Background
Hepatocellular carcinoma (HCC) is the main type of primary liver cancer and the fifth most common malignant cancer worldwide. Its poor prognosis makes it the third leading cause of cancer-related mortality [
1‐
3]. Only about 30% of patients are eligible for curative therapies (e.g. resection and transplantation) and the disease recurs frequently following liver resection [
4]. Sorafenib, an oral multikinase inhibitor, is effective in the treatment of advanced HCC [
5]. However, sorafenib therapy is limited by side effects and lack of long-term efficacy.
The tumor necrosis factor (TNF)-related apoptosis inducing-ligand (TRAIL) is a member of the TNF cytokine family. TRAIL is currently in clinical development as a potential novel anticancer therapeutic because it selectively induces apoptosis in cancer cells [
6‐
11]. After TRAIL-binding TRAIL-R1, also called Death Receptor 4 (DR4), [
12] and TRAIL-R2 (DR5) [
13,
14] initiate apoptosis following formation of the death-inducing signaling complex (DISC): trimerization of TRAIL-R1 and/or TRAIL-R2 leads to recruitment of FADD and cytoplasmic caspase-8 to the intracellular death domain (DD) of both receptors. Caspase-8 recruitment to the DISC activates this protease, which triggers a caspase cascade and, ultimately, apoptotic death of susceptible cells. Two other receptors, TRAIL-R3 and TRAIL-R4, do not induce apoptosis; they lack a functional intracellular death domain [
15‐
17] and have been suggested to inhibit TRAIL-induced apoptosis by competing with TRAIL-R1 and TRAIL-R2 for TRAIL-binding. TRAIL-R4 has also been shown to inhibit apoptosis through ligand-independent association with TRAIL-R2 via the preligand assembly domain (PLAD) [
18] or by NF-κB activation upon TRAIL-R4 overexpression [
17]. The fifth TRAIL-receptor, osteoprotegerin, is a soluble receptor and is mainly involved in the regulation of bone metabolism [
19,
20].
Apart from representing potential therapeutic targets for novel, TRAIL-based therapies, the two TRAIL receptors and their expression pattern may be both prognostic and predictive for patient survival. However, the currently available data is controversial in this regard. For instance, in renal cell carcinoma high TRAIL-R2 and low TRAIL-R4 expression correlated with poorer overall survival [
21]. In breast cancer, expression of TRAIL-R2 was associated with TRAIL-R4 positivity and correlated with poorer survival [
22]. In contrast, in colorectal cancer Ullenhag et al. could not detect any correlation of TRAIL-R1 and TRAIL-R2 expression status with patients survival [
23].
Caspase-8 is crucial for triggering apoptosis by death receptors since its recruitment to and activation at the DISC is the decisive step for the initiation of the caspase cascade [
24]. Besides apoptosis induction non-apoptotic functions of caspase-8 have been discussed, although these non-apoptotic signaling pathways and molecular targets have not been defined yet [
25]. Bcl-xL and Mcl-1 belong to the anti-apoptotic B-cell lymphoma-2 (Bcl-2) family of proteins [
26]. High expression of Bcl-XL has been associated with more aggressive tumor biology and/or drug resistance to various chemotherapeutic agents in hematologic and solid malignancies [
26]. Inhibition of Bcl-xL induces apoptosis and suppresses growth of hepatoma cells in combination with sorafenib [
27]. Mcl-1 is overexpressed in about 50% of HCC tissues [
28] but on the other hand deletion of Mcl-1 triggers hepatocarcinogenesis in mice [
29]. Recent studies have demonstrated that TRAIL expression is altered in HCC in comparison to normal liver tissue, but there are contradictory data about the expression of the different TRAIL receptors in HCC cells and tissues [
30‐
34]. Thus, we analyzed TRAIL receptors and the apoptosis regulatory proteins caspase-8, Bcl-xL and Mcl-1 in correlation with HCC grading and survival.
Discussion
In this study we assessed the expression of TRAIL receptors, caspase-8, Bcl-xL and Mcl-1 in 157 patients with hepatocellular carcinoma and normal liver tissue using tissue microarrays, and correlated the expression with clinico-pathological parameters. Survival analysis was carried out for patients who underwent liver resection.
TRAIL-R1 was significantly downregulated in less differentiated HCC. However, TRAIL-R1 expression did not correlate with patient survival after liver resection. Kriegl et al. reported a significant membrane staining of TRAIL-R1 in HCC compared to normal liver tissue and a longer survival of HCC patients undergoing partial hepatectomy with TRAIL-R1 membrane positive versus negative tumors [
31]. However, our immunohistochemical analysis detected considerable cytoplasmic but not membrane TRAIL-R1 staining [
31,
37]. Having established the specificity of our TRAIL-R1 (and TRAIL-R2) antibodies in TRAIL-R1- (and TRAIL-R2-) transfected cells, cytoplasmic staining prevailed also in this setting [
22]. Using the highly specific antibodies for TRAIL-R1 and TRAIL-R2, HCC cell lines also displayed strong cytoplasmic, rather than membrane, staining which was confirmed by flow cytometry (data not shown). Upon TRAIL death receptor upregulation by chemotherapeutic drugs, membrane staining of both receptors could be detected in HCC cell lines which was paralleled by enhanced surface receptor as detected by flow cytometry [
10]. These control experiments support the sensitivity and high specificity of our TRAIL receptor antibodies for both cytoplasmic and membrane staining. Our data are in line with reports on strong cytoplasmic rather than membrane staining of both TRAIL-R1 and TRAIL-R2 in primary HCC tissue [
33]. Correlation analyses of TRAIL-R1 expression and survival in other tumor entities revealed contradictory results. In colorectal cancer both low [
38] and high [
39] TRAIL-R1 expression correlated with poorer survival. Ullenhag et al. found no correlation between TRAIL-R1 expression level and survival in colorectal cancer patients [
23].
In our study both TRAIL-R2 and TRAIL-R4 were upregulated in dedifferentiated HCCs. However, for none of the TRAIL receptors expression correlated with patient survival. In previous studies high expression of TRAIL-2 [
40] was also associated with less differentiated tumors and implied poorer survival in breast cancer [
22,
41], renal cell carcinoma [
21], and NSCLC [
40]. In the report by Kriegl et al., TRAIL-R2 membrane staining correlated with better survival of HCC patients after partial liver resection [
31]. However, as stated above, in our cohort no relevant TRAIL-R2 membrane staining could be detected in HCC tissues. In summary, TRAIL receptor expression patterns seem to vary between different tumor entities and, therefore, their correlation with survival data may depend on tumor type and clinical setting (adjuvant, curative and palliative treatment).
Downregulation of TRAIL-R2
in vivo may mirror the selection pressure by antitumor immune responses (e.g. by TRAIL-expressing NK cells). On the other hand, TRAIL-R2-positive tumor cells may have developed TRAIL resistance downstream of the receptor level, thereby allowing for tumor cell proliferation despite TRAIL death receptor expression. Nevertheless, many chemotherapeutic drugs sensitize resistant tumor cells to TRAIL-induced apoptosis via enhancement of proapoptotic regulators of the extrinsic and intrinsic pathway [
8,
10,
42]. Thus, HCCs with high TRAIL-R2 expression should be eligible for combinatorial TRAIL-based therapies. Previously, we could show that TRAIL-R2 expression was highly correlated with TRAIL-R4 positivity in breast cancer [
22]. TRAIL-R4 overexpression correlated with poorer survival in breast [
22] and prostate cancer [
43]. Applying TRAIL-R2-specific agonists (e.g. the TRAIL-R2-specific antibody lexatumumab) may bypass the anti-apoptotic effects of high TRAIL-R4 expression and allow for effective tumor treatment [
11]. It has been discussed that therapeutic implications of TRAIL-based therapies might be limited by toxicity to non-transformed human hepatocytes [
44,
45]. Yet, we previously showed that there is a large therapeutic window which allows effective TRAIL-based cancer therapy [
10].
Analysis of the two anti-apoptotic Bcl-2 family members Bcl-xL and Mcl-1 revealed low expression of Bcl-xL in normal liver tissue, which was not-significantly upregulated in G2 and G3 tumors (data not shown). Expression of Mcl-1 was also increased in G3 tumors as compared to G1/2 tumors and normal tissue; however no correlation with survival could be detected (data not shown).
