The Expression of FAP in Hepatocellular Carcinoma Cells is Induced by Hypoxia and Correlates with Poor Clinical Outcomes

Patients and specimens

Paraffin-embedded specimens were obtained from a cohort of 138 HCC patients after surgery. Another 11 fresh tissue samples were also collected immediately after surgical resections, then frozen and stored at -80 °C, for RNA and protein isolation. All patients were pathologically confirmed HCC and received their primary surgical resections between the year 2010 and 2015, in the 5th affiliated hospital, Sun Yat-sen University, Guangdong Province, People's Republic of China. Follow-up were performed by four independent physicians unaware of the study. The follow-up endpoint was assessed as the time to one of the defining events: (i) disease recurrence, (ii) disease-related death. The median follow-up for the entire cohort was 33.2 months (95%CI, 26.2-40 months, range 1.1-80.2 months). The TNM staging were determined according to the 2010 American Joint Committee on Cancer (AJCC) staging system. The classification of early and advanced tumors was done as previous studies ^{12, 13}. Overall survival (OS) was defined as the interval between surgical resection and the date of death or last follow-up. The time-to-recurrence (TTR) was defined as “from the date of surgery until tumor recurrence, death or last follow-up”. All samples were obtained with informed consent, and the study was approved by the Institutional Ethical Review Board of the 5th affiliated hospital, Sun Yat-sen University.

Immunohistochemistry staining (IHC)

IHC staining determination of HIF-1α and FAP expression was performed as previous report ¹⁴. Before IHC staining, the tissue sections were HE-stained and reviewed by two experienced pathologists. HIF-1α was detected using a rabbit monoclonal antibody against HIF-1α (Boster, PB0245, China), while FAP was detected using a rabbit monoclonal antibody against FAP (Boster, BM5121, China). Briefly, paraffin-embedded specimens' cylinders were sectioned into 4μm-thickness and baked at 70°C for 60minutes, and the paraffin sections of the HCC specimens were de-paraffinized in xylene and immediately rehydrated via gradient alcohol. Then the sections were bathed in the blocking buffer (3%hydrogen peroxide) for endogenous peroxidase for 15minutes. The antigenic unmasking was done by citric acid under high temperature and high pressure, and cooling was made naturally. Then, non-specific binding sites were blocked with the goat serum. The tissue sections were incubated with anti- HIF-1α antibody at a dilution of 1:200 and anti-FAP antibody at a dilution of 1:400 for 12 hours at 4°C. Subsequently, anti-rabbit IgG-HRP-linked secondary antibody was applied for 30 minutes at 37°C. Then, the sections were developed with 3-diaminobenzidinetetrahydrochloride (DAB). Mayer's hematoxylin was applied for 5 minutes as counterstain. Negative control included in all assays: slides incubated with secondary antibody only (Figure S1). Positive control used in this study is the mesenchymal fibroblasts stained positive for FAP (Figure 1E), we also refer to the positive results in various cancer cells including HCC from previous studies ^{5, 15, 16}. Photographs of representative fields were captured under ZEISS N-ACHROPLAN microscope (ZEISS, Germany). The integrated optical density (IOD) measurement was performed as the method previously reported ^{17, 18}. The IODs of the taken pictures were measured and analyzed in the Image-Pro Plus v6.0 software (Media Cybernetics Inc, Bethesda, MD, USA). The process is shown in the supplementary materials (Figure S2). In brief, three parameters were obtained (area sum, mean density, and IOD). The mean IOD and mean area were calculated from all taken photographs. Subsequently, the mean IOD was divided by the mean area as the relative IOD of specimen. The relative IOD was used for the analysis in this study.

X-tile determinaton of low/high expression

X-tile analysis was used for assessment of HIF-1α and FAP expression which was represented by optimal cut-off points of the relative IODs based on patients' outcome ¹⁹. The cut-off points were made to determine the low and high expression of HIF-1a and FAP. Statistical significance was assessed as the cut-off score derived from the 138 cases by a standard log-rank method, with P values.

Cell lines and cell culture agents

The HCC cell lines (Huh7, Bel-7402, and HepG2) were cultured in DMEM medium (Gibco, USA), supplemented with 10 % fetal bovine serum, 50 U/mol penicillin and 50 mg/ml streptomycin (Sigma-Aldrich, USA) and maintained in a humidified incubator at 37 °C with 5 % CO2. All cell lines were obtained from Cancer Center, Sun Yat-sen University, Guangzhou, Guangdong, People's Republic of China, and their background information has been described ^20-22. Cellular hypoxia was performed by use of Cobalt Chloride (CoCl2, Sigma-Aldrich, USA). A concentration of 100 μM CoCl2 was used for hypoxia induction, meanwhile PBS was used as control.

