FAKTyr397 and p38Thr180/Tyr182 in shRNA-CD151-HCCLM3, HCCLM3-mock, and HCCLM3 cells treated with laminin 5 as described.30 We found that the phosphorylation level of FAKTyr397 was higher in HCCLM3-mock cells than that in shRNA-CD151-HCCLM3 cells. However, there was no difference in the expression of FAK and p38 or the phosphorylation level of p38Thr180/Tyr182 (Supporting Information Fig. 2). Then, we investigated the expression of Akt, GSK-3β, Snail, and MMP9 and the phosphorylation level of AktSer473 and GSK-3βSer9 in HCCLM3, Hep3B, and modified counterparts. In shRNA-CD151-HCCLM3 cells, the phosphorylation level of AktSer473 and GSK-3βSer9 and the expression of Snail and MMP9 in HCCLM3 cells were significantly reduced in comparison with HCCLM3-mock cells (Fig. 2A). On the contrary, when the expression of CD151 in Hep3B and shRNA-CD151-HCCLM3 cells was up-regulated through transfection of pcDNA3-CD151 plasmids, the phosphorylation level of AktSer473 and GSK-3βSer9 and the expression of Snail were strongly enhanced. Interestingly, the expression of MMP9 in the modified HCC cells was restored as well (Fig. 2B and Supporting Information Fig. 3A). These results indicate that overexpression of CD151 probably enhances the expression and secretion of MMP9 via the PI3K/Akt/GSK-3β/Snail signal. It has been reported that both mitogen-activated protein kinase (MAPK) and PI3K signaling can enhance the expression of MMP9 by inducing the accumulation of Snail in nuclei.31 Here, to further elucidate the PI3K/Akt/GSK-3β/Snail signal in MMP9 expression, first HCCLM3 cells were treated with 2-morpholin-4-yl-8-phenylchromen-4-one (LY294002; a specific PI3K inhibitor) or 1,4-diamino-2,3-dicyano-1,4-bis(2-aminophenylthio)butadiene [U0126; 25 μmol/L; a blockade of extracellular signal-regulated kinase 1/2 (ERK1/2)–MAPK], or the endogenic Akt was down-regulated by specific shRNA. The expression of MMP9 was reduced markedly in HCCLM3 treated with LY294002 (Fig. 2C) and shRNA-Akt-HCCLM3 cells (Fig. 2D) but only slightly reduced in HCCLM3 treated with U0126 (Fig. 2E). Immunoblotting showed that Snail and the phosphorylation level of GSK-3βSer9 were also down-regulated in HCCLM3 treated with LY294002 and shRNA-Akt-HCCLM3 cells, whereas the expression of GSK-3β was up-regulated. U0126 had no effect on the expression of GSK-3β and Snail in HCCLM3 cells (Fig. 2C-E). Therefore, the results indicate that the transcription of MMP9 induced by Snail probably depends on PI3K signaling pathways rather than MAPK signaling pathways in HCC cells. GSK-3 is a critical downstream molecule of the PI3K/Akt cell survival pathway whose activity can be inhibited by Akt-mediated phosphorylation at Ser9 of GSK-3β.32 The inhibition of expression of GSK-3β in HCCLM3 cells by siRNA or the blockade of its activity by GSK-3 inhibitor XV had no effect on CD151 or Akt expression but slightly up-regulated Snail and MMP9 expression and the phosphorylation level of AktSer473 (Fig. 2F,G). To further assay the functional role of GSK-3β in the PI3K/Akt/GSK-3β/Snail signal, we used the constitutively active GSK-3β mutant S9A, in which Ser9 was replaced with alanine, for transfection into HCCLM3 cells. The expression of Snail and MMP9 and the phosphorylation level of AktSer473 were markedly reduced, whereas CD151 and Akt expression remained unchanged (Fig. 2H). When the expression of Snail in HCCLM3 was inhibited by siRNA interference, the expression of CD151 and Akt remained stable, but the phosphorylation level of AktSer473 and GSK-3βSer9 and MMP9 expression were reduced (Fig. 2I). Finally, we interfered with the expression of MMP9 in HCCLM3 cells with shRNA. As anticipated, none of the aforementioned signal molecules was altered, other than the expression of MMP9 (Fig. 2J). On the basis of the aforementioned experiments, in which we formed a zone-by-zone blockade of the PI3K/Akt/GSK-3β/Snail signal in HCCLM3 cells, we concluded that CD151 promoted secretion of MMP9 via the PI3K/Akt/GSK-3β/Snail signal in HCCs.
