HCC tumor samples, matched adjacent non-tumor hepatic tissues and clinical data from 156 patients who underwent curative surgical resection between 1989 and 2009 were obtained from the Department of Pathology, National University Hospital of Singapore (NUH). This study was approved by the National University of Singapore (NUS)-Institutional Review Board (IRB 09-112).

The tissues were deparaffinized and rehydrated before antigen retrieval was performed by heating the sections for 30 min in antigen unmasking solution (Vector Laboratories Inc., Burlingame, CA, USA) using the microwave oven. The sections were then treated with 3% hydrogen peroxide, washed in phosphate-buffered saline with Tween-20 (PBST) (1X PBS, 0.1% Tween-20), and incubated with primary antibodies overnight at 4°C. Rabbit polyclonal antibody against HDAC1 from Abcam Inc. (Cambridge, MA, USA) was used at a 1:8,000 dilution, while rabbit polyclonal antibody against HDAC2 from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA, USA) was used at a 1:200 dilution. Sections were then washed and incubated with a goat anti-rabbit IgG conjugated with avidin-biotinylated horseradish peroxidase (Dako, Glostrup, Denmark), before final washing and incubation for 1 min with DAB substrate. Counterstaining was conducted using hematoxylin solution (Sigma-Aldrich, St. Louis, MO, USA).

Each sample on the tissue microarray slides was scored based on the intensity of HDAC staining in the nuclei of the hepatocytes, as the presence of nuclear HDACs is relevant to their transcription-modifying activities. A score of 0 indicated no staining while a score of 1, 2 and 3 represented low, moderate and intense staining respectively.

The deceased and living status of patients was determined using the in-house hospital database as well as the registry of births and deaths, while the recurrence status was determined using the in-house hospital database. Results are expressed as mean ± SD for continuous variables or percentages for dichotomous variables.

As for descriptive statistics, overall frequencies and distribution of demographic and clinicopathological parameters by HDAC1 and HDAC2 expression status were calculated by either 2 by 2 or 2 by K tables with Chi-squared test and Fisher's exact test with the frequency weights (fweights) option to take into account duplicated observations of tumors and adjacent non-tumor. And two sample mean comparison test between biomarkers positive and negative statuses were made using ANCOVA test with the frequency weights (fweights) option to take into account duplicated observations of tumors and adjacent non-tumor. The comparisons between HDAC1 and HDAC2 expression status between tumor and non-tumor tissue were made using McNemar test. The presence of HDAC1 expression in tumor and non-tumor tissue was identified using the cut-off score value of ≥1, and no expression of HDAC1 in tumor and non-tumor score was identified using '0' and treated as a reference group. In addition, the levels of HDAC1 expression in tumor and non-tumor tissue were identified as low level of expression with a score cut-off 0 to 1 point, and identified as high level of expression with a score cut-off >2 points in tumor and non-tumor tissues. Similar comparisons were applied to the HDAC2 gene expression.

Associations of HDAC1 and HDAC2 gene expression with disease-specific mortality and recurrence were assessed using standard univariate and multivariate Cox proportional hazard model with robust standard error estimates to allow for intragroup correlation for tumor and adjacent non-tumor tissue measurements in the same patient. Kaplan-Meier's survival curves with log-rank tests for comparisons of survival curves were also constructed. The duration of follow-up in this study was 10 years with median survival time of 6.96 (95% CI, 4.85–7.46).

Standard Cox proportional hazards model, which assumes that the hazard ratio is constant over time, was applied initially. The proportionality assumption of the Cox regression model was assessed graphically and with the use of Schoenfeld residuals (21–22) using 'estat phtest' STATA command after fitting a model with stcox. However, violations of the Cox proportional hazards assumption were observed for both HDAC1 and HDAC2 gene expression in tumor and non-tumor tissue to predict mortality and disease-free survival (DFS), as discussed in the Results section below. Hence, the competing risk regression model was used to estimate the impact of HDAC1 and HDAC2 expression (as main covariates of interest) on the probability of mortality due to HCC, in which disease recurrence was treated as a competing event. Similarly, death due to HCC was treated as a competing event when estimating the impact of HDAC1 and HDAC2 expression on the probability of recurrence due to HCC. The results (effect sizes) are expressed as sub-hazard ratios (SHR). In addition, we adjusted for various demographic, clinicopathological factors, including tumor staging. The crude cumulative probability of mortality while accounting for the dependence of the cumulative probability function (CPF) on the hazards of other competing event (i.e., disease recurrence) was calculated. We then compared the resulting curves between two groups based on the absence vs. presence of HDAC2 expression in tumor and non-tumor tissue. With the same approach, we also compared the resulting curves between two groups based on low expression (score of 0–1) vs. high expression (score of ≥2) of HDAC1 in tumor and non-tumor tissue.

