Expression, Cellular Distribution, and Prognostic Relevance of TRAIL Receptors in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is the fifth most common cancer worldwide and shows a rising incidence. Unfortunately, whereas curative surgical or local ablative therapies are feasible only in 30% of patients, at this time systemic chemotherapy alone does not provide a curative option and HCC still represents the third most frequent cause of cancer-related death with a survival averaging 10% at 5 years (1).

Recently, the clinical course of HCC patients has improved, thanks to the use of the Raf/vascular endothelial growth factor inhibitor sorafenib (2). This success has stimulated intensive research intended to unveil the role played by other different signaling pathways likely to influence survival and proliferation of liver cancer cells. It is expected that a comprehensive description of these pathways will allow treatment schemata tailored on corresponding molecular targets in individual tumors (3, 4).

Induction of apoptosis through the interaction of tumor necrosis factor–related apoptosis-inducing ligand (TRAIL) with its receptors on the surface of cancer cells is a well-described mechanism of tumor surveillance (5). The in vivo importance of loss of sensitivity to TRAIL-mediated apoptosis is shown by clinical studies showing a correlation between TRAIL receptor expression, poor prognosis, and tumor recurrence (6). In addition, TRAIL knockout mice exhibit enhanced metastasis formation (7), and we recently showed that expression of the TRAIL-binding soluble decoy receptor OPG correlates with tumor stage in patients affected by colon carcinoma (8).

Following the recognition of the physiologic importance of TRAIL signaling, growing experimental evidence accumulated on the fact that TRAIL induces apoptosis in cancer cells but not in normal cells (9). For this reason, TRAIL has been developed as a promising alternative therapeutic strategy, and many types of recombinant TRAIL or agonistic antibodies targeting TRAIL receptors have been made available for the clinic (5).

However, the role of the TRAIL system in pathogenesis of HCC is not clear, and no study had yet assessed the relevance of TRAIL receptor expression as a prognostic marker in this tumor (10).

We found that loss of TRAIL receptors represents a common feature of HCC and an independent prognostic marker identifying a subset of tumors that have lost sensitivity to receptor-mediated apoptosis.

Patients with a diagnosed HCC who had been treated with liver transplantation or partial hepatectomy at the University Clinic Munich Groβhadern between 1985 and 2008 were considered for analysis. To avoid high selection bias due to inclusion criteria for transplantation, for survival analyses, patients who received a liver transplantation were excluded. Survival data of patients were retrieved from the database of the Munich Cancer Registry (MCR; http://www.tumorregister-muenchen.de). Tissue samples were obtained from paraffin blocks archived at the Institute of Pathology of the University of Munich. A tissue microarray containing tumor samples as well as matched surrounding nontumor tissue was established as previously described (11).

Sections (5 μm) of tissue microarray blocks were used for immunohistochemical staining. Anti–TRAIL-R1 monoclonal goat antibody (Santa Cruz Biotechnology, Inc.) and anti–TRAIL-R2 monoclonal rabbit antibody (Calbiochem) were applied as primary antibody. For antigen retrieval, sections were pretreated by boiling in a microwave oven two times at 15 minutes at 750 W in Target Retrieval Solution (Dako). Endogenous peroxidase was blocked by incubation in 7.5% hydrogen peroxide for 10 minutes. Vectastain ABC Elite Universal (Vector Laboratories) kit was taken for antibody detection, and AEC (Zytomed Systems) was used as a chromogen. Slides were counterstained with hematoxylin (Vector). Positive staining for TRAIL-R1 or TRAIL-R2 was categorized according to its cellular distribution and regardless of the intensity of the signal as follows: positive staining in cytoplasm only, positive staining on cell membranes only, and positive staining in both cell membrane and cytoplasm. According to the rationale that TRAIL receptors are functionally active if situated on the surface of cell membranes, for statistical analysis, tumors exhibiting TRAIL receptor staining on cell membranes (“membrane staining” group) were compared with tumors that showed no TRAIL receptor staining or in which TRAIL receptors were confined to the cytoplasm only (“no membrane staining” group). The efficacy of TRAIL receptor evaluation by immunohistochemistry to discriminate between membrane and cytoplasmatic staining was confirmed by costaining of TRAIL receptors and membrane-bound E-cadherin by using confocal microscopy (Supplementary Data). Additionally, we conducted a semiquantitative analysis of HCC samples by assigning a score ranging from 0 to 3+ to TRAIL receptor staining. Immunohistochemical evaluation of TRAIL receptors was confirmed by quantitative PCR analysis of 39 samples showing a significant correlation of TRAIL receptor staining and the intensity of their immunohistochemical signal (χ² test: TRAIL-R1, P = 0.021; TRAIL-R2, P = 0.01).

