Recent progress in predictive biomarkers for metastatic recurrence of human hepatocellular carcinoma: a review of the literature

Abstract

Molecular markers (biomarkers) for hepatocellular carcinoma (HCC) metastasis and recurrence could provide additional information to that gained from traditional histopathological features. A large number of biomarkers have been shown to have potential predictive significance. One important aspect of this is to detect the transcripts of tumor-associated antigens (such as AFP, MAGEs, and CK19), which are proposed as predictive markers of HCC cells disseminated into the circulation and for metastatic recurrence. Another important aspect is to analyze the molecular markers for cellular malignancy phenotype, including DNA ploidy, cellular proliferation index, cell cycle regulators, oncogenes, and tumor suppressors (especially p53 gene), as well as telomerase activity. Molecular factors involved in the process of HCC invasion and metastasis, including adhesion molecules (E-cadherin, catenins, ICAM-1, laminin-5, CD44 variants, osteopontin), proteinases responsible for the degradation of extracellular matrix (MMPs, uPA system), as well as angiogenesis regulators (such as VEGF, intratumor MVD), have also been shown to be potential predictors for HCC metastatic recurrence and clinical outcomes. One important new trend is to widely delineate biomarkers with genomic and proteomic expression with reference to predicting metastatic recurrence, molecular diagnosis, and classification, which has been drawing more attention recently. Body fluid (particularly blood and urine) testing for biomarkers is easily accessible and more useful in clinical patients. The prognostic significance of circulating DNA in plasma or serum and its genetic alterations is another important direction. More attention should be paid to these areas in the future. As understanding of tumor biology deepens, more and more new biomarkers with high sensitivity and specificity for HCC metastatic recurrence could be found and routinely used in clinical assays. However, the combination of the pathological features and some of the biomarkers mentioned above seems to be more practical up to now.

Introduction

Lack of control of metastatic foci and recurrence is the most prevalent cause of death in patients with cancer, including hepatocellular carcinoma (HCC). It is important for tumor control to identify the factors that predispose patients to death. Much effort has been made to predict cancer behavior; however, specific predictive markers for metastasis and recurrence are still lacking. Conventionally, predictive factors for cancer relapse after surgical treatment consist of staging with the tumor node metastasis system (TNM) and grading by tumor cellular differentiation. Morphological features of the tumor, both gross and histological, have been found to be significantly associated with HCC recurrence and patient survival (Qin and Tang 2002); however, they cannot accurately predict the outcome of all HCC patients. A complementary way is to analyze molecular markers (biomarkers) for their prognostic significance with reference to cancer recurrence, which could provide additional information to that gained from traditional histopathological features. Biomarkers for molecular biological substaging may provide an opportunity to identify those patients with the most aggressive forms of the disease. In recent years, besides the traditional tumor-associated markers, such as alpha-fetoprotein (AFP), with the development of molecular biological techniques and new discoveries in cancer biology, many biomarkers related to invasion, metastasis, recurrence, and survival have been explored (Korn 2001). Some of them have been used clinically for the prediction and detection of cancer in patients who are under regular surveillance for HCC recurrence. In this review, we summarize progress regarding some important aspects in this area during the last 5 years.

Tumor-associated antigens

Alpha-fetoprotein (AFP)

Serum AFP is useful not only for diagnosis, but also as a predictive marker for tumor invasiveness and recurrence of HCC. Patients with high AFP levels at diagnosis tended to have greater tumor size, bilobar involvement, massive or diffuse types, and portal vein thrombosis (Tangkijvanich et al. 2000). Nonetheless, a correlation could not be established between increased AFP and Okuda’s stages, degree of tumor differentiation, or extrahepatic metastasis.

Lens culinaris agglutinin A-reactive fraction of AFP (AFP-L3) has been shown to be a more useful indicator of distant metastasis for HCC. AFP-L3-positive tumor is an advanced tumor regardless of small tumor size and lower serum AFP levels, especially portal vein invasion and lower tumor classification, and more significantly often and early distant metastasis. Periodic examination of AFP-L3 may be useful for the early detection of recurrent HCC (Ando et al. 2003; Hayashi et al. 1999; Oka et al. 2001; Okuda et al. 2002) and predicting postoperative survival of HCC patients (Okuda et al. 1999). Compared with imaging techniques, it has a 9–12 month lead-time in early recognition, with a sensitivity of 56% and a specificity of over 95% (Li et al. 2001).

AFP mRNA has been proposed as a predictive marker of HCC cells disseminated into the circulation and for metastatic recurrence in many reports (Matsumura et al. 1999; Minata et al. 2001; Miyamoto et al. 2001; Wong et al. 1999; Wong et al. 2001; Yamamoto Y et al. 2001a), but its clinical significance remains controversial (He et al. 1999; Kienle et al. 2000). The patients with positive AFP mRNA in peripheral blood have a higher possibility of extrahepatic metastasis. If the positive AFP mRNA could change to negative after adjuvant treatment, the patients’ overall and disease-free survival rates are much better than those with permanent positive AFP mRNA. In a prospective study, Ijichi et al. found that patients with consistent positive AFP mRNA showed the highest recurrence rate (85%) and a trend to distant or multiple recurrences (Ijichi et al. 2002). It seems that AFP mRNA in the peripheral blood of HCC patients in the perioperative period might be a predictive marker for early intrahepatic recurrence and distant metastasis after HCC resection (Matsumura et al. 1999; Minata et al. 2001). All HCC patients with persistently elevated AFP mRNA levels died from intrahepatic/extrahepatic metastasis, liver recurrence, or persistent HCC within 1 year after surgery (Wong et al. 1999; Wong et al. 2000). Minata et al. found that the presence of AFP mRNA before surgery was significantly correlated with tumor size. Metastatic recurrence was associated with the postoperative detection of AFP mRNA, but not with the preoperative and/or perioperative detection. AFP mRNA could be detected in some cases that had low serum AFP levels at recurrence. The recurrence-free period after the detection of AFP mRNA varied from 1 month to 12 months (Minata et al. 2001). However, some recent reports indicated that circulating AFP mRNA was transiently detected in cirrhosis with no predictive value for HCC development and recurrence after curative surgery (Gross-Goupil et al. 2003; Witzigmann et al. 2002).

