Abstract
Background: The poor prognosis of hepatocellular carcinoma (HCC) and the lack of specific screening markers underline the need for new biomarkers for human hepatocarcinogenesis. Materials and Methods: We investigated 10 postulated biomarkers for HCC (AFP, GPC3, OPN, IGF1, HGF, SPINK1, KPNA, FUCA1, CgA, HSP90) with microarray gene expression analysis and real-time polymerase chain reaction (RT-PCR) in HCC tissues of different etiologies. Results: Four candidate genes (FUCA1, HGF, IGF1, CgA) showed low median fold changes (fc) of expression compared to corresponding non-malignant liver tissues (fc range=0.2-0.8; maximum 15% of samples). The classic biomarker, alpha-fetoprotein (AFP), was significantly over-expressed (fc=2.4) in 30% of tumors. High tumor AFP expression was associated with significantly elevated serum AFP concentrations (>90% of cases). Five genes (OPN, SPINK1, GPC3, HSP90, KNPA2) showed significantly higher expression than AFP in 64% to 82% of samples (median fc range=2.9-8.3). RT-PCR analyses gave similar results. Conclusion: Unlike previous studies, our results did not confirm FUCA1, HGF, IGF1 or CgA as potential markers for HCC. In contrast, OPN, SPINK1, GPC3 and KNPA2 were significantly over-expressed in HCC tissues. These genes may be useful in developing future biomarkers and therapeutic strategies for HCC.
Hepatocellular carcinoma (HCC) is the fifth most common cancer and the third most common cause of death by cancer worldwide (1). In the last few decades, the incidence of HCC has increased, likely due to the growing worldwide prevalence of chronic hepatitis B or C. In particular, the emergence of the hepatitis C virus (HCV) in developed countries can account for approximately half of the increase in HCC (1, 2).
Currently, surgical resection and liver transplantation offer the best potential treatments for HCC (3, 4). However, those treatments are mostly only available when tumors are detected early. Thus, a curable tumor resection is only a realistic treatment option in a minority of patients that develop HCC as a late effect of chronic viral hepatitis with advanced end-stage liver disease. These patients currently account for 15-25% of HCC cases. The poor prognosis of HCC and its increasing incidence underline the importance of discovering novel screening biomarkers to detect tumors at an earlier, curable stage.
Alpha-fetoprotein (AFP) is currently the most common, classical tumor marker used for HCC evaluations. In 1956, AFP was discovered by Bergstrand and Czar (5). The earliest reports on the potential function of AFP as a diagnostic marker for HCC were published in the 1960s (6). AFP is a glycoprotein of approximately 70 kDa, synthesized during early embryonic development and, subsequently, in fetal hepatocytes. The AFP gene has been linked to chromosome 4, which harbors part of the albuminoid gene superfamily (7). Significantly elevated serum AFP levels were associated with various, non-malignant liver diseases, including viral hepatitis or liver cirrhosis; thus, AFP had low sensitivity (20-60%) and specificity (76-96%) as a biomarker for HCC, particularly in the early stages (8, 9). Other studies have shown significantly lower positive predictive values for AFP in different etiological subtypes of HCC, particularly in patients with virally-induced HCC (10). Based on these data, screening with AFP is currently not recommended by the German Association for the Study of the Liver (11). The absence of a specific biomarker for HCC with higher sensitivity and specificity than AFP underlines the need to establish other biomarkers for human hepatocarcinogenesis.
Modern gene expression analysis allows the detection of specific genetic patterns and the molecular pathways involved in HCC (12, 13). This approach may facilitate the detection of better screening markers for HCC. Recent studies that employed this technique have described several genes and gene products that were significantly overexpressed in tumor tissues and showed potential as new biomarkers for HCC (14, 15).
In the present study, we analyzed gene expression levels for 10 currently postulated biomarkers for HCC, including AFP, glypican 3 (GPC3), osteopontin (OPN), insulin-like growth factor-1 (IGF1), hepatic growth factor (HGF), specific serine protease inhibitor kazal type 1 (SPINK1), karyopherin α (KPNA), fucosidase, alpha-L- 1 (FUCA1), chomogranin A (CgA) and heat shock protein-90 (HSP90) to determine their potential use in screening or as prognostic indicators in patients with an elevated risk for HCC.
