The authors have declared that no competing interests exist.
SOX9 has been previously shown to be involved in hepatocellular carcinoma (HCC) and other types of cancer. However, prognostic studies so far involved rather small cohorts or lack external validation and experimental data. In this study, we firstly determined the histological expression pattern of SOX9 in human HCC by immunohistochemistry (n = 84) and evaluated its prognostic value. External cohorts of publicly available datasets were used to validate its prognostic relevance in HCC (n = 359) and other types of cancer including breast (n = 3951), ovarian (n = 1306), lung (n = 1926) and gastric cancer (n = 876). Functional SOX9 knock-down studies using siRNA and cancer stem cell models were generated in a panel of liver and breast cancer cell lines. High level of SOX9 was associated with poor survival even after adjustment for other prognostic factors in multivariate analysis (HR = 2.103, 95%CI = 1.064 to 4.156,
Hepatocellular carcinoma (HCC) is still a major global problem in health care systems regarding that HCC is the third leading cause of cancer-related deaths worldwide.[
Therefore, there is a great medical need for novel prognostic factors to adequately identify and classify HCC patients who are at higher risk for shortened survival. Further, this might be helpful in better patient stratification and thus monitoring for disease progression.[
The transcription protein SRY-box 9 (SOX9) is a member of the high-mobility-group box class DNA-binding proteins and centrally involved in human development processes.[
Our retrospective HCC screening cohort consists of eighty-two patients with primary, histologically confirmed HCC who underwent a curative liver resection at the Medical University of Graz, Austria. Tissues used in the study were retrieved from the Institute of Pathology, Medical University of Graz, Austria. We included consecutive patients diagnosed with HCC between January 1988 and December 2009, where sufficient amount of tissue for immunohistochemical analysis was available. None of the patients received neoadjuvant therapy or preoperative local treatment and all underwent tumour resection. Postoperative surveillance was performed at the Division of Gastroenterology or Division of Oncology, Medical University of Graz, Austria, including routine clinical and laboratory examinations every third month, computed tomography scans of the abdomen, and radiographs of the chest every third month. After five years, the examination interval was extended to 12 months. The 7th edition of the American Joint Committee on Cancer (AJCC) / International Union Against Cancer (UICC) TNM system was used to classify the patients.[
The second HCC validation cohort consisted of 359 cases of the cancer genome atlas (TCGA) and data have been retrieved by the publicly available
For SOX9 immunohistochemistry, deparaffinised tissue sections were pre-treated in sodium citrate buffer (0.01 mM) for 40 min in a pressure cooker, followed by 20 min at room temperature. Slides were briefly washed with PBS. Blocking was done with 3% H2O2 in 90% Methanol for 20 min, followed by brief washing with PBS. Additionally, the Lab Vision™ Ultra V Block (Thermo Fisher, Waltham, MA) was applied and incubated at room temperature for 5 min followed by brief washing with PBS. Slides were incubated overnight at 4°C with polyclonal rabbit antibody to human SOX9 (ab76997, Abcam, Cambridge, MA) diluted 1:100 in DAKO Antibody Diluent S2022 (DAKO, Glostrup, Denmark), followed by another brief washing step with PBS. Detection of binding of SOX9 antibodies was performed with Lab Vision™ UltraVision™ LP Detection System (Thermo Fisher) and with Chromogene Substrate AEC (Dako) for 6 min. After immunostaining, sections were counterstained with hematoxylin and mounted with Aquatex (Merck, Darmstadt, Germany). Primary antibodies were omitted and replaced by diluent for negative control.
The frequency of SOX9-positive cells was estimated by 2 independent, experienced pathologists on a semi-quantitative scoring system, where immunoreactivity was categorized as 0%, 1–10%, 11–50%, and 51–100% positive tumour cells. Staining was assessed by counting the percentage of positive cells per 10 high-power fields (HPF) in the area of the cancerous tissue.
