Abstract
Early detection of hepatocellular carcinoma (HCC) is critical for successful treatment and favorable prognosis. To identify novel HCC biomarkers, we used the WHV/c-myc transgenic (Tg) mice, an animal model of hepatocarcinogenesis. By analyzing their gene expression profiling, we investigated differentially expressed genes in livers of wild-type and Tg mice. The cytochrome P450, family 17, subfamily A, polypeptide 1 (CYP17A1), a hepatic P450 enzyme, was revealed to be overexpressed in the liver tissues of Tg mice at both preneoplastic and neoplastic stages. Mouse-to-human validation demonstrated that CYP17A1 mRNA and protein were also significantly increased in human HCC tissues compared with paired nontumor tissues (P = 0.00041 and 0.00011, respectively). Immunohistochemical studies showed that CYP17A1 was overexpressed in 67% (58 of 87) of HCC, and strong staining of CYP17A1 was observed in well-differentiated HCCs. Consistent with this, the median serum levels of CYP17A1 were also significantly higher in patients with HCC (140.2 ng/mL, n = 776) compared with healthy controls (31.4 ng/mL, n = 366) and to those with hepatitis B virus (57.5 ng/mL, n = 160), cirrhosis (46.1 ng/mL, n = 147), lung cancer (27.4 ng/mL, n = 109), and prostate cancer (42.1 ng/mL, n = 130; all P < 0.001). Notably, the elevations were seen in most AFP-negative HCC cases. Altogether, through mouse-to-human search and validation, we found that CYP17A1 is overexpressed in HCCs and it has great potentiality as a noninvasive marker for HCC detection. These results provide a rationale for the future development and clinical application of CYP17A1 measurement to diagnose HCC more precisely. Cancer Prev Res; 9(9); 739–49. ©2016 AACR.
Introduction
Hepatocellular carcinoma (HCC) is currently the fifth most common cancer in men and the seventh most common cancer in women worldwide, with increasing incidence in economically developed regions, such as Western Europe and the United States (1, 2). Most cases of HCC are associated with cirrhosis related to chronic hepatitis B virus (HBV) and hepatitis C virus (HCV) infection, and with a very high mortality rate (3–5). Because early HCC is usually asymptomatic, most patients are frequently detected in advanced stages, leaving few therapeutic options by the time of diagnosis. So far, alpha-fetoprotein (AFP) is the standard molecular marker widely used in clinical practice, in conjunction with hepatic ultrasonography, to monitor and diagnose HCC (3, 6). However, AFP levels were normal (<20 ng/mL) in up to 40% of patients with HCC, particularly during the early stages, showed a low sensitivity for detection of HCC (7–9).
Apart from AFP, other markers, including des-gamma carboxyprothrombin (10), glypican-3 (11), and golgi protein 73 (GP73) (12, 13) have been proposed for HCC detection. For instance, a recent large cohort study has shown that GP73 is an accurate serum marker and the combined measurement of GP73 and AFP as a standard clinical test for HCC is recommended (13). However, GP73 might not be a general HCC serum marker but might have utility as a marker of HCV-related HCC, as reported by Riener and colleagues (14). Thus, before these markers can be applied clinically, multicenter cohort studies and further validation phases are still needed (15). Meanwhile, discovering novel biomarkers and building a more sensitive detection system are expected.
The DNA microarray is a high-throughput approach of simultaneously analyzing the expression of thousands of genes and have been applied in many research fields, including discovery of cancer biomarkers (16). To investigate the differential gene expression profile of early HCC, the animal model might be useful, because it is very difficult to collect liver samples from patients before HCC formation or at the very earliest stage of HCC. The WHV/c-myc transgenic (Tg) mouse is a model of hepatocarcinogenesis in which the c-myc oncogene is activated by adjacent woodchuck hepatitis virus (WHV) DNA sequences (17). WHV/c-myc mice stably develop HCC with a relatively short latent period of 8 to 12 months, with a high (near 100%) tumor incidence. Additionally, the histological features of HCCs in the mice are very similar to human HCCs (17).