Materials and Methods
Acquisition of samples. Surgical specimens (both tumorous and corresponding non-malignant liver tissue) were obtained from 33 patients with HCC that underwent surgical treatment for histologically-proven HCC (10 HCV-positive, 4 hepatitis-B virus (HBV)-positive, 19 non-virally-induced; Table I).
In accordance with the Declaration of Helsinki, we obtained Ethics Committee approval. Parts of the resected samples were used for further genomic analyses, after obtaining written informed consent from the patients.
Comparison of AFP gene expression levels and AFP serum levels. Serum AFP levels were measured in all patients before liver surgery. In all cases, a minimum of 5 ml serum was stored and AFP was analyzed with the chemiluminescence technique (CLIA, Immulite AFP; Siemens Healthcare, Erlangen, Germany). Based on the fact that chronic viral infection or inflammation may lead to elevated AFP serum levels, an AFP elevation ≥200 ng/ml was considered sufficiently significant for suspecting hepatocellular carcinoma (HCC).
Histopathological evaluation. For each tissue sample, we obtained clinical, virological and pathological reports of tumor typing, staging (based on UICC criteria) and grading, in addition to patient clinical data. Hematoxylin-eosin staining was performed for detecting features, like bile canalicular structure and Mallory hyaline bodies. Additional histological and immunohistochemical stainings with a panel of antibodies (HEP-PAR-1, AFP, CK7, CA19-9, CD10, CEA) were routinely performed for all resected tissue biopsies to confirm the histological diagnosis of HCC and to exclude other types of malignancies.
Preparation of labeled cRNA and hybridization to oligonucleotide arrays. We used the SuperScript Choice system (Invitrogen, Carlsbad, CA, USA) to transcribe the purified, extracted total RNA to obtain double stranded cDNA. The first strand cDNA synthesis was primed with a T7-(dT24) oligonucleotide primer, after second-strand synthesis, in vitro transcription was performed in the presence of biotin-11-CTP and biotin-16-UTP (Enzo Diagnostics, city, county, country) to produce biotin labeled cRNA. Hybridization was performed by incubating (18-20 h) 200 μl of the sample with an human GeneChip (Hu133A; Affymetrix, Santa Clara, CA, USA), which contained 22,283 probe sets for known genes or expressed sequence tags (ESTs). A Gene Array scanner G2500A (Hewlett Packard, Palo Alto, CA, USA) was used to scan the chips according to procedures developed by Affymetrix.
Statistical analysis. The statistical analyses and presentation of data were performed in accordance with the MIAME criteria. The data will be published on the following web page: http://www.paracelsus-kliniken.de/scheidegg/fuer-fachkreise/forschung/gene-profiles.
Preliminary data analysis. We conducted a raw data analysis with the Affymetrix microarray suite software (MAS v5.0.1, Santa Clara, CA, USA). Statistical analyses and post-processing were performed with GeneSpring (v6.1; Silicon Genetics, Redwood City, CA, USA) and GeneExplore software (v1.1; Applied Maths, Sint-Martens-Laten, Belgium). Expression values were then log2-transformed on the basis of the signal:noise log ratio, which is given by comparing array results between tumor and non-malignant tissues. A p-value<0.05 (t-test) and a fold-change ≥2 in 60% or more of all analyzed samples were considered significant.
Selection of ten candidate screening markers for HCC. After searching the literature (Pubmed©, NCBI, Bethesda, MD, USA), we identified ten recently postulated screening biomarkers that we considered appropriate for individual expression analysis in the 33 HCC samples. The expression of these genes was compared with expression in corresponding, non-malignant liver tissues. Therefore, our screening panel included probes for the following genes: AFP, OPN, SPINK1, GPC3, HSP90, FUCA1, HGF, IGF1, CgA and KNPA2.