All statistical analyses were performed using Stata statistical software (version 11.1; StataCorp LP, College Station, TX, USA).

The human colon cancer cell line, HCT116, and human hepatocarcinoma cell lines, HEP3B, HEPG2 and PLC5, were purchased from the American Type Culture Collection (ATCC; Rockville, MD, USA). The HCT116 cells were cultured in McCoy's 5A (Sigma-Aldrich) medium supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). The HEP3B, HEPG2 and PLC5 cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Sigma-Aldrich) supplemented with 10% fetal bovine serum (Gibco). The cells were maintained in an incubator at 37°C in a 5% CO₂ humidified atmosphere.

Twenty-four hours before transfection, the cells were counted and plated. Twenty picomoles of Invitrogen's Stealth Select siRNA (a pool of three different sequences targeting the same gene) or universal control siRNA with matching percentage GC content and Lipofectamine RNAiMax reagent (Invitrogen) was used according to the manufacturer's instructions. The siRNA sequences along the HDAC1 mRNA sequence (5′–3′) were: aacgaauugccugugaggaa gaguu (siRNA 1A), gcaugacucauaauuugcugcucaa (siRNA 1B), and caguauucgauggccuguuugaguuc (siRNA 1C). The siRNA sequences along the HDAC2 mRNA sequence (5′–3′) were: ucu aacagucaaaggucaugcuaaa (siRNA 2A), gaagauccagacaagagaa uuucua (siRNA 2B), and ggugauggagauguaucaaccuagu (siRNA 2C).

Cells were washed in 1X PBS, trypsinized, and collected. Both the live adherent and dead floating cells were collected and pelleted at 2,500 rpm for 5 min at 4°C. They were washed in 1X PBS, fixed in cold 70% ethanol and stored in −20°C for overnight. The fixed cells were centrifuged at 2,500 rpm for 5 min at 4°C and washed once in 1X PBS. They were then resuspended in 500 µl propidium iodide (PI) staining solution (0.1% Triton X-100, 0.2 mg/ml RNaseA, 0.02 mg/ml PI, in 1X PBS) and incubated for 15 min at 37°C in the dark. The PI-stained cells were filtered through a 40-µm filter before passing through a Beckman Coulter Epics Altra at the Flow Cytometry Unit at the National University Medical Institute. Ten thousand cells were analyzed for each sample to generate a cell cycle profile. Analysis was carried out using WinMDI software.

To measure cell proliferation, the colorimetric assay was performed using WST-1 reagent (Roche Applied Science, Indianapolis, IN, USA). Twelve hours post-transfection, the cells were trypsinized, counted and replated at 500 cells/well in a 96-well plate. Twenty-four hours after replating, 10 µl of WST-1 reagent was added to each well containing 100 µl of media. The plate was returned to the incubator for 4 h before the absorbance was read at 460 nm using a spectrophotometer. A 'blank' well with only the media and WST-1 reagent but without cells was used as negative control.

Between 24 to 48 h after transfection or drug treatment, cells were harvested, counted and plated in triplicate at 1,000 cells/well in a 6-well plate. After 10 to 14 days, the wells were washed in 1X PBS and the colonies were stained with crystal violet solution. The plates were then scanned and the images were analyzed using the ImageJ (NIH) software to measure the number of colonies formed in each well.

Immunohistochemistry was performed on a tissue microarray constructed from 156 pairs of HCC tissues and their matched adjacent non-tumor tissues to examine the expression of HDAC1 and HDAC2. Scoring was based on the HDAC staining intensity of the hepatocyte nuclei, where HDACs are known to function, with 0 being undetectable and 3 being the most intensely stained. An IgG-isotype antibody was used as the negative control. The overall cumulative mortality rate of patients is shown in Fig. 1. The scoring intensities for HDAC1 and HDAC2 in tumor tissues are shown in Fig. 2, respectively, and the scoring intensities for the respective non-tumor tissues are shown in Fig. 3. The overall frequencies and distribution of demographic and clinicopathological parameters by status of HDAC1 and HDAC2 expression status are shown in Table 1A and B.