Melanoma-associated antigen gene (MAGE)

The human MAGE gene family encodes tumor-specific antigens recognized by autologous cytotoxic T lymphocytes. Patients with tumors expressing at least one MAGE gene showed a better recurrence-free survival rate (Liu et al. 1999; Suzuki et al. 1999). Detection of MAGE transcripts with a follow-up survey in peripheral blood mononuclear cells (PBMC) to detect the circulating tumor cells is also a feasible and reliable assay for the early prediction of the relapse and prognosis of the HCC patients (Mou et al. 2002). Patients with persistent MAGE-1 and/or MAGE-3 mRNA-positive or with negative turned to positive, have a shorter survival time, and all of them died of metastasis and/or recurrence (Mou et al. 2002). A multimarker assay with cancer-specific markers such as MAGE-1 and MAGE-3 in combination with AFP may be a promising diagnostic tool for monitoring HCC patients with better sensitivity and specificity (Miyamoto et al. 2000).

Cytokeratin 19

The cytokeratins are the intermediate filament proteins characteristic of epithelial cells. In human cells, some 20 different cytokeratin isotypes have been identified. Their expression is remarkably tissue-specific, suggesting that the type of cytokeratins present in the cells is related to their biological function. Epithelial cells express between two and ten cytokeratin isotypes and the consequent profile which reflects both epithelial type and differentiation status may be useful in tumor diagnosis. In normal human liver, hepatocytes express CK8 and CK18, while bile duct cells also contain CK7 and CK19. CK19, the smallest member of the cytokeratin family, is clinically used as a discrimination marker of HCC and intrahepatic cholangiocarcinoma (ICC) (Uenishi et al. 2003a). However, some HCCs have been reported to express biliary markers even with a typical HCC growth pattern and morphologic appearance (Uenishi et al. 2002). Recent studies have indicated that cytokeratins, including CK19, play an important role in the regulation of cell migration and invasion. Uenishi et al. found that CK19 expression is one of the independent predictors of postoperative recurrence. Significantly higher incidences of both early recurrence and extrahepatic disease, especially lymph node metastasis, are found in patients with CK19-positive HCC than those with CK19-negative HCC (Uenishi et al. 2003b). Very recently, in collaboration with our institute—using proteomic techniques–Ding et al. found that CK19 was overexpressed in the MHCC97-H cell strain with higher metastatic potential and those human HCC tissues with overt intrahepatic metastases. The serum level of CK19 fragment CYFRA21-1 in the metastatic nude mice model bearing human HCC (LCI-D20) increased with tumor progression, particularly when pulmonary metastases occurred. These results demonstrated that overexpression of CK19 in HCC is related to metastatic behavior, and serum CK19 level might reflect pathological progression and may be a useful marker for predicting tumor metastasis of HCC patients (Ding et al. 2003). Higher levels of CK19 mRNA both in bone marrow and in the peripheral blood have also been shown to be a significantly predictive maker for breast cancer recurrence (Nogi et al. 2003), and for nasopharyngeal carcinoma micrometastasis (Lin et al. 2002), respectively.

Cell proliferation-related markers

Cell cycle regulators

The normal cell cycle is closely regulated by cyclins, in combination with its corresponding cyclin-dependent kinase (cdk). Overexpressions of cyclin A, cyclin D, cyclin E, and cdc2 play a crucial modulating role in cell cycle progression and cell proliferation of HCC. They have also been found to correlate with tumor relapse, and are independent predictive markers for their recurrence and prognosis (Ito et al. 2000a; Ohashi et al. 2001).

p27Kip1 belongs to the Cip1/Kip1 family of cyclin-dependent kinase inhibitors, and is a negative regulator of cell-cycle progression and a potential tumor suppressor gene. In most tumor types, reduced p27 expression correlates with poor prognosis, making p27 a novel and powerful predictive marker for prognosis and tumor relapse in various kinds of cancer, such as lung, breast, colon, liver, and prostate (Fiorentino et al. 2000; Ito et al. 1999; Philipp-Staheli et al. 2001; Thomas et al. 2000).

Telomerase activity

The ribonucleoprotein telomerase extends telomeres in cancer cells and has been proposed as a prognostic marker for cancer. Telomerase activity can be identified as an independent predictor for recurrence after resection of HCC. The relative telomerase activity (RTA) of patients with early recurrence is significantly higher than those without recurrence (Kobayashi et al. 2001). The determination of peripheral blood telomerase activity can also be used as a molecular marker for the detection of circulating HCC cells in the blood of patients, which reflects hematogenous micrometastasis (Tatsuma et al. 2000).