Validation of expression data by real-time polymerase chain reaction (RT-PCR). After microarray analyses, we used RT-PCR to validate the relative changes in gene expression observed for the most consistently over-expressed genes in HCC tissues (SPINK1, OPN, GPC3, KPNA). RT-PCR was performed with the LightCycler© system (Roche Diagnostics, Mannheim, Germany) on cDNA isolated from tumor tissues and corresponding non-malignant liver tissues. The gene-specific primers corresponded to the coding region of each gene. These primers were designed with the OLIGO software, supplied by Biomers.net (Ulm, Germany) and had the following sequences:
SPINK1: 151U; GCCTTGGCCCTGTTGAGTCTA
SPINK1: 273L; CACGCATTCATTGGGATAAGTATTT
GPC3: 1808U; CAGCAGGCAACTCCGAAGG
GPC3: 1929L; TGGGCACCAGGCAGTCAGT
KPNA: 328U; GAAAACCGCAACAACCA
KPNA: 501L; GCCCAAGAAGGACACAAAT
OPN: 696U; GGACAGCCGTGGGAAGG
OPN: 810L; TCAATCACATCGGAATGCTCA
RT-PCR reactions were performed with the LC RNA Amplification Kit SYBR Green I, (Roche Diagnostics, Mannheim, Germany). Amplification was followed by melting curve analysis. Relative values for the initial target concentration in each sample were determined with the LightCycler software, v3.5 (Idaho Technology Inc., Salt Lake City, UT, USA). The relative change in gene expression was computed by pairwise comparisons of tumor samples to samples of adjacent normal tissue for each patient.
Results
Gene expression profiling of HCC. Based on the 22,283 probe sets present on the chip, we found that, on average, 42.6% (HCC) and 39.8% (corresponding non-malignant liver tissue) of genes were expressed in the liver tissue samples.
Out of approximately 13,000 genes expressed in the HCC and adjacent non-malignant tissue samples, approximately 1,200 genes were either up- or down-regulated significantly. All statistically different genes (p<0.05) with a fold change of at least a 2-fold increase or decrease in at least 60% of the HCC tumor samples were used to generate a databank of 1,085 genes/ESTs. Out of these, 445 were up-regulated and 640 were down-regulated. With this data set, we implemented a supervised learning method (SLM) based on neuronal networking to obtain a specific gene expression profile for HCC that enabled rapid, reproducible differentiation between malignant and nonmalignant liver tissues in all cases (Figure 1).
Potential screening marker expression detected with microarray analysis. Four of the chosen candidate genes (FUCA1, HGF, IGF1, CgA) were not significantly overexpressed. Their median fold change (fc) ranged from 0.2 to 0.8 and they were over-expressed in a maximum of 15% of all analyzed tumor tissues. These genes showed significantly lower expression than AFP expression; therefore, they were not considered in further analyses.
AFP gene expression was increased, with a median fc of 2.4-fold, and it was significantly over-expressed in 30% of all tumor samples. In addition, five genes (OPN, SPINK1, GPC3, HSP90, KNPA2) showed significant over-expression (median fc between 2.9 and 8.3) in 64% to 82% of samples (Table I).
Potential screening marker expression detected with RT-PCR analysis. To validate the five most highly over-expressed candidate genes, we confirmed the expression levels of AFP, OPN, SPINK1, GPC3 and KNPA2 in malignant and corresponding non-malignant tissue samples with quantitative RT-PCR. In general, the changes in gene expression measured with the microarray technique reflected the results obtained with RT-PCR. However, the dynamic range of the expression levels estimated by RT-PCR was significantly higher (10- to 70-fold) than the ranges estimated in the microarray analysis (Figure 2).
Gene expression and serum AFP levels. To determine whether the changes in AFP gene expression found in HCC tissues reflected changes in protein expression, we measured serum AFP levels in all patients. We found that serum AFP levels ranged between 5 and 170,000 U/l. Moreover, in 30.3% of all patients, AFP levels were significantly elevated (≥200 U/l). In fact, 91% of serum AFP levels significantly correlated (p=0.001) with gene expression levels estimated with gene expression analysis (Figure 3).