Protein extraction and western blot

Total protein from tissue specimens and cells was extracted by sample buffer (62.5 mmol/L Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, and 5% 2-β-mercaptoethanol, Sigma-Aldrich, USA). The concentrations of protein were then determined using the BCA protein assay kit (Beyotime, China). Western blot analysis was performed as described ¹⁴. Briefly, after being separated in SDS-PAGE gels, proteins were transferred to nitrocellulose membranes (Bio-Rad, USA). The membranes were then incubated with different primary antibodies at 4°C, overnight. Afterwards, respective secondary antibodies were applied at room temperature for 1hour. ECL development solution was manipulated for the visualization of the expression of different protein. The antibodies used are listed as follow: anti-HIF-1α (dilution at 1:1000, A11945, Abclonal, China), anti-FAP (dilution at 1:800, BM5121, Boster, China), anti-E-cadherin (dilution at 1:1000, 3195, CST, USA), anti-Twist (dilution at 1:1000, ab175430, Abcam, England), anti-Snail (dilution at 1:1000, ab53519, Abcam, England).

RNA isolation and quantitative RT‑PCR

Total RNA was extracted using RNAiso Plus (Takara Bio, China) following the manufacturer's instructions. The cDNA was synthesized from total RNA by TaKaRa One Step RNA PCR Kit (AMV) (Takara Bio, China). Relative expressions of different genes were detected by the PowerUp SYBR Master Mix Applied Biosystems (Invitrogen, USA). Each PCR reaction was run in at least 3 independent experiments using the Eppendorf Realplex system (Eppendorf AG, Germany). The comparative Ct method (ΔCt) was applied to calculate the relative expression levels, which was then normalized by the expression of β-actin. The primer sequences were either designed on the NCBI website (http://www.ncbi.nlm.nih.gov/tools/primer-blast) and the PrimerBank website (http://pga.mgh.harvard.edu/primerbank/). The primers used are listed as follows:

FAP: forward 5'ATGAGCTTCCTCGTCCAATTCA3'; reverse 5'AGACCACCAGAGAGCATATTTTG3'; HIF-1-α: forward 5'GAACGTCGAAAAGAAAAGTCTCG3'; reverse 5'CCTTATCAAGATGCGAACTCACA3'; β-actin: forward 5'CATGTACGTTGCTATCCAGGC3'; reverse 5'CTCCTTAATGTCACGCACGAT3'.

Statistical analysis

Statistical analyses were performed with the statistical software package SPSS16.0 (SPSS, Inc., Chicago, IL. USA) and the GraphPad Prism version 6.00 for Windows (GraphPad Software, La Jolla California USA). Optimal cutoff point for survival analyses of HIF-1α and FAP expression was determined by the X-tile software version 3.6.1 (Yale University School of Medicine, New Haven, CT, USA). The results were presented as the mean ± standard deviation (SD). Pearson correlation test was applied between the expression of HIF-1a and FAP in IHC. Kaplan-Meier Survival analysis and the log-rank test were used to plot the disease-free survival curve in HCC patients. COX proportional hazards regression, univariate survival analysis and correlations between different factors were carried out in the statistical software package SPSS16.0 (SPSS, Inc., Chicago, IL. USA). It was considered statistically significant when the p value < 0.05.

The correlation between HIF-1α and FAP expressions in HCC tissues

The staining by IHC revealed that FAP in HCC patient tissues were common in both mesenchymal cells and tumor cells. The representative low and high expression pictures of HIF-1α and FAP were shown. Specifically FAP expression within tumor cells was associated with tumoral hypoxia, which was indicated by the expression of HIF-1α (Figure 1). The relative IODs of HIF-1α and FAP from the 138-patient cohort were determined as the expression within cancer cells (examples shown as figure 1A-D, but not the stroma of the tissues (example shown as figure 1E). As in Figure 1F the expression levels of HIF-1α and FAP were significantly correlated (Pearson r² = 0.2753, p < 0.0001).

Fresh frozen specimens from 11 patients were also analyzed to further investigate the expression of HIF-1α and FAP in western blot and Real-time Quantitative PCR (q-RT-PCR). HIF-1α and FAP were up-regulated in advance-staged HCC tumors, compared to early lesions, which indicated their roles in progression of HCC. Figure 2A and 2B showed that expression of HIF-1α and FAP increased significantly in staged II-IV patients (HIF-1α, p=0.0120; FAP, p=0.0011). The mRNA levels were also significant in accordance (HIF-1α, p=0.0152; FAP, p=0.0307).