For assessment of tumors, tumor grade, vessel invasion, number of lesions, and tumor size were considered. Tumor grading was evaluated according to the WHO criteria (12) and was based on the area showing the highest grade. Vascular invasion was defined according to macroscopic or microscopic evidence of blood vessel invasion. Tumor size and number of lesions were assessed by macroscopic pathologic investigation of resected livers. Matched nontumor tissues were investigated for anatomopathologic features describing underlying liver disease (grade of fibrosis/cirrhosis, portal inflammation, piecemeal necrosis, and steatosis). Histologic evaluation of tumors or surrounding nontumor tissue was done on slides stained with H&E and evaluated by a senior pathologist and two of the authors (L.K. and E.N.D.T.) who were blinded to tissue annotations and prognostic data. Liver fibrosis, portal inflammation, and piecemeal necrosis were assessed according to the Ishak score for chronic hepatitis (13). Steatosis and lobular infiltration were evaluated according to the nonalcoholic fatty liver disease activity score and staging system (14).

All statistical analyses were run using SPSS (version 17; SPSS, Inc.). For frequency data, exact χ² tests were used. Significance level was adjusted using the Bonferroni correction for repeated tests. Observed (unadjusted overall) survival was first estimated with the Kaplan-Meier method and tested with the log-rank procedure. Relative survival was computed by the ratio of the observed survival rate to the expected survival rate (15). The expected survival time of age- and gender-matched individuals was calculated from the life tables of the normal German population. Relative survival was thus used as an estimate for disease-specific survival. Survival was then investigated with a Cox proportional hazards regression model. Hazard ratios (HR) and 95% confidence intervals (CI) of these four Cox proportional hazards regressions are presented. For survival analysis, the significance level was set at 5%.

One hundred and fifty-seven patients with a diagnosed hepatic cellular cancer who underwent liver resection or liver transplantation due to HCC could be identified. Of these patients, 55 were treated with liver transplantation, whereas 102 underwent partial hepatectomy and were considered for survival analysis. Of these patients, 101 were also registered in the database of the MCR, where survival was available for 93 patients who underwent liver resection at the University Hospital Grosshadern between 2001 and 2007. The demographic clinicopathologic features of the patients are listed in Table 1. For patients who underwent liver resection who had been considered for analysis of survival, the median follow-up was 27 months. When stratified by survival status, the median follow-up for survivors was 46 months compared with 19 months for patients who died. Almost all patients died due to tumor recurrence, as shown by the overlap of the overall and relative survival of these patients (Fig. 2A).

TRAIL-R1 stained positive in all nontumor tissues surrounding liver cancer lesions. In 8% of these samples, TRAIL-R1 stained positive in cytoplasm only. Membrane staining was positive in all other tissue samples (92%). Instead, 3% of HCC samples stained negative for TRAIL-R1, 30% showed cytoplasmatic staining only, and 66% stained positive on cell membranes (Fig. 1). When TRAIL-R2 was examined, no staining was detected in 13% of nontumor tissues surrounding tumor lesions; 37% of nontumor tissues showed cytoplasmatic staining only, and 50% had a positive staining on cell membranes. In tumor tissues, 46% of samples showed no signal for TRAIL-R2, 36% of samples stained positive in cytoplasm, and only 18% of samples stained positive on cell membranes (Fig. 1). The efficacy of immunohistochemical staining to discriminate between membrane and cytoplasmatic staining was confirmed by costaining of TRAIL receptors and membrane-bound E-cadherin by using confocal microscopy (Supplementary Data). Altogether, the fraction of samples exhibiting receptor staining on cell membranes was remarkably lower in tumor tissues versus that detected in surrounding HCC lesions (TRAIL-R1: 92% in normal cells versus 66% in HCC; TRAIL-R2: 50% in nontumor cells versus 18% in cancer cells; exact χ² test: TRAIL-R1, P = 0.0148; TRAIL-R2, P = 0.0150). Similarly, 45% of nontumor samples versus 17.6% of tumor samples exhibited positive membrane staining for both TRAIL receptors; conversely, only 2.8% of nontumor samples exhibited double-negative staining on the cell membrane for TRAIL receptors versus 30.7% of tumor samples. Therefore, overall positive staining of TRAIL-R2 was remarkably lower in tumor samples than in matched normal tissue; the fraction of tumor samples exhibiting TRAIL-R1 or TRAIL-R2 staining on the surface of cell membranes was significantly lower than that of matched surrounding nontumor tissues.