Role of CD151 in Vascular Remodeling and Neoangiogenesis via Modulation of MMP9 Secretion.
The Matrigel assay was used to confirm the role of MMP9 induced by CD151 from the supernatant of HCCLM3, shRNA-CD151-HCCLM3, shRNA-MMP9-HCCLM3, HCCLM3-mock, and Hep3B cells in neoangiogenesis and vascular remodeling.33 Significantly more integrated capillary-like structures (an endothelial function crucial to angiogenesis) were found in HCCLM3 and HCCLM3-mock cells versus Hep3B cells, and this coincided with the level of CD151 in the supernatant (Fig. 3A). When the expression of CD151 was inhibited in shRNA-CD151-HCCLM3 cells, the capacity for neoangiogenesis and vascular remodeling was impaired (Fig. 3A). Similarly, shRNA-MMP9-HCCLM3 cells showed a markedly impaired capacity for neoangiogenesis and vascular remodeling (Fig. 3A). Interestingly, supplementation of shRNA-CD151-HCCLM3 and shRNA-MMP9-HCCLM3 groups with supernatant from HCCLM3 restored the ability of HUVECs to form capillaries (Supporting Information Fig. 4A and B), and this indicates that MMP9 is involved in CD151-dependent neoangiogenesis and vascular remodeling. The aortic ring assay12 demonstrated more neoangiogenesis when aortic rings were cultured in the supernatant collected from HCCLM3 and HCCLM3-mock cells. However, the microvascular sprouting ability was impaired when they were cultured with the supernatant from Hep3B, shRNA-CD151-HCCLM3, and shRNA-MMP9-HCCLM3 cells, and this suggests that CD151 probably has an important role in the formation of capillaries and vascular remodeling in vitro through secretion of MMP9 (Fig. 3B,D). In order to exclude the possibility of neoangiogenesis through the secretion of angiogenic factors, such as VEGF or bFGF, we compared the concentrations of VEGF and bFGF in the supernatant of shRNA-CD151-HCCLM3 and HCCLM3 cells by ELISA. The concentrations of VEGF and bFGF were 173.4 ± 5.9 and 32.6 ± 3.7 pg/mL in HCCLM3 cells, respectively, and 164.1 ± 7.4 and 32.1 ± 2.3 pg/mL in shRNA-CD151-HCCLM3 cells, respectively. The differences were not significant (P > 0.05), and this suggests that overexpression of CD151 does not affect the secretion of VEGF and bFGF. A mouse cornea micropocket angiogenesis model was successfully developed. In the HCCLM3 and HCCLM3-mock groups, the areas of neoangiogenesis were 1.4 ± 0.2 and 1.5 ± 0.1 mm2, respectively, which were larger than those for shRNA-CD151-HCCLM3, Hep3B, and shRNA-MMP9-HCCLM3 cells (0.7 ± 0.2, 0.5 ± 0.1, and 0.3 ± 0.1 mm2, respectively, P < 0.001; Fig. 3C,E), and this provided powerful evidence for the role of CD151 in neoangiogenesis.
CD151-Dependent Neoangiogenesis and Metastasis in a Spontaneous Metastasis Assay.