Oncogenes and tumor suppressor genes

p53 gene

p53 protein plays a central role in cellular responses, including cell-cycle arrest and cell death in response to DNA damage. p53 dysfunction can induce abnormal cell growth, increased cell survival, genetic instability, and drug resistance. Mutations in the p53 gene are the most frequently reported somatic gene alteration in human cancer. The association of p53 mutation or positive immunohistochemistry staining with higher grade and more advanced stage has been noted for cancers of various origins. In addition, p53 mutation is considered as a strong marker predicting an increased risk of local relapse, distant metastasis, treatment failure, and overall- and disease-free survival in many kinds of human carcinomas, such as breast (Amornmarn et al. 2000; Blaszyk et al. 2000; Overgaard et al. 2000; Takahashi et al. 2000), colorectal (Diez et al. 2000; Kahlenberg et al. 2000), esophageal (Ireland et al. 2000), head and neck (Tamas et al. 2000), lung (Murakami et al. 2000; Mitsudomi et al. 2000; Moldvay et al. 2000), ovarian (Shahin et al. 2000), and prostate (Quinn et al. 2000) cancers, as well as sarcoma (de Alava et al. 2000). An increased intracellular concentration of the p53 protein, although not identical, is sometimes seen in tumors with p53 mutation, and has been correlated with poor prognosis in some tumor types. Several studies have shown a relationship between the nuclear accumulation of p53 protein and poor disease-free and overall survival of cancer patients (Leibovich et al. 2000; Osaki et al. 2000). However, many reports failed to show the independent predictive value of p53 in carcinomas of tongue (Kantola et al. 2000), breast (Ferrero et al. 2000; Reed et al. 2000), stomach (Kaye et al. 2000), lung (Schiller et al. 2001), ovarian (Gadducci et al. 2000), bladder (Fleshner et al. 2000), colorectal (Gallego et al. 2000), endometrial (Fanning et al. 2002), cervical (Graflund et al. 2002), prostate (Stackhouse et al. 1999), liver (Saffroy et al. 1999), and ampullary cancers (Ajiki et al. 2001), as well as non-Hodgkin’s lymphoma (Nieder et al. 2001).

Similarly, there are many very controversial results regarding the prognostic value of p53 overexpression or p53 gene mutation in HCC patients. Some studies showed that p53 mutation was involved in determining dedifferentiation, proliferating activity, and tumor progression (Itoh et al. 2000a), and was strongly related to the invasiveness of HCC, and also influenced the postoperative course (particularly recurrence within 1 year) (Jeng et al. 2000). Mutations in the p53 gene or positive immunostaining for mutant p53 protein expression could be used as a significant indicator of poor prognosis, associated with shortened overall survival and shortened cancer-free survival of HCC patients after surgery. HCCs with p53 mutations have high malignant potential, and a p53 mutation in the primary lesion is useful as an indicator for the biological behavior of recurrent HCCs, which in turn is a useful independent prognostic factor affecting survival after recurrence (Heinze et al. 1999; Sheen et al. 2003; Sugo et al. 1999). In a prospective study, we found the 3-year and 5-year overall survival rates of HCC patients with positive p53 nuclear accumulation were much lower than those of the HCC patients with negative p53 expression. In univariate and multivariate Cox analysis, p53 overexpression was the most significant factor that was associated with the overall survival rates of HCC patients after resection. Its significance was even greater than that of factors such as tumor size, vascular invasion, and tumor capsule, though they were also related to overall survival (Qin et al. 2002).

The presence of serum anti-p53 antibody has also been shown to be potentially useful for predicting the recurrence of cancer and survival of patients with breast, ovarian, liver, and colorectal cancer (Suzuki et al. 2001; Takeda et al. 2001). p53 mutations in plasma DNA could also be detected in cancer patients, and may be used as an early marker to indicate recurrence or distant metastasis (Shao et al. 2001).

Oncogenes and their receptors

Aberrations of many tumor promoter genes, such as ras, c-myc, and c-fms have been shown to be indicators of malignant potential and poor prognosis in HCC. Disease-free survival in patients with c-myc amplification is significantly shorter than in those without amplification (Kawate et al. 1999). Hepatocyte growth factor/scatter factor (HGF/SF) is one of the most important humoral mediators of liver regeneration. It is potentially related to molecular mechanisms of hepatocarcinogenesis via a paracrine system involving its cellular receptor, c-met. Up-regulation of c-met plays an important role in the development and progression of HCC, and may be a prognostic marker (Tavian et al. 1999).

Among the erb-B receptor family members, c-erbB-2 (Her-2/neu) is a well-established predictive marker of faster tumor growth, increased metastasis, and shorter disease-free survival in several human tumor types, especially in breast cancer (Emi et al. 2002; Schulze 2003). However, c-erbB-3, but not c-erbB-2, plays an important role in the progression of HCC, and could predict the recurrence and disease-free survival of HCC patients (Ito et al. 2001; Prange et al. 2001).

Cellular adhesion molecules

E-cadherin and beta-catenins

Epithelial cadherin (E-CD) is a member of the cadherin family of cell adhesion molecules and has been implicated as an invasion suppressor molecule in vitro and in vivo. Since E-CD mediates cell-cell adhesion by associating with catenins, it is an important molecule in the development and maintenance of the epithelial phenotype. Loss of E-CD expression has been postulated to facilitate tumor cell dissociation and metastasis. This impaired adhesion status is clearly correlated with a haematogenous spread of the primary tumor cells, is an independent predictive marker for lymph node or distant metastasis, and for disease recurrence of many kinds of cancers including breast, colorectal, gastric, colorectal, esophageal, prostate, etc. (Aoki et al. 2003; Asgeirsson et al. 2000; Bankfalvi et al. 2002; Chow et al. 2001; Ikeguchi et al. 2000; Loric et al. 2001; Rhodes et al. 2003; Tanaka et al. 2003). It is also a significantly poor prognostic factor for survival (Shnayder et al. 2001; Sugio et al. 2002). Examination of E-CD expression in biopsy tumor specimens may be useful for predicting micrometastasis and/or tumor cell microinvolvement (Nakajo et al. 2001; Rodrigo et al. 2002). However, in some studies, no relationship could be found between the expression of E-CD and recurrence, distant metastases, lymph node stage, vascular invasion, primary tumor size, disease-free survival of beast cancer (Parker et al. 2001), and papillary thyroid carcinoma (Kapran et al. 2002) patients.