Cytochrome P450, family 17, subfamily A, polypeptide 1 (CYP17A1), also known as Cytochrome P450c17 or CYP17, is the microsomal p450 enzyme that catalyzes both 17 α-hydroxylase and 17,20-lyase activities with a critical role in biosynthesis of steroid hormones (18, 19). The 17 α-hydroxylase activity is essential for the generation of glucocorticoids such as cortisol, but both the hydroxylase and 17,20-lyase activities are required for the production of androgenic and oestrogenic sex steroids (20). CYP17A1 is also an important target for the treatment of prostate cancer that proliferate in response to androgens (21–23). Deletion of the mouse CYP17A1 gene causes early embryonic lethality, suggesting an essential function of steroid products of CYP17A1 in embryonic development (24). Interestingly, the expression of CYP17A1 is age-related; it presents in the liver and testis of immature rats (1 and 3 weeks old), but not in those of adults; the significant decrease in the serum levels of steroids produced by CYP17A1 with increasing age was also observed (25). However, whether abnormal expression of CYP17A1 is associated with hepatocarcinoma is unclear.
In the present study, we performed gene microarray to investigate the hepatic gene expression profiles of WHV/c-myc Tg mice and wild-type (WT) C57/BL6 mice at the age of 5 and 11 months. Microarray data revealed that CYP17A1 was significantly upregulated in Tg mice compared with WT mice at both stages. Further assaying of CYP17A1 uncovered that its levels are significantly increased in HCC tissues and sera of HCC patients. Additionally, the diagnostic accuracy of CYP17A1 and AFP was also compared in the same patients. Our data suggest that CYP17A1 could serve as a potential serum marker for HCC detection.
Materials and Methods
Animals
WHV/c-myc Tg mice were generously provided by Dr. Etiemble (Institut Pasteur, France). C57/BL6 WT mice were purchased from the Shanghai Experimental Animal Center. Liver examination and H&E staining were performed, as previously described (17). Serum alanine aminotransferase (ALT), alkaline phosphatase (AKP), and aspartate aminotransferase (AST) levels in Tg and WT mice were measured by the 7020 Clinical Analyzer (Hitachi, Tokyo, Japan). All procedures were approved by the Laboratory Animal Care and Use Committees of Kunming University of Science and Technology.
Specimens
All the human sera and HCC tissues, including adjacent nontumorous controls, were obtained from the 1st Affiliated Hospital of Kunming Medical University over a 3-year period (2013–2016). Written informed consent was obtained from each subject. Tumor staging was determined using the United Network of Organ Sharing–modified pathologic tumor–node–metastasis (TNM) staging system for HCC (26) and early HCC was defined as stages I to II. The histological grade of differentiation was assessed on H&E-stained sections according to Edmondson and Steiner criteria (27). AFP was tested using commercially available immunometric assays at hospital. The clinically acceptable normal AFP level is defined as <20 ng/mL. Detailed information about the samples used in this study is listed in Supplementary Table S1. The study protocol was approved by the Institutional Guidelines Committee of Kunming University of Science and Technology and conformed to the ethical guidelines of the Declaration of Helsinki.
Sample preparation and GeneChip hybridization
Total RNA was extracted from the livers of 5 and 11 months old of Tg and WT mice, respectively, using an RNeasy kit (QIAGEN). Procedures for cDNA synthesis, Biotin-labeling and hybridization were carried out according to the manufacturer's protocol (Affymetrix). The Mouse Genome 430 2.0 Expression Arrays (Affymetrix) which offers over 39,000 transcripts on a single array were used for hybridization. The data were processed and analyzed using GCOS1.2 software instructed by the manufacturer (Affymetrix).
Quantitative real-time polymerase chain reaction (real-time qPCR)
RNAs were isolated using the TRIzol reagent (Invitrogen). Reverse transcription was performed with SuperScript II reverse transcriptase (Invitrogen). Real-time qPCR was conducted on an ABI 7500 Fast Sequence Detection System using SYBR Green PCR Master Mix (Applied Biosystems). The following primer sets were used: (i) for mouse c-myc: myc-F, 5′-GAACCCGTGAGGTGGAAGAA-3′ and myc-R, 5′-TGGGTAAAGGCGGGGAAAAGT-3′; (ii) for mouse cyp17A1: cyp-F, 5′-GCCCAAGTCAAAGACACCTAAT-3′ and cyp-R, 5′-GTACCCAGGCGAAGAGAATAGA-3′; (iii) for mouse GAPDH: gap-F, 5′-GCCTCCAAGGAGTAAGAAAC-3′ and gap-R, 5′-GAAATTGTGAGGGAGATGCT-3′; (iv) for human cyp17A1: cyp-F, 5′-TTCGTATGGGCACCAAGACT-3′ and cyp-R, 5′-GTTGTTGGACGCGATGTCTA-3′; (v) for human GAPDH: gap-F, 5′-GTTCGACAGTCAGCCGCATC-3′ and gap-R, 5′-GGAATTTGCCATGGGTGGA-3′. Each reaction was repeated independently at least three times in triplicate.