After the subcutaneous injection of HCCLM3, HCCLM3-mock, shRNA-CD151-HCCLM3, Hep3B, and shRNA-MMP9-HCCLM3 cells into nude mice, all groups successfully formed tumors (Fig. 4A). The tumor volumes of HCCLM3-derived and HCCLM3-mock–derived xenografts were 6.4 ± 1.4 and 5.4 ± 1.2 cm3, respectively, significantly larger than those of shRNA-CD151-HCCLM3, Hep3B, and shRNA-MMP9-HCCLM3 (2.4 ± 0.3, 2.6 ± 0.6, and 2.4 ± 0.4 cm3, respectively, P < 0.01; Fig. 4A). However, there was no significant difference in the tumor volume among shRNA-CD151-HCCLM3–derived, Hep3B-derived, and shRNA-MMP9-HCCLM3–derived xenografts (P > 0.05; Supporting Information Fig. 5). In order to exclude the differences in the tumor volume from the proliferation variation, five HCC cell–derived xenografts were assessed by immunostaining with antibody to Ki-67, a widely accepted marker of cell proliferation. However, there were no significant differences in the expression of Ki-67 among the five HCC cell–derived xenografts (Supporting Information Fig. 6). Notably, the MVDs of HCCLM3-derived and HCCLM3-mock–derived xenografts were 135.2 ± 16.4/0.74 mm2 and 139.2 ± 7.9/0.74 mm2, which were larger than those of the shRNA-CD151-HCCLM3 (45.2 ± 17.0/0.74 mm2), Hep3B (37.2 ± 12.7/0.74 mm2), and shRNA-MMP9-HCCLM3 groups (44.8 ± 16.9/0.74 mm2; Fig. 4B,C), and they coincided with the levels of expression of CD151 and MMP9 (Supporting Information Fig. 7A). More importantly, the shRNA-CD151-HCCLM3–derived, shRNA-MMP9-HCCLM3–derived, and Hep3B-derived xenografts also contained masses of necrotic tissues (Fig. 4B), and this suggests that CD151 mediates the expression of MMP9 and has a key role in neoangiogenesis. To explore the role of CD151 in the expression and secretion of VEGF, five HCC cell–derived xenografts were immunostained with antibody to VEGF, and no significant difference was noted (Supporting Information Fig. 8). Fluorescence stereomicroscopy showed that there were more lung metastasis lesions in the HCCLM3-mock group than in the shRNA-CD151-HCCLM3 group or the shRNA-MMP9-HCCLM3 group (Fig. 4D). Serial sections confirmed that the pulmonary metastasis rates and metastatic tumor clusters per mouse were 100% (5/5) and 132.8 ± 4.0 in the HCCLM3 group and 100% (5/5) and 134.0 ± 8.0 in the HCCLM3-mock group but were 20% (1/5) and 33.6 ± 19.6 in the shRNA-CD151-HCCLM3 group, 20% (1/5) and 5.6 ± 12.5 in the Hep3B group, and 20% (1/5) and 24.0 ± 22.8 in the shRNA-MMP9-HCCLM3 group (P < 0.05; Fig. 4E,F and Supporting Information Fig. 7B). The numbers of lung metastatic loci in xenografts were also consistent with their expression of CD151, MMP9, and MVD (Fig. 4C-F and Supporting Information Fig. 7A,B), and this demonstrates that CD151-dependent neoangiogenesis is modulated through modification of MMP9 expression and is involved in the metastasis of HCC.
CD151high HCC Samples Harbored More Neoangiogenesis and Coincided with a High Level of Expression of MMP9.