The expression level of E-CD has also been shown to inversely correlate with histological grade and prognosis, and might make some contribution to the early recurrence of HCC (Huang et al. 1999; Endo et al. 2000; Matsumura et al. 2001). In the progression of combined HCC and cholangiocarcinoma (cHCC-CC), the reduced expression of E-CD is significantly correlated with the tumor grade of the HCC and CC components, intrahepatic metastasis (IM) of HCC, IM of CC, and vascular invasion of CC (Asayama et al. 2002).

As a protein associated with the cytoplasmic region of E-CD, it is now apparent that deregulation of beta-catenin signaling is an important event in the genesis of a number of malignancies including HCC (Morin 1999). HCCs with a nonnuclear type (membranous expression) of beta-catenin overexpression have poorer cellular differentiation, more invasiveness, and intrahepatic metastasis, and the patients have significantly shorter disease-free survival time (Cui et al. 2001; Hsu et al. 2000; Suzuki et al. 2002; Wong et al. 2001a). This might due to beta-catenin playing an important role in promoting tumor progression by stimulating tumor cell proliferation and reducing the activity of cell adhesion systems (Inagawa et al. 2002).

Laminin-5

More than twelve laminin isoforms are presently known. The various isoforms have cell- and tissue-specific expression and are differentially recognized by integrins. Laminin-5 (Ln-5) promotes static adhesion and hemidesmosome formation, and can also stimulate cell migration and/or invasion after having been cleaved by matrix metalloproteinases (MMPs). Ln-5, and in particular its gamma2 chain, has been found to be preferentially expressed in the cytoplasm of epithelial human cancer cells located at the advancing edge of the tumor. Such a distribution, which is restricted only to malignancies, suggests that the gamma2 chain may be implicated in epithelial cancer growth, invasion, and metastasis (Giannelli and Antonaci 2000; Lohi 2001; Patarroyo et al. 2002). Ln-5 gamma2 chain expression is significantly correlated with depth of invasion, lymph node metastasis, distant metastasis, and recurrence of colorectal (Masaki et al. 2003), esophageal carcinoma (Yamamoto H et al. 2001), pancreatic ductal adenocarcinoma (Takahashi et al. 2002), and breast cancer (De Iorio et al. 2001). Its detection is useful for the prediction of recurrence and poor prognosis for these kinds of cancers.

Ln-5 expression is also associated with a more metastatic phenotype of HCC, and could be an important finding both as an unfavorable prognostic factor and as a diagnostic marker for detecting micrometastasis in peritumoral tissues. In particular, the gamma2 chain is strongly associated with the occurrence of metastasis and worse prognosis (Giannelli et al. 2003).

Intercellular adhesion molecule-1 (ICAM-1)

ICAMs are structurally related members of the immunoglobulin supergene family. Of the five ICAMs identified, ICAM-1 is the most extensively studied. ICAM-1 is expressed constitutively at low levels on endothelial cells, some lymphocytes, and monocytes. The serum concentration of intercellular adhesion molecule-1 (sICAM-1) in patients with HCC is associated with disease progression and prognosis. Higher sICAM-1 levels are more frequently observed in patients with multiple lesions, intrahepatic metastasis, and poorer prognosis. Detecting sICAM-1 is of important value in predicting tumor recurrence after surgery (Sun et al. 1999b). However, Parasole’s results suggest that the serum level of sICAM-1, as well as sIL-2R, IL-6, and anti-p53, has no further useful prognostic value over the CLIP score for HCC (Parasole et al. 2001).

CD44

CD44, in its standard form as well in its isoforms, is a cell-surface adhesion glycoprotein which occurs in a wide variety of non-neoplastic and neoplastic cells. Their major physiological role is to maintain organ and tissue structure via cell-cell and cell-matrix adhesion, but certain variant isoforms can also mediate lymphocyte activation and homing, and the presentation of chemical factors and hormones. CD44 expression, particularly its variants, appears to be an indicator of invasive and metastatic behavior, and may be useful in predicting metastasis or clinical recurrence of many kinds of cancers, such as renal (Gilcrease et al. 1999), prostate (Vis et al. 2000), thyroid (Bohm et al. 2000), tongue (Masuda et al. 2000), head and neck (Maula et al. 2003), bladder (Woodman et al. 2000; Stavropoulos et al. 2001), and colorectal (Bhatavdekar et al. 2001; Masaki et al. 2001) carcinomas, as well as oligodendrogliomas (Buccoliero et al. 2003). In HCC, up-regulation of CD44 isoforms, including CD44 s, CD44v5, CD44v6, CD44v7–8, and CD44v10, have been found to correlate with high histological grade, being the highest in poorly differentiated HCC. Being positive for one or more CD44 isoforms is one of the most useful independent factors for overall survival (Endo et al. 2000). High CD44v6 expression is significantly correlated with the presence of vascular invasion and p53 overexpression. However, some reports showed that expression of CD44 was not found to correlate with survival, local recurrence or metastatic ability (Sneath et al. 2000; Sokmen et al. 2001).

Osteopontin (OPN)

OPN, a secreted adhesive glycoprotein, is not only expressed in bone, but also in epithelium and several carcinomas, and is associated with tumorigenesis and cancer progression (Agrawal et al. 2001; Yeatman et al. 2003). Since OPN has been shown to be over-expressed in a variety of human tumors and is present in elevated levels in both the blood and tumors of some patients with metastatic cancers of breast (Tuck et al. 2001), head and neck squamous cell carcinomas (Le et al. 2003), and hormone-refractory prostate carcinoma (Hotte et al. 2002), it may provide prognostic information (Furger et al. 2001; Hotte et al. 2002; Rudland et al. 2002; Tuck et al. 2001). Recently, Carlinfante et al. found that the divergent pattern of OPN/BSP expression even could be an important determinant for the different characteristics of different types of bone metastasis, i.e., lytic vs sclerotic. Significantly more breast cancer bone metastases exhibited high OPN expression as compared with prostate tumor bone metastases (Carlinfante et al. 2003). However, Sulzbacher et al. found that OPN expression did not provide predictive information about the outcome of osteosarcoma patients (Sulzbacher et al. 2002).

