The study was approved by the Research Ethics Committee of 302nd Hospital of PLA, Beijing, China. Informed consent was obtained from all of the patients. All specimens were handled and made anonymous according to the ethical and legal standards.

A total of 130 patients with primary HCC who underwent a curative liver resection at the 302nd Hospital of PLA, Beijing, China, were included in this retrospective study. Tissues used in the study were retrieved from the tissue bank of the Department of Pathology in the 302nd Hospital of PLA. These patients were diagnosed as HCC between 2001 and 2006. None of the patients recruited in this study had chemotherapy or radiotherapy before the surgery. HCC diagnosis was based on World Health Organization (WHO) criteria. Tumor differentiation was defined according to the Edmondson grading system. Liver function was assessed using the Child-Pugh scoring system. Tumor staging was determined according to the sixth edition of the tumor-node-metastasis (TNM) classification of the International Union against Cancer. The clinicopathological features of 130 patients are summarized in Table I. In addition, 30 self-pairs of HCC specimens (5 TNM stage I, 8 TNM stage II, 12 TNM stage III, and 5 TNM stage IV) and adjacent nonneoplastic liver tissues were snap-frozen in liquid nitrogen and stored at −80°C following surgery for real-time quantitative RT-PCR assay and western blot analysis.

The median follow-up period was 8.6 years. Postoperative surveillance included routine clinical and laboratory examinations every third month, computed tomography scans of the abdomen, and radiographs of the chest every third month. After 5 years, the examination interval was extended to 12 months.

Immunohistochemical staining was carried out following the protocol of our previous study 17 18 19 . The primary antibody against SOX9: rabbit polyclonal antibody (Santa Cruz Biotechnology, Inc. USA), dilution 1:50. The specificity of the primary antibody has been validated by the previous studies of Müller et al. 20 and Lü et al. 21 . Secondary antibody for the detection of primary antibody: anti-rabbit IgG (Sigma, St. Louis, MO, USA). The negative controls were processed in a similar manner with PBS instead of primary antibody. Further, positive SOX9 expression confirmed by western blotting was used as positive controls for immunostaining.

Following a hematoxylin counterstaining, immunostaining was scored by two independent experienced pathologists, who were blinded to the clinicopathological parameters and clinical outcomes of the patients. The scores of the two pathologists were compared and any discrepant scores were trained through re-examining the stainings by both pathologists to achieve a consensus score. The number of positive-staining cells showing immunoreactivity in the nucleus for SOX9 in ten representative microscopic fields was counted and the percentage of positive cells was calculated. The percentage scoring of immunoreactive tumor cells was as follows: 0 (0%), 1 (1-10%), 2 (11-50%) and 3 (≫50%). The staining intensity was visually scored and stratified as follows: 0 (negative), 1 (weak), 2 (moderate) and 3 (strong). A final score was obtained for each case by multiplying the percentage and the intensity score. Therefore, tumors with a multiplied score exceeding 5 (median of total scores for SOX9) were deemed to be low expressions of SOX9; all other scores were considered to be high expressions of SOX9.

The Western blot protocol and semiquantitative analysis were carried out following the protocol of Xu et al 22 . SOX9 antibody (rabbit polyclonal antibody, dilution 1:50, Santa Cruz Biotechnology, Inc. USA) was used, and GAPDH antibody (CW0266, dilution 1:1,000, CoWin Biotech) was used as internal control.

To measure the mRNA expression levels of SOX9, total RNA was extracted from frozen liver tissues using TriZol reagent (Invitrogen) following the manufacturer's instructions. Two micrograms of total RNA was subjected to reverse transcription to synthesize cDNA using the ProtoScript M-MuLV Taq RT-PCR Kit (New England Biolabs), according to the manufacture's instruction, followed by real-time PCR using the TransStart Green qPCR SuperMix (TransGen Biotech). The primer sequences of SOX9 were forward primer, 5'-CGA ACG CAC ATC AAG ACG A-3', reverse primer, 5'-AGG TGA AGG TGG AGT AGA GGC-3'. The transcription of GAPDH was used as an internal control for normalization. SOX9 expression levels were calculated relative to GAPDH using the delta-delta computed tomography method 23 .

The software of SPSS version13.0 for Windows (SPSS Inc, IL, USA) and SAS 9.1 (SAS Institute, Cary, NC) was used for statistical analysis. Fisher's exact test and the X² test were performed to assess associations between SOX9 expression and clinicopathological parameters. The Kaplan-Meier method was used for survival analysis, and differences in survival were estimated using the log-rank test. A multivariate survival analysis was performed for all parameters that were significant in the univariate analyses using the Cox regression model. Differences were considered statistically significant when P was less than 0.05.

Immunohistochemical analysis revealed that SOX9 staining was mainly localized in the nucleus of HCC cells (Figure 1a). SOX9 expression was absent or sporadic in adjacent nonneoplastic liver tissues (Figure 1b). In addition, we found 98 (75.38%) of 130 HCC tissues with high SOX9 expression and 32 (24.62%) of 130 HCC tissues with low SOX9 expression, while 6 (4.62%) of 130 adjacent nonneoplastic liver tissues with high SOX9 expression and 124 (95.38%) of 130 adjacent nonneoplastic liver tissues with low SOX9 expression. Thus, the SOX9 immunostainings in HCC tissues were significantly higher than those in the adjacent nonneoplastic liver tissues (P ≪ 0.01).

To confirm SOX9 protein expression by an independent method, Western blot analysis was performed using 30 self-pairs of HCC and adjacent nonneoplastic liver tissues. The distinct overexpression of SOX9 protein in HCC tissues compared with adjacent nonneoplastic liver tissues was also detected (P ≪ 0.01, Figure 2a and b), as well as significantly increased mRNA level by quantitative RT-PCR (P ≪ 0.01, Figure 2c). The expression levels of SOX9 protein and mRNA in HCC tissues with high stage (III-IV) were both significantly stronger than those with low stage (I-II; for protein and mRNA: both P = 0.02; Figure 2b and c).

To evaluate whether SOX9 protein expression was associated with clinicopathological features of patients with HCC, we correlated immunohistochemical SOX9 staining results with T stage, tumor grade, presence of cirrhosis, underlying liver disease including alcohol abuse, viral hepatitis B and C, sex, and age (Table 1). As the results, we found that more SOX9 positive cells in the higher tumor stage (T3 ~ 4) and tumor grade (G3) than in the lower tumor stage (T1 ~ 2, P = 0.03) and tumor grade (G1 ~ 2, P = 0.01), respectively.

Five-year disease-free survival was observed in 30 (23.08%) patients, whereas in 100 (76.92%) patients, disease recurred, and 88 (67.69%) even died during a 5-year follow-up period. We observed a trend that 5-year disease-free survival in the group with high SOX9 expression was significantly poorer than that in the group with low SOX9 expression (P ≪ 0.01, log-rank test; Figure 3a). Additionally, the Kaplan-Meier plot of 5-year overall survival curves stratified by SOX9 expression was shown in Figure 3b. A significant relationship was found between SOX9 expression and 5-year overall survival (P ≪ 0.01, log-rank test, Figure 3). Futhermore, in a multivariate Cox model, including tumor size, tumor stage, tumor grading, presence of cirrhosis, gender, age, and SOX9 staining, we found that SOX9 expression was an independent poor prognostic factor for both 5-year disease-free survival (hazards ratio [HR] = 2.621, 95% confidence interval[CI] = 1.548-5.829, P = 0.01, Table 2) and 5-year overall survival (HR = 3.825, CI = 1.638-7.612, P = 0.003, Table 2) in HCC.