We investigated the expression of CD151, MMP9, and CD34 by immunohistochemical double staining in a tissue microarray composed of primary tumors of 327 HCC patients. Representative cases of immunohistochemical double staining of all three markers are shown in Fig. 5A-D. Correlation analysis showed that HCC with CD151high expression tended to have high MMP9 expression and MVD and vice versa (rCD151 vs. MMP9 = 0.663, P < 0.001, and rCD151 versus MVD = 0.610, P < 0.001; Fig. 5E). An association between the expression of CD151 and MMP9 was further investigated by RT-PCR and immunoblotting in 60 HCC samples. Semiquantitative analysis of gel bands showed that the expression of MMP9 was tightly associated with the expression of CD151 at the mRNA and protein levels (Fig. 5F). To further assay the role of VEGF-A (VEGF) in CD151-dependent angiogenesis, we also investigated the expression of VEGF by immunohistochemical staining in 327 HCC patients. Representative cases of immunohistochemical staining are shown in Supporting Information Fig. 9A. Similarly to a previous study,34 high tissue VEGF levels predicted poor OS and higher cumulative recurrence rates in HCC patients (unpublished data), whereas correlation analysis showed that there was a slight association between the expression of CD151 and VEGF (rCD151 vs. VEGF = 0.115, P = 0.037). Therefore, we infer that overexpression of CD151 probably up-regulated the expression of MMP9 and subsequently facilitated the formation of new vessels in HCCs.
Concomitant Overexpression of CD151, MMP9, and MVD-CD34 Staining as an Adverse Prognostic Marker for HCC Patients.
Overexpression of CD151 or a high MVD alone was correlated with a poor prognosis for HCC patients.6, 27 To evaluate the prognostic significance of the overexpression profile of CD151, MMP9, and MVD together, immunohistochemical double-staining analysis of CD151, MMP9 expression, and MVD-CD34 staining was performed. Simultaneously higher levels of CD151, MMP9 expression, and MVD were observed in HCC tissues with a malignant phenotype (e.g., microvascular invasion, larger size, and dedifferentiation; Supporting Information Table 1). However, other clinical characteristics, including age, sex, hepatitis B surface antigen background, liver cirrhosis, preoperative treatment, preoperative serum alpha-fetoprotein, Child-Pugh score, tumor encapsulation, and TNM stage, were not directly related to the concomitant overexpression of the three markers (Supporting Information Table 1).
The 3-, 5-, and 7-year OS in the whole population was 67.3%, 54.1%, and 44.3%, respectively, and the cumulative recurrence rates were 36.7%, 45.6%, and 48.6%, respectively. Univariate analysis revealed that the tumor size (>5 cm), multiple tumors, vascular invasion, and a high TNM stage were predictors for low OS and high cumulative recurrence. Tumor differentiation was associated with OS. Other characteristics had no prognostic significance for OS and cumulative recurrence (Table 1). Expression of CD151, MMP9, or MVD was also found to be correlated with OS and cumulative recurrence rates (Table 1). The 3-, 5-, and 7-year OS in the CD151low group was significantly higher than that in the CD151high group (80.5% versus 52.3%, 66.7% versus 39.9%, and 56.9% versus 30.1%, respectively). The 3-, 5-, and 7-year cumulative recurrence rates in the CD151low group were significantly lower than those in the CD151high group (17.8% versus 58.2%, 29.9% versus 63.4%, and 33.9% versus 65.4%, respectively). The 3-, 5-, and 7-year OS in the MMP9low group was significantly higher than that in the MMP9high group (80.4% versus 54.3%, 63.2% versus 45.1%, and 52.2% versus 36.6%, respectively). The 3-, 5-, and 7-year cumulative recurrence rates in the MMP9low group were significantly lower than those in the MMP9high group (29.4% versus 43.9%, 42.9% versus 48.1%, and 48.5% versus 48.7%, respectively). The 3-, 5-, and 7-year OS in the MVDlow group was significantly higher than that in the MVDhigh group (77.3% versus 57.3%, 60.7% versus 47.6%, and 50.9% versus 37.8%, respectively). The 3-, 5-, and 7-year cumulative recurrence rates in the MVDlow group were significantly lower than those in the MVDhigh group (31.3% versus 42.1%, 41.1% versus 50.0%, and 46.4% versus 50.6%, respectively). When evaluating the combined effect of CD151, MMP9, and MVD on the prognosis of HCC, we classified patients into three subgroups according to their CD151, MMP9, and MVD-CD34 density: group I had high expression of all three markers, group II had high expression of one or two of the three markers, and group III had low expression of all three markers. We found that the 3-, 5-, and 7-year OS in group I was 50.9%, 39.1%, and 30.0%, respectively, significantly lower than the OS for groups II and III (Fig. 6A). The 3-, 5-, and 7-year cumulative recurrence rates in group I were 58.2%, 63.6%, and 64.5%, respectively, which were significantly higher than those for groups II and III (Fig. 6B).