In HCC, OPN mRNA overexpression is found to associate closely with AFP elevation, p53 mutation, larger tumors, high-grade, late-stage, early tumor recurrence, and/or metastasis and a lower 10-year survival rate. OPN mRNA overexpression predicts a higher early recurrence rate, and could serve as an unfavorable prognostic factor and a useful marker for predicting early recurrence in early-stage HCC (Pan et al. 2003). Gotoh et al. found both OPN transcript and OPN protein were significantly overexpressed in HCC with capsular infiltration, compared with HCC without capsular infiltration. Although OPN overexpression was not related to vascular invasion or intrahepatic metastasis, OPN was suggested to play a role in HCC, especially in cancer-stromal interactions (Gotoh et al. 2002). Recently, based on the metastasis-related expression signatures study, we revealed an important role of OPN in HCC metastasis. Overexpression of OPN is correlated with primary HCC with metastatic potential and with invasiveness of liver tumor-derived cell lines in vitro, and an OPN-neutralizing antibody can efficiently block invasion and metastasis of highly metastatic HCC cells in vitro and in vivo. These identify OPN as a molecular marker to define HCC patients with metastatic potential (Ye et al. 2003). A similar result was obtained by another two studies (Agrawal et al. 2002; Yeatman et al. 2003).

Extracellular matrix degradation proteinases

The extracellular matrix (ECM) holds cells together and maintains the three-dimensional structure of the body. It also plays critical roles in cell growth, differentiation, survival, and motility. Degradation of ECM mediated by the concerted action of several proteinases, including members of the serine, cysteine, aspartate, and matrix metalloproteinases (MMPs) families, is crucial for malignant tumor growth, invasion, metastasis, and angiogenesis (Brinckerhoff et al. 2000). The MMP and the plasminogen activation system (PA) play crucial roles in the process of cancer invasion and metastasis. Their expression levels are correlated to recurrence and survival after HCC resection (Sakamoto et al. 2000).

Matrix metalloproteinases (MMPs)

To date, 23 MMP genes have been identified in humans and many are implicated in cancer. Their activities are balanced by the tissue inhibitors of metalloproteinases (TIMPs). Elevated levels of distinct MMPs can be detected in tumor tissueserum, or even urine of patients with advanced cancer, and their role as predicting marker in cancer metastatic recurrence has been studied (Gerhards et al. 2001; Ranuncolo et al. 2003; Rao 2003; Sasaki et al. 2002; Vihinen et al. 2002). MMP-2 is the most important enzyme in the process of ECM remodeling involved in tumor invasion and metastasis, and could serve as a marker for shortened recurrence-free survival or relative overall survival in patients with carcinomas of breast (Talvensaari-Mattila et al. 2003), lung (Cai et al. 2002; Sienel et al. 2003), head and neck (Ondruschka et al. 2002), thyroid (Patel et al. 2002), tongue (Yoshizaki et al. 2001), superficial transitional cell (Hara et al. 2001), and bladder (Papathoma et al. 2000).

The contribution of MMPs and TIMPs to HCC invasion, metastasis, and recurrence still remains controversial. The expression of MMP-2 in the tumor tissues, particularly at the invasive front, improves tumor progression and metastasis. MMP-2, as well as MMP-9, MT1-MMP, and TIMP-1 and TIMP-2 is increased in nonencapsulated tumors. Since their levels appear to reflect the potential and ongoing activity of vascular invasion, it could be used as an important index to judge the invasion, metastasis, and recurrence of HCC (Chung et al. 2002; Sakamoto et al. 2000; Theret et al. 2001). Ishii et al. found that MMP-2 and MT1-MMP were predominantly involved in the carcinogenesis and progression of HCC, while MMP-7 and MMP-9 were involved in capsular infiltration and portal vein invasion (Ishii et al. 2003). Serum and tissue levels of TIMP-2 were significantly higher in HCC patients without metastasis than in those with. The imbalance of MMP-2/TIMP-2 could represent an important prognostic factor in patients with HCC (Giannelli et al. 2002). HBx promotes tumor cell invasion by a mechanism involving the upregulation of MT1-MMP and COX-2 and provides new insights into the mechanism of the viral protein’s involvement in tumor metastasis and recurrence of HCC (Lara-Pezzi et al. 2002). However, some reports indicated that serum and tissue levels of MMP-2 were not statistically different between patients with and without metastasis (Giannelli et al. 2002); MMP-2 and MMP-9 levels in HCC patients were comparable to patients with chronic liver disease without HCC, were not related to the size of HCC tumor or the grade of histological differentiation, and have little value for assessing or monitoring HCC patients (Kuyvenhoven et al. 2003; Murawaki et al. 2000). Recently, TIMP-2 has been reported to be essential for cell-mediated activation of MMP-2, and appears to have a tumor-promoting role (Ross et al. 2003).