To assess whether TRAIL receptor staining in tumor samples or in matched nontumor tissue samples could be associated to specific clinicopathologic features, membrane staining of TRAIL-R1 and TRAIL-R2 was correlated with clinicopathologic parameters including age and gender of patients, etiology of the underlying liver disease, presence of severe fibrosis or liver cirrhosis, as well as specific pathologic features of tumors such as grading, size and number of lesions in the liver, or the presence of signs of microscopic or macroscopic blood vessel invasion (Table 2). After adjusting significance levels according to the Bonferroni correction for repeated tests, a significant association between membrane staining for TRAIL-R1 in tumor samples and smaller tumors (<5 cm, P = 0.004) was found. In addition, a borderline significance was found between higher degree of liver fibrosis (Ishak score ≥5) and TRAIL-R1 staining on cell membrane (P = 0.008). No relevant correlation was found between TRAIL receptor staining on cell membranes in surrounding nontumor cells and these clinicopathologic variables (data not shown).

To avoid selection bias due to inclusion criteria for transplantation, analyses of survival were conducted in patients undergoing partial hepatectomy only. In these patients (Fig. 2A), the overall and relative survival curves showed a substantial coincidence. Therefore, in further analyses, overall survival was considered as representative for tumor-related survival in these patients.

A Kaplan-Meier analysis showed vascular invasion (Fig. 2B) and TRAIL receptor staining on cell membranes (Fig. 2C and D) as significant determinants of survival in these patients. Patients showing TRAIL-R1 staining on cell membranes had a better prognosis versus patients bearing tumors without TRAIL-R1 membrane staining (5-year overall survival was 52% versus 27%, respectively). Overall 5-year survival of patients exhibiting TRAIL-R2 membrane staining was 43%, whereas that of patients with no TRAIL-R2 staining was 15%. Instead, analysis of survival conducted using semiquantitative criteria to assess TRAIL receptor staining intensity [scoring ranging from 0 (no signal) to 3+] failed to show any difference in survival, neither if the overall staining regardless of the cellular distribution was considered nor when the analysis was conducted only on membrane- positive tumor samples. A difference in survival could be seen only if patients were stratified according to the presence or absence of TRAIL-R1 staining in tumor samples (P = 0.02, log-rank test; data not shown).

A Cox multivariate analysis taking into account age, gender, tumor grading, tumor size, the presence of multifocal lesions, or signs of blood vessel invasion showed that membrane staining of TRAIL-R1 and TRAIL-R2 independently influences the survival of HCC patients after partial hepatectomy. Calculated HRs for TRAIL-R1 and TRAIL-R2 were 2.3 (P = 0.01) and 2.4 (P = 0.02), respectively (Table 3). In contrast, no correlation was found between TRAIL receptor staining in nontumor tissues surrounding tumor lesions and survival of patients (TRAIL-R1: P = 0.89; TRAIL-R2: P = 0.11). When survival of patients was assessed according to the simultaneous staining of TRAIL receptors on cell membranes, patients with double-positive staining for TRAIL-R1 and TRAIL-R2 showed a better prognosis in the Cox multivariate analysis in comparison with patients with no membrane staining for TRAIL receptors (HR, 5.72; 95% CI, 2.10-15.5; P = 0.001) and in comparison with patients bearing tumors with positive membrane staining for one receptor only (HR, 2.85; 95% CI, 0.99-8.28; P = 0.05; Fig. 3).