Western blotting
The tissue or serum samples were lysed in RIPA lysis buffer (Thermo) containing 1% protease inhibitor cocktail (Sigma-Aldrich). Western blot analysis was performed by the standard method with the following primary antibodies: the rabbit polyclonal antibody against CYP17A1 (Proteintech) diluted to 1:1,000, the rabbit polyclonal antibody against AFP (Proteintech) diluted to 1:2,000, and the mouse monoclonal antibody against β-actin (Sigma-Aldrich) diluted to 1:2,000. Protein bands were detected using an enhanced chemiluminescence reagent (Promega) and quantified using Quantity One software (Bio-Rad).
Tissue microarray and immunohistochemical staining
A tissue microarray (catalog No. OD-CT-DgLiv01-0050200) containing 87 pairs of human HCC tissues was purchased from Outdo Biotech. The immunohistochemical staining of the tissue microarray was performed using an anti-CYP17A1 antibody (Proteintech) diluted to 1:200 in accordance with the commercial protocol (Outdo Biotech). The protein signals were evaluated by assessing staining intensity using a BX51 microscope (Olympus) and quantified by Image-pro plus 6.0 software (Media Cybernetics). Clinical–pathological data were examined by two pathologists under double blindness. The weak or strong expressions of CYP17A1 were defined according to the product of staining intensity and percentage of positively stained cells (strong cases had a score of ≥2 while weak cases had a score of <2), as described previously (28).
Sandwich ELISA
Ninety-six-well MaxiSorp plates (Nunc) were coated with 0.5 μg/mL anti-CYP17A1 goat polyclonal antibody (Santa Cruz Biotechnology) in PBS buffer overnight at 4°C. The plates were blocked with BSA Diluent/Blocking Solution (KPL) for 10 minutes at room temperature. Diluted Serum samples (1:100 in BSA Solution) were added and incubated 1 hour at room temperature. After washing the unbound material with Wash Solution (KPL), bound CYP17A1 was detected using anti-CYP17A1 rabbit polyclonal antibody (Proteintech) at a dilution of 1:2,000 followed by incubation with peroxidase-conjugated anti-rabbit IgG solution (KPL) and ABTS Peroxidase Substrate Solution (KPL). To quantify the serum CYP17A1, a calibration curve of purified CYP17A1 protein (Abgent) was performed and used as the standard. The optical density (OD) was measured at 405 nm using a microplate reader (Thermo Fisher Scientific). Each sample was measured three times by quadruplicates.
Statistical analysis
Quantitative data were presented as the mean ± SEM or median values (25th–75th percentiles), and statistical significance was determined by the Student t test or Mann–Whitney U test, where appropriate. Differences between the CYP17A1 expressions in paired liver tissues were analyzed by the Wilcoxon matched pairs test. Associations between CYP17A1 expression and clinicopathological parameters were assessed by the χ2 test. To determine the optimal cutoff values for CYP17A1 and AFP, receiver operating characteristic (ROC) curves were constructed. The areas under the ROC curve (AUC) were calculated and compared as described previously (29). P value of <0.05 was considered statistically significant.
Results
The WHV/c-myc Tg mouse model was used for identifying HCC biomarkers
The c-myc transgene was specifically expressed in the liver and tumor tissues of WHV/c-myc Tg mice, as confirmed by semiquantitative reverse transcription PCR (Fig. 1A). Noticeably, liver cancer developed more rapidly in males than in females of Tg mice (Fig. 1B), as the same phenomenon that was frequently observed in human. The histopathological examination showed that the livers derived from Tg mice at the 5 months exhibited mild to moderate hepatocyte dysplasia. The liver tumors were visualized at 11 months of Tg mice and showed well-differentiated and trabecular types of HCC. In contrast, the liver tissues from C57/BL6 WT mice showed normal characteristics (Fig. 1C, left). Serum levels of ALT, AKP, and AST, which are the biochemistry parameters of liver damage, were strikingly higher in Tg mice at tumor stage (11 months) than those in the control group (P = 0.0045, 0.0072, and 0.035, respectively), while these parameters were not changed in 5 months of Tg mice compared with those in WT controls (Fig. 1C, right). As shown in Fig. 1D, to search biomarkers for HCC, we designed the study. In the process of biomarker screening, we performed the study mainly based on the following criteria:
A good candidate should highly express in the early stage of liver tumor so that it can be used for early detection to offer opportunity to cure or treat HCC. Using the Affymetrix Mouse Genome 430 2.0 Expression Arrays, we investigated the differential gene expression profiles between the livers at preneoplasia (5 months) or with HCC (11 months) derived from Tg mice and the livers from age-matched WT mice. Upregulation or downregulation was defined using the assigned cutoff expression ratio of 2 fold change. There were 306 and 2,663 genes differentially expressed in Tg mice at 5 months (preneoplastic stage) and 11 months (neoplastic stage), respectively (Fig. 2A). Then, the hierarchical clustering was performed to analyze these differentially expressed genes. As indicated in Fig. 2B, pattern 5 represents a group of genes (n = 54) which have higher expression in Tg mice compared with age-matched WT controls at both preneoplastic and neoplastic stages.
