Loss of Histone H3 Lysine 36 Trimethylation Is Associated With an Increased Risk of Renal Cell Carcinoma-specific Death

Thai H Ho; Payal Kapur; Richard W Joseph; Daniel J Serie; Jeanette E Eckel-Passow; Pan Tong; Jing Wang; Erik P Castle; Melissa L Stanton; John C Cheville; Eric Jonasch; James Brugarolas; Alexander S Parker

Disclosures

Mod Pathol. 2016;29(1):34-42. 

In This Article

Results

Impact of SETD2 Expression on Overall Survival From the Cancer Genome Atlas Clear Cell Renal Cell Carcinoma Data Set

The impact of SETD2 DNA copy number loss or mRNA expression on overall survival is unknown. 3p loss of heterozygosity occurs in >90% of clear cell renal cell carcinoma cases, and a loss-of-function SETD2 mutation in the remaining allele would lead to biallelic inactivation of SETD2. We assessed the impact of SETD2 copy number loss alone, concurrent SETD2 copy number loss and SETD2 mutation and no SETD2 copy number loss or mutations on overall survival (Figure 1a, Supplementary Figure S1 http://www.nature.com/modpathol/journal/v29/n1/suppinfo/modpathol2015123s1.html?url=/modpathol/journal/v29/n1/full/modpathol2015123a.html). We did not observe an association (overall log-rank test P=0.25) between overall survival and SETD2 copy number alterations (hazard ratio=0.66, 95% confidence interval=0.37–1.18) or concurrent copy number loss with SETD2 mutation (hazard ratio=0.9, 95% confidence interval=0.44–1.82). To determine whether overall survival was associated with SETD2 mRNA expression, we dichotomized mRNA (high or low) expression at the median RNA-Sequencing by Expectation Maximization value (Figure 1b). As with SETD2 copy number alterations, we did not observe an association between high SETD2 mRNA expression and overall survival (low SETD2 expression hazard ratio (95% confidence interval), 1.28 (0.91–1.80) (P=0.16, log-rank test)).

Figure 1.

Kaplan–Meier survival plots for SETD2 DNA and RNA in The Cancer Genome Atlas kidney renal clear cell carcinoma data set. (a) Analysis of SETD2 copy number and mutations. (b) Analysis of SETD2 mRNA dichotomized as high or low expression based on median RNA-Sequencing by Expectation-Maximization value. Two samples were missing for overall survival.

Correlation of Mutant SETD2 Genotype With Loss of H3K36me3 Phenotype

Setd2-knockout mice lack global H3K36me3,[3] and in humans, loss-of-function SETD2 mutations in various tumors are associated with loss of H3K36me3.[3,5,6] To evaluate H3K36me3 expression as a dichotomized variable using an immunohistochemistry assay in which loss of H3K36me3 correlates with SETD2 mutations, we generated targeted SETD2 deletions in the 786-O renal cell carcinoma cell line using zinc-finger nucleases. In a clone, 786-O SETD2 zinc finger nuclease, a four-base deletion in SETD2 was confirmed by Sanger sequencing (Supplementary Figure S2 http://www.nature.com/modpathol/journal/v29/n1/suppinfo/modpathol2015123s1.html?url=/modpathol/journal/v29/n1/full/modpathol2015123a.html). Consistent with our prior study in an independent SETD2 zinc-finger nuclease clone, SETD2 deletion disrupts the expression of H3K36me3 by western blotting and immunohistochemistry assays (Figures 2a and b).[7] To validate H3K36me3 expression as a dichotomized variable, H3K36me3 was scored by two pathologists (PK and MLS), blinded to the SETD2 genotype. Of the 26 SETD2-genotyped tumors, 21 (81%) were classified as H3K36me3 positive or negative, 4 as focal negative, and 1 as weak positive (Supplementary Table S1 http://www.nature.com/modpathol/journal/v29/n1/suppinfo/modpathol2015123s1.html?url=/modpathol/journal/v29/n1/full/modpathol2015123a.html, Figures 2c–f). Of the 15 tumors with a SETD2 wild-type genotype, 12 were classified as positive, 2 as focal negative, and 1 as weak positive. Of the 11 tumors with a SETD2 mutant genotype, 9 were classified as negative and 2 as focal negative. Overall, 81% of the tumors using immunohistochemistry classifications of negative or positive correlated with the known SETD2 genotype; the SETD2 genotype–H3K36me3 phenotype concordance improves to 100% after exclusion of the five heterogeneous staining tumors classified as weak positive or focal negative.

Figure 2.

Analysis of H3K36me3 in isogenic SETD2 renal cell carcinoma cell lines and The Mayo Clinic Nephrectomy Registry. The SETD2 wild-type 786-O cell line was transfected with zinc finger pairs that generate a SETD2 deletion. Single cell-derived clones were analyzed by fragment length analysis to identify clone SETD2 zinc finger nuclease. SETD2 was sequenced and confirmed to have a four-base deletion. (a) Western blotting confirming depletion of H3K36me3 with histone H3 and actin as controls. (b) Immunohistochemical staining of H3K36me3 comparing isogenic SETD2 cell lines. Representative images for H3K36me3 immunohistochemical staining classifications: (c) positive, (d) negative, (e) weak positive, and (f) focal negative in nephrectomy samples. Scale bar 50 μm (original magnification 400 ×; inset 1000 ×). Kaplan–Meier estimate of (g) renal cell carcinoma-specific death and (h) progression-free survival in patients with H3K36me3-negative and -positive tumors. H3K36me3 indicates histone 3 lysine 36 trimethylation.