Discussion
The poor prognosis and disappointing treatment options for patients with non-resectable HCC underline the need for better screening parameters and therapeutic targets. To date, AFP is the only tumor marker established for routine clinical screening and treatment observations in patients with HCC. However, consistent with other published data, our results demonstrated that AFP had very low specificity in the 33 tumors analyzed in the present study. Only 30% of all tumors showed over-expression of AFP, with a median fold change (fc) of 2.4. Also, in contrast to other groups (10), we observed an AFP over-expression more frequently in virally-induced tumors (36-40%). Among the patients that showed AFP overexpression in the microarray analysis, 91% also showed elevated serum AFP levels (≥200 U/l). Only two cases showed elevated serum AFP levels without microarray detection of significant gene over-expression. Taken together, our results showed that AFP gene expression levels correlated very well with serum levels, but AFP was only useful as a screening marker for HCC in about one third of all patients.
Modern diagnostic approaches, like gene expression profiling with DNA or oligonucleotide arrays or whole genome sequencing, are useful for discovering potential candidate genes that may serve as biomarkers or therapeutic targets. In the present study, we used oligonucleotide arrays to analyze the expression of 10 candidate genes that were discussed in the literature as potential screening and prognostic markers in human hepatocarcinogenesis. To confirm the expression data, we used RT-PCR and melting curve analysis to analyze the marker genes whose over-expression was most likely linked to HCC.
Interestingly, four of the genes postulated as screening or prognostic markers for HCC (FUCA1, CgA, HGF, IGF1) (16-194) were found to be inferior to AFP in their association with HCC. The expression changes (fc=0.2-0.8) were significantly lower than the change in AFP expression and expression was notably different from controls in only 0-15% of the analyzed cases. In our opinion, these genes may not be useful for routine clinical screening or for detecting early HCC.
The other five genes (OPN, SPINK1, GPC3, KPNA2, HSP90) included in our microarray analysis were found to be superior to AFP in their association with HCC. They showed significantly higher gene expression changes (fc=2.9-8.3) than AFP and expression was notably different from controls in 64-82% of the cases analyzed. Moreover, these data (except HSP90) were validated in the RT-PCR analysis. The expression levels determined with RT-PCR correlated well with the gene expression data. However, as shown previously by our group and other colleagues (20, 21), the dynamic range of the RT-PCR data was about 10-fold to 70-fold higher than that of the microarray analysis.
Among the five genes we identified, the most overexpressed, independent of the etiological risk factor (i.e., HCV, HBV or alcohol), was GPC3 (fc=8.3 in 76% of tumors). GPC3 was first described by Filmus and colleagues (22) (originally called OCI-5). They discovered it in an undifferentiated rat epithelial cell line. The first reports about up-regulation of GPC3 in HCC were published in 1997 (23). In the following years, other studies confirmed those reports with cDNA or microarray techniques (24), which provided the basis for performing immunohistochemical studies to elucidate the value of GPC3 as a screening marker for HCC (25). In those studies, GPC3 showed high sensitivity (>70%) and specificity (up to 90%) for discriminating HCC from other tumors or non-malignant hepatopathies.
Importantly, in addition to its usefulness as a marker for the early diagnosis of HCC, GPC3 also functions as an oncogene that promotes the growth of HCC by stimulating canonical Wnt signaling (26). Due to the high expression of GPC3, particularly in malignant transformed hepatocytes, compared to other molecular targets in hepatocarcinogenesis, current studies are developing anti-GPC3 strategies for treating HCC (27).
OPN was also significantly up-regulated (fc=6.2) in over 60% of the analyzed HCC samples. OPN (also called secreted phosphoprotein 1) is not a typical oncogene; it is mutated in activated cancer cells and it is regulated by a variety of stimuli, including the Wnt/Tcf signaling pathway, steroid receptors, growth factors, tumor necrosis factor-alpha and transcription factors, such as AP-1 and the RAS proto-oncogene (28). OPN is a ligand for the CD44 receptor, binds to alphaV-containing integrins and plays an important role in malignant cell attachment and tumor invasion (29). OPN is involved in several potential mechanisms in human carcinogenesis, which require the regulation of different signaling cascades and seems to be a potent anti-apoptotic factor via inhibition of caspase 3 activation (30); it also induces the activation of promatrix metalloproteinase-2 (pro-MMP2, mediated by nuclear factor kappaB) and it induces the secretion of urokinase-type plasminogen activator (31). In a previous gene expression profiling study, our group was the first to show that OPN was the most highly over-expressed gene in cholangiocarcinoma. In the present study, our data corroborated those of recent studies by showing that OPN may also be a useful marker for hepatocarcinogenesis. In addition, it was previously shown that OPN expression correlated with early recurrence, poor prognosis and metastasis in HCC (32).