A good candidate should cause less pain and improve the noninvasive detection of liver cancer. Hence, we carried out bioinformatics analysis to predict the presence and location of the secretory signal peptide in the amino acid sequences of all candidates in pattern 5 using the SignalP 3.0 server (30). We found 16 signal peptide-triggered secretory proteins that may possibly be identified in HCC sera and be developed as serum biomarkers. Further, real-time qPCR validation was performed, and 11 candidate signatures that are consistently upregulated at both preneoplastic and neoplastic stages were selected (Figs. 2C, 3A, and 3B). Remarkably, one of our candidate, GP73, has already been extensively reported for HCC detection (12–14, 31, 32).
It should be easy and practicable to build a sensitive detection system for a good candidate. Among the selected candidates, CYP17A1 was markedly upregulated, the array signals indicated that expression intensity of CYP17A1 is about 3.5-fold higher in Tg than in WT mice at both 5 and 11 months (Fig. 3A). The increased levels of CYP17A1 mRNA and protein were also validated by real-time qPCR and Western blotting (Fig. 3B). Because its antibody is conveniently available and it is relatively inexpensive to develop a sandwich ELISA to exactly measure CYP17A1 concentration in serum, we decided to choose CYP17A1 for further analysis.
The WHV/c-myc Tg mouse model was used for identifying HCC biomarkers. A, RT-PCR analysis of the expression of c-myc transgene in liver tumors and different tissues of Tg mice. 10d, 10 days; NL, normal liver from WT mice. GAPDH was used as an internal control. B, cumulative incidence of liver tumors in male and female Tg mice. C, histology of liver tissues obtained from Tg and WT mice at the age of 5 and 11 months (left). m, months; scale bar, 100 μm. Right, analysis of serum ALT, AKP, and AST levels in these mice. Statistical significances between the groups were calculated using the Student t test. *, P < 0.05; **, P < 0.01. D, experimental strategy for discovering candidate biomarkers using the mouse model of hepatocarcinogenesis.
The WHV/c-myc Tg mouse model was used for identifying HCC biomarkers. A, RT-PCR analysis of the expression of c-myc transgene in liver tumors and different tissues of Tg mice. 10d, 10 days; NL, normal liver from WT mice. GAPDH was used as an internal control. B, cumulative incidence of liver tumors in male and female Tg mice. C, histology of liver tissues obtained from Tg and WT mice at the age of 5 and 11 months (left). m, months; scale bar, 100 μm. Right, analysis of serum ALT, AKP, and AST levels in these mice. Statistical significances between the groups were calculated using the Student t test. *, P < 0.05; **, P < 0.01. D, experimental strategy for discovering candidate biomarkers using the mouse model of hepatocarcinogenesis.
Searching of candidate biomarkers that are highly expressed at both pre- and neoplastic stages in the livers of Tg mice. A, the number of differentially expressed genes (an increase or decrease by at least 2-fold) in 5m (preneoplastic stage) and 11m (HCC) of Tg mice livers compared with age-matched WT controls. B, hierarchical clustering and expression patterns of these differentially expressed genes. C, real-time qPCR validation of selected candidates which could be developed as novel biomarkers for HCC detection. Statistical significances between the groups were calculated using the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Searching of candidate biomarkers that are highly expressed at both pre- and neoplastic stages in the livers of Tg mice. A, the number of differentially expressed genes (an increase or decrease by at least 2-fold) in 5m (preneoplastic stage) and 11m (HCC) of Tg mice livers compared with age-matched WT controls. B, hierarchical clustering and expression patterns of these differentially expressed genes. C, real-time qPCR validation of selected candidates which could be developed as novel biomarkers for HCC detection. Statistical significances between the groups were calculated using the Student t test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.
