Two-hundred and eighty-four subjects participated in this study from January 2013 to March 2014, including 74 AFP negative hepatocellular carcinoma patients, 60 AFP positive HCC patients and 70 liver cirrhosis patients, these patients all had chronic hepatitis B infection. There were 40 chronic hepatitis B patients and 40 healthy controls. Serum plasma samples were obtained according to the guidelines given by the HUPO Plasma Proteome Project (18), and 4 ml of peripheral blood was collected from each subject. Diagnoses of chronic hepatitis B, liver cirrhosis and hepatocellular carcinoma were performed according to the Asian Pacific Association for the Study of the Liver (APASL), the European Association for the Study of the Liver (EASL) and the American Association for the Study of Liver Diseases (AASLD) (19–22). This study was approved by the Ethics Committee of Chongqing Medical University. Written informed consent was obtained from all participants before the treatment. Patient demographics and clinicopathological data are summarized in Table I.

Ten randomly chosen individual samples from each group were mixed to create three sample pools (Fig. 1). The most abundant proteins were depleted using an immunodepletion kit (GE Healthcare, Shanghai, China) as per the manufacturer’s instructions. Immunodepleted plasma was subjected to protein concentration assays using a 2-D Quant kit (GE Healthcare). Protein (100 μg) was precipitated from each pooled group, dissolved in dissolution buffer, denatured, cysteine blocked, digested with 2 μg of sequencing grade modified trypsin and labeled using iTRAQ reagents [LC, 113 tag; AFP(−) HCC, 114 tag; and AFP(+) protein, 115 tag] provided by an iTRAQ kit (AB Sciex Analytical Instrument Trading Co., Shanghai, China) (Fig. 1). For the parallel study, the same sample set was labeled with the iTRAQ reagents 116, 117 and 118, respectively (Fig. 1). Peptides from each sample set were mixed prior to subsequent analysis.

Labeled peptides were fractionated by immobilized-pH-gradient isoelectric focusing (IPG-IEF), as previously described (23,24). Briefly, samples were dissolved in a Pharmalyte (GE Healthcare) and urea solution, rehydrated on a pH 3–10 IPG strip, and then subjected to IEF focusing at 68 kV/h with an IPGphor system (GE Healthcare). Peptides were then extracted from the gel using an acetonitrile (ACN) and formic acid solution (25). The fractions were lyophilized, and purified with SPE Discovery DSC-18 columns (Supelco Inc., Bellefonte, PA, USA). The purified peptides were re-lyophilized and stored at −20°C until use.

Purified peptide fractions were reconstituted in solvent A [water/ACN (98:2 v/v) with 0.1% formic acid] and separated using a C18-PepMap column (Thermo Fisher Scientific, Beijing, China) with a solvent gradient of 2–100% Buffer B (0.1% formic acid and 98% acetonitrile) in Buffer A at a flow rate of 0.3 μl/min. The peptides were electrosprayed using a nanoelectrospray ionization source at an ion spray voltage of 2300 eV and analyzed by a NanoLC-ESI-Triple TOF 5600 system (AB Sciex). The mass spectrometer was set in the positive ion mode at a mass range of 300–1800 m/z. The two most intensely charged peptides above 20 counts were selected for MS/MS at a dynamic exclusion of 30 sec (25).

Data were processed by ProteinPilot v2.0 (AB Sciex) and compared with the International Protein Index (IPI) Human database v3.77. Cysteine modified by methane thiosulfate (MMTS) was specified as a fixed modification. Protein identification was based on a threshold of protein score >1.3. For quantitation, at least two unique peptides with 95% confidence and a P-value <0.05 were required.

The Gene Ontology was analyzed by PANTHER (http://www.pantherdb.org/) on biological processes, protein classes and molecular functions. The signaling pathway analysis was performed by using the STRING (http://string-db.org/) program.