The human HCC cell lines HepG2 and Hep3B and the human breast cancer cell lines MCF7, BT474 and SUM159 (ATCC) were used and their origin was proven by DNA identity STR-analysis at the Cell bank of the Core Facility of the Medical University of Graz, Austria (Kit: Promega, PowerPlex 16HS System; Cat.No. DC2101). The HCC cell lines HepG2 and Hep3B as well as the luminal A breast cancer cell lines MCF-7 and BT474 were purchased from the American Type Culture Collection (ATCC) and the basal-like cell line SUM159 was obtained from Asterand (Detroid, MI). HepG2 and Hep3B cells were maintained in Minimum Essential Medium with L-Glutamine (Thermo Fisher Scientific, Waltham, MA USA), 10% fetal bovine serum gold (FBS; Biochrom, Cambridge, UK) and 1% penicillin/streptomycin (Gibco, Darmstadt, Germany; for all used cell lines: Penicillin: 10000 Units/ml, Streptomycin: 10.000 μg/ml, Gibco). MCF-7 cells were grown in MEM with Earle's salts containing 2 mmol/l L-glutamine (PAA, Pasching, Austria), 1% sodium pyruvate (PAA), 1% penicillin/streptomycin (Gibco) and 10% FBS gold (Biochrom). SUM159 cells were maintained in Ham`s F12 containing 2 mmol/l L-glutamine (PAA), 2 mmol/l HEPES buffer (Gibco), 5 μg/ml insulin actrapid (Novo Nordisk, Vienna, Austria), 1 μg/ml hydrocortisone (Sigma-Aldrich, Vienna, Austria) and 5% FBS gold (Biochrom). BT474 were cultured in RPMI 1640 (with L-Glutamine, Gibco), 20% FBS gold (Biochrom), 1% penicillin/streptomycin (Gibco) and 10 μg/ml insulin Actrapid (Novo Nordisk). Cells were incubated in a 5% CO2 humidified atmosphere at 37°C.
To measure differences in parental adherent growing cells and tumour spheres, which are considered to be enriched of stem/progenitor cancer cells, we established a spheroid growth model as previously described.[
To test whether loss of function of SOX9 expression influence cellular growth rates of HepG2 cells, we performed SOX9 knock-down experiments. HepG2 cells were transiently transfected with short interfering RNA (siRNA) to knock-down SOX9 (SOX9; 20 nM, Hs_SOX9_3 FlexiTube siRNA, cat. No. SI00007609, Qiagen) using fast forward transfection procedure according to HiPerFect Transfection Reagent (Qiagen) protocol. As a reference, the AllStars Negative Control (20nM, Qiagen) was used.
After transfection, we measured the cellular growth rate at every 24 hours over a period of time of 96 hours by applying the WST-1 proliferation assay (Roche Applied Science, Mannheim, Germany). In detail, after standard trypsinisation, 5000 HCC cells per well were seeded in a 96-well culture plate. After transfection with a siRNA against SOX9 or respective control (AllStars Negative Control, Qiagen), cells were incubated in 100 μl of normal growth medium for 96 h and the WST-1 proliferation reagent (Roche Applied Science) was added every 24 hours according to manufacturer’s recommendations. At each time point, after 3 hours WST-1 incubation colorimetric changes were measured using a SpectraMax Plus (Molecular Devices) at a wavelength of 450 nm with a reference wavelength at 620 nm.