Mouse-to-human validation of CYP17A1 mRNA and protein expression. A, the expression intensity of CYP17A1 was determined through microarray analysis. B, the expression of CYP17A1 mRNA and protein in Tg and Wt mice was confirmed by real-time qPCR (left) and Western blotting (right). *, P < 0.05; **, P < 0.01, as evaluated using the Student t test. β-Actin was used as the loading control. C, the relative mRNA levels of CYP17A1 in 33 pairs of human HCC (T) and adjacent non-tumor (N) tissues were analyzed by real-time qPCR. The ratio of CYP17A1 mRNA (T/N) in each pair is indicated by a column. Differences between CYP17A1 mRNA levels in paired liver tissues were analyzed by the Wilcoxon matched pairs test. D, Western blot analysis of CYP17A1 protein expression in 24 pairs of HCC (T) and adjacent non-tumor (N) tissues (top) and quantification of these signals (bottom). An increase by 2-fold (dash line) was chosen as a threshold to calculate the percentage of upregulated samples.
Mouse-to-human validation of CYP17A1 mRNA and protein expression. A, the expression intensity of CYP17A1 was determined through microarray analysis. B, the expression of CYP17A1 mRNA and protein in Tg and Wt mice was confirmed by real-time qPCR (left) and Western blotting (right). *, P < 0.05; **, P < 0.01, as evaluated using the Student t test. β-Actin was used as the loading control. C, the relative mRNA levels of CYP17A1 in 33 pairs of human HCC (T) and adjacent non-tumor (N) tissues were analyzed by real-time qPCR. The ratio of CYP17A1 mRNA (T/N) in each pair is indicated by a column. Differences between CYP17A1 mRNA levels in paired liver tissues were analyzed by the Wilcoxon matched pairs test. D, Western blot analysis of CYP17A1 protein expression in 24 pairs of HCC (T) and adjacent non-tumor (N) tissues (top) and quantification of these signals (bottom). An increase by 2-fold (dash line) was chosen as a threshold to calculate the percentage of upregulated samples.
Expression levels of CYP17A1 are increased in human HCC tissues
To evaluate the expression of CYP17A1 in human HCC tissues, we analyzed CYP17A1 mRNA levels in 33 pairs of HCC specimens by real-time qPCR. Compared with the paired nontumor tissues, CYP17A1 mRNA levels were significantly upregulated in HCC tissues (P = 0.00041). Among the 33 pairs of samples, 23/33 (70%) showed CYP17A1 overexpression (T/N > 2), only 4 cases showed CYP17A1 downregulation (T/N < 0.5), and 6 cases showed no change (0.5 < T/N < 2; Fig. 3C).
Next, we probed the amount of CYP17A1 protein in 24 HCC tissues by Western blotting. The results showed that the expression of CYP17A1 protein was increased in 18 out of 24 (75%) HCC cases, while the others (6/24, 25%) displayed no change or decrease (Fig. 3D). These data demonstrated that CYP17A1 protein is also overexpressed in human HCC tissues.
Immunohistochemistry of CYP17A1 in human HCC tissues
CYP17A1 expression at the protein level was further studied in 87 pairs of human HCC tissues by immunohistochemistry using a tissue microarray. Compared with paired nontumor tissues, the CYP17A1 protein levels were significantly increased in HCC tissues (P = 0.00093). Among the 87 pairs of samples, 58/87 (67%) of the HCC samples showed higher staining scores for CYP17A1 compared with paired nontumor tissues (T>N; Fig. 4A). Representative immunostainings for CYP17A1 from the tissue microarray were shown in Fig. 4B. The detailed information for CYP17A1 expression determined by tissue microarray analysis was shown in Supplementary Table S2 and Supplementary Fig. S1. Notably, a strong staining pattern of CYP17A1 was observed in well-differentiated HCCs (Fig. 4C).
Tissue microarray analysis of CYP17A1 expression. A, the CYP17A1 protein levels were analyzed by immunohistochemical staining of a tissue microarray containing 87 pairs of human HCC (T) and matched non-tumor tissues (N). The ratio of CYP17A1 protein (T/N) in each pair is indicated by a column. B, representative immunostainings showing the expression intensity of CYP17A1 in paired HCC tissues. C, representative immunostainings showing the expression intensity of CYP17A1 in different clinical grades of HCC. Scale bar, 100 μm.
Tissue microarray analysis of CYP17A1 expression. A, the CYP17A1 protein levels were analyzed by immunohistochemical staining of a tissue microarray containing 87 pairs of human HCC (T) and matched non-tumor tissues (N). The ratio of CYP17A1 protein (T/N) in each pair is indicated by a column. B, representative immunostainings showing the expression intensity of CYP17A1 in paired HCC tissues. C, representative immunostainings showing the expression intensity of CYP17A1 in different clinical grades of HCC. Scale bar, 100 μm.