The clinical use of sorafenib (2) has intensified the research of novel molecular targets for cancer therapy (16, 17). Impairment of TRAIL-mediated apoptosis due to loss of TRAIL receptors or inhibition of the interaction with the respective ligands has been documented in several tumor entities (18–20), and several antibodies specifically targeting TRAIL receptors have been developed as anticancer therapy (5). Increasing evidence has shown also that TRAIL receptors play a role in the pathogenesis of several liver diseases (21, 22). Yet, no systematic immunohistochemical study has been conducted to assess the importance of the TRAIL system and its prognostic value in HCC or in the diseased liver [reviewed by Herr et al. (21)].

In our cohort, almost all tumor samples had a positive staining for TRAIL-R1, whereas 46% of tumor samples stained negative for TRAIL-R2 (Fig. 1). Loss of TRAIL-R2 in tumor cells agrees with previous reports on other tumor entities (20, 23). Instead, the only work published thus far investigating TRAIL receptors in HCC samples reported overall positive staining for both TRAIL receptors in all 60 tumor samples examined (24); the discrepancy is unlikely to be due to suboptimal staining of our tissue samples because the wide majority of the costained samples from matched samples of surrounding nontumor tissue showed positive staining for TRAIL receptors (Fig. 1). Differences in the size of the collective and the underlying liver disease of patients recruited in these two cohorts might represent an explanation for this difference. In particular, in this previous study, 55 of the 60 patients were HBV positive, whereas in our collective there was a predominant presence of patients affected by a toxic/metabolic disease. This reflects the different epidemiology of liver cirrhosis in China and in the region of Germany where our study was conducted and might be responsible for the differences observed. Future studies done on a wider patient collective will be needed to confirm these data and unveil possible effects of underlying liver diseases in determining TRAIL receptor expression. For statistical analysis, we first discriminated the fraction of tumors exhibiting TRAIL receptor staining on cell membranes with the rationale of assessing the fraction of receptors exposed to the action of circulating TRAIL. According to this criterion, 44% of tumor samples showed negative membrane staining for TRAIL-R1 and 92% were negative for TRAIL-R2 (Fig. 1). This suggests that, as recently shown in vitro (25), internalization of TRAIL receptors in HCC might play a central role in determining loss of the functional fraction of these receptors as a distinctive feature of cancer cells.

To assess whether loss of membrane-bound TRAIL receptors might have a functional significance in the progression of HCC, we assessed their effect on survival. Five-year survival of patients bearing tumors exhibiting TRAIL-R1 or TRAIL-R2 staining on cell membranes doubled that of the other patients undergoing the same treatment (Fig. 2; Table 3). Stratification based on TRAIL receptor intensity staining scores showed instead no statistical significance in the analysis of survival. A significant survival advantage was shown only if patients were stratified on the basis of overall TRAIL-R1 staining versus the absence of immunoreactive products, this condition possibly reflecting the absence of membrane-bound TRAIL receptors in this subset. Patients bearing tumors with double-positive membrane staining for both TRAIL receptors survived significantly longer in comparison with patients not only showing double-negative membrane staining but also bearing tumors with positive membrane staining for only one receptor. This suggests that membrane-bound TRAIL-R1 and TRAIL-R2 might have an additive and favorable effect on the survival of these patients (Fig. 3).

