Decreased GATA5 mRNA expression associates with CpG island methylation and shortened recurrence-free survival in clear cell renal cell carcinoma

One hundred and thirty-five kidney tissues, including 77 ccRCCs and 58 paired adjacent normal renal tissue samples, were included in this study. RCC tissues were obtained from open or laparoscopic nephrectomies and partial resection. Paired tissue samples with adjacent normal tissue (adN) were obtained from a subgroup of our 77 ccRCCs cohort. Adjacent normal tissues, i.e. morphologically normal kidney were isolated with minimum of 0.5 cm to 2 cm distant from the primary tumor lesion. Samples were snap-frozen in liquid nitrogen immediately following surgery and stored at −80°C. Ethical approval of the university ethical committee (Prof. H. D. Tröger, Hannover Medical School, Carl-Neuberg-Str. 1, Hannover, Germany) and informed consent from all patients were obtained. Localized disease was defined as pT ≤ 2, lymph node involvement and metastasis negative (N0/M0), and grading (G) G1 and 1–2, whereas advanced RCC was defined as pT ≥ 3, N1 and/or M1, and G2-3 and G3. Patients with G2 were assigned to the intermediate risk group and were not considered as a parameter for low vs. high grade group comparisons. Follow up data were available for 35 patients, and RFS was defined as the interval up to the time that disease progression could be detected by computer tomography scans. Clinical and histopathological parameters are summarized in Table 1.

Isolation of total RNA from tissue specimens and from renal proximal tubular epithelial cells (RPTEC) as controls, cDNA synthesis, and quantitative real-time PCR analysis (qRT-PCR) were carried out as described previously [12]. Duplicate measurements for qRT-PCR analysis were performed using 384 sample plates, an automated liquid handling system (FasTrans, AnalyticJena, Jena, Germany), and the ABI 7900 Fast Sequence Detection System as described previously [12]. Experimenters were blinded to any patient clinicopathological or survival information. The TaqMan expression assays used were GATA-5 (Hs00388359_m1), HPRT1 (Hs99999909_m1), GUSB (Hs00939627_m1), and RPL13A (Hs03043885_g1) (all assays were from Life Technologies, Foster City, CA, USA). HPRT1, GUSB, and RPL13A were included as endogenous references. The cDNA obtained from RPTEC primary cell transcripts served as biological controls. For each qRT-PCR run, blank and no-template controls were included. Relative expression levels were calculated using the delta-deltaCT (ΔΔCt) method [13, 14], and the SDS 2.3 Manager and dataAssist V2.0 software (Life technologies) as described previously [12]. The endogenous controls, HPRT1, RPL13A and GUSB, were combined by dataAssist V2.0 software and “arithmetic mean” was used as a method of normalization.

Natural logarithms of relative expression (lnRQ) values were used for statistical calculations. All statistics were done using the statistical software, R 2.15.2 [15]. The paired t-test was used for statistical analyses of expression differences in paired tumor and adjacent normal tissues samples, whereas univariate Cox regression models were used for statistical analysis of RFS. The threshold for dichotomization of expression values was calculated using the selected rank statistics of R package, which provides the minimum p-value for log rank statistics [15]. P-values < 0.05 were considered to be statistically significant. The Kaplan-Meier method was used for survival analyses.

The analyses of relative GATA5 mRNA expression levels revealed significantly decreased expression in tumor specimens (TU; mean lnRQ = −1.7; ±SD = 1.63) compared with the corresponding adN (mean lnRQ = 1.73; ±SD = 1.32; p < 0.001; paired t-test). Figure 1A illustrates the differences in expression values observed for paired tumor and adN, indicating a strong reduction of up to 31-fold in the expression levels, largely in tumor tissues. The comparison of the distribution of relative expression values between both tissue groups showed only a small overlap (Figure 1B).

We compared lnRQ expression values and natural logarithms relative methylation (lnRML) values of GATA5 in all samples (Figure 2). Regression analysis revealed an inverse relationship between relative expression levels and methylation of GATA5 (coefficient of regression = −0.41, p < 0.001). The comparison of paired tissues, indicated by solid lines, shows that high methylation values with concurrent low expression is frequently observed in tumor tissues, whereas corresponding adN samples largely had high expression levels and low methylation.

Logistic regression analysis for comparison of tumor subgroups detected a significant difference in mRNA expression levels only for the tumor diameter (p = 0.02, odds ratio = 0.63; 95% CI: 0.42–0.93) whereas other clinicopathological parameters like sex, gender, distant metastasis, lymph node metastasis and tumor grade exhibited no statistically significant association.

Cox regression survival analyses using a statistically calculated optimum cut off value for relative GATA5 mRNA expression (lnRQ = −3.52) showed that a lower expression status was associated with increased risk for shorter time to disease recurrence (p = 0.023, hazard ratio (HR) = 0.25, 95% CI: 0.07–0.82; Table 2). Within 30 months, four out of five patients (80%) whose tumor specimens demonstrated expression values below the cut off value were identified with disease recurrence (Figure 3). The status of localized and advanced disease (p = 0.03, HR = 4.18; 95% CI: 1.15–15.2), status of metastasis (p = 0.009, HR = 4.27; 95% CI: 1.43–12.8), and tumor grade (p < 0.001, HR = 9.48; 95% CI: 2.92–30.8) were also shown to be associated with RFS (Table 2). Pairwise bivariate Cox regression analyses were first carried out to investigate whether an association between expression status and clinicopathological parameters and RFS exists. In bivariate statistical models, considering the statuses of advanced disease, metastasis, and tumor grade as covariates, we found in each case that GATA5 expression was not associated with RFS while mRNA levels were detected as a significant parameter in bivariate Cox regression models including the statuses of lymph node metastasis, age and gender (Table 3). Moreover, we carried out a multivariate analysis demonstrating that mRNA expression of GATA5 (p = 0.12, HR = 0.32; 95% CI: 0.07–1.34) is not significantly associated with RFS including the covariates status of advance disease (p = 0.29, HR = 0.18; 95% CI: 0.01-4.41), status of metastasis (p = 0.14, HR = 6.60; 95% CI: 0.55–78.68), tumor grade p = 0.002, HR = 0.09; 95% CI: 2.46–56.14, age (p = 0.36, HR = 2.54; 95% CI: 0.05–2.85) and gender (p = 0.39, HR = 2.03; 95% CI: 0.39–10.38).