Hepatocellular carcinoma (HCC) is one of most malignant and aggressive human tumors. Transforming growth factor-β1 (TGF-β1) and its coreceptor CD105 have been shown to contribute to HCC malignant progression. TGF-β1 and CD105 have also been implicated in angiogenesis, but their role in the vascularization of HCC has not been investigated. To fill this gap, we studied the effect of TGF-β1 and CD105 on HCC-derived endothelium. By using immunomagnetic beads, we isolated and cultured endothelial cells (ECs) from HCC (HCC-EC) and adjacent nonneoplastic tissue (nNL-ECs) obtained from 24 liver biopsies. HCC and nNL biopsies were also analyzed by immunohistochemistry for the expression of CD105, TGF-β1, Ve-cadherin (Ve-cad), CD44, β-catenin, and E-cadherin. Compared with nNL-ECs, HCC-ECs had higher expression of CD105, enhanced spontaneous motility, and greater capacity to migrate in response to TGF-β1 (5 ng/mL), particularly in the presence of a fibronectin matrix. The chemotactic effect of TGF-β1 was blocked by anti-CD105 antibodies and correlated with the grade of HCC malignancy. Histologic examination of HCC biopsies showed that HCCs with the worse malignant features had the highest expression of TGF-β1, CD105, and angiogenic markers (Ve-cad and CD44). Because CD105 was highly expressed in microvessels at the tumor periphery and TGF-β1 staining was only found in neoplastic hepatocytes, we conclude that HCC-derived TGF-β1 may act as a chemoattractant for CD105-expressing ECs and as a promoter of tumor angiogenesis. Thus, drugs that selectively target the TGF-β1/CD105 axis may interfere with HCC-related angiogenesis and HCC progression. [Cancer Res 2008;68(20):8626–34]
Hepatocellular carcinoma (HCC) is one of the most common neoplasms worldwide and the third cause of death among those patients with malignant tumors ( 1, 2). One of the principal risk factors for HCC is chronic infection by hepatitis B or C virus (HBV and HCV), although other important risk factors, such as Aflatoxin B1 and alcoholism, have been implicated in its pathogenesis ( 3– 5). HCC is richly vascularized as a result of intense angiogenesis ( 6). Microvascular density (MVD), proliferation index, and the grade of tumor cell differentiation have been shown to correlate with HCC malignancy and aggressive clinical behavior. Moderately or poorly differentiated HCC with high microvascular density have a worse prognosis than well-differentiated and less vascularized HCC (grade 1, G1). Highly undifferentiated HCCs with intense angiogenesis are very invasive, metastatic, and ultimately lethal.
Surgical removal of the primary tumor and liver transplantation are the mainstay therapy when HCC is confined to the liver ( 7). Unfortunately, conventional chemotherapy in HCC is not effective. A possible explanation for this failure is the abnormal architectural organization of the HCC vasculature, which causes poor blood flow and blood stagnation, leading to inadequate delivery of circulating drugs to cancer cells ( 8). Normalization of the tumor vasculature with antiangiogenic agents has been shown to enhance the efficacy of anticancer drugs and improve survival rates in patients with colon and lung cancer ( 8). Thus, a potential treatment of HCC could be the combined use of antiangiogenic drugs with chemotherapy ( 9). To that end, clinical trials with antiangiogenic agents have already started ( 9, 10). Preclinical studies have shown that expression of vascular endothelial growth factor (VEGF), a potent angiogenic factor ( 11), correlates with HCC progression and metastatic spread ( 12, 13), but the molecular mechanisms regulating angiogenesis in HCC have not been fully investigated.
