Tissue samples were collected from 86 ccRCC patients who underwent radical nephrectomy at the Department of Urology, Medical University of Gdansk, Poland, between January 2011 and September 2013. The clinical data of patients is presented in Table I. The study was approved by the local ethics committee; written consent was obtained before the surgery from each patient.

Samples were obtained according to our previous report (10) with some modifications. Briefly, dissected tissue samples of primary ccRCC tumor (n=86, named T), normal kidney (n=86, named C as controls) and adrenal gland or the whole lymph node (n=12, named M), were collected in the operating room and placed immediately in approximately five volumes of RNAlater (Ambion Inc., Austin, TX, USA).

We utilized the same methodology as previously described (10) with some modifications. In short, ExtractMe Total RNA kit (DNAGdansk, Gdansk, Poland) was used for RNA extraction, followed by spectro-photometric (NanoDrop ND-1000; Thermo Fisher Scientific, Fitchburg, WI, USA) and electrophoretic (Bioanalyzer 2100 apparatus; Agilent Technologies, Santa Clara, CA, USA) analysis of RNA. After DNA removal (Turbo DNA-free kit; Ambion, Austin, TX, USA), 2 μg total RNA was reversibly transcribed with the use of RevertAid reverse transcriptase (Fermentas-Thermo Fischer Scientific) and 0.5 μg dT₁₈ primers (Sigma-Aldrich, Munich, Germany) in a total volume of 20 μl.

Quantitative real-time PCR (qPCR) technique was applied to measure mRNA level of RASSF1A and RASSF1C genes in all samples after normalization to the GUSB reference gene level in each sample (10). List of primers and details regarding time-temperature protocol are described in Table II. All reactions were run in duplicate using 1 μl of 4X diluted cDNA and SensiFast Sybr™ No-ROX kit (BioLine, London, UK) chemistry in a total volume of 10 μl. Each gene assay was run on a separate plate (StepOnePlus apparatus; Life Technologies-Applied Biosystems, Grand Island, NY, USA) with non-template control (water instead of cDNA) and 10X diluted pooled cDNA as a run-to-run precision control. Data was acquired by StepOne Software ver. 2.2 and geometric mean of Ct (threshold cycle) values were used for comparable expression analysis. The Livak's method was utilized for quantification: Ratio = 2^−ΔΔCt to obtain raw expression data for each sample, followed by calibration to average expression data of control samples (fold change; control sample = 1).

DNA was isolated from each specimen (~2 mg) to a total volume of 20 μl followed by bisulfide modification according to manufacturer's protocol (DNA Methylation-Direct™ kit; Zymo Research, Irvine, CA, USA). Final elution was performed with the use of 50 μl of ddH₂O, and the DNA concentration was assessed with NanoDrop ND-1000. For the generation of a dilution series of control DNA standards, fully methylated (named MD) and unmethylated (UMD) human genomic DNA (HCT116 cell line, DKO strain, Human Methylated & Non-methylated DNA Set; Zymo Research) were utilized.

Methylation was assessed in samples with the use of methylation specific-high resolution melting (MS-HRM) (9). Primers sequences were designed using MethPrimer software (http://www.urogene.org/methprimer/); primers, reaction mixtures and time-temperature conditions are listed for each promoter region interrogated in Table I. MS-qPCR reactions were set on StepOnePlus (Life-Technologies) apparatus and after conventional 42-cycle amplification, post-PCR products were checked using HRM application (details in Table I) with the use of HRM software ver. 3.1 (Life-Technologies). For each run, matched DNA from T, C and M samples were set; standard dilutions of MD and UMD were made to 100, 50, 25, 10 and 0% of MD in UMD and used in the same PCR plate as well as no template control. The final methylation results divided samples into two groups: homogenous and heterogenous methylated DNA; homogenous DNA was further semi-quantitatively divided into intervals, according to MD standards (0–10%; 10–25%; 25–50% and 50–100%). Heterogenously methylated DNA samples were excluded from the analysis, due to limitation of the method (9,11).