Further, we analyzed the correlations between CYP17A1 expression and clinicopathological features of HCC. The HCC samples were classified into grades I–III by pathologists. Interestingly, we observed that the expression of CYP17A1 protein gradually decreased as the disease progressed from well-differentiated grade I to poorly differentiated grade III (Supplementary Table S3 and Supplementary Fig. S2), demonstrating that CYP17A1 expression is inversely associated with HCC progression (P = 0.036). Taking together, immunohistochemical staining of liver tissue microarray showed that CYP17A1 is overexpressed in HCC tissues, and it could be developed as a histochemical marker for early HCC detection.
Serum levels of CYP17A1 in human
Bioinformatics analysis predicted that CYP17A1 is a signal peptide-triggered classical secretory protein (Supplementary Fig. S3). A sandwich ELISA was then developed and a calibration curve of purified CYP17A1 was made to exactly measure its concentration in serum (Supplementary Fig. S4). The levels of serum CYP17A1 were detected in populations containing 366 healthy individuals, 160 patients with hepatitis B virus (HBV), 147 patients with cirrhosis, 109 patients with lung cancer, 130 patients with prostate cancer, and 776 patients with HCC (Fig. 5A). We found that the median serum CYP17A1 levels were 140.2 ng/mL (range, 77.4–205.9) in HCC cases, 31.4 ng/mL (range, 7.1–58.8) in healthy individuals, 57.5 ng/mL (range, 36.5–81.8) in HBV, 46.1 ng/mL (range, 39.4–51.5) in cirrhosis, 27.4 ng/mL (range, 10.8–35) in lung cancer, and 42.1 ng/mL (range, 31.7–54.1) in prostate cancer. The results demonstrated that serum levels of CYP17A1 were significantly higher in HCC cases (P = 4.85E−99, 1.17E−34, 6.11E−43, 4.01E−48, and 5.16E−39 for HCC versus healthy control, HBV, cirrhosis, lung cancer, and prostate cancer, respectively). In addition, we found that HCC patients who had tumors with diameters ≤2 cm had higher serum CYP17A1 levels (176.8 ng/mL; range, 138.1–214) than those who had tumors with diameters >2 cm (74.5 ng/mL; range, 31.9–124, P = 5.79E–61; Fig. 5B). We also analyzed serum CYP17A1 levels at different clinical stages of HCC according to the TNM staging system. Median serum CYP17A1 levels were 176.5 ng/mL (range, 133.4–246.3) in HCC stage I–II patients and 116.1 ng/mL (range, 49.1–161.4) in HCC stage III–IV patients. The CYP17A1 levels in stages I–II were significantly higher than those in stages III–IV (P = 9.60E−33; Fig. 5C). Furthermore, comparisons of CYP17A1 levels were made in patients with various HCC and non-HCC liver diseases. The results showed that CYP17A1 levels in HCC plus HBV, HCC plus cirrhosis, or HCC plus HBV, and cirrhosis were significantly higher than those in either non-HCC liver disease background, such as HBV, cirrhosis, and HBV plus cirrhosis (all P < 0.001; Fig. 5D).
Analysis of CYP17A1 protein in sera. A, serum CYP17A1 levels were detected by the sandwich ELISA in different patients and healthy individuals. Boxplots of these different groups are displayed, black circles indicate extreme points. ***, P < 0.001 for HCC vs. all other groups. B, association of serum CYP17A1 levels with tumor size. C, association of serum CYP17A1 levels with clinical stages. D, serum CYP17A1 levels in HBV- or cirrhosis-background liver diseases. All statistical analyses were carried out by the Mann–Whitney U test. ***, P < 0.001.
Analysis of CYP17A1 protein in sera. A, serum CYP17A1 levels were detected by the sandwich ELISA in different patients and healthy individuals. Boxplots of these different groups are displayed, black circles indicate extreme points. ***, P < 0.001 for HCC vs. all other groups. B, association of serum CYP17A1 levels with tumor size. C, association of serum CYP17A1 levels with clinical stages. D, serum CYP17A1 levels in HBV- or cirrhosis-background liver diseases. All statistical analyses were carried out by the Mann–Whitney U test. ***, P < 0.001.
Sensitivity and specificity of CYP17A1 and AFP in HCC detection
We next examined the correlations of CYP17A1 and AFP. The Western blot analysis showed that CYP17A1 protein was demonstrably expressed in the sera of AFP-negative (AFP < 20 ng/mL) HCC cases (Fig. 6A). Then, serum CYP17A1 levels in 267 AFP-negative and 509 AFP-positive (AFP ≥ 20 ng/mL) HCC cases were compared. There was no statistical difference between median CYP17A1 levels in AFP-negative (140.3 ng/mL) and AFP-positive (141.5 ng/mL) HCC cases (P = 0.072, Fig. 6B), but both had higher CYP17A1 levels compared with healthy controls (P = 3.65E−70 and 1.35E−95, respectively).