Table 1. Univariate and Multivariate Analyses of Factors Associated with Survival and Recurrence
The Cox proportional hazards regression model was used.
Abbreviations: AFP, alpha-fetoprotein; CI, confidence interval; HBsAg, hepatitis B surface antigen; HR, hazard ratio; NA, not adopted; NS, not significant.
Patients were classified into three groups according to their CD151, MMP9, and MVD-CD34 density: group I (n = 110) had high expression of all three markers, group II (n = 99) had high expression of one or two of the three markers, and group III (n = 118) had low expression of all three markers. Individual variables were first adopted for their prognostic significance by multivariate analysis, and then the combination of CD151, MMP9, and MVD was analyzed.
Age (<52 versus ≥52 years)
Sex (male versus female)
HBsAg (negative versus positive)
Liver cirrhosis (yes versus no)
Preoperative treatment (yes versus no)
Child-Pugh score (A versus B)
Serum AFP (<20 versus ≥20 ng/mL)
Tumor size (>5 versus ≤5 cm)
Tumor differentiation (I/II versus III/IV)
Tumor number (multiple versus single)
Tumor encapsulation (none versus complete)
Microvascular invasion (yes versus no)
TNM stage (I/II versus III)
CD151 density (<50% versus ≥50%)
MMP9 density (<40% versus ≥40%)
MVD-CD34 density (<261/0.74 mm2 versus ≥261/0.74 mm2)
Individual clinicopathological features that showed significance by univariate analysis were adopted as covariates in a multivariate Cox proportional hazards model, and then combined variables were further analyzed. Multivariate Cox proportional hazards analysis also showed that overexpression of CD151, MMP9, and MVD together was independent of other prognostic markers (large size, microvascular invasion, and multiple tumors) for both OS (P < 0.001) and cumulative recurrence (Table 1; P < 0.001).
Traditionally, tetraspanin CD151 may activate Rac and Cdc42 by facilitating the integrins and growth factor receptor signals or redistribute integrins by endocytosis and/or trafficking, with the end result being the promotion of motility and metastasis of tumor cells.4, 35, 36 In the present study, we consistently observed that overexpression of CD151 facilitated tumor-associated neoangiogenesis in HCC and apparently did so by engaging MMP9 as an agent via the PI3K/Akt/GSK-3β/Snail signal, and thus it promoted the progression of HCC.