During HCC progression, liver endothelial cells (ECs) undergo significant phenotypic changes and acquire several markers, such as CD31, CD34, and UEA-1 lectin, that are not present in normal sinusoidal ECs ( 14– 16). Among these, endoglin/CD105 functions as a coreceptor for transforming growth factor-β1 (TGF-β1) and modulates TGF-β1 signaling by interacting with TGF-β receptors I and/or II ( 17). CD105 has proved to be an excellent HCC vascular marker compared with CD34 or VEGF ( 18, 19). Yu and colleagues, however, found that CD105 was not only present in tumor vessels but also in the sinus endothelium of cirrhotic tissue ( 20). CD105 and TGF-β1 have also been shown to be significantly elevated in the serum and urine of patients with HCC cancer or liver cirrhosis ( 21– 23). In vitro studies and work with experimental animal models have implicated TGF-β1 and CD105 as critical regulators of angiogenesis ( 24, 25). The role of the TGF-β1/CD105 system in HCC angiogenesis has, however, remained unclear because existing reports in this field have mostly focused on immunohistochemical analysis of CD105 expression without providing functional analysis of resident EC behavior. To gain further insight into the role of CD105 in HCC angiogenesis, we investigated 24 patients with different grades of HCC malignancy. We studied the expression of CD105 and other EC markers in cultured ECs derived from HCC (HCC-ECs) and adjacent nonneoplastic liver (nNL-ECs), as well as the capacity of HCC-ECs and nNL-ECs to proliferate and migrate in response to TGF-β1. CD105 and TGF-β1 expression was concurrently analyzed by immunohistochemistry in the same specimens used for ECs isolation. Our findings support the hypothesis that CD105 and TGF-β1 are functionally involved in HCC angiogenesis. We propose that HCC-derived TGF-β1 promotes angiogenesis through migratory stimulation of CD105-expressing ECs at the periphery of the invading tumor. These findings identify the TGF-β1/CD105 system as a regulator of HCC angiogenesis and as a possible new target of antiangiogenic therapy in patients with HCC.
Materials and Methods
Patients and HCC biopsies. Neoplastic and normal liver tissue was obtained from 24 patients who underwent curative resection of HCC in the Department of General Surgery of Spedali Civili. Patients did not receive any chemotherapeutic treatment before surgery. This study was approved by the ethical committee board of the hospital. Fresh tissue (0.2–1 g) sampled from the tumor and the adjacent nNL was obtained from the Department of Pathology of the hospital. About half of the specimens were used for EC isolation, whereas the rest were fixed in 10% buffered formalin and embedded in paraffin for histologic sectioning and immunoperoxidase studies. HCC biopsies were also stained with H&E for routine histologic evaluation to define the grade of malignancy. HCCs of decreasing levels of differentiation were graded as G1, G2, and G3 according to Edmondson classification ( 26). Additional features, such as tumor size (calculated as maximal tumor diameter of the primary lesion), venous invasion, secondary microsatellite lesions, proliferation index, and MVD, were also taken into consideration. For each biopsy, the active rate of cell proliferation was calculated using the Ki67 marker (dilution 1:50; Dako), whereas MVD was measured according to a previously reported standard procedure ( 27). Briefly, sections immunostained with anti-CD34 antibodies (dilution 1:30; NeoMarkers, Inc.) were scanned at low magnifications to identify three areas with the greatest degree of vascularization (hotspots) within the tumor or at the tumor/nNL interface. These areas were scored by counting CD34-positive profiles of microvessels at high magnification (40×). MVD was calculated by averaging vessel counts for each specimen.