Urokinase plasminogen activator (uPA)

uPA, its receptor (uPAR) and inhibitor (plasminogen activator inhibitor-1, PAI-1) also play a key role in tumor invasion and metastasis. In several tumor types, elevated levels of these molecules are associated with a poorer prognosis (Fox et al. 2001), and may be predictive of early relapse (Borstnar et al. 2002; Dazzi et al. 2003). uPA activity may be the most sensitive factor affecting HCC invasion in the plasminogen activation system and is a strong predictor for the recurrence and prognosis of HCC (Zheng et al. 2000; Itoh et al. 2000b). PAI-1 is a multifaceted proteolytic factor. It not only functions as an inhibitor of the protease uPA, but also plays an important role in signal transduction, cell adherence, and cell migration. Thus—and an apparent paradox considering the name of PAI-1—although it inhibits uPA during blood coagulation, it actually promotes tumor invasion and metastasis. In many malignancies including HCC, elevated PAI-1 is associated with tumor aggressiveness and poor patient outcome (Zheng et al. 2000).

Regulators of tumor angiogenesis

Angiogenesis, the process leading to the formation of new blood vessels from a preexisting vascular network, is necessary for tumor growth, invasion, and metastasis. The molecular mechanisms leading to angiogenesis in tumor cells are not completely known, but the modifications over time of the net balance between pro- and antiangiogenic factors appear to be responsible for the angiogenic activity of tumors.

Circulating angiogenesis regulators (both pro- and antiangiogenic) have been evaluated not only as diagnostic and/or prognostic factors but also as predictive factors in cancer patients. They could be used to determine the risk of developing cancer, screen for early detection, distinguish benign from malignant disease, distinguish between different types of malignancies, determine prognosis, predict the response to therapy, and monitor the clinical course (Kuroi et al. 2001). As a hypervascular tumor, angiogenesis plays a significant role in the progression of HCC (Poon et al. 2002). In recent years, many angiogenesis regulators, such as vascular endothelial growth factor (VEGF), basic fibroblast growth factor (bFGF), platelet-derived endothelial cell growth factor (PD-ECGF), thrombospondin (TSP), angiogenin, pleiotrophin and endostatin (ES) levels, as well as intratumor microvessel density (MVD) have been evaluated and found to relate to HCC prognosis.

Intratumor microvessel density (MVD)

Intratumor MVD is a direct reflection of tumor angiogenesis. It can be visualized by immunohistochemical staining with antibodies of anti-CD34, factor VIII, and alpha smooth muscle actin (Morinaga et al. 2001). MVD is found to be a significant and independent prognostic indicator in patients of many kinds of cancers in most retrospective studies. In 2000, Gasparini analyzed 43 retrospective studies on the prognostic significance of intratumor MVD for breast cancer patients. He found more than 75% of the authors reported positive results on the association of MVD with clinical outcome of the patients; and more than 85% of them, including those with multivariate analysis, found that intratumor MVD is an independent prognostic variable. However, determination of intratumor MVD is quite a crude surrogate marker of angiogenesis and cannot supply functional information on the molecular pathways involved in the angiogenic activity of each single tumor (Gasparini 2000).

In HCC, MVD level is closely associated with tumor capsule status, tumor size (HCC 2–5 cm diameter has the highest MVD level), intrahepatic recurrence, and disease-free survival, and can be a predictive marker for disease-free survival. In the authors’ institute, three types of intratumor microvessels, including capillary-like, sinusoid-like, and mixed-type, are found in HCC. The MVD level is an independent factor of disease-free survival in small-HCC patients, and is a predictive marker for early recurrence (Sun et al. 1999a). Poon et al. obtained a similar result in a prospective study. Multivariate analysis showed that tumor MVD-CD34 was the only significant factor predictive of disease-free survival in patients with HCC ≤5 cm. For HCCs larger than 5 cm, MVD did not have a significant prognostic influence (Poon et al. 2002).

Vascular endothelial growth factor (VEGF)

VEGF, the most potent endothelial cell mitogen and also a regulator of vascular permeability, is emerging as a powerful new prognostic tool. A substantial number of studies have demonstrated a strong association between elevated tumor expression of VEGF and advanced disease or poor prognosis in various cancers. Circulating VEGF seems to be a reliable surrogate marker of angiogenic activity and tumor progression in cancer patients. It may be useful in predicting and monitoring tumor response to anticancer therapies and in follow-up surveillance for tumor relapse (Ng et al. 2001; Karayiannakis et al. 2002; Poon et al. 2001a). In HCC, a high serum VEGF level significantly correlates with absence of tumor capsule, presence of intrahepatic metastasis and microscopic venous invasion, and advanced stage. It is a significant independent predictor of tumor invasiveness, tumor recurrence, disease-free survival, and overall survival in patients with resectable HCC (Chao et al. 2003; Li et al. 1999; Niu et al. 2000; Poon et al. 2001b).

However, the debate continues on whether serum VEGF level is a true reflection of tumor angiogenic activity in cancer patients, which originates from the finding that most VEGF in the serum is released from platelets during clotting. This unsettled issue has become a major obstacle regarding its clinical application (George et al. 2000; Jelkmann 2001; Poon et al. 2003). Some authors suggested that serum VEGF might reflect platelet count rather than tumor VEGF secretion (Benoy et al. 2002; Vermeulen et al. 1999). Recently, Poon et al. found serum VEGF165 levels correlated significantly with platelet counts. When corrected for platelet count, serum VEGF165/platelet associated significantly with tumor cytosolic VEGF165 concentration, which in turn is related to VEGF165 mRNA expression in the tumors. Advancing tumor stage was associated with a significant increase in tumor cytosolic VEGF165 concentration, tumor VEGF165 mRNA expression, serum VEGF165/platelet, and serum VEGF165 levels (Poon et al. 2003). This provides strong evidence that supports the use of serum VEGF level as an indirect estimate of tumor VEGF expression.

Genomic aberrations and cancer profiling

With the rapid progress of the human genome project, the recent development of high-throughput approaches, such as comparative genomic hybridization (CGH), cDNA microarray analysis, and proteomics, etc. provides an opportunity to take a genome-wide approach to predict clinical outcomes in a variety of cancers based on the molecular classification of a similar pathological group of cancers with different prognoses.