In our cohort of patients, membrane staining for TRAIL-R1 but not for TRAIL-R2 in tumor samples was associated with a smaller size of tumors, possibly reflecting a favorable effect of TRAIL-R1 expression on tumor growth. However, this could not be observed for TRAIL-R2. Furthermore, no other statistically significant association between membrane staining status of each receptor and the other pathologic features of tumors could be shown. To further investigate the role of TRAIL receptors in the pathogenesis of HCC, we therefore subsequently assessed the correlation between TRAIL receptor membrane staining and clinicopathologic variables related to the liver diseases underlying HCCs. Because most primary liver cancers arise on the ground of liver diseases (1), we had hypothesized that TRAIL receptor loss could represent a progressive process related to the severity of liver fibrosis, the presence of steatosis, leukocyte infiltration, or their etiologies. However, no significant correlation between TRAIL receptor status in nontumor tissues and these pathologic features or patient survival could be found. These data suggest that loss of TRAIL receptors might not concur with the underlying liver disease in the pathogenesis of HCC as a precarcinogenetic condition, but might occur later as an additional feature of malignant cells. This agrees with two recent studies showing that TRAIL receptor loss does not play a critical role in tumor development in APC mutant mice (26) and does not affect the formation of primary squamous cell carcinomas (27), but had a decisive effect on metastasis formation (26, 27). Recently, Takeda and colleagues have shown that adoptive transfer of TRAIL-expressing natural killer cells prevents recurrence of HCC after partial hepatectomy (28, 29). Together with these studies, our data support the idea that, rather than influencing tumor formation, loss of TRAIL receptors might affect tumor progression at a later stage due to the selection of cell clones resistant to circulating TRAIL and to immune-mediated mechanisms controlling clearance of metastatic cells.

Although the mechanism by which TRAIL receptors are lost in tumor cells is not fully understood, recent evidence showed that genetic loss or mutation of TRAIL receptors is a rare event in cancer cells, averaging 1% for TRAIL-R2 in HCC (18, 30). Furthermore, several compounds (25, 31–34) proved to increase TRAIL receptor expression and synergize with TRAIL administration to trigger apoptosis (35), suggesting that epigenetic gene silencing or TRAIL receptor internalization is a potentially reversible cause for the loss of functional TRAIL receptors. Our data showing a correlation between membrane staining and prognosis represent a clinical correlate of these studies. Therefore, patients bearing tumors without membrane staining for TRAIL receptors might profit from the administration of agents capable of increasing their expression or functional localization on cell membrane and eventually restoring the efficacy of endogenous TRAIL (29). Interestingly, it has been shown that sorafenib synergizes with TRAIL to induce apoptosis in cancer cells (36). This suggests that the association of these agents to sorafenib might represent a profitable strategy in the treatment of HCC.

Another obvious consequence of the frequent functional loss of TRAIL receptors in HCC is that many liver tumors might not respond to administration of the specific agonistic antibodies recently made available for the clinic targeting either TRAIL-R1 such as mapatumumab or TRAIL-R2 such as lexatumumab (5). However, because loss of these receptors had a strong and additive prognostic relevance, patients lacking a TRAIL receptor are likely not to respond to the administration of the corresponding antibody, but “double-positive” patients will probably profit more from the targeting of both receptors.

Furthermore, agonistic antibodies targeting TRAIL receptors might synergize to induce apoptosis in combination with agents capable of increasing TRAIL receptor expression (35).

Of clinical relevance is the lack of association between TRAIL receptor staining and liver disease, which could be observed in our analysis. Concerns on the safety of a therapeutic use of TRAIL have originated from in vitro reports showing an increased sensitivity to TRAIL in the diseased liver, likely mediated by enhanced expression of TRAIL receptors (37). Our data have a reassuring meaning related to the possibility that enhanced TRAIL receptor expression might represent a cause for increased susceptibility of normal cells to apoptosis mediated by TRAIL administered for therapeutic purposes. Future research should be directed to prospectively assessing the significance of TRAIL receptors in cancer cells. Importantly, the effect of several drugs, which are now undergoing investigation in clinical trials, including sorafenib and histone deacetylase inhibitors, should be examined about the possibility that their effect might be correlated to their influence on TRAIL receptor expression.

In summary, loss of TRAIL receptors is a common feature of HCC. Their localization on cell membranes seems to be a determinant of survival. Future therapeutic protocols might profit from the characterization of TRAIL receptor staining and cellular distribution.

No potential conflicts of interest were disclosed.

We thank Dominique Schühmann for retrieving pathologic material and contributing to the establishment of the tissue microarray used for these experiments.

Grant Support: Deutsche Forschungsgemeinschaft grant TO 605/2-1 (E.N. De Toni).

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.