Comparison of the performances of CYP17A1 and AFP in HCC diagnosis. A, Western blot analysis of CYP17A1 expression in the sera of AFP-negative (AFP < 20 ng/mL) HCC cases. Lane 2 is an AFP-positive control to indicate the exact position of protein bands of AFP-negative cases. B, serum CYP17A1 levels in AFP-negative and AFP-positive (AFP ≥ 20 ng/mL) HCC cases. ***, P < 0.001; NS, not significant (P > 0.05), as determined by the Mann–Whitney U test. C, ROC curves comparing the performances of CYP17A1, AFP, and the combination of both in all HCC vs. healthy controls (left) or in early HCC (stages I–II) vs. healthy controls (right). D, comparative analysis of serum CYP17A1 and AFP levels in HCC. Samples 1–238: CYP17A1 elevated, AFP normal range (30.7%); samples 239–267: CYP17A1 and AFP normal range (3.7%); samples 268–340: CYP17A1 normal range, AFP elevated (9.4%); samples 341–776: both CYP17A1 and AFP elevated (56.2%).
Comparison of the performances of CYP17A1 and AFP in HCC diagnosis. A, Western blot analysis of CYP17A1 expression in the sera of AFP-negative (AFP < 20 ng/mL) HCC cases. Lane 2 is an AFP-positive control to indicate the exact position of protein bands of AFP-negative cases. B, serum CYP17A1 levels in AFP-negative and AFP-positive (AFP ≥ 20 ng/mL) HCC cases. ***, P < 0.001; NS, not significant (P > 0.05), as determined by the Mann–Whitney U test. C, ROC curves comparing the performances of CYP17A1, AFP, and the combination of both in all HCC vs. healthy controls (left) or in early HCC (stages I–II) vs. healthy controls (right). D, comparative analysis of serum CYP17A1 and AFP levels in HCC. Samples 1–238: CYP17A1 elevated, AFP normal range (30.7%); samples 239–267: CYP17A1 and AFP normal range (3.7%); samples 268–340: CYP17A1 normal range, AFP elevated (9.4%); samples 341–776: both CYP17A1 and AFP elevated (56.2%).
Further, the ROC curves were plotted to define the optimal cutoff values and to identify the sensitivity and specificity for serum CYP17A1 and AFP in differentiating patients with HCC versus healthy controls (Fig. 6C, left). The AUC for CYP17A1 was 0.91, with a sensitivity of 86.9%, specificity of 76.8%, and an optimal cutoff point of 60.2 ng/mL, while the AUC for AFP was 0.78, with a sensitivity of 65.6%, specificity of 65.6%, and a cutoff value of 20 ng/mL. CYP17A1 had a better AUC compared with AFP (P = 0.0034). The combination of CYP17A1 and AFP achieved the best overall results, with an AUC of 0.94, a sensitivity of 90.1%, and a specificity of 80.3% (Supplementary Table S4). We then evaluated sensitivity and specificity of CYP17A1 in patients with early HCC (TNM stages I–II) versus healthy controls (Fig. 6C, right). The AUC for CYP17A1 was 0.97, with a cutoff of 60.2 ng/mL, leading to a sensitivity of 98.4% and a specificity of 76.8%. The AUC for AFP in the same patients was 0.78, with a cutoff of 20 ng/mL, leading to a sensitivity of 66.6% and a specificity of 65.6%. When we compared the two ROC curves, CYP17A1 was significantly better than AFP (P = 0.0009). The combined measurement of CYP17A1 and AFP increased the AUC to 0.99, with a sensitivity of 99.7% and a specificity of 83.1% (Supplementary Table S5). We also evaluated the performance of CYP17A1 and AFP in patients with HCC versus those non-HCC subjects (patients with HBV or cirrhosis), and the results showed that CYP17A1 as determined by the ROC curves was better than AFP (Supplementary Fig. S5 and Supplementary Tables S6 and S7).
Using the optimal cutoffs derived from the ROC curves, we observed that CYP17A1 was positive in 238 of 267 (89.1%) of HCC patients with normal AFP levels. More specifically, as shown in Fig. 6D, both markers were elevated together in about 56.2% of HCC patients, and only 3.7% of cases were negative for both markers. These results suggest that CYP17A1 has great potentiality as a novel marker for HCC, and combined use of the two markers, CYP17A1 and AFP, can significantly increase the overall sensitivity and specificity for HCC diagnosis.