Association of H3K36me3 With Pathological Characteristics, Renal Cell Carcinoma-specific Death, and Progression-free Survival

Of the initial 1465 slides, 1454 (99.2%) had successful immunohistochemistry staining for H3K36me3. The 1454 slides were stratified into two categories of H3K36me3 results: positive (1066; 73.3%) or negative (388; 26.7%) (Supplementary Figure S3 http://www.nature.com/modpathol/journal/v29/n1/suppinfo/modpathol2015123s1.html?url=/modpathol/journal/v29/n1/full/modpathol2015123a.html). Tumors classified as H3K36me3 negative were associated with larger tumor size, higher grade, and increased tumor necrosis (Table 1). Our mean duration of follow-up is 8.1 years (median, 8.3 years; range, 0–23.5 years; with 6 patients missing follow-up).

When H3K36me3 was modeled as a dichotomous variable, we observed evidence of an increase in risk of renal cell carcinoma-specific death and progression-free survival with loss of H3K36me3. In Figures 2g and h after adjusting for age, patients with H3K36me3-negative tumors were two times more likely to experience renal cell carcinoma-specific death and progression than patients with H3K36me3-positive tumors (hazard ratio (95% confidence interval), 2.23 (1.77–2.81); P<0.0001 and hazard ratio (95% confidence interval), 2.12 (1.74–2.60); P<0.0001, respectively). After excluding those with heterogeneous staining (weak positive, focal negative), patients with H3K36me3-negative tumors were approximately three times more likely to experience renal cell carcinoma-specific death and progression (Supplementary Table S2 http://www.nature.com/modpathol/journal/v29/n1/suppinfo/modpathol2015123s1.html?url=/modpathol/journal/v29/n1/full/modpathol2015123a.html). As loss of H3K36me3 is associated with greater tumor size, grade, and necrosis, we evaluated the prognostic value of H3K36me3 expression after adjusting for the SSIGN score, an externally validated prognostic scoring system (Supplementary Table S3 http://www.nature.com/modpathol/journal/v29/n1/suppinfo/modpathol2015123s1.html?url=/modpathol/journal/v29/n1/full/modpathol2015123a.html). Adjustment for the SSIGN score results in attenuation of the association of H3K36me3 with risk of renal cell carcinoma-specific death (hazard ratio (95% confidence interval), 1.26 (0.99–1.60); P=0.06). Subsequently we stratified the tumors by SSIGN groups (0–3, 4–7, and 8+) into low-, intermediate-, and high-risk groups, respectively. In the high-risk group (SSIGN 8+), we did not observe an association between H3K36me3 expression and renal cell carcinoma-specific survival (Table 2). However, H3K36me3-negative tumors in the low- and intermediate-risk SSIGN groups had a worse renal cell carcinoma-specific survival and progression-free survival. In Figure 3, we provide the stratified Kaplan–Meier curves for the low-, intermediate-, and high-risk SSIGN groups. Exclusion of those with heterogeneous staining (Table 3) improves upon the prognostic ability; patients with H3K36me3-negative tumors in the low- and intermediate-risk SSIGN were more likely to experience renal cell carcinoma-specific death (hazard ratio (95% confidence interval), 2.18 (1.09–4.36); P=0.03 and hazard ratio (95% confidence interval), 1.45 (1.01–2.07); P=0.04. Our data indicate that loss of H3K36me3 expression is associated with a higher risk of renal cell carcinoma-specific death and progression; this association remains significant after correcting for age and SSIGN score.

Figure 3.

Kaplan–Meier Estimate of renal cell carcinoma-specific death and progression-free survival in patients with H3K36me3 negative and positive tumors by low-, intermediate-, or high-risk Mayo SSIGN scores. (a) Renal cell carcinoma-specific survival. (b) Progression-free survival. H3K36me3 indicates histone 3 lysine 36 trimethylation; SSIGN, stage, size, grade, and necrosis.

Association of H3K36me3 With Epigenetic Tumor Suppressors BAP1 and PBRM1

The molecular pathogenesis of clear cell renal cell carcinoma is linked to chromosome 3p copy number loss and subsequent biallelic inactivation of the 3p tumor-suppressors VHL, SETD2, BAP1, and PBRM1. We previously evaluated the prognostic impact of BAP1 and PBRM1 to define epigenetic subtypes of clear cell renal cell carcinoma.[12,13] To examine the association of H3K36me3 with BAP1 and PBRM1 protein expression, we evaluated the expression of all three markers dichotomized as positive vs negative in clear cell renal cell carcinoma samples with available staining (Supplementary Figure S3 http://www.nature.com/modpathol/journal/v29/n1/suppinfo/modpathol2015123s1.html?url=/modpathol/journal/v29/n1/full/modpathol2015123a.html). H3K36me3-negative tumors were almost three times more likely to be PBRM1-negative than were H3K36me3-positive tumors (odds ratio (95% confidence interval), 2.98 (2.14–4.17); P<0.0001; Supplementary Figure S4 http://www.nature.com/modpathol/journal/v29/n1/suppinfo/modpathol2015123s1.html?url=/modpathol/journal/v29/n1/full/modpathol2015123a.html). We did not observe an association between H3K36me3 and BAP1 expression (P=0.31).

processing....