The plasma levels of CRP, SAA, AFP and C9 were measured in 284 cases using commercial ELISA kits in accordance with the manufacturer’s instructions. The cut-off value of CRP was determined using the receiver-operator characteristic curve (ROC) curve, which was twice the SD above the average of the control individuals. Human CRP ELISA kit (ab99995) and Complement C9 Human ELISA kit (ab137972) were purchased from Abcam (Cambridge, UK). The human SAA ELISA kit (ELH-SAA-001) was purchased from Ray Biotech (Norcross, GA, USA).

IHC evaluation of C9, SAA and CRP was performed with a commercial tissue microarray (BC03117; Us Biomax Inc., Rockville, MD, USA) containing 48 unique HCC samples and 22 liver cirrhosis tissues. Paraffin-embedded liver sections were deparaffinized, rehydrated and subjected to heat-induced antigen retrieval in 0.01 M sodium citrate buffer for 5 min (26). Next, 3% H₂O₂ was added to quench the activity of endogenous peroxidase for 5 min. After BSA blocking, the sections were incubated overnight with primary antibodies for CRP (1:100), SAA (1:100) and C9 (1:100) (Abcam). The EnVision system with horseradish peroxidase (DakoCytomation, Glostrup, Denmark) was used for IHC visualization (26). Gill’s hematoxylin was used to counterstain slides according to methods previously described (25,26). Monoclonal antibodies against human CRP (ab32412), serum amyloid A (SAA, EPR4134), matrix metalloproteinase 2 (MMP2, EPR1184), matrix metalloproteinase 9 (MMP9, EP1254), signal transducer and activator of transcription 3 (STAT3, E121-21), phosphorylated STAT3 (pY705-STAT3, EP2147Y) and β-actin (EP1123Y) were purchased from Abcam. Horseradish peroxidase (HRP)-conjugated secondary antibodies were obtained from Santa Cruz Biotechnology (Dallas, TX, USA).

The stable HBV-transfected cell line HepG2.2.15, the human HCC cell line HepG2 (ATCC, Manassas, VA, USA) and the BEL7402 cell line (Cell Bank of the Chinese Academy of Medical Science, Beijing, China) were cultured in high-glucose DMEM that was supplemented with 100 μg/ml streptomycin, 0.1% non-essential amino acids, 100 IU/ml penicillin, 1.0 mM sodium pyruvate, 2 mM glutamine and 10% FBS at 5% CO₂ and 37°C (27).

Cell lines were transfected with 100 nm of CRP-specific siRNA (HSS175221, HSS102299 and HSS102300) or a negative control plasmid (12935-400) using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). Cell viability was determined with a trypan blue exclusion assay and only cells that had over 95% live cells by the trypan blue exclusion assay were used for subsequent assays.

For wound healing assays, cells were cultured in 6-well plates until they reached 100% confluence. A 200-μl pipet tip was used to scratch the cell monolayer, followed by washes with the growth media to remove debris. The resultant gap was monitored for up to 24 h via a microscope.

Invasion assays were completed by a Cell Invasion assay kit (Cell Biolabs, Inc., Beijing, China). As determined by the trypan blue exclusion, ~1×10⁵ viable and siRNA-transfected cells were seeded onto the upper chamber of a 24-well plate with polycarbonate membrane inserts, and the number of cells that invaded through the ECM Matrix gel was determined 24 h after seeding by CyQuant GR fluorescent dye (560 nm).

Cells were lysed with RIPA buffer, and a 2-D Quantification kit (GE Healthcare) was used to determine the protein concentration. Protein samples were electrophoretically separated by SDS-PAGE and then transferred onto PVDF membranes. Membranes were blocked with BSA in Tris-buffered saline solution with Tween-20 (TBS-T), overnight at 4°C and then incubated with the primary antibodies (1:500–1:1,000 dilution) for 3 h at room temperature. Then membranes were incubated with HRP-conjugated secondary antibodies at a dilution of 1:5,000 after three wahses with TBST buffer. Finally, membranes were visualized with the ChemiDoc MP imaging system (Bio-Rad Laboratories, Hercules, CA, USA).