Genomic aberrations

Molecular genetic analyses have clarified that the accumulation of genomic changes provides important steps in carcinogenesis and have identified a number of valuable genetic markers for certain cancers. The association of these genomic aberrations with the progression and prognosis of cancer has drawn more and more attention. To date, allelic loss of 1p, 1q21–23, 2p21–16.3, 3p24-p25, 8p22, 8p23, 9p21, 9q, 10, 13q12, 16q 17p13.3, and 22q13 have been proposed as being related to the survival and prognosis of cancer patients (Bisgaard et al. 2001; Hirano et al. 2001; Simoneau et al. 2000; Tada et al. 2001; Washburn et al. 2000). Loss of heterozygosity (LOH) on chromosomes 13q, 16q, and 17p has been associated with the progression of HCC. Multivariate Cox survival analysis identified the presence of LOH on 16q and the number of chromosomes with LOH as the most significant independent negatively predictive factors for metastasis-free survival of HCC patients after curative resection (Nishida et al. 2002).

Using CGH techniques, we compared the differences of chromosomal aberrations between the primary HCC tumors and their matched metastatic lesions, and found chromosome 8p deletions might contribute to HCC metastasis (Qin et al. 1999). This result was further confirmed by comparison between nude mice models of HCC with different metastatic potentials (Qin et al. 2001). In addition, a more accurate location was identified on 8p23.3, 8p11.2 (Zhang et al. 2003). These findings provide new targets for exploring new predictive markers for the recurrence and prognosis of HCC. Itano et al. used restriction landmark genomic scanning (RLGS) to detect unknown genetic alterations in HCC, and found that the disease-free survival rate for patients ≥16 changed RLGS spots was significantly lower than that for patients with fewer changed RLGS spots (≤15 spots). In multivariate analysis, the number of changed spots was proven to retain an independent prognostic value. This suggests that the number of changed RLGS spots may be a useful biological marker for recurrence of HCC (Itano et al. 2000).

One important trend is the prognostic value of circulating DNA in plasma or serum, and its genetic alterations in cancer patients. Small amounts of DNA circulate in both healthy and diseased human plasma/serum, and increased concentrations of DNA are present in the plasma of cancer patients. Characteristics of tumor DNA have been found in genetic material extracted from the plasma of cancer patients. These features include decreased strand stability, the presence of specific oncogene or tumor suppressor gene mutations, microsatellite alterations, Ig rearrangements and hypermethylation of several genes. The results obtained in many different cancers have opened a new research area indicating that plasma DNA might eventually be a suitable target for the development of noninvasive diagnostic, prognostic, and follow-up tests for cancer (Anker et al. 2000; Silva et al. 2002; Taback et al. 2001). However, there is still controversy over prognostic significance (Nunes et al. 2001). LOH on chromosome 14q (D14S62 and D14S51) detected in tumor tissue has been shown to be closely related to metastasis and recurrence in breast cancer. We found that LOH at D14S62 and D14S51 could also be detected in plasma DNA, and they had great potentials to be clinically used to predicate metastasis and recurrence after HCC resection (Niu et al. 2003).

Genes expression profiling

cDNA microarray technology has offered an opportunity to probe disease-related gene expressions at a global genome scale. This approach has allowed the successful molecular classification of several human malignant tumors regarding their stage, metastatic recurrence potentials, prognostic outcome or response to therapy (Ramaswamy et al. 2003), such as breast cancer (Huang et al. 2003; van de Vijver et al. 2002; Woelfle et al. 2003); cervical lymph node metastasis in oral squamous cell cancer (Nagata et al. 2003), non-small cell lung cancers (Kikuchi et al. 2003), and clear cell renal carcinoma (Takahashi et al. 2001); it has also offered the possibility of identifying metastasis-associated genes (Clark et al. 2000; Huang et al. 2003). Using this approach and supervised classification, van ‘t Veer et al. identify a gene expression signature strongly predictive of patients with a poor or good prognosis (van ‘t Veer et al. 2002). The gene-expression profile is a more powerful predictor of the outcome of disease in young patients with breast cancer than standard systems based on clinical and histological criteria (van de Vijver et al. 2002). Its predicting accuracy for metastatic recurrence could be about 90% (Huang et al. 2003). Recently, expression profiling using microarrays has been applied to PBMC samples from patients with advanced colorectal cancer as a surrogate system for biomarker analyses in clinical oncology studies (DePrimo et al. 2003). However, the prognostic value of expression profiling for the outcome of cancer patients remains controversial. Recently, Ntzani et al. systematically assessed the predictive performance of 30 eligible studies identified from MEDLINE (1995 to 2003) for major clinical outcomes (death, metastasis, recurrence, response to therapy) and the correlation of gene profiling with other clinicopathological correlates of malignant disorders, and found that the prognostic performances of expression profile were variable. Smaller studies showed better sensitivity and specificity for clinical outcomes than larger studies. Thus, larger studies with appropriate clinical design, adjustment for known predictors, and proper validation are needed for this highly promising technology (Ntzani et al. 2003). A selected combination of gene profiling and clinical/pathological parameters could reach the best prediction of recurrence (Bettuzzi et al. 2003).

A few reports have dealt with the gene expression profiles of primary HCC samples (Iizuka et al. 2002; Li et al. 2002; Okabe et al. 2001; Xu et al. 2001a; Xu et al. 2001b). However, little is known about the gene expression profiles associated with metastatic feature of HCC. Through the investigation of the genome-wide expression profile via cDNA microarray consisting of 23,000 genes, Cheung et al. found a total of 90 clones to be correlated with intrahepatic metastasis of HCC (Cheung et al. 2002). Recently, we analyzed the expression of 9,180 genes in primary HCC tumors from 40 patients without or with accompanying intrahepatic metastases with cDNA microarray technique. Using a supervised machine-learning algorithm approach to classify patients based on their gene expression signatures for potential to develop metastasis and for the prognosis of patients, we generated, for the first time, a molecular signature that correctly classified patients with or without accompanying intra-hepatic metastases and have identified genes that are mostly relevant to the prediction outcome including patient survival. The expression profiles also show that primary metastasis-free HCC is distinct from primary HCC with metastasis, and metastatic tumors are indistinguishable from their primary tumors, regardless of tumor size, tumor encapsulation and patient’s age (Ye et al. 2003).