Histology and immunohistochemical studies. HCC and nNL samples from 24 patients were fixed in 10% of formalin, embedded in paraffin, and then processed for histology and immunohistochemistry. H&E-stained serial sections of 5-μm thickness were used for histologic evaluation of tumor grade as described above. The remaining sections were immunostained for proteins of interest according to standard protocols ( 28). Briefly, sections were transferred to glass slides coated with polylysine, deparaffinized in 100% xylene, and rehydrated in graded ethanol. After heat-induced antigen retrieval, endogenous peroxidase was inhibited by incubating tissue sections with 3% hydrogen peroxidase for 15 min at room temperature, whereas a specific epitope binding was blocked by incubation for 20 min with 20% human serum. All samples were then processed by the avidine-biotin peroxidase complex method according to manufacturer's recommendations (LASB kit DakoCytomation). The monoclonal antibodies (mAbs) used for this study were anti-CD105 (dilution 1:50; Santa Cruz Biotechnology), anti-CD34 (1:30; NeoMarkers, Inc.), anti–Ve-cadherin (Ve-cad; 1:30; Santa Cruz Biotecnology), anti-CD44 (1:30; Novocastra Laboratories), anti–TGF-β1 (1:200; Santa Cruz Biotecnology), anti–β-catenin (β-cat; 1:50; Zymed Laboratories), and anti–E-cadherin (E-cad; 1:50; Zymed Laboratories).
Flow cytometry studies. The cell phenotypes were characterized by flow cytometry [fluorescence-activated cell sorting (FACS)]. To minimize the effect of tissue culture conditions on variability of CD105 expression ( 29), HCC-ECs or nNL-ECs were always analyzed by FACS at passage 3 and 70% to 80% of the confluence. These conditions were generally obtained after 4 to 5 d of incubation, seeding 2 × 105 cells onto T25 flasks coated with collagen type I (Coll-1) and human plasma fibronectin (FN; both purchased from Sigma). FACS was performed as previously described ( 30). Roughly, 2 × 105 cells were incubated with mAbs against the following human antigens: CD44, Ve-cad, CD31, and von Willebrand factor purchased from BD PharMingen, Ulex europaeus (UEA-1) lectin (Sigma), and CD105 PE-conjugated from Serotec. The controls were matched isotype mouse immunoglobulins. All the incubations were done at 4°C. After incubation, cells were washed with PBS containing 0.1% bovine serum albumin (BSA) and then incubated with mAbs for 30 min in the presence of normal FCS. FACS analysis was done with FACSCalibur flow cytometer and Cell Quest Software (BD PharMingen). Each analysis included at least 30,000 events for each gate. The percentage of positive cells was assessed after the correction for the percentage reactive to an isotype control conjugate of FITC or phycoerythrin. With the isotype controls for fluorescin isothiocynate or phycoerythrin, the gates for phenotype analysis were set so that the bottom left panel contained at least 98% of all analyzed cells. Alternatively, the phenotypic analysis of the cells was also performed by immunocytochemistry after the avidin-biotin peroxidase complex as described above.
Isolation and culture of HCC-ECs and nNL-ECs. Human HCC-ECs and nNL-ECs were isolated and cultured according to a previously described method ( 31). Briefly, HCC and nNL specimens of ∼500 mg in weight were finely minced with scissors and then digested for 2 to 3 h at 37°C in DMEM (Sigma) containing 0.25% collagenase D (Boehringer) and 0.2% BSA. The cell suspension was washed with PBS (Euroclone Celbio), passed through a 20-μm pore size filter, and then plated onto Petri dishes coated with 1 μg/cm2 Coll-1 and 1 μg/cm2 FN. Cells were cultured in EBM growth medium (EBM-GM) consisting of EBM medium supplemented with 10% FCS and EC mitogens (EGM bullet kit, LONZA). One week later, the primary cultures were detached by tripsinization and the HCC-ECs and nNL-ECs were positively purified by using magnetic Dynabeads coated with anti-CD31 mAbs and seeded onto Coll-1 + FN–coated dishes in EBM-GM. All the EC cultures were routinely maintained in EBM-GM on Coll-1 + FN–coated T25 flasks and passed weekly at 1:2 splitting dilution. Results presented in this study were conducted with HCC-ECs and nNL-ECs harvested after three to five in vitro passages.