Protein lysates were prepared with Mammalian Cell Extraction kit (BioVision, Milpitas, CA, USA). The lysates (10 μg) were loaded to a 10% Mini-Protean TGX gel (Bio-Rad Laboratories, Hercules, CA, USA), resolved by SDS-PAGE and transferred to a PVDF membrane using the Trans-Blot Turbo system (Bio-Rad Laboratories). Membranes were stained with 0.1% Ponceau S to ensure equal loading after transfer, and subsequently blocked with 5% albumin fraction V in TBS buffer with 0.1% Tween-20 (TBST) for 1 h at room temperature (RT). After washing with TBST, membranes were incubated (overnight, 4°C) with specific primary antibodies in 2% albumin/TBS: 1:2,000 rabbit anti-RASSF1A (#bs-1234R; Bioss, Inc., Woburn, MA, USA); mouse polyclonal anti-RASSF1C (1:1,000) (#ab24419; Abcam, Cambridge, UK) and anti-GAPDH peroxidise-conjugated IgM 1:50,000 (#G9295; Sigma-Aldrich, St. Louis, MO, USA). After triple washing with TBST, blots were incubated (2 h, RT) with horseradish peroxidase-conjugated secondary antibodies: 1:15,000 anti-rabbit IgG or anti-mouse IgG (Sigma-Aldrich). Following triple washing with TBST, immunoreactive bands were detected on medical X-ray film (Agfa HealthCare, Mortsel, Belgium) using chemiluminescent peroxidase substrate (Sigma-Aldrich). Densitometric analyses of immunoreactive protein bands was performed with Quantity One software (Bio-Rad Laboratories) and calculated as units = Intensity/mm². After normalization to GAPDH protein units for each sample, the semi-quantitate results for either tumor or metastasized samples were obtained as a ratio = mean units_T/M/mean units_C for RASSF1A or RASSF1C proteins.

Formalin-fixed paraffin-embedded tissue sections (6 μm) from side tissues were deparaffinized and hydrated through xylenes and graded alcohol series. After antigen retrieval using hot acidic citrate buffer (Epitope Retrieval Solution pH 6.0; Leica Biosystems Newcastle Ltd., Newcastle upon Tyne, UK) samples were blocked for endogenous peroxidase activity by using 3% hydrogen peroxide for 10 min. Sections were then incubated with 2.5% normal horse serum [ImmPRESS anti-rabbit Ig (peroxidase) polymer detection kit; Vector Laboratories, Inc., Burlingame, CA, USA] to block non-specific binding of immunoglobulin. Immunohistochemical (IHC) staining was performed using anti-RASSF1A rabbit anti-human polyclonal antibody (1:100) (#bs-1234R; Bioss, Woburn, MA, USA) or mouse polyclonal anti-RASSF1C (1:100) (#ab24419; Abcam). After 2-h incubation with primary antibodies at room temperature, slides were washed in PBS and incubated with an appropriate secondary antibody [ImmPRESS anti-rabbit Ig (peroxidase) polymer detection kit or ImmPRESS anti-mouse Ig (peroxidase) polymer detection kit] for 30 min. Slides were rinsed in PBS and immunoreactive cells were visualized by addition of 3,3′-diaminobenzidine solution (DAB peroxidase substrate kit; Vector Laboratories) and counterstained with hematoxylin. Sections were then dehydrated, mounted in DPX mounting medium and viewed under a Nikon Eclipse E800 light microscope with Lucia G software. The specificity of the IHC staining was determined by a negative control, which was prepared under the same conditions as mentioned above, replacing primary antibodies with 2.5% normal horse serum [ImmPRESS anti-rabbit Ig (peroxidase) polymer detection kit].

Statistics was performed with the use of GraphPad Prism ver. 6.05 (GraphPad Software, Inc., San Diego, CA, USA) and Statistica ver. 10c (Statsoft Inc., Tulsa, OK, USA). Non-parametric Mann-Whitney U and Kruskal-Wallis ANOVA tests were used to compare clinical and molecular data since most data did not pass D'Agostino and Pearson omnibus test. Fisher's 2×2 exact test was used to analyse relationships between the subgroups. Spearman's correlation or multivariate regression were utilized for testing the associations between two or three variables. The Cox-Mantel proportional hazard regression model was used to evaluate the effect of explorative variables on survival of ccRCC patients. First, univariate Cox regression analysis for every single variable was performed. Secondly, variables with a P-value <0.05 were included into multivariate Cox regression analysis with a variable selection via backward elimination. All associations were presented as hazard ratios (HR) with their 95% confidence interval (CI) and P-values (12). Variables for overall survival (OS) and progression-free survival (PFS) rates were calculated separately. Kaplan-Meier estimations were performed to describe survival rates.

Of 86 ccRCC patients (mean age, 62.1±11.2 years) (Table II), 37 were diagnosed as stage I (T1-2N0M0), 8 as stage II (T2N0M0), 12 as stage III (T1-2N1M0 or T3N0-1M0) and 29 as stage IV (T4N0-2M0 or T1-4N2M0 or T1-4N0-2M1). At the time of surgery 47.7% ccRCC patients were diagnosed with local or distant metastases. According to Fuhrman's nuclear grading 4 patients were grade 1, 32 grade 2, 23 grade 3 and 26 were grade 4. None of the patients had undergone chemotherapy or radiotherapy before the nephrectomy. The mean follow-up period was 21 months (range, 3–48), to date 45 patients were alive (52); all deaths (except for 1 patient) were related to ccRCC progression. Median overall survival (OS) rate was 12 months. During follow-up metastases occurred in 38 (44) patients while the median progression free-survival (PFS) rate was 6 months.