Discussion
In this study, we used a mouse hepatocarcinogenesis model, WHV/c-myc Tg mice, to screen HCC markers through genome-wide expression profiling analysis. We identified the differentially expressed genes in the livers between Tg and Wt mice to mine insights and indicators associated with progression of liver cancer. Among the overexpressed candidate markers in Tg mice at both preneoplastic and neoplastic stages, CYP17A1 was found as a secretory protein. The overexpression of CYP17A1 was demonstrated in approximately 70% human HCC tissues, and its serum levels were also elevated in the patients with HCC. Our findings prove that the WHV/c-myc Tg mouse model is helpful to the discovery of human HCC biomarkers.
Notably, the immunostaining of CYP17A1 in 87 pairs of HCC tissues revealed that strong stainings of CYP17A1 exist in well-differentiated HCC tissues compared with poorly differentiated HCC (Fig. 4C). Furthermore, higher serum CYP17A1 levels have been seen in HCC patients with small tumors (diameters ≤ 2 cm) and early stages (Fig. 5B and C). These results highlight increased expression of CYP17A1 at early stage of liver tumorigenesis, suggesting a potential role for CYP17A1 to host molecular regulatory program in triggering HCC development and progression. In the 5′-untranslated region of the CYP17A1 gene, a (−34) T to C transversion polymorphism can lead to overproduction of liver steroid biosynthesis (33, 34). It has been reported that this high-activity allele associated with increased circulating levels of estrogens and androgens may affect HCC risk (35). Thus, as a key factor participating in the steroidogenic pathway, CYP17A1 might regulate HCC progression by involving in liver metabolic disorder and affect susceptibility for developing HCC in men and women (36–38). However, the detailed function and mechanism of CYP17A1 in hepatocellular carcinogenesis is still unclear. It is conceivable that the role of CYP17A1 in HCC development will be exciting areas to explore.
AFP has been the most widely utilized serum marker for HCC, but the sensitivity of AFP is not satisfactory, it would arouse suspicion to consider this biomarker acceptable for clinical diagnosis of HCC (39, 40). By searching for articles examining the test characteristics of AFP for detecting HCC, a systematic review presented that using a cutoff value of AFP >20 ng/mL, the sensitivity is 41% to 65%, specificity is 80% to 94% (41).Indeed, in our study, among the 776 HCC sera, AFP levels were within normal range (<20 ng/ml) in 34.4% (n = 267) of the cases, AFP >20 ng/mL were in 65.6% (n = 509) of cases (Fig. 6B and D). By comparing the diagnostic performances of CYP17A1 and AFP in the same patients, we found that the sensitivity and specificity of CYP17A1 were better than AFP in differentiating both overall HCC and early HCC (Supplementary Tables S4–S7). Moreover, serum CYP17A1 levels above the optimal cutoff were positive in 89.1% HCC patients with negative AFP (Fig. 6D), indicating the potential utility of CYP17A1 in AFP-negative HCC cases. Therefore, combined monitoring CYP17A1 could be a valuable complementary tool in the diagnosis of HCC.
The National Cancer Institute's Early Detection Research Network (EDRN) has developed guidelines on phases of evaluating a biomarker for cancer detection. A five-phase criterion has been established (42). Our current work is phase I–II research, the biomarker discovery and evaluation of its ability to detect cancer; therefore, a larger and robust cohort and 3 to 5 phases of validation are still needed in future studies.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design:F. Wang, Y. Dong
Development of methodology: F. Wang, J. Huang, Z. Zhu, X. Ma, Y. Dong
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Wang, J. Huang, Z. Zhu, X. Ma, Y. Dong
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Wang, J. Huang, Z. Zhu, X. Ma, L. Cao, Y. Zhang, W. Chen, Y. Dong
Writing, review, and/or revision of the manuscript: F. Wang, W. Chen, Y. Dong
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Huang, Z. Zhu, X. Ma, L. Cao, Y. Zhang, Y. Dong
Study supervision: W. Chen
Acknowledgments
We are grateful to Dr. Linlin Tian and Dr. Yazhen Zhu for helpful discussions and technical assistance. We thank Prof. Xueshan Xia and Prof. Chris Green for helpful comments.
Grant Support
This work was supported by the grants from project of the young academic and technical leaders reserve talent of Yunnan Province (2015HB024) to Y. Dong and research startup foundation for KMUST talented scholars recruitment (KKSY201626003) to F. Wang.
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.