Introduction
Biliary tract cancers (BTC) comprise tumors of the extra- and intrahepatic bile ducts and the gallbladder cancer (1). The prognosis is poor as the only curative treatment, resection, is feasible in about 50% of patients only due to advanced stage and co-morbidities precluding surgery (2). Additionally, the 5-year survival rate is 13–44% only, even with optimal conditions of tumor cell-free margins and absence of lymph node dissemination (3). For non-resectable BTC, palliation using chemo- radio-chemotherapy or photodynamic tumor ablation could achieve median survival times of up to 18–28 months (4–6). As yet these therapies cannot stop BTC tumor cell proliferation in the long-term. Therefore, the identification of molecular mechanisms of tumor growth, of tumorigenic sub-populations of cells and of predictors of therapeutic response is needed to develop better strategies for treatment of BTC (7,8).
In this context, it was recently proposed that cancer cells achieve a malignant phenotype by re-activating early embryonic differentiation programs that physiologically regulate invasion, migration, angiogenesis and differentiation, thus adopting characteristics of normal stem cells (9–11). Particularly the re-expression of embryonic transcriptions factors like c-Myc, Sox-2, Oct3/4, KLF4 or Nanog is now considered as a mediator of oncogenesis in solid tumors (11,12).
The identification of cell surface markers expressed on tumor initiating cell populations (e.g. CD24, CD44) (13–15) has led to the description of so called stem cell-like signatures in various human cancers, including hepatocellular carcinoma [CD44+, CD90+, CD133+, ALDH (aldehyde dehydrogenase)], neuroblastomas (Oct3+, Sox-2+, Nanog+, KLF4+) and other disease entities (15). Although these markers are neither specific for a certain cancer histology nor for all stem cells per se, these profiles are independently associated with a poorer outcome in patients with e.g. colorectal cancer (16) and, in part, also in BTC (17). Yet, a detailed analysis of stem cell markers in BTC is still missing and only recently cells with stem cell-like properties (such as CD24+, CD44+, EpCAM+) were isolated from extrahepatic BTC (18). In line with these results, an aggressive phenotype was observed in BTC cells subsequent to overexpression of the stem cell-like genes c-MYC, Sox-2, OCT3/4, and KLF4 (19).
In this study, we investigated the expression of markers of putative stem- or progenitor cells in several human biliary tract cancer cell lines in vitro, in a nude mouse xenograft model in vivo and in a human TMA and correlated these expression patterns with cell proliferation and histological differentiation as well as clinicopathological characteristics.
Materials and methods
Cell culture and human BTC model
Bile duct carcinoma cell lines CCLP-1, Egi-1, SkChA-1, TFK-1 and gallbladder cancer cell lines MzChA-1, MzChA-2, GBC were cultured as described previously (20) using Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (v/v) fetal bovine serum (FBS; PAA Laboratories, Pasching, Austria) at 37°C and 5% CO2. Cell lines and human tissue samples are collectively referred to as biliary tract cancer (1) in the following [see (20) for original tumor grading].
Quantitative real-time RT-PCR
For quantification of mRNA levels, cells cultured in 10 cm diameter Petri dishes were lysed using TRIzol reagent (Invitrogen, Lofer, Austria) and total RNA was isolated according to the manufacturer’s instructions. Total RNA was subjected to DNase treatment, reverse transcription and the cDNA was analyzed using quantitative real-time reverse transcription PCR (qRT-PCR) as described previously (20) using a Rotorgene 6000 real-time PCR thermocycler (Qiagen, Hilden, Germany) - see Table IA for primer sequences. The relative expression of each transcript (X) was calculated using the formula 2−ΔCt (21,20), in which ΔCt=Ct(x)−Ct(housekeeping gene) with β-actin as the housekeeping gene (22).
Xenograft model
Cell lines were harvested, resuspended in sterile physiologic NaCl solution and injected into the flank of 4 to 6-week old NMRI mice (each 3x106 cells per mouse/5 mice per cell line; Harlan Winkelmann GmbH, Borchen, Germany) (20,23,24). Animals were kept in a light- and temperature-controlled environment and provided with food and water ad libitum. Animals were sacrificed at mean day 45 (range 39–46) by cervical dislocation and tumor samples were fixed in 10% phosphate-buffered formalin. Ethical approval was granted before any animal experiment by the Regional Government of Lower Franconia, Würzburg, Germany (no. 54–2531.31–9/06).