For detection of SOX9 expression in transiently transfected HepG2 cells after siRNA treatment and to compare expression values of SOX9 in adherent growing cells and tumour spheres, total RNA was isolated using a standard TRIzol protocol according to manufacturer’s instructions. For detection of mRNA expression levels after stable transfection experiments, 1 μg of total RNA was reverse transcribed by using QuantiTect Reverse Transcription Kit (Qiagen, Hilden, Germany) according to manufacturer’s protocol. Quantitative RT-PCR was carried out in technical duplicates of biological triplicates using commercially available primers specific for SOX9 (Hs_SOX9_1_SG_SG QuantiTect primer assay, Qiagen), Nanog (fw:
All statistical analyses were performed using SPSS version 17.0 software (SPSS, Chicago, IL) or MedCalc Statistical Software version 16.8.4 (MedCalc Software bvba, Ostend, Belgium;
Clinico-pathological patient characteristics including sex, age, tumour stage, tumour grade, growth pattern and presence of liver cirrhosis can be found in
Clinicopathological parameter | No. of patients (n = 82)(%) |
---|---|
Male | 63/82 (76.8%) |
Female | 19/82 (23.2%) |
≤65 | 41/82 (50.0%) |
>65 | 41/82 (50.0%) |
T1 | 13/82 (15.9%) |
T2 | 24/82 (29.3%) |
T3 | 33/82 (40.2%) |
T4 | 8/82 (9.8%) |
Missing | 4/82 (4.9%) |
G1 | 22/82 (26.8%) |
G2 | 49/82 (59.8%) |
G3 | 10/82 (12.2%) |
Missing | 1/82 (1.2%) |
Trabecular | 37/82 (45.1%) |
Trabecular-tubular | 24/82 (29.3%) |
Trabecular-solide | 9/82 (11.0%) |
Tubular | 2/82 (2.4%) |
Fibro, trabecular-tubular-solide | 5/82 (6.1%) |
Fibro-lamellar | 3/82 (3.7%) |
Solide | 2/82 (2.4%) |
Yes | 52/80 (63.4%) |
No | 28/80 (34.1%) |
Positive | 28/82 (34.1%) |
Negative | 54/82 (65.9%) |
The cut-off value of positivity for SOX9 staining was set to 5% positive tumour cells/per 10 HPF. Using this definition, we found 28 of 82 (34.1%) cases positive for SOX9 expression and 54 (65.9%) cases negative for SOX9 expression. In general, SOX9 was located in the nucleus and mainly displayed a strong staining pattern. Majority of SOX9 positive HCC cases were negative for SOX9 expression in large portions of tumour tissue and SOX9 positive staining’s were found only in single cells (
To evaluate whether SOX9 protein expression was associated with either clinico-pathological parameters or outcome of patients with HCC, we correlated immunohistochemical SOX9 staining results with tumour stage, age, tumour grade, sex and presence of inclusion bodies, Mallory-denk bodies (MDB), intrahepatic B cells (IHB) or cirrhosis. SOX9 positivity was only significantly associated with male gender (
Clinico-pathological parameters | SOX9 positive staining (%) | SOX9 negative staining (%) | P |
---|---|---|---|
Male | 26 (31.7%) | 37 (45.1%) | |
Female | 2 (2.4%) | 17 (20.7%) | |
.641 |
|||
≤65 | 15 (18.3%) | 26 (31.7%) | |
>65 | 13 (15.9%) | 28 (34.1%) | |
.225 |
|||
Yes | 20 (25.0%) | 32 (40.0%) | |
No | 7 (8.8%) | 21 (26.3%) | |
.748 |
|||
Low tumour stage (T1+T2) | 13 (16.7%) | 24 (30.8%) | |
High tumour stage (T3+T4) | 13 (16.7%) | 28 (35.9%) | |
.075 |
|||
Low grade (Grade 1) | 11 (13.6%) | 11 (13.6%) | |
High grade (Grade 2 & 3) | 17 (21.0%) | 42 (51.9%) | |
.583 |
|||
No MDB | 15 (20.3%) | 23 (31.1%) | |
MDB | 12 (16.2%) | 24 (32.4%) | |
.078 |
|||
No IHB | 21 (28.4%) | 27 (36.5%) | |
IHB | 6 (8.1%) | 20 (27.0%) | |
.583 |
|||
No inclusion bodies | 15 (20.3%) | 23 (31.1%) | |
One or both inclusion bodies | 12 (16.2%) | 24 (32.4%) | |
.339 |
|||
No inclusion | 13 (17.6%) | 18 (24.3%) | |
IHB | 2 (2.7%) | 5 (6.8%) | |
MDB | 8 (10.8%) | 9 (12.2%) | |
IHB and MDB | 4 (5.4%) | 15 (20.3%) |
aχ2 test; MDB = Mallory-Denk body; IHB = intracellular hyaline bodies
Overall, three-year overall survival was observed in 44 (53.7%) of 82 patients. We divided HCC patients according to the cut-off value of 5% into the groups SOX9 positive (n = 28) and SOX9 negative (n = 54), respectively. For the 3-year overall survival, death occurred in 20 patients (37.0%) of 54 in the SOX9 negative group and 18 (64.3%) of 28 died in the SOX9 positive group (
A: Kaplan-Maier plot for 3-year survival in patients with HCC showing immunohistochemical positive SOX9 expression versus negative SOX9 expression (n = 84). B: Kaplan-Maier plots for 3-year survival calculated from a dataset derived from the Cancer Genome Atlas in patients with HCC showing high SOX9 expression versus low SOX9 expression.