An earlier study reported that homophilic interactions of tetraspanin CD151 up-regulated the expression of MMP9 in human melanoma (MelJuSo) cells through the FAK/p38/MAPK/JNK/c-Jun pathway.17 In contrast to the results with MelJuSo cells,17 we found that overexpression of CD151 in HCC cells up-regulated the expression of MMP9 by facilitating the PI3K/Akt/GSK-3β/Snail signal in HCC cells. One of the reasons for this inconsistency may reside in the special structural and functional characteristics of the tetraspanins.4 These proteins can assemble themselves into complexes consisting of a core structure surrounded by other specific proteins. This complex formation provides a great deal of variability, which in turn allows for specificity and functional differences to occur in different cell types.4, 37 Tetraspanin complexes can also present different functional profiles at different cell development stages, even though they may share several common components.4, 35
To our knowledge, the present study is the first to clearly demonstrate that overexpression of CD151 promotes MMP9 expression via the PI3K/Akt/GSK-3β/Snail cascade. A review of the relevant literature on these molecules indicates that all are implicated in the neoangiogenesis and invasiveness of cancer cells. Recent studies demonstrated that CD151 gene delivery activated the PI3K/Akt pathway, induced cell migration, survival, and production of proangiogenic factors such as nitric oxide, and also promoted neovascularization after myocardial infarction in rats.11 On the contrary, mouse lung endothelial cells from CD151null mice displayed a marked reduction in pathological angiogenesis-related endothelial events, which apparently were mediated by modulation of the molecular organization of laminin-binding integrins.12 An important downstream molecule of the PI3K/Akt pathway, phosphorylated GSK, mediates the effects of Akt on cell growth, proliferation, protection from proapoptotic stimuli, and stimulation of neoangiogenesis.32 The expression of Snail, a zinc-finger transcription factor, correlates with cancer invasion and poor prognosis in HCC patients and is induced by the MMP family.38 In the present study, overexpression of CD151 was correlated positively with up-regulated AktSer473, Snail, and MMP9, and direct evidence has been provided for the involvement of Akt and Snail in MMP9 expression induced by overexpression of CD151. In another study, silencing of CD151 in HCCM3 up-regulated the adhesion molecule E-cadherin, and this suggested that CD151 was involved in the epithelial-to-mesenchymal transitions (unpublished data). Overexpression of CD151 prompted the accumulation of Snail in the nucleus, rather than overexpression of Snail in cytoplasm in HCC cells and HCC patients (Supporting Information Fig. 9B,C). GSK-3β, an endogenous inhibitor of Snail transcription, can be inactivated by phosphorylated AktSer473 and is involved in the epithelial-to-mesenchymal transitions of cancer cells.39 The present study has shown that inhibition of GSK-3β up-regulates the expression of MMP9, and this indicates that the Akt/GSK-3β/Snail signal affects the expression of MMP9. In summary, overexpression of CD151 promotes the expression of MMP9 in HCCs, apparently primarily through the PI3K/Akt/GSK-3β/Snail signal (Fig. 7).
A variety of molecules, such as VEGF, have been implicated in the process of angiogenesis.40 Interestingly, in the present study, we found that the expression of VEGF was hardly relevant to CD151-dependent neoangiogenesis, and this was consistent with previous reports.11, 13 Instead, MMP9 had a crucial role in CD151-dependent angiogenesis and remodeling of vessels in vitro. More importantly, the consistency of the expression level of CD151 and its relationship with MMP9 expression and angiogenesis were confirmed in an animal model. Even more significantly, we identified a role for the CD151/MMP9/angiogenesis cascade in the clinical setting of HCCs. HCC patients with CD151high were inclined to harbor higher levels of expression of MMP9 and more neoangiogenesis. Moreover, the subgroup presenting with CD151highMMP9highMVDhigh had the worst prognosis of all of the patients examined. This finding strongly supports the notion that overexpression of CD151/MMP9/angiogenesis is intimately involved in the metastasis of HCC. On the basis of the available existing data, although we cannot completely exclude a role for other angiogenic factors, such as VEGF and MMP2, in neoangiogenesis of HCC, we hold that the CD151/MMP9/angiogenesis cascade probably is one of the factors controlling tumor angiogenesis in HCC. This provides a perspective on how tumor cells can induce tumor neoangiogenesis and how they are implicated in metastasis.
In conclusion, we have examined the role of CD151-dependent tumor angiogenesis in the progression of HCCs. CD151-dependent tumor angiogenesis may be mediated by MMP9 via the PI3K/Akt/GSK-3β/Snail pathway. More importantly, our findings highlight the possibility of CD151 being used as a high-priority target for antiangiogenesis therapy in HCC.
The authors thank Dr. Yong-Xiang Jiang for construction of the mouse cornea micropocket angiogenesis model. They also thank Professor Fei Yuan and Dr. Chen-Li Feng for assaying neoangiogenesis in the cornea and Dr. Yi-Zhou He for drawing the working model.