Tissue preparation and immunochistochemistry
All specimens were fixed in 4% buffered formalin, routinely processed and embedded in paraffin wax. Immunohistochemistry was done using routine diagnostic methods as published recently (25). In short, immunohistochemical stainings were carried out using an autostainer system (Dako, Vienna, Austria) according to the manufacturer’s recommendations. Antigen retrieval was performed by heat induced epitope retrieval in pH 9.0 antigen retrieval buffer (Dako) at 95°C for 40 min. Table IB lists the used antibodies, pretreatment conditions and dilutions. Tonsils and lymph nodes served as positive controls; control experiments were negative using PBS replacement of primary or secondary antibodies and same processing as described above (not shown).
Interpretation of immunohistochemistry
The stained slides were digitalized using the ImageAccess 9 Enterprise software (Imagic Bildverarbeitung, Glattbrugg, Switzerland). Images were evaluated using the particle analysis module with optimized binarization method and assessed by two independent investigators (R.K. and D.N.).
Patient characteristics and tissue microarray
Only patients who underwent surgical resection between 1997 and 2010 with complete histopathological records at the Institute of Pathology (Salzburg) were included in the investigations. None of these patients received neoadjuvant therapies. Follow-up data were received from the clinical database and the tumor registry of the Paracelsus Medical University/Salzburger Landeskliniken. Overall, 34 paraffin-embedded tissue samples with primary BTC were included (mean age of 65.90±12.6 years, sex: 12 male/22 female) with complete comprehensive clinical information which were subsequently used as covariables (for detail see Table II). All cases were re-classified according the 7th TNM classification of 2010 (26). In contrast to published data (27), cases with intrahepatic BTC (58.8%, n=20) were more frequent in our sample pool. In short, intrahepatic BTC showed preferential mass forming growth pattern and were larger than perihilar BTC, whereas perihilar BTC had a more advanced disease stage according TNM (details see Table II).
Tissue microarrays were prepared from paraffin-embedded tissue blocks. First, we designed a grid using commercial drawing software Microsoft Office 2007 Visio consisting of 2 mm black circles leaving 2 mm space between them and printed the grid on plain paper. The grid was fixed to stainless steel moulds and dried cores of paraffin-embedded BTC tissue blocks obtained using a 2 mm Harris Uni-Core punch (Ted Pella Inc., Redding, CA, USA) were aligned to the grid. Once all cores were attached to the mould, melted paraffin was gently poured into the mould and the TMA was handled according to routine histopathological procedures from this point on. Additionally, to exclude heterogeneous distribution pattern of the investigated stem cell markers, immunohistochemical staining was performed on tumor center and tumor border areas corresponding to conventional H&E stained slides.
Ethics
All analyses on human BTC samples were carried out according to the guidelines of the Paracelsus Medical University Salzburg/Salzburger Landeskliniken and were approved by the local ethics committee (415-EP/73/37-2011).
Statistical analysis
Statistical analysis was performed with SPSS 18.0 (SPSS Inc., Chicago, IL, USA). Kendall’s rank two-tailed test and Spearman rank correlation test was used for the correlation analysis. The χ2 test, Mann-Whitney U test or Student’s t-test was used to compare data of nominal, ordinal or interval level, respectively. Univariate ANOVA (analysis of variance) was used to test for differences between groups of tissue samples [using least significant difference (LSD) test post hoc test to adjust for multiple comparisons]. For survival analysis, cases with missing date of death were excluded. Univariate survival analysis was performed using the Kaplan-Meier method comparing the survival curves with the log-rank test. For multivariate survival analysis, the Cox proportional hazards model was used. Differences were considered significant at p<0.05.
Results
Expression pattern of stem cell markers identify distinct groups of cell lines
The mRNA of all stem cell markers could be detected at low levels throughout all tested human BTC cell lines in vitro, with highest level for CD44 compared to lower levels in all other BTC cell lines, Fig. 1A. The levels of mRNA in vitro mostly correspond to the protein expression levels in vivo (compare Fig. 1A and B). Generally, the expression data on mRNA and protein levels revealed a heterogeneous pattern with cell lines expressing rather low levels of e.g. CD133 and CD44 (Egi-1, MzCha-2, SkChA-1) compared to the others.
Based on the observed protein expression heterogeneity, hierarchical cluster analysis was used to test whether these patterns could identify distinct subgroups of cells lines. As shown in Fig. 2A, protein expression of CD133 and CD44 classified the cell lines into two groups of which the first group (GBC, MzChA-1, TFK-1 and SkChA-1) is characterized by a CD133+/CD44+/Nanog+/Sox-2− expression pattern. The second group (MzChA-2, Egi-1 and CCLP-1) displayed an inverse expression pattern (Fig. 2B). Further comparison between these groups indicated that the first cluster (CD133+/CD44+/Nanog+/Sox-2−) shows decreased proliferation markers (Ki67, cyclin D1) and a higher level of differentiation markers (Ck7, Ck8/18, E-cadherin).
Correlation analysis of stem cell marker expression and differentiation in vitro and in vivo
In the next step, we compared the expression of stem cell markers with those of differentiation in these human BTC cell lines which were comprehensively analyzed for immunochemical and histological characteristics in a previous study (20). As shown in Fig. 3, correlation analysis revealed significant association with morphological and molecular markers of proliferation and differentiation as follows: overall, similar correlations are found based on mRNA data (in vitro; Fig. 3A) and protein data sets for these cell lines (in vivo/xenografts; Fig. 3B). As expected from cluster analysis (Fig. 2), the expression of the differentiation markers Ck7 and Ck19 is significantly associated with CD133 and CD44 expression (Fig. 3A and B, respectively). In xenograft tumors, expression of proliferation markers (cyclin D1 and Ki67) are preferentially low in cell lines with high expression of CD44. Interestingly, the expression level of Sox-2 is positively associated with vimentin and nuclear β-catenin expression, while negatively correlated with the differentiation markers Ck8/18 (not shown) and Ck19.
Expression of stemness markers in human BTC
The expression of the mentioned stem cell markers were investigated in human BTC using a tissue microarray consisting of a total of 34 cases including 20 intrahepatic, 11 perihilar, and 3 cases of extrahepatic BTC (see Table II for details). As shown in Fig. 4A, overall protein expression exhibited the highest levels of Nanog followed by CD133 as well as CD44, Bmi-1, Nestin and Sox-2. Fig. 4A also indicates that the expression of these stem cell markers is not dependent on the anatomical location (intra- versus perihilar or extrahepatic), except for Nestin (only perihilar versus extrahepatic). Analysis of stemness marker expression levels depending on the growth type also revealed no significant differences, yet a trend to increased expression of Nanog can be seen for periductal compared to a mass-forming growth type (Fig. 4B). Detailed analysis depending on the intratumoral localization indicated significant higher expression of Bmi-1, CD133 and Nanog in the tumor centers, while Nestin and CD44 are preferentially expressed in the tumor front sections (all p<0.05, paired t-test; Fig. 4C).
Correlation analysis of the marker expression and selected differentiation/proliferation characteristics is shown in Fig. 3C: positive correlation between the proliferation marker Ki67 and stemness marker expression can be found for CD133 and Bmi-1, while CD44 and Nanog expression is associated with high levels of the cell cycle inhibitor p27. Additionally, other cell cycle associated proteins were positively associated with the stemness marker expression. Nuclear β-catenin is indicative of active Wnt signaling, a signaling pathway implicated in the generation and maintenance of the cancer stem cell population. Interestingly, expression of Bmi-1 CD133 and Nanog in our set of samples is significant positively correlated with nuclear β-catenin. Interestingly, markers of differentiation (cytokeratins/vimentin) showed no consistent and partially opposite relationship with the stem cell markers. Significant correlation are preferentially found in the tumor center (in detail see Fig. 3C). Based on the findings from xenograft stemness marker expression, the TMA cases were classified using hierarchical cluster analysis based on the marker expression of CD133 and CD44. As shown in Fig. 5A, three distinct clusters could be identified characterized by significant differences in expression of these markers. The marker Bmi-1 showed a significant different expression in the three groups similarly to that of CD133. Interestingly, the proliferation marker Ki67 is significantly higher expressed in the cluster which shows high levels of CD133 and Bmi-1 (Fig. 5B).