In the present study, SPINK1 (also called PSTI or TATI) was significantly over-expressed (median fc=5.7), particularly in HCV-induced HCC (up to 80%), compared to non-malignant liver tissues. Recent studies postulated the oncogenic mechanism and role of SPINK1 and it was proposed as an early tumor marker for hepatobiliary cancer (33). This protein was first described in 1982 (34) after detection in patients with ovarian cancer. In that study, it was found to inhibit trypsins and other proteases, which led to its description as a tumor-associated trypsin inhibitor (TATI). At the end of the 1980s, TATI was proved to be identical to SPINK1 and PSTI.
As a potent proteinase inhibitor, the expression of TATI in malignant tissues appears to represent the proteolytic and metastatic activity of the tumor. Stenman and colleagues identified and isolated two tumor-associated trypsins (TAT-1 and -2) that were targeted by TATI. These enzymes were shown to activate matrix metalloproteases (MMPs) in vitro and in vivo in different tumor cell lines and in tumors, respectively (35). Extracellular matrix re-modeling and proteolysis mediated by MMPs is associated with cancer invasion and metastasis in advanced tumor stages. Consequently, TATI may play a crucial role in tumor cell dissemination and invasion. Other research groups postulated that TATI directly blocked apoptosis by inhibiting caspase-independent or serine protease-dependent apoptosis, particularly in virally-induced hepatocarcinogenesis (36). The postulated viral induction of SPINK1 up-regulation may explain the up-regulation of this gene observed in the present study (37), particularly in the group of patients with HCV-induced HCC. This observation suggested that HCV and HBV might affect treatment options for both HCC and chronic viral hepatitis, one of the most important risk factors for human hepatocarcinogenesis.
In our data, KNPA2 (also called Importin α2) was over-expressed in over 80% of all HCCs (median fc=3.3). Currently, KNPA2 is postulated to be a new tumor marker for different adenocarcinomas, particularly lung cancer (38). Although its function remains under investigation, particularly in human carcinogenesis, one hypothesis is that KNPA2 may be involved in the nuclear transport of proteins. An immunohistochemical analysis of tissue samples by Yoshitake et al. demonstrated (39) cancer-specific over-expression of KNPA2 in up to 36% of HCCs.
HSP90 was also significantly up-regulated (fc=2.9) in 76% of our HCC samples. HSP90 is an essential chaperone that facilitates the function and integrity of a wide range of potential oncogenes (40). HSP90 expression modulates transcription factors (e.g., HIF-1, STAT3) and intracellular signaling pathways (e.g., Akt/Erk pathway), which lead to cell growth and proliferation (41). HSP90 up-regulation, in combination with cyclin-dependent kinase 4 (CDK4) activation, was predicted to contribute to HCC development. Moreover, HSP90 over-expression was associated with a poor prognosis in patients with HCC (42). Recently, several studies have shown that HSP90 may be a relevant therapeutic target in different tumors, but particularly in HCC (43).
Compared to AFP that encodes for the current “gold standard” for serological screening, the present study suggested that five other genes (OPN, SPINK1, HSP90, KNPA2, GPC3) showed a greater potential as screening or prognostic markers for human hepatocarcinogenesis. Despite the small sample size of the current study, our data underlined the importance of these five genes in the development of human HCC. Currently, GPN3, HSP90 and OPN are under investigation as potential therapeutic targets that may contribute to our fight against HCC in the future.
Acknowledgements
This study was supported by the fortüne-program of the University of Tuebingen, No. F1281305.
Footnotes
-
Conflicts of Interest
The Authors declare no conflict of interest.
- Received December 15, 2014.
- Revision received January 10, 2015.
- Accepted January 16, 2015.
- Copyright© 2015 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved