3.1 Aspects on p53 from the TCGA
The clinicopathological characteristics of the analyzed patients and their tumor in the TCGA are shown in suppl. Table 1. The TCGA data were investigated for somatic mutations of the p53 gene (TP53) in ccRCC. Only 5/290 (1.7%) of the samples had somatic mutations. Mutations could be found over all tumor stages without a clear association with advanced tumor stages (suppl. Figure 1). However, patients with p53 somatic mutations had a significant worse OS (p<0.001), CSS (p<0.001), and PFI (p<0.001) (suppl. Figure 2).
Next, the associations of mRNA expression levels of TP53 and clinicopathological features were analyzed. TP53 mRNA levels were higher in stage III and stage IV ccRCC (p=0.0024). Moreover, TP53 mRNA was higher expressed in patients with distant metastases, lymph node metastases, higher T stages or higher histological grades (Suppl.Figure 3). However, the analyses of the survival endpoints demonstrated no monotone association of TP53 mRNA levels with OS, CSS or PFI: Patients with mediumTP53 mRNA expression showed better survival than patients with low or high expression levels (Suppl. Figure 4).
Additionally, the protein data of p53 in the TCGA were analyzed. Higher p53 protein levels were associated with worse survival outcome in univariate analysis (HR;95% CI; p-value (logrank test): OS 2.17 (1.21 - 3.9) 0.010; DSS 2.41 (1.13 - 5.13) 0.024; PFI 1.75 (0.96 - 3.16) 0.067, suppl. Figure 5). After adjustment for TNM stages and grading p53 Protein levels could not be confirmed as independent prognostic factor in ccRCC (HR;95% CI; p-value (logrank test): OS 1.74 (0.94 – 3.24) 0.079; DSS 1.92 (0.86 – 4.27) 0.111; 1.50 (0.85 – 2.66) 0.164).
Copy number variations (CNV) in the TP53 gene were observed only in a minority of the TCGA ccRCCs. There were 53/454 (11.7%) tumors that had CNV. Deletions could be found in all tumor stages (Figure 1). Interestingly, amplifications were more often found in low than in high tumor stages (Stage I+II vs III+IV p=0.0073; M0 vs M1 p=0.013; T1+T2 vs T3+T4 p=0.002; N0 vs N1 p=0.61; G1+G2 vs G3+G4 p=0.17; Figure 1).
Expression data of twelve TP53 exon positions were available on Xena (https://xenabrowser.net). Three exon positions (chr17:7565097-7565332:-, chr17:7576525-7576657:-, chr17:7580643-7580745) showed low expression for all samples (RPKM<3) and were therefore not considered in the subsequent analyses. The evaluation of the expression levels of the remaining nine positions demonstrated that all positions had significant higher expression levels in tumor than in adjacent normal kidney tissue. Survival analysis demonstrated that for eight of the nine exons investigated, increased expression was significantly associated with OS, CSS, and PFI (all p-values<0.001; suppl. Figures 6-8). In contrast, increased expression of chr17:7571720−7573008:− was significantly associated with worse overall survival, disease specific survival, and recurrence free interval (all p-values p<0.001; suppl. Figure 6-8). The comparison of the clinicopathological features (UICC stages, TNM and Fuhrmann grading) according to the expression of the nine exon positions demonstrated no significant associations for all CNV (analyses not shown).
In multivariable analyses, all p53 results were corrected for TNM stages and Fuhrmann grading. RNA seq data, p53 protein expression, and copy number variations did not show a significant association with cancer specific survival outcome in the TCGA cohort Table 1. The expression of nine of twelve evaluable TP53 exons demonstrated an association for independent better cancer specific survival while chr17:7571720-7573008:- was associated with worse cancer specific survival only in univariable analysis Table 1.
3.2 TMA Validation of pp53 and p53 in ccRCC
To validate the prognostic relevance and to analyze the role of activated p53 in ccRCC, the expression of p53 and phosphorylated p53 (pp53) was examined in a TMA of 274 ccRCC patients. The clinicopathological characteristics of the analyzed patients and their tumor in the TMA are shown in suppl. Table 2. The median follow-up time was 89 months (IQR 25th – 75th percentile 21 – 152 months) and 72/253 patients (26%) had died at the time of analysis. Fourteen ccRCCs were not evaluable due to loss of tumor spots during the antigen retrieval and staining process.
All comparisons are shown in Table 2A and Table 2B. p53 and pp53 expression was found in only 33/242 (12%) and 98/242 (37%) evaluable ccRCCs, respectively. There was a statistical difference for the association of p53 expression with the frequency of lymph node metastases (4/33 (12) vs. (6/222 (3%), p=0,026) and there was also a significant association of pp53 intensity x frequency with lymph node metastases. Otherwise, there was no difference in tumors with vs. without p53 expression. Likewise, there was no difference in ccRCCs with vs. without expression of phosphorylated (activated) p53 except for T stages (p=0.044). These differences did not proof to be significant after correction for multiple comparisons. The univariable comparison of CSS of ccRCC patients with vs. without p53 or pp53 expression with Kaplan Meier analyses revealed no statistically provable difference (p53 p=0,943; pp53 p=0,381) (suppl. Figure 9 A and B). Next, patients with non-metastatic and metastatic ccRCC (mccRCC) were analyzed separately for differences in CSS according to p53 or pp53 expression. There was no difference in CSS of patients with ccRCCs expressing pp53 vs. no pp53 (log rank p=0.524) and there was also no statistically provable difference in CSS for tumors with p53 expression vs. no p53 expression (p=0.102) in non-metastatic ccRCCs. Similarly, there was no statistical difference for both p53 (p=0,316) and pp53 (p=0,726) expression in mccRCC.
Lastly, the product of staining frequency and intensity of p53 and pp53 was used as a continuous variable to investigate the association with CSS. The combined p53 staining frequency and intensity (see methods and materials part 2.2) was associated with CSS in univariable analysis (HR 1,01 95% CI: 1,00 – 1,03, p=0,019). There was no statistical association with CSS for pp53 staining intensity x frequency (p=0,731). In multivariable analysis, p53 staining intensity x frequency lost its significance after adjusting for TNM stages and Fuhrmann grading. Then conditional interference tree analyses were applied to calculate a possible cutoff value based on the score frequency x intensity (IntMax-Score). However, no systematic cutoff could be determined based on this systematic statistical approach for both pp53 and p53.
3.3 In-vitro studies: Functional loss of p53 in ccRCC
3.3.1 Irradiation induces p53 accumulation and activation in ccRCC cell lines
The ccRCC cell lines 786-0, Caki-1, RCC4, A-498 and the kidney cell line RC-124 were irradiated with 2 Gray. The protein content of p53 and pp53 was determined with western blot analyzes. As demonstrated in Figure 2 irradiation induced a significant increase ((range: 1.67 fold (786-0) – 2.78 fold (RCC4); Figure 2A)) of p53 levels in all cell lines. Following irradiation, p53 is activated by several posttranslational modifications including phosphorylation at serine 15 (pS15)[8]. Phosphorylation of p53 is an accepted sign of activation [8]. A considerable p53 phosphorylation (range: 1.53 fold (RCC4) – 6.16 fold (A-498) Figure 2B) was observed in all cell lines.
3.3.2 Migration and proliferation are similar after irradiation in irradiated ccRCC cell lines and controls.
Furthermore, we examined migration and proliferation in ccRCC cell lines and the non-malignant cell line RC-124. No difference in proliferation could be observed after irradiation between RCC cell lines and control Figure 3A. Similarly, after irradiation there were no differences in migration of all ccRCC cell lines in comparison to controls Figure 3B.
3.3.3 Transcriptional activity of p53 after irradiation in ccRCC cell lines and controls
Next, a reporter gene assay was applied to examine p53 transcriptional activity. No difference in p53 transcriptional activity could be detected as demonstrated in Figure 4.
3.4 Isoforms of p53 in clinical specimens of ccRCC
Next, the levels of p53 isoforms were investigated in cancer and normal tissue specimens of 55 ccRCC patients. The isoforms D40 a and D40 g could not be detected in all ccRCC tumor samples of this patient cohort.
The comparison of clinicopathological characteristics is demonstrated is demonstrated in Table 3. In summary, the only significant result after a-error (p=0,05) Bonferroni correction (p=0,05/45=0,0011) was the difference in tumor sizes for the occurrence of D133p53a in cancer and normal tissue.