To externally validate our findings in an independent cohort, we explored the Cancer Genome Atlas dataset (TCGA data, n = 359 patients with HCC) and used the same grouping proportions (66% of patients defined as low and 34% of patients as high expression) for categorizing patients. As shown in
(A) SOX9 knock-down was assessed by expression analysis of mRNA by qRT-PCR after siRNA treatment against SOX9. The dark grey bar shows a significant decrease in SOX9 expression in HepG2 cells. (b) HepG2 cells show reduced cellular growth after SOX9 knock-down after 96 hours follow-up.
To test for universality of SOX9 expression as prognostic factor in other cancers, we additionally analysed unbiased publicly available datasets of breast (n = 3951), ovarian (n = 1306), lung (n = 1926) and gastric cancer (n = 876). Regardless of the type of cancer, high SOX9 expression levels always and consistently prevailed as strong predictor of poor 5-years relapse free survival (breast,
(A) 5-year relapse free survival rates in breast cancer patients. (B) 5-year progression free survival in ovarian cancer patients. 5-year survival rates in patients with (C) lung cancer and (D) gastric cancer.
As SOX9 has been previously discussed as a factor involved in stem cell biology, we generated tumour spheres in a panel of liver and breast cancer cell lines (n = 5) under low attachment condition. This
Representative examples of tumour spheres in hepatocellular carcinoma (HepG2)(D) and breast cancer (MCF7)(A) (D) cells. SOX9 expression has been assessed by qRT-PCR in two liver cancer cell lines (HepG2 (B) and Hep3B (C)) and in three breast cancer cell lines (MCF7 (E), BT474 (F) and SUM159 (G)) comparing parental adherent or tumour spheres. In all five tested cell lines, SOX9 was significantly higher expressed in tumour spheres (between 1.6-fold and 57-fold increased) than in parental adherent cells. Expression levels of two cancer stem cell marker (Nanog (H) and Oct4 (I)) have been determined all five cancer cell lines in adherent cells and tumour spheres. Nanog and Oct4 were significantly up-regulated in tumour spheres in comparison to normal adherent tumour cells.
To confirm that our tumour sphere models are enriched with putative cancer stem cells, we measured the established stem cell markers Nanog and Oct4 in the same tumour spheres. Both transcription factors were up-regulated in tumour spheres compared to normal adherent growing tumour cells (
SOX9 has been implicated as a cancer stem cell marker with the ability to drive cancer growth and promote cancers metastatic ability[
The positivity for SOX9 expression was independent of the presence of cirrhosis, one of the major risk factors for HCC. On the one hand we could not find any association between SOX9 expression and tumour stage nor age which is in line with others[
In conclusion, our study supports the hypothesis that SOX9 is involved in liver cancer proliferation and HCC stem cell capabilities. Furthermore, SOX9 is a universal prognosis marker across several types of cancer. Therefore, therapeutic strategies targeting SOX9 and it´s up-stream effectors might be helpful in treating SOX9-dependent tumours.
We gratefully thank Friedrich Jamnig for composing and formatting the figures. The data shown here are in part based upon data generated by the TCGA Research Network (TCGA,