Correlation of HIF-1α and FAP expression with clinicopathological features of HCC patients

The cut-off points determining HIF-1α and FAP expression level (low and high expression) in the 138 HCC patients cohort were based on the relative IODs and clinical prognosis (OS and TTR), which were defined in the X-tile program when the p values were minimum with optimal standard log-rank method plotted for both OS and TTR. Accordingly, the cohort was divided into low HIF-1α expression group (65 cases) and high HIF-1α expression group (73 cases); meanwhile, low FAP expression group (74 cases) and high FAP expression group (64 cases).

Based on such results, we further explored the association between the different clinicopathological factors and the expression level of HIF-1α and FAP in the 138 HCC patients' cohort. The results showed that HIF-1α expression were associated with serum AFP (p=0.0004), TNM (p=0.0001), tumor size (<0.0001), vascular invasion (<0.0001), and did not significantly associated with other factors (Table 1). Meanwhile, FAP expression was associated with serum AFP (p=0.003), TNM (p<0.0001), tumor size (p=0.0321), vascular invasion (<0.0001) (Table 1).

HIF-1α and FAP expression for HCC patient's clinical outcomes analysis

Kaplan-Meier analysis based on HIF-1α and FAP expression levels were then performed to plot the OS and TTR curves. Results revealed that mean OS interval in high HIF-1α expressing HCC patients was 36.5 months, compared with 70.7 months in those expressing low levels of HIF-1α (p < 0.0001, log-rank test; Figure 3A). The mean TTR interval in high HIF-1α expressing HCC patients was 25.2 months, compared with those expressing low levels of HIF-1α was 56.5 months (p < 0.0001, log-rank test; Figure 3 B). For FAP, the mean OS interval in high FAP expressing HCC patients was 34.8 months, compared with 71.1 months for those expressing low levels of FAP (p < 0.0001, log-rank test; Figure 3C). The mean TTR for HCC patients expressing high levels of FAP was 21.6 months, compared with those expressing low levels of FAP was 58.6 months (p < 0.0001, log-rank test; Figure 3D).

Clinical outcome analysis based on the combined HIF-1α and FAP expression

We sub-classified the 138 HCC patients into 3 groups according to their expression of HIF-1α and FAP. The 3 groups of patients were defined as follows, 1) patients with low HIF-1α expression and low FAP expression; 2) patients with low HIF-1α expression and high FAP expression, or high HIF-1α and low FAP expression; 3) patients with high expressions for both HIF-1α and FAP. Then we performed Kaplan-Meier survival analysis to investigate the correlation of such combination sub-groups and patients' clinical outcomes.

As depicted in figure 4, the patient sub-group with the low HIF-1α/low FAP expression (group 1) experienced the longest OS (mean 74.3 months) and TTR (mean 64.2 months). On the contrary, the high HIF-1α/high FAP expression sub-group (group 3) had the shortest OS (mean 28.9 months) and TTR (mean 18.2 months). Meanwhile, patients in the sub-group in-between (low HIF-1αexpression and high FAP expression, or high HIF-1α and low FAP expression) had moderate OS (mean 53.9 months) and TTR (mean 35.4 months).

Association of HIF-1α and FAP expression with patient survival

The survival analyses were then performed in the SPSS16 software. Univariate analysis revealed that liver cirrhosis, TNM stage, tumor size, tumor number, vascular invasion, serum AFP, HIF-1α, FAP, and combined HIF-1α/ FAP was associated significantly with OS and TTR, respectively (p value for OS were <0.0001, <0.0001, 0.0184, <0.0001, <0.0001, 0.0002, <0.0001, <0.0001, and <0.0001; p value for TTR were <0.0001, <0.0001, 0.01047, 0.00315, <0.0001, 0.01262, <0.0001, <0.0001 and <0.0001). Lymph node metastasis was associated significantly with post-surgical TTR (p=0.04963) (Table 2). By incorporating factors that indicated significance in the univariate analysis, multivariate analysis was performed through the Cox multivariate proportional hazard regression model for analysis. In brief, liver cirrhosis, tumor number, HIF-1α/FAP combination were the independent prognostic factors for OS. In addition, liver cirrhosis, tumor size, vascular invasion, and HIF-1α/FAP combination were the independent prognostic predictors of TTR (Table 2).

FAP was up-regulated in HCC cells exposed to hypoxia and correlated with epithelial-mesenchymal transition (EMT)

Three HCC cell lines (HepG2, Huh7 and MHCC97H) were treated with CoCl₂-induced hypoxia, as compared with the normal oxygen group. Western blot was adopted to confirm the change of different protein expression levels. The up-regulation of HIF-1α indicated cellular hypoxia after the treatment. FAP increased in parallel with that of HIF-1α and three EMT markers (E-cadherin, Snail and TWIST) (Figure 5).