Association of stem cell marker expression with clinicopathological data
Using univariate ANOVA (LSD post hoc test), the expression of the investigated stemness markers was compared for different groups of samples stratified either for the different UICC stages, grading, the T-, N-, M-status, growth type and the intra-tumoral localization (see Fig. 6). Here, the analysis revealed only significant different expression of stemness markers for T-status (CD44) and M-status (Nestin) as well as tumor localization (CD133 and Nestin) regarding the tumor subregions (emphasizing predominantly the impact of the tumor front). It is notably that clinicopathological data summarizing different cofactors such as UICC stage (T-, N- and M-status as well as tumor localization) and tumor grading (different morphological aspects) did not reveal a clear dependency of the stemness marker expression pattern.
Impact of differential stemness marker expression on overall survival
Survival analysis using the Kaplan-Meier method as well as log-rank test showed that patients with mass-forming growth types of BTC have a tendency to poorer survival characteristics compared to those with a periductal growth pattern (Fig. 7A). When patients were stratified for low versus high expression levels of the stemness markers, a clear yet not significant trend of high expression towards shorter survival times could be found for CD133 and CD44 when these markers were used individually (Fig. 7B and C). Combination of all stem cell markers revealed a similar trend of poorer prognosis in patients with tumors of higher marker expression (Fig. 7D) which was statistically significant for the protein expression values (p=0.017, Fig. 7E).
Discussion
Referring to the recently updated hallmarks of cancer by Hanahan and Weinberg (28) the cancer stem cell model is now accepted as a valid concept of carcinogenesis. This model implies that either a subpopulation of cancer cells acquires a stem cell-like phenotype or dormant tissue stem cells are activated and gain oncogenic properties, associated with enhanced tumorigenic potency of self-renewal, migration and metastatic dissemination. On this background, numerous studies investigated stem cell characteristics in diverse human malignancies giving evidence that most if not all tumors showed a specific stem cell signature (15). Therefore, the current study aimed to investigate the expression of protein markers in human BTC associated with the stemness phenotype and its clinicopathological consequences. In brief, analysis of in vitro samples, xenograft tumor and a human BTC TMA revealed a heterogeneous expression pattern of stemness marker expression. Additionally, survival analyses of these samples indicate a trend to a worse prognosis for cases with higher expression of CD133 and CD44.
In the first part of our investigations we detect all of the employed stem cell markers Bmi-1, Sox-2, Nestin, CD133, CD44 and Nanog showing differential expression pattern in vitro and in vivo, whereby mRNA and protein expression were mostly concordant. Compared to data of stem cell markers in BTC, we confirm the upregulated expression of CD44 and CD133 (17–19,29), but also of the other investigated stemness markers. We expand existing limited data by linking the expression of these stem cell markers with morphology (ductal-solid, positive) and molecular markers of proliferation (negative) and differentiation (dedifferentiation: positive/transdifferentiation negative). As published earlier, dedifferentiation is preferentially linked to stemness characteristics, which can be readily recognized at the level of morphology emphasizing the role of basic histology (30). However, not the hyperproliferative tumor fraction, but rather the resting tumor compartment shows higher expression of these stem cell markers, partly explaining the fact that conventional therapeutic approaches such as chemo- and radiotherapy frequently cannot eradicate the tumor cells completely leading to subsequent recurrence (29,31). Nevertheless, Sasaki et al (32) were able to show that Bmi-1 is expressed in intrahepatic BTC independent of histologic differentiation or location and knockdown of Bmi-1 in BTC cell lines decreased proliferation/colony formation (32). In line with these results, the groups of BTC samples generated by cluster analysis based on CD133 and CD44 marker expression show a concordant expression of Bmi-1 and the proliferation marker Ki-67 (Fig. 5B). Such protein associated subgrouping of primary BTC could be relevant for targeted therapeutic approaches in future (32), as it was shown that expression analysis of nuclear β-catenin and CD133 in CRC were able to detect high risk cases in patients CRC subgroups of stage IIA for therapeutic decisions (33).