Proliferation assay. HCC-ECs and nNL-ECs (at passage 4) were harvested from the culture flask by treatment with trypsin (Sigma) in PBS (0.05% w/v); the trypsin solution was made up fresh by diluting a 100× frozen aliquot. After enzyme inactivation and centrifugation, cells were resuspended in EBM + 0.2% BSA at a final density of 2 × 104 cells/mL. Five milliliters of each EC suspension (105 cells) were transferred to T25 flasks previously coated with Coll-1 + FN. After a 7-d incubation in EBM-GM (medium was changed once), HCC-EC and nNL-EC were detached and counted by hemocytometer. To evaluate the growth response to mitogens, 0.5 mL of HCC-ECs or nNL-ECs suspension (104 cells) was seeded into each well of a 24-multiwell plate coated with Coll-1 + FN; after cell adhesion, medium was aspirated and replaced with EBM-GM supplemented with VEGF-A (0.5–5 ng/mL; LONZA) or human TGF-β1 (0.01–50 ng/mL; EDM Chemicals, Inc.). After 72 h, the wells were washed and the cells were fixed and stained as described ( 31). The cells were counted with a calibrated ocular eyepiece in 10 different fields at 400× magnification. Every test was run in triplicate, and the experiments were repeated twice.
Migration assay. Corning Costar Transwell (Celbio) supports were used to test spontaneous and TGF-β1–induced ECs migration. 6.5-mm Transwell 5-μm pore size polycarbonate membrane inserts were coated with Coll-1 or with Coll-1 + FN, as previously described ( 32). For each test, 105 cells in 200 μL of EBM + 0.2% BSA were routinely placed on the top of the membrane insert (the upper compartment of the well). To evaluate spontaneous migration, 500 μL of control EBM + 0.2% BSA medium was added to the lower compartment of the wells. To evaluate TGF-β1–induced HCC-ECs and nNL-ECs migration, different concentrations of the molecule (0.1–50 ng/mL) were added in the lower compartment of each well. A substance placed in the lower compartment of the well acts as chemoattractant, and the cells move from the surface through the membrane against a concentration gradient. The effect of mAbs against the human extracellular domain of CD105 (Serotec) on spontaneous and TGF-β1–induced ECs motility was tested, either by adding mAbs (1:10 to 1:200 dilution) in the lower compartment of the well or by priming ECs first with anti-CD105 mAbs before seeding them on top of the transwell membrane. Priming of ECs was obtained by preincubating the cells for 1 h at 37°C (105 cells) resuspended in 200 μL of EBM + 0.2% BSA in the presence of different concentrations of anti-CD105 mAbs. Mouse IgGs (Serotec) were used as control mAbs isotype. Migration assay was carried out for 6 h at 37°C in 5% CO2. Then the membrane inserts were removed, fixed in 10% formalin, and stained with Wright's solution ( 32). Cells attached to the upper surface of the filter were removed with a swab, and the cells migrated across the membrane were counted by microscopically examining the lower surface. Reported data represent the total number of cells found in 10 different fields for each membrane at 400× magnification. Each determination was done in duplicate.
Statistical analysis. Quantitative data were analyzed by Student's t test, Kruskal-Wallis test, or Welch test (ATMS). Statistical significance of differences was set at P < 0.05.
CD105 expression is increased on cultured HCC-ECs, but not on nNL-ECs. We first investigated the expression of CD31, CD105, CD44, and Ve-cad on cultured ECs isolated from biopsies of HCC and adjacent nNL. As shown in Table 1 , CD31+ cells were successfully isolated with anti-CD31 mAb-coated immunomagnetic beads ( 31) from all 24 patients enlisted in this study. The percentage of isolated CD31+ cells varied among HCC and nNL specimens and rarely exceeded 10% of the total cells immunoselected. CD31+ cells were generally isolated and cultured more easily from normal liver than corresponding HCCs, although a sufficient number of HCC-EC cultures was established for comparative studies with nNL-ECs (nNL-ECs, 19 of 24; HCC-ECs, 15 of 24). Patients for whom nNL-ECs were expanded, but isolation and culture of HCC-ECs failed, were not included in the in vitro study. Ultimately, we performed in vitro studies on ECs from 14 of 24 patients (four of eight with G1 grade HCC, five of nine with G2 HCC, and five of seven with G3 HCC; Table 1). After isolation, CD31+ cells were analyzed to confirm their endothelial origin. Live cultures of HCC-ECs and nNL-ECs showed the typical morphology of ECs: they appeared as polygonal flat cells, sometimes with an elongated spindle shape, and formed a uniform monolayer at confluence (Supplementary Fig. S1). Labeling studies showed that isolated cells expressed typical EC markers, such as CD31 and vWf, and bound the endothelial-specific UEA-1 lectin (data not shown). The HCC-EC and nNL-EC cultures were further characterized for the expression of CD105, CD31, Ve-cad, and CD44 by FACS ( Fig. 1A ). Compared with nNL-ECs, CD105 expression was significantly higher in all HCC-EC cultures, and the highest expression was observed in HCC-ECs derived from G3 followed by G2 and G1 tumor specimens. In contrast, no difference was observed in the expression of CD31, Ve-cad, and CD44, which were all highly expressed in both HCC-ECs and nNL-ECs.
Effect of TGF-β1 on HCC-ECs and nNL-ECs growth and migration. Because CD105 is considered a marker of EC proliferation ( 33), we compared the rate of HCC-ECs proliferation to that of their nNL-ECs counterpart. Under the same culture conditions, HCC-ECs showed a longer doubling time (48–56 hours) compared with nNL-ECs (36–48 hours). Thus, CD105 expression in cultured HCC-ECs did not correlate with a higher proliferative rate compared with nNL-ECs. In a proliferation assay, HCC-ECs had a lesser growth capacity compared with nNL-ECs, and the addition of TGF-β1 (0.1–50 ng/mL) did not enhance proliferation, but eventually, at 1 to 5 ng/mL, TGF-β1 reduced proliferation (data not shown). VEGF-A (5 ng/mL) promoted the growth of both nNL-ECs and HCC-ECs. The addition of TGF-β1 did not substantially affect VEGF-A stimulatory activity, although sometimes an improvement of HCC-ECs proliferation was recorded (see Supplementary Fig. S2). We then investigated cell motility and whether TGF-β1 may act as a chemoattractant for ECs. We found that spontaneous motility of HCC-ECs and nNL-ECs was increased by the presence of an FN matrix, and this effect was substantially more pronounced for HCC-ECs compared with nNL-ECs ( Fig. 1B). Generally, the HCC-EC cultures derived from G3 specimens showed the highest spontaneous migration capacity. The addition of 0.1 to 1 ng/mL of TGF-β1 in the lower well did not provoke any effect on HCC-ECs or nNL-ECs migration (data not shown). Concentrations of TGF-β1 ranging from 5 to 10 ng/mL produced a sharp enhancement of HCC-ECs migration but had no effect on nNL-ECs migration ( Fig. 1C). The highest effect of TGF-β1 was produced on G3-derived HCC-ECs that were seeded on Coll-1 + FN; these cultures had the highest level of CD105 expression. Concentrations of TGF-β1 at >10 ng/mL were significantly less effective (data not shown).
Antibodies against CD105 block TGF-β1 induced HCC-ECs migration. To verify whether the chemotactic effect of TGF-β1 on HCC-ECs was mediated by its coreceptor CD105, we repeated the migration assay experiments with blocking anti-CD105 antibodies, which were added to the lower compartment of a Transwell together with TGF-β1. Anti-CD105 antibodies (dilution, 1:50) not only blocked the effect of TGF-β1 but also reduced the spontaneous migration of HCC-ECs ( Fig. 1D). A dose dependence activity of anti-CD105 was found, and mAbs dilutions of 1:100 were ineffective in blocking spontaneous migration, such as control mouse IgG (data not shown). The highest inhibition was observed on G3-derived HCC-ECs, which expressed the highest level of CD105 ( Fig. 1D). Blockade of TGF-β1–induced migration was more efficient if the HCC-ECs were first primed with anti-CD105 antibody before the migration test. Anti-CD105 counteracts the stimulatory effect of TGF-β1, also at 1:100 dilution (Supplementary Fig. S3). Anti-CD105 did not substantially affect nNL-ECs migration.