Supernatant HBV-DNA was quantified by RT-PCR using a commercial HBV detection kit (Fosun Diagnostics, Shanghai, China) on a Roche LightCycler instrument (Roche Molecular Systems, Alameda, CA, USA). Elecsys HBsAg II and HBeAg quantitative assay kits were used to detect HBsAg and HBeAg titers, respectively, using the Roche Cobas e601 electrochemical luminescence analyzer (Roche Diagnostics GmbH, Mannheim, Germany) (28). To measure the expression levels of downstream IFN-stimulated genes and type I IFN in transfected cells, we used gene-specific primers for GAPDH (Hs02758991_g1), OASL (Hs0 0984390_m1), Mx1 (Hs0 0895608_m1), ISG15 (Hs01921425_s1), OAS2 (Hs0 0942643_m1), EIF-2α (Hs00230684_m1), IFNβ1 (Hs01077958_s1), OAS1 (Hs00973637_m1), PKR (Hs00169345_m1), IFNα1 (Hs00855471_g1) and OAS3 (Hs00196324_m1) (Life Technologies). The 2^−ΔΔCT method (29) was used to analyze the relative changes in gene expression. All experiments were performed in triplicate.

SPSS software v13.0 (SPSS Inc., Chicago, IL, USA) was used to perform statistical analysis. Quantitative variables are presented as the mean and standard deviation (±SD). Comparisons between groups were analyzed by the Student’s t-test or a Mann-Whitney U test. Qualitative variables are presented as counts and percentages, which were analyzed with the χ² test. ROC curve analysis of CRP was performed to determine the diagnostic accuracy of CRP expression levels and 2×2 tables were used to evaluate sensitivity and specificity. Correlations between CRP and HBV DNA were determined using a Spearman’s rank correlation analysis. P<0.05 was considered significant.

We used iTRAQ-based MS to analyze serum proteins from the AFP(−) HCC, AFP(+) HCC and LC groups (Fig. 1). We confidently identified and quantified 510 proteins. The top 30 upregulated and the top 30 downregulated proteins are shown in Table II. We further defined the DEPs using a ±1.3-fold cut-off in accordance with commonly adopted iTRAQ-based MS conventions (30,31). Use of this cut-off is based on the assumption that the estimated overall data variation from duplicate experiments is ≤30%. Gene Ontology analysis with PANTHER suggested that the majority of the DEPs were enzymes or signaling molecules, followed by cell development regulators and immune-related proteins (Fig. 2A–C). Using STRING analysis, we identified CRP as the most important node in the DEP network because it had the greatest connectivity (Fig. 2D).

To determine the reliability of the iTRAQ analysis data, we selected samples from the same sample set we analyzed in the iTRAQ experiments and employed ELISA assays to test the plasma levels of several of the most upregulated proteins, including CRP, SAA and C9. We examined 90 plasma samples from 30 LC, 30 AFP(−) HCC and 30 AFP(+) HCC individuals (Fig. 3A). The ELISA measured CRP, SAA and C9 levels were consistent with the iTRAQ results, as the plasma levels of all three DEPs were significantly higher in HCC subjects than in LC subjects (P<0.05, Fig. 3A).

We further tested the serum levels of CRP, C9 and SAA in all 284 plasma samples by ELISA (Fig. 3B). We found that average CRP concentrations in the AFP(−) HCC (3932±277 ng/ml) and AFP(+) HCC (4860±384.3 ng/ml) groups were significantly higher than the CRP concentrations in the LC (2637±282.4 ng/ml, P<0.001), CHB (1810±177.9 ng/ml, P<0.0001) and healthy control (424.9±23.95 ng/ml, P<0.0001) groups (Fig. 3B). Similar trends were also observed in the SAA and C9 assays (Fig. 3B).