In the second part of our study we investigated the expression pattern of these stem cell markers in a TMA of resected primary BTC. The markers were detected to a heterogeneous extent as seen in the experimental in vitro and in vivo setting, whereby the different levels possibly reflect the influence of the extracellular matrix. Interestingly, we could not detect differences between intrahepatic, perihilar and extrahepatic anatomic location (except for Nestin) as would be expected from studies investigating cell cycle-associated components (34–36). Similarly, a recent publication revealed a significantly different expression of CD133 depending on the tumor growth pattern, but not on intrahepatic and perihilar tumor localization of BTC (37). However, we found a significant preferential localization of CD133, Bmi-1 and Nanog in the tumor center, whereas CD44 and Nestin tend to be expressed in the tumor front. This is in line with earlier findings that intra-tumoral gradients are important for tumorigenesis as well as progression [for review see (38,39)]. Experimental investigation revealed that tumor stem cell population is heterogeneous reflecting full morphologic and phenotypic heterogeneity and forcing whole tumor analysis (40). In own previous studies, we were able to show that members of the hedgehog pathway as well as markers of cell cycle-associated proteins were topographically differentially distributed in ductal adenocarcinoma of the pancreas (41) reflecting the intratumoral heterogeneity as well as the associated complexity of tumorigenesis. Our findings that not anatomy, but intra-tumor compartments are linked to expression pattern must be seen under the limitation of small number of extrahepatic cases and need confirmation in a larger set of BTC samples.
Interestingly, we found similar correlation of the stem cell markers with markers of differentiation and cell-cycle-associated proteins in the BTC TMA and in our in vitro and in vivo experiments, whereby the in vitro/in vivo situation rather reflects a long-term quiescent than an actively cycling (primed) stem cell as seen in the BTC TMA analysis described (42,43). Moreover, these results do not challenge the indispensable explanatory worth of clinical samples (44) yet they highlight the value of preclinical models as a starting point for investigations as well as models for development of novel therapeutic approaches (45).
In the third part of our study we investigated whether the expression of stem cell markers is linked to clinicopathological data (such as staging, grading and survival). Interestingly, depending on the intra-tumoral localization, only a few stem cell markers (such as CD133, CD44 and Nestin) showed an association with clinicopathological data (T and M status as well as tumor localization) emphasizing the importance to investigate all tumor subregions and indicating that the classical WHO grades for BTC do not reflect the biological and clinical characteristics of BTC. Using Kaplan-Meier survival analyses, high expression of CD133 and CD44 goes along with a non-significant trend to lower overall survival. This result is supported by published data in different human cancers: especially gastric [Sox-2, Oct34 (46,47), Nanog (48), Bmi-1 (49)], colorectal [c-MYC, Sox-2, OCT3/4, LIN28, KLF4, and Nanog (50)], liver [Bmi-1 (51)] and esophageal squamous cell cancer [Oct3/4 and Sox-2 (52)] as well as of human non-gastrointestinal cancers {breast [Sox-2 (53)], lung [Bmi-1 (54)] and brain [Bmi-1 (55)]}. Additionally, these studies could partially demonstrate an association of stem cell marker expression with clinical and pathomorphological data such as stage, and lymph node status. Such relationships could not be found in our sample set, possible due to the rather small sample size. Nevertheless, survival analyses showed that, besides tumor growth pattern, the combination of CD133 and CD44 likely is as a prognostic marker significantly associated with worse survival. This is in line with the data of Shimada et al who demonstrated that intrahepatic cholangiocarcinoma expressing the CD133 marker has a significantly worse survival than patients without expression of CD133 (29). The initial isolation and characterization of cancer stem cells in extrahepatic cholangiocarcinoma by Wang et al (18) was based on a marker combination of CD44+, CD24+, EpCAMhigh. Cells encompassing this marker phenotype were present in human extrahepatic cholangiocarcinoma to an extent of 0.39–2.27% and showed high tumorigenic potential in a NOD/SCID mouse model. It remains to be analyzed, whether these markers have a more significant prognostic value than those investigated in the present study. Finally, our findings are supported by recent published investigations on primary cell cultures of gall bladder cancer showing that CD133+ as well as CD133+/CD44+ tumor cell population were more resistant to chemotherapeutic reagents, possessed higher colony-formation ability in vivo and higher tumorigenicity in nude mice than their antigen-negative counterparts (37,56,57).
In conclusion, analysis of expression patterns of putative stem cell markers in vitro and in a xenograft model classified the human BTC into subgroups indicating the heterogeneity of carcinogenesis in BTC. Additionally, the prognostic relevance of the markers identified in this study to significantly impact on overall survival must be proven in further experimental and morphological studies of human tumor specimens of BTC in a larger patient cohort.