Enhancement of CD105 and TGF-β1 expression in HCC biopsies correlates with angiogenic and malignant features of the tumor. A portion of the HCC and nNL biopsies used for ECs isolation were analyzed by immunohistochemistry to determine in situ expression of CD105, TGF-β1, and other vascular and cell adhesion markers (Ve-cad, CD44, β-cat, E-cad) and to correlate these observations with HCC malignant features. Tumor size, venous invasion, secondary liver lesions, Ki67 expression, and MVD were also evaluated. Table 2 summarizes the results obtained from 24 HCC patients. The most undifferentiated HCCs (G3 and G2) showed the highest proliferation index (Ki67 expression), MVD, venous invasion, and formation of secondary tumor nodules. G1 biopsies were more comparable with nNL biopsies, except for their higher MVD. Figure 2 shows the staining pattern of Ve-cad, CD105, and TGF-β1 in HCC and the adjacent nNL area. Anti–Ve-cad antibody stained specifically liver vessels but did not discriminate HCC vessels from nNL vessels. Ve-cad staining was stronger in HCC than nNL tissue, and its intensity in the tumor stroma correlated with the grade of HCC malignancy. CD105 stained both HCC and nNL vessels and was most intense in G3 and G2 biopsies. However, compared with Ve-cad, CD105 staining was associated more distinctively with vessels at the tumor periphery or close to the tumor capsule ( Fig. 3 ). CD105 staining was significantly reduced or even absent in the central part of tumor; this was particularly noticeable in G3 biopsies ( Fig. 3). Staining for TGF-β1 was most intense in the most undifferentiated HCC (G3), followed by G2 and G1 tumors, whereas it was substantially absent in the normal tissue ( Fig. 2). TGF-β1 staining was mainly localized in the cytoplasm of neoplastic hepatocytes. CD44, another tumor vascular marker ( 34), was most highly expressed in the most malignant G3 and G2 tumors but, unlike Ve-Cad and CD105, did not stain HCC vessels. It was often localized in tumor cells near tumor vessels ( Fig. 4 ). Finally, we analyzed the expression of β-cat and E-cad because of their involvement in cell-cell adhesion, cell migration, and HCC malignancy ( 35– 37). Figure 4 shows that β-cat and E-cad were abnormally expressed in HCC compared with normal liver. Whereas β-cat staining of normal liver tissue showed a peripheral cell membrane staining, β-cat staining of HCC highlighted the cytoplasm or nuclei of the neoplastic hepatocytes. E-cad staining was found at the periphery of nNL hepatocytes and was significantly weaker in HCC, and its decrement in HCC was correlated to the grade of malignancy ( Fig. 4).
The rich vasculature of HCC is believed to play an important role in the progression and responsiveness to treatments of this malignant neoplasm. Recent reports suggest that endoglin/CD105 and its physiologic ligand TGF-β1 participate in HCC-induced angiogenesis ( 18– 22), but the mechanisms involved in TGF-β1/CD105-mediated HCC angiogenesis are poorly understood. To gain new insights into this process and in light of previous reports lacking direct investigation on separated ECs, here, we studied the role of CD105 and TGF-β1 on the migration and proliferation of human ECs isolated from HCC and adjacent nNL tissue. Our studies were performed on nNL and HCC specimens obtained from 24 patients with different grades of HCC malignancy. The same material was also evaluated by immunohistochemistry to determine whether CD105 and TGF-β1 expression correlated with HCC angiogenic and malignant features.
HCC-ECs and nNL-ECs were successfully isolated from most of the nNL and HCC specimens. HCC-ECs expanded somewhat less efficiently than their nNL-ECs counterparts, which confirmed the observation that tumor endothelium has a reduced capacity to expand in vitro ( 31). However, we were able to successfully isolate and culture 14 distinct HCC-EC strains from a total of 24 patients enlisted in this study. We decided to prepare ECs directly from nNL and HCC liver and rejected the idea of using commercial preparation of cultured ECs, like human umbilical vascular EC, because of the known heterogeneity of ECs derived from different organs ( 38).
Using this new approach, we were able to show that the HCC-derived TGF-β1, signaling through CD105 directly mobilizes HCC-ECs but has no effect on nNL-ECs. Our studies explain, at least in part, why HCCs with the highest expression of TGF-β1 and CD105 have the highest angiogenic and malignant features. These findings provide a mechanistic explanation to previous reports, implicating CD105 ( 33) and TGF-β1 as mediators and markers of HCC progression ( 21, 22, 39).
Our conclusions are based on the following observations: (a) CD105 was highly expressed on cultured HCC-ECs, whereas it was weakly expressed in nNL-ECs; (b) CD105 expression correlated with the grade of HCC malignancy (G3 > G2 > G1); (c) HCC-ECs derived from G3 biopsies, which had the highest level of CD105 expression, had the highest capacity to spontaneously migrate in a migration assay, particularly on FN, and chemotactically respond to TGF-β1, whereas nNL-ECs were unresponsive to TGF-β1; (d) mAbs against CD105 reduced or completely blocked spontaneous and TGF-β1–induced migration; (e) immunohistochemical analysis of 24 HCC biopsies showed that CD105 staining was markedly increased, particularly on vessels at the tumor periphery; (f) TGF-β1 was highly expressed only in HCC cells and absent in nNL hepatocytes. Moreover, expression of both TGF-β1 and CD105 correlated with increased expression of MVD, Ve-cad, and CD44 expression. The expression of these angiogenic markers correlated with malignant features of HCC, such as abnormal nuclear localization of β-cat, decrease expression of E-cad, and increment of Ki67 and secondary neoplastic nodules.
In addition to providing a mechanistic explanation for how HCC cells may recruit ECs during angiogenesis, the direct evidence that isolated HCC-ECs overexpress CD105, whereas nNL-ECs counterpart do not, supports the idea that CD105 could be a good marker of “activated” HCC angiogenic endothelium, thus potentially discriminating “active angiogenesis” versus “postangiogenesis.” This idea has been discussed in a recent report, where high expression of CD105 in sinusoid ECs of nonneoplastic cirrhotic liver was found ( 20). However, these authors did not examine CD105 expression on live culture of HCC-ECs, as we have performed here. By immunohistochemistry, we confirmed that nNL sinusoid ECs may also express CD105 but not as much as the vessels surrounding or infiltrating the tumor periphery. Moreover, CD105 staining was practically absent in the tumor core, particularly in most malignant HCC biopsies. Similarly to CD105, Ve-cad stained HCC-ECs very intensively, both in vitro and in vivo, but unlike CD105, Ve-cad in tumor biopsies did not discriminate between vessels at tumor periphery or in the HCC core. CD44, another proposed vascular marker ( 34, 40), in the same biopsies, did not stain liver vessels at all.
Although we cannot exclude that the scant presence of CD105-positive vessels in the HCC core could be due to a vasculogenic mimicry process associated with these tumor cells, particularly pronounced in the most malignant biopsies ( 41), altogether, our data support the idea that CD105 could be a marker of HCC-activated vessels at the invading edge of the growing tumor ( 42). The close association in HCC biopsies of CD105+ vessels with TGF-β1+ cancer cells suggests that the TGF-β1/CD105 system stimulates angiogenesis in vivo through a chemotactic mechanism comparable with the one responsible for the in vitro migration of HCC-ECs. The existence of a chemotactic gradient of TGF-β1 originating in the tumor is strongly suggested by the additional observation that unlike HCC cells, which stain strongly for TGF-β1, normal hepatocytes show no significant expression of TGF-β1 by immunohistochemistry. The high level of TGF-β1 found in the serum ( 22, 23) and urine of HCC patients ( 21) further supports the idea that HCC produces large quantities of TGF-β1, which may play an important role in the angiogenic mobilization of ECs. The conclusion that TGF-β1 may promote HCC angiogenesis in vivo is also supported by our immunohistochemical findings, indicating a positive correlation between TGF-β1 expression and angiogenesis in the most malignant HCC. It is also important to note that TGF-β plays a crucial role in mural cell recruitment and, although still controversial, in inhibiting their proliferation ( 43). We think that the high production of TGF-β1 by HCC may help to induce vascular abnormalities, possibly by reducing the number of mural cells in blood vessels. Malignant tumors usually have abnormalities in pericytes content, and lack of pericytes leads to abnormal vascular morphogenesis ( 44). The preliminary results of our group, using antibodies against NG2 a pericyte marker, seem to indicate a significant reduction of mural cells in the most malignant HCC biopsies. 6
It has been suggested that TGF-β1 and other members of the TGF-β family of cytokines may exert a dual role in tumor progression ( 45). TGF-β1 may suppress the initial steps of tumor progression, but it may also enhance tumor growth at later stages ( 46). Our results are compatible with this hypothesis and provide a possible interpretation for these changes in TGF-β1 activity. The transformation of TGF-β1 into a cancer-promoting factor may be related to the switch of its production from stromal cells to cancer cells. Our studies suggest that when hepatocytes acquire the capacity to produce TGF-β1, they also become strongly angiogenic. This TGF-β1–driven angiogenic switch would enhance the ability of HCC to grow and metastasize, thereby promoting its progression ( 47).
Our immunohistochemistry results also showed that TGF-β1 overexpression in the most malignant HCCs was associated with translocation of β-cat from cell membrane to nuclei. This is consistent with previous studies indicating that the Wnt/β-cat pathway is regulated by members of the TGF-β family ( 48). Abnormal TGF-β1 production by transformed hepatocytes may disrupt the mechanisms by which Wnt/β-cat regulates cell differentiation and proliferation ( 49), as suggested by the observation that the cell proliferation marker Ki67 was most highly expressed in the least-differentiated HCC cells, which also stained the strongest for TGF-β1. However, regardless of its regulatory growth effects in cancer cells, HCC-derived TGF-β1 is likely to significantly alter the angiogenic equilibrium of the liver parenchyma through chemotaxies-mediated HCC neovascularization.
In summary, our studies provide evidence for the existence of a novel mechanism by which HCC become vascularized. Our in vitro observations with HCC-ECs and nNL-ECs, taken together with immunohistochemical studies performed on the same specimens from which ECs were derived, indicate that HCC stimulate angiogenesis, at least in part, by producing TGF-β1, which in turn stimulates the migration of HCC-ECs, but not nNL-ECs. This process is particularly pronounced in the least differentiated HCCs, which express the highest levels of TGF-β1 and are vascularized by CD105-rich ECs. Besides the difficulty in proposing an inhibitor of TGF-β1 as a strategy to interfere with HCC progression ( 50), the observation that TGF-β1–mediated stimulation of HCC-ECs migration can be effectively blocked with anti-CD105 antibody raises new questions regarding the potential use of CD105 as a target for the treatment of HCC. More studies are, however, needed to determine whether these findings can be translated into a new therapeutic approach for the treatment of HCC.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: Guido Berlucchi Foundation, Center for the Study, Prevention and Therapy of Hepatic Diseases, Spedali Civili Brescia, and Fondazione IRCCS Istituto Neurologico “C. Besta” grant LR6.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
↵6 Unpublished observation.
- Received April 1, 2008.
- Revision received July 16, 2008.
- Accepted August 7, 2008.
- ©2008 American Association for Cancer Research.