As the main initiator caspase of the TRAIL pathway, caspase-8 is located in the cytosol to be recruited to the TRAIL DISC after ligand binding to TRAIL-R1/R2. Loss or downregulation of caspase-8 has been proposed as a possible mechanism of apoptosis resistance in tumor cells [
46]. In our cohort, high cytosolic caspase-8 expression correlated with better survival independently from tumor grade, possibly reflecting the higher apoptotic potential of these tumor cells. Interestingly, we could demonstrate nuclear staining of caspase-8 in HCCs but not in normal hepatocytes. The staining intensity of nuclear caspase-8 correlated with grade of malignancy but also with poorer patient survival. Due to the strong correlation between nuclear expression of caspase-8 and tumor grading, multivariate Cox regression analysis could not detect an influence of nuclear caspase-8 on survival independent from the tumor grade. However, patient number with a nuclear caspase-8 score ≥10.3 might be too small (n = 10) for a multivariate analysis of the two parameters, high nuclear caspase-8 and tumor grading. Thus, our data need to be scrutinized in a larger cohort. Although high nuclear and cytosolic caspase-8 expression have an opposed effect on patient survival, high nuclear and cytoplasmic caspase-8 expression is not mutually exclusive, since 9 out of 56 patients (16%) and 3 out of 14 patients (21%) with a high nuclear caspase-8 score of ≥7 and ≥10.3, respectively, had also an equally high cytoplasmic caspase-8 expression level. Most of these patients had WHO grade 3 tumors (78% for a score ≥7, 100% for a score of ≥10).
Whereas the role of cytosolic caspase-8 as a factor in triggering apoptosis via death receptors has been well examined [
24,
47,
48], nuclear translocation of caspase-8 has so far not been described in HCCs. In contrast, nuclear localisation of caspase-8 has been found in apoptotic neurons [
49]. Since these cells were undergoing apoptosis, caspase-8 was suspected to shuttle to the nucleus to exert cleavage of the DNA repair enzyme PARP2, a hallmark of apoptotic cell death. In contrast to apoptotic neurons, in our study nuclear caspase-8 was detected in nearly all tumor cells of poorly differentiated HCCs (Figures
2 and
3E) and nuclear caspase-8 expression did not correlate with the apoptosis rate (r = 0.078, p = 0.420). This may indicate a non-apoptotic function of caspase-8 in HCCs. Enhancement of tumor cell migration and inhibition of Fas-induced apoptosis has been recently described as a non-apoptotic function of caspase-8 in different experimental cancer cell lines, which was not dependent on its catalytic activity but on Src-mediated phosphorylation of Tyr380 in a linker region between the small and large caspase-8 subunits [
50,
51]. Metastasis formation of non-apoptotic neuroblastoma cells was enhanced by recruitment of caspase-8 to the cellular migration machinery [
52]. Interestingly, in our cohort, high nuclear expression of caspase-8 correlated with a higher proliferation index of tumor cells (Ki67, r = 0.282, p = 0.0004, whereas the cytosolic expression of caspase-8 did not (r = 0.089, p = 0.274). A recent study has shown that caspase-8 can be sumoylated at lysine 156 leading to a 75 kDa isoform (p75) and that sumoylation of caspase-8 by SUMO-1 is associated with nuclear localization of caspase-8 [
53] suggesting that nuclear expression of caspase-8 in our study might be a result of sumoylation. Interestingly, SUMO-1 is overexpressed in HCCs [
54] and expression profiling has shown that HCC patients with shorter survival show higher expression of genes involved in sumoylation [
55,
56]. Although the physiological relevance of sumoylated caspase-8 is unclear, recent studies suggest that sumoylation of caspase-8 does not impair cytoplasmic caspase-8 activation, but that sumoylated nuclear caspase-8 (p75) can presumably cleave other, so far undefined, specific nuclear substrates [
53]. However, using a cleavage-specific antibody for caspase-8, we could not detect activated caspase-8 in the nuclei of tumor cells in our cohort.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
SB and RK participated in the design of the study and wrote the manuscript together with TMG. JS established the staining protocols for all antibodies and carried out the immunohistochemical studies together with SB and EB. EB collected clinical data. UH performed the statistical analysis. PS participated in the design and coordination of the study. WS conceived of the study, and participated in its design and coordination and helped to draft the manuscript. PS, SS and KB provided the histoarrays and revised the manuscript. HW developed all TRAIL-receptor-specific antibodies employed in this study, oversaw the establishing of the staining protocols for all antibodies, initiated and designed the study together with TMG, and revised the manuscript. All authors read and approved the final manuscript.