EVI1 as a Prognostic and Predictive Biomarker of Clear Cell Renal Cell Carcinoma

Abstract

The transcription factor EVI1 plays an oncogenic role in several types of neoplasms by promoting aggressive cancer features. EVI1 contributes to epigenetic regulation and transcriptional control, and its overexpression has been associated with enhanced PI3K-AKT-mTOR signaling in some settings. These observations raise the possibility that EVI1 influences the prognosis and everolimus-based therapy outcome of clear cell renal cell carcinoma (ccRCC). Here, gene expression and protein immunohistochemical studies of ccRCC show that EVI1 overexpression is associated with advanced disease features and with poorer outcome-particularly in the CC-e.3 subtype defined by The Cancer Genome Atlas. Overexpression of an oncogenic EVI1 isoform in RCC cell lines confers substantial resistance to everolimus. The EVI1 rs1344555 genetic variant is associated with poorer survival and greater progression of metastatic ccRCC patients treated with everolimus. This study leads us to propose that evaluation of EVI1 protein or gene expression, and of EVI1 genetic variants may help improve estimates of prognosis and the benefit of everolimus-based therapy in ccRCC.

Keywords: EVI1; clear cell renal cell carcinoma; everolimus; genetic association; mTOR.

Figure 1

EVI1 expression in ccRCC tumors and venous tumor thrombi are associated with features of disease progression and poorer patient outcome. (A) Representative images of immunohistochemical detection of EVI1 in primary tumors (left panels) and venous tumor thrombi (right panels) from the cohort of 39 ccRCC cases (Supplementary Table S1). The top panel insets include magnified images showing nuclear positivity in cancer cells; weaker staining in the cytoplasm is also appreciated in some cases, which is consistent with observations in other cancer types [8,13]. (B) Grid showing the proportions of EVI1 IHC positivity in tumors relative to lymph node status in the same cohort. The odds ratio (OR) and corresponding p-value (Fisher’s exact test) for the association between EVI1 positivity and cancer-affected lymph node are shown. (C) Kaplan–Meier curves showing the association between EVI1 positivity and poorer survival in the same cohort. The multivariate (including age and gender) Cox regression overall survival (OS) hazard ratio (HR) estimate, 95% CI, and p-value are shown. The estimations for age and gender (male as reference) in this model were, respectively: HR = 1.05, 95% CI = 0.99–1.10, p = 0.07; and HR = 0.20, 95% CI = 0.07–0.54, p = 0.002.

Figure 2

Frequent chromosome 3q26 EVI1/MECOM gain in the CC-e.3 KIRC/ccRCC subtype, gene expression association with poorer outcome in this subtype, and with RHEB and RPTOR influencing progression. (A) Graph showing the proportions of EVI1/MECOM genomic alterations (as depicted in the inset) in TCGA KIRC primary tumor subtypes (CC-e.1-3). The percentage of tumors with genomic gain in each subtype is shown. (B) Kaplan–Meier curves showing the association between EVI1 overexpression and poorer PFI in the TCGA KIRC CC-e.3 cohort. This set was divided in two groups using the average expression value of EVI1 as threshold (low or high EVI1 tumor expression, being normally distributed). The multivariate (including age, gender, and tumor stage (I-II and III-IV) Cox regression HR estimate, 95% CI, and log-rank p-value are shown. (C) Kaplan–Meier curves showing the association between overexpression of EVI1 and RHEB (left panel) or RPTOR (right panel) with poorer PFI in the TCGA KIRC CC-e.3 cohort. This set was divided in four groups using the average expression value of EVI1 and RHEB or RPTOR as thresholds (low or high EVI1 and low or high RHEB/RPTOR tumor expression). The log-rank p-values are shown.

Figure 3

Ectopic oncogenic EVI1 overexpression confers resistance to everolimus. (A) Graphs showing viability of RCC cells (Y-axis) transiently transfected and selected with GFP or GFP-EVI1^Del190–515 expression constructs, and exposed to different concentrations of everolimus for 72 hours (X-axis). The two cell line conditions (GFP and GFP-EVI1^Del190–515) are indicated in the insets and the estimated IC₅₀ values are shown in the graphs. Each measure shows the mean and standard deviation of quintuplicate values. The curve fitting regression was computed using the log value versus normalized response. (B) Western blot results from the three RCC cell lines and two conditions, treated with DMSO or everolimus (20 nM), and analyzed for the levels of total S6 and pS6, total AKT1 and pAKT1, and loading control (tubulin, TUBA; or vinculin, VCL). The solid arrows (top left panel) indicate increased levels of pS6 in GFP-EVI1^Del190–515 over-expressing 786-O and ACHN cells. The dashed arrows indicate increased levels of total S6 in GFP-EVI1^Del190–515 over-expressing ACHN cells. Molecular weight markers are shown on the left side and expressed in kiloDalton (kDa).

Figure 4

Common EVI1 genetic variation is associated with response to everolimus of metastatic ccRCC. (A) Kaplan–Meier curves of OS based on rs1344555 C/C (n = 35) against C/T (n = 16) plus T/T (n = 1) genotypes. The univariate Cox regression HR estimate, 95% CI, and log-rank p are shown. (B) Graph showing the association between rs1344555 genotypes and AKT1 expression in metastatic RCC. The Kruskal–Wallis test p-value is shown. (C) Box plots showing the EVI1 eQTL at rs11718241 in primary tumors from complete TCGA KIRC cohort (top panel) and from the CC-e.3 cohort (bottom panel). The Wilcoxon test p-values are shown. (D) Kaplan–Meier curves of OS of metastatic ccRCC based on rs75316749 A/G (n = 5) against A/A (n = 45) genotypes.