As shown in Fig. 1A, RASSF1A mRNA level in T and M samples were ~4 and 5 times lower vs. C samples, respectively. The mRNA levels of RASSF1C were ~3 times higher in tumor in comparison to either C or M samples, respectively (Fig. 1B).

After setting-up the threshold value based on median values of RASSF1A or RASSF1C expression levels in control samples decreased level of RASSF1A mRNA was noted in 66 of T and 75 of M samples, whereas RASSF1C expression was upregulated in 76 T and 33 M samples.

Patients with higher nuclear grades (Fuhrman's 3+4) and metastatic ccRCC (TNM 3+4) showed downregulation of RASSF1A (Table II). The mRNA levels of RASSF1A were ~3 and 4 times lower in TNM 3+4 and Fuhrman's 3+4 than in TNM 1+2 or Fuhrman's 1+2 groups, respectively (Fig. 1C and D). No correlations between clinical data and RASSF1C gene expression were found (Table II and Fig. 1C and D).

According to the analysis of MD/UMD standards the results of HRM-MS-qPCR were qualified into four grades: 1, 0–10% methylation; 2, 10–25%; 3, 25–50%; 4, 50–100% (data not shown). We assessed >25% methylation as hypermethylation status finding 35 (60%) unmethylated and 23 (40%) hypermethylated T samples. A significant negative correlation between increasing grades of methylation and RASSF1A mRNA levels was found (Fig. 3). In hypermethylated T samples ~3.5 times lower RASSF1A mRNA level than in non-methylated T cases was observed. Hypermethylation of RASSF1 promoter was associated only with high TNM status (Table III).

RASSF1A promoter hypermethylation was found in 5/8 of metastasized samples in which RASSF1A mRNA levels were ~15 times lower than in non-methylated metastasized samples (Fig. 2). No correlations between methylation status of metastases and clinical data were found.

RASSF1A and RASSF1C protein levels were assessed by western blot analysis in paired 58 ccRCC and control samples, and 12 metastases analyzed for RASSF1A hypermethylation. The average RASSF1A protein levels in T and M samples were ~5 and 20 times lower than in control tissue, respectively (Figs. 4 and 1E). After setting-up threshold level 44/58 T (76) and 10/12 M (83) samples showed decreased RASSF1A protein level (Fig. 1E and Table III). Decreased RASSF1A level in tumor samples was associated with higher Fuhrman's grade and high TNM grades (Table III).

RASSF1C protein level was on average ~6 times higher in T vs. C samples with similar levels in M and C (Figs. 4 and 1F). However, high RASSF1C level was found only in 10/58 (20) T and 3/12 (25) M samples (Fig. 1F and Table III).

Positive correlations were observed between mRNA and protein levels (rs=0.66 for RASSF1A and rs=0.69 for RASSF1C; P<0.001, Spearman's test).

The analysis of possible indirect association between RASSF1A promoter methylation → decreased mRNA level → decreased protein level revealed significant relationship between three measured variables (multivariable regression; P<0.001, b=−0.63; Fig. 5).

Immunohistochemical staining for RASSF1A and RASSF1C proteins was performed on twelve paired T and C samples and six M samples. As presented in Fig. 6, both proteins showed intense immunoreactivity in epithelial cells of healthy kidney whereas in tumor and metastasized samples RASSF1A was barely noticeable (Fig. 6B–D). Strong immunoreactivity of the RASSF1C protein was found in all studied samples (Fig. 6F–I).

We found that overall and progression-free survival were strongly associated with higher TNM and Fuhrman's grades of ccRCC patients (Fig. 7A, B, D and E). Patients with decreased RASSF1A mRNA levels showed significantly shorter OS and PFS rates than patients with high RASSF1A mRNA levels (Fig. 7C and F). Furthermore, higher RASSF1A promoter methylation status and lower RASSF1A protein levels were associated with shorter OS (Fig. 7G and H).

It was noted, that the levels of RASSF1C were not associated with either OS or PFS rates of ccRCC patients (plots not shown).

Multivariate analysis using the Cox proportional hazard model indicated that the classification based on RASSF1A mRNA level was the independent predictor of OS and PFS in ccRCC patients when assessed by Fuhrman's histological grade (Tables IV and V).