IHC analyses of tissue microarrays coincided with the results of the serum tests, indicating that CRP, SAA and C9 were significantly overexpressed in tumor tissues when compared to cirrhotic tissues (P<0.05) (Fig. 3C).

Our analysis indicated that CRP was the most important node in the DEP analysis, and IHC analysis confirmed its overexpression in clinical samples. Therefore, we performed ROC analyses to address whether HCC diagnostic tests could be improved by using CRP expression or combined CRP and AFP expression as biomarkers for HCC diagnosis. According to our ROC results, the cut-off value to confirm a positive diagnosis of HCC was 2361.82 ng/ml for CRP. The sensitivity and specificity for AFP(−) HCC subjects, AFP(+) HCC subjects, all subjects combined were 72.97 and 60%; 70 and 75.71 and 82.09 and 73.33%, respectively (Fig. 4A–C). The positive predictive values and negative predictive values for all subjects combined, AFP(−) HCC subjects and AFP(+) HCC subjects were 73.3 and 82.1%; 65.9 and 67.7%; and 71.2 and 74.6%, respectively. The plasma CRP levels of both the AFP(−) HCC group and all HCC patients combined had significantly higher predictive accuracy than the plasma CRP levels of the LC group. The area under ROC curve (AUC) of CRP in all HCC patients and AFP(−) HCC patients was 0.8642 (95% CI, 0.819–0.902, P<0.01), and 0.724 (95% CI, 0.643–0.795, P=0.0001), respectively (Fig. 4A and C). When combining CRP with AFP, the AUC was 0.873 (95% CI, 0.829–0.910, P<0.01), which was statistically higher than the AUC of AFP alone (AUC=0.812, 95% CI, 0.763–0.871, P<0.05) (Fig. 4D). It is important to emphasize that 65.9% of the patients with AFP(−) serum in the present study had elevated CRP.

To study the role of CRP in tumor cell motility, we silenced CRP expression in HCC cell lines with CRP-specific siRNA. According to western blot analysis, we found that each of the CRP-specific siRNAs significantly downregulated CRP expression in both HepG2 and BEL7402 cells (Fig. 5A). The assay showed that CRP-specific siRNAs significantly increased the invasion ability of BEL7402 cells and HepG2 by 47–52 and 40–49%, respectively (P<0.05) (Fig. 5B). Additionally, we found that CRP-specific siRNA transfection of BEL7402 cells and HepG2 resulted in a 50–60 and 40–55% increase in wound healing, respectively (P<0.05) (Fig. 5B).

To examine the functional consequences of aberrant CRP expression, we analyzed the expression levels of the STAT3, pY705-STAT3, MMP2 and MMP9 proteins. These proteins are all known to participate in the pathogenesis of tumor metastasis (32,33). We found that CRP-siRNA transfection resulted in significantly increased expression of pY705-STAT3, MMP2 and MMP9 when compared to control siRNA in both HepG2 and BEL7402 cells (Fig. 6).

Considering the prevalence of HBV infection in HCC patients, we investigated whether CRP expression level is associated with HBV viral load. We observed that CRP plasma levels were positively correlated with HBV-DNA copy number (Pearson r=0.571, P<0.05) (Fig. 7A). Additionally, CRP expression levels were higher in the plasma of HBV-DNA positive patients than in HBV-DNA negative individuals (P<0.05) (Fig. 7A). To assess the effect of CRP expression level on HBV replication, we transfected HepG2.2.15 cells, which stably produce HBV-DNA, with CRP specific siRNA. CRP specific siRNA transfection resulted in successful knockdown of both intracellular and extracellular CRP (Fig. 7B). We found that the HBV-DNA titer decreased nearly 2-fold in the CRP-knockdown groups when compared the control group (Fig. 7C). Furthermore, we determined that CRP knockdown significantly suppressed the transcription of type I IFNs, which are known anti-viral cytokines, and resulted in significant inhibition of genes downstream of IFN, including OAS1, OAS2, OAS3, RNase L and ISG15 (P<0.05) (Fig. 7D).