Proteins expression profiling

Proteomics is another technology rapidly changing our approach to understand tumor biology. It permits the analysis of thousands of modified or unmodified proteins simultaneously, and has become increasingly popular for identifying biomarkers for early detection, classification and prognosis of cancers, as well as pinpointing targets for improved treatment outcomes. It is also possible to identify changes in expression of proteins that potentially may be related to tumor progression, invasion and metastasis, and prognosis. Proteomic profiling can be applied to tissue samples as well as body fluids (e.g. serum, urine, etc.), and it can provide surrogate markers of disease processes, potential response to treatment, possibility of recurrence and metastasis for cancers including HCC (Kennedy 2001). Recently, differential proteomic analysis was conducted on two HCC cell strains (MHCC97-H and MHCC97-L) with different metastatic potentials, established from the same parent cell line MHCC97 in our institute. CK19 was identified as a potential predictive marker for HCC metastasis (Ding et al. 2003).

One new trend is the combination of annotation/protein sequence analysis, transcript profiling, immunohistochemistry, and immunoassay, which provides a more powerful approach for delineating candidate biomarkers of cancer with potential clinical significance (Welsh et al. 2003).

Others

It has been shown that HBx enhances tumor cell invasion, both in vivo and in vitro, which is mediated by an upregulation of MT1-MMP expression, and in turn activates MMP-2. Thus, HBx could be a predictive marker of tumor metastasis and recurrence of HCC (Lara-Pezzi et al. 2002). Serum HBeAg positivity has also been shown to be an independent risk factor for recurrence and an independent unfavorable factor for survival time. The prognosis after liver resection for HCC is worse in HBeAg-positive patients than that in HBeAg-negative patients (Kubo et al. 2002). The long-term survival rate of patients seronegative for HBsAg after liver resection for HCC is greater than that of patients seropositive for HBsAg (Wu et al. 1999).

Many other factors, including higher levels of urinary TGF-beta 1, heat shock protein-27 (HSP-27) (King et al. 2000) and glutamine synthetase (GS) expression in the tumor (Osada et al. 2000), increased levels of cyclooxygenase-2 (COX-2) in nontumor liver tissue (Kondo et al. 1999), preoperative serum IL-10 (Chau et al. 2000), and HFE mutations (Pirisi et al. 2000) or down-regulation of DRH1 (Yamamoto Y et al. 2001b), have also been shown to be powerful prognostic indicators for shorter disease-free survival and poor prognosis of HCC. The RECK (reversion-inducing-cysteine-rich protein with Kazal motifs) gene suppresses the invasive and metastatic activities of cancers, has negative effects on the invasiveness of HCC, and can be regarded as a promising prognostic biomarker for HCC (Furumoto et al. 2001). The expression of Fas, a pro-apoptotic receptor, and Fas ligand (Fas L) play a role in apoptosis of cancer cells, and is associated with the prognosis of cancer patients (Ito et al. 2000b). Valosin-containing protein (VCP, also known as p97) level is an indicator for disease-free survival in each early- (I and II) and advanced- (III and IV) stage group of pathologic tumor-node-metastasis classification. VCP expression level has prognostic significance for disease-free and overall survival of patients with HCC (Yamamoto S et al. 2003).

Questions and prospects

In summary, the evaluation of markers in the molecular classification of cancer patients, as indicators of tumor progression or recurrence, is main recommendation for future work. Biological markers that are both sensitive and specific for tumor metastasis are increasingly becoming available and are routinely monitored during the regular follow-up of patients treated for cancer. Obtained by a simple blood test, these markers provide an inexpensive non-invasive means for the early detection of recurrence (or progression). Currently, the longitudinal behavior of the marker is viewed as an indicator of early disease progression, and is applied by a physician in making clinical decisions. However, to date, none of them has been proved to be specific enough, and most of the studies for specific molecular parameters were correlative and retrospective. Methodologies, sample sizes, and definitions differ. Consideration should be given to the design of prospective clinical trials in evaluating the prognostic significance of these markers.

Biomarkers can be detected both in tissue and body fluids (serum, urine, bile, etc.). Body fluid (particularly blood and urine) testing is easily accessible and useful in clinical patients, and is more important in “predicting” the possibility or “early diagnosis” of recurrence and metastasis. The measurement of circulating levels of predictive markers has several advantages over direct assessment of their levels in tumor tissues: (a) it does not require a tumor specimen and thus is theoretically applicable to every cancer patient; (b) it is technically simpler; and (c) repeated measurements can be made in the same patient before and after anticancer treatments (Poon et al. 2003). Thus, future work should be focused on serum or urine markers.

The prognostic significance of circulating DNA in plasma or serum, and its genetic alterations, is an important trend that deserves attention. Molecular signatures (gene or protein profiling) provide another new way to delineate new biomarkers genome-wide, which has attracted more and more attention in recent years. Thus, we can assume that—with the continuing progress of the human genome project, the development of new molecular and cytogenetic techniques, and a more complete understanding of tumor biology—more and more new prognostic markers with high sensitivity and specificity will soon be found and used in clinical assays. However, the combination of some items, i.e., pathological features and some of the biomarkers mentioned above, seems to be more practical at present.