Overall, we observed that the levels of both HDAC1 and HDAC2 in tumor tissues were significantly higher than those noted in adjacent non-tumor tissues (P<0.0001, Table II). In addition, the levels of HDAC1 and HDAC2 expression (i.e., expression level ≥2) were significantly high in the tumor tissues when compared to the adjacent non-tumor tissues (P<0.0001 for both HDAC1 and HDAC2 by McNemar test) (Table II). HDAC1 expression in tumor tissues (but not HDAC2) was higher in the moderately and poorly differentiated tumors compared to that in the well-differentiated tumors (P=0.028).

The competing risk regression method was used to estimate the impact of HDAC1 and HDAC2 expression (as main covariates of interest) on mortality and disease recurrence. The final fitted competing risk regression model showed that the presence of HDAC2 expression (score ≥1) in tumor tissues is a significant independent predictor of mortality after adjusting for other significant risk factors while taking into account disease recurrence as the other competing risk event (Table III and Fig. 4A). The final adjusted SHR estimate was 9.24 (95% CI, 1.76–48.43; P=0.009) (Table III). The final adjusted SHR in the final competing risk model was adjusted for other clinically important and statistically significant variables in the univariate testing. These adjusting variables were age at diagnosis, tumor stage, presence of cirrhosis, presence of multiple lesions, the presence of vascular invasion, and presence of HDAC1 expression in tumor and non-tumor tissues.

The prediction of mortality by levels of HDAC1 expression (i.e., score 0–1 vs. ≥2) as a main covariate of interest showed a tendency towards higher risk of mortality in patients with high HDAC1 expression in tumors (Table IV and Fig. 4B), although the presence of HDAC1 expression alone was not found to be a significant predictor of mortality. The SHR estimate for high HDAC1 expression levels (score ≥2) in the tumor tissues was 2.48 with 95% CI, 0.88–7.00. However, this value did not reach statistically significant levels. As for the prediction of disease recurrence using multivariate competing risk regression approach, neither the presence of HDAC1 or HDAC2 (i.e., score ≥1), nor the levels of HDAC1 or HDAC2 expression (i.e., score 0–1 vs. score ≥2) prognosticated HCC recurrences in our population (Tables V and VI).

There was no significant interaction between HDAC1 and HDAC2 in prediction of both cancer recurrence and cancer-specific mortality in this study when HDAC1 and HDAC2 were treated as either presence or absence categories or different levels of expression in the competing risk model.

The expression of HDAC1 and HDAC2 was knocked down by siRNAs both individually and in combination in three liver cancer cell lines, HEP3B (Fig. 5A and B), HEPG2 (Fig. 6A), PLC5 (Fig. 6C) and a colorectal cancer cell line, HCT116 (Fig. 6E). To investigate the effects of HDAC1 and HDAC2 on cell proliferation, WST-1 assay was performed on HEP3B cells treated with HDAC1 and HDAC2 siRNAs in combination or individually, and compared with scrambled siRNA-treated (Scr) and untransfected control cells (Ctrl). Silencing of both HDAC1 and HDAC2 in HEP3B cells significantly reduced cell growth compared to the controls. However, knockdown of either HDAC1 or HDAC2 alone did not affect cell proliferation (Fig. 5C). In addition, colony formation was reduced when both HDAC1 and HDAC2 were knocked down in HEP3B cells, but not when HDAC1 and HDAC2 were silenced individually (Fig. 5D). Similarly, significant reduction in colony formation was only observed when combined knockdown of HDAC1 and HDAC2 was performed in the HEPG2 (Fig. 6B), PLC5 (Fig. 6D) and HCT116 (Fig. 6F) cell lines.

The effects of HDAC1 and HDAC2 on cell death was assessed by PI staining of fixed HEP3B cells treated with HDAC1 and HDAC2 siRNAs and control siRNA as above. Flow cytometric analysis showed that combined knockdown of both HDAC1 and HDAC2 significantly increased the sub-G1 fraction by 22.5% at 96 h post-transfection, compared to ~2% in the scrambled siRNA and untreated controls, indicating increased cell death (Fig. 7A and B). In contrast, knockdown of either HDAC1 or HDAC2 alone did not significantly affect cell death. The expression of various proteins involved in apoptosis was investigated by western blotting (Fig. 7C). Consistent with the flow cytometric data on cell death, cleavage of caspase-3 as well as its substrate PARP were observed in cells in which both HDAC1 and HDAC2 were knocked down up to 120 h post-transfection, but not in cells treated with either HDAC1 or HDAC2 siRNA. Taken together, our data suggest that the combined targeting of HDAC1 and HDAC2 in hepatocellular and colon cancer cells is more effective in reducing their growth potential and inducing cell death than selective targeting of either HDAC.