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


Mod Pathol. 2016;29(1):34-42. 

In This Article


Although recurrent DNA alterations in 3p tumor suppressors occur in The Cancer Genome Atlas clear cell renal cell carcinoma cohort, the prognostic impact of these alterations is limited after adjusting for the SSIGN score.[14] Systematic molecular profiling of tumors such as chondrosarcomas, glioblastomas, and leukemias identified recurrent molecular alterations that phenotypically converge on functional deregulation of the H3K36me3 axis.[4,5,15] Related to this, multi-region sequencing of clear cell renal cell carcinoma tumors has identified parallel evolution of distinct SETD2 mutations that phenotypically converge on loss of methyltransferase activity.[16] In this study, we evaluated deregulation of the H3K36me3 axis using an immunohistochemistry assay in which loss of expression correlates with a SETD2-mutant genotype in cell lines and human clear cell renal cell carcinoma tumors. Our notable observations include: (1) alterations of SETD2 DNA and mRNA expression in clear cell renal cell carcinoma tumors from The Cancer Genome Atlas clear cell renal cell carcinoma cohort are not associated with overall survival; (2) loss of H3K36me3, a posttranslational histone modification, is associated with worse outcomes (renal cell carcinoma-specific survival, progression free survival); (3) the association with progression free survival remained significant after adjusting for age and SSIGN score; and (4) these associations remain apparent among the subset of clear cell renal cell carcinoma patients already determined to be in the intermediate-risk group based on the Mayo SSIGN score.

The association of H3K36me3-negative tumors with greater size, grade, and necrosis suggests a link between loss of H3K36me3 and renal cell carcinoma tumor progression. H3K36me3 regulates DNA repair, alternative splicing, and chromatin remodeling, and these functions are linked to chromatin 'readers' with proline–tryptophan–tryptophan–proline domains that interact with methylated lysine residues.[17–19] Taken together, evidence supports a model in which loss of H3K36me3 or SETD2 'writer' function may alter the cancer phenotype through deregulation of these chromatin 'readers'. There are likely other posttranslational mechanisms that downregulate H3K36me3. In gliomas and chondroblastomas, mutations at lysine 36 in the histone H3.3 variant H3F3A are also associated with loss of H3K36me3.[15,20] We hypothesize that deregulation of the H3K36me3 axis may have a role in progression in other tumors.

There are limitations to this investigation that should be considered when evaluating our observations regarding loss of H3K36me3 and clear cell renal cell carcinoma outcome. First, we did not assess tumor genotype by massive parallel sequencing of all tumors in our cohort. Instead, we focused on a mechanistically relevant end point of loss of H3K36me3 expression after validation in cell line and tumors with a defined SETD2 genotype.[7] DNA sequencing can detect single-nucleotide variants and allelic fractions; however, in the absence of germline or normal controls, germline polymorphisms can confound mutation calls. Furthermore, it was estimated in a sequencing study of multiple cores from primary renal cell carcinoma tumors that a minimum of three distinct cores are required for accurate tumor genotyping.[21] In contrast, immunohistochemistry may be used to screen for functional loss of SETD2 without concurrent evaluation of germline DNA. Second, challenges exist in how to interpret the identification of tumors with heterogeneous staining (focal negative, weak positive). In our training cohort after exclusion of five tumors with heterogeneous staining, there is a 100% concordance between H3K36me3-positive/negative and SETD2 wild-type/mutant genotypes. On the basis of our immunohistochemistry training cohort of SETD2-genotyped tumors, two focal-negative tumors were SETD2 wild-type and two were SETD2 mutant; one SETD2 wild-type tumor was weak positive. The inclusion of heterogeneous staining samples in the training cohort reduces the concordance to 81% and attenuated the prognostic impact of H3K36me3 immunohistochemistry in the Nephrectomy Registry cohort; however, the association between H3K36me3 immunohistochemistry-negative tumors and worse outcomes remained significant.

Notwithstanding the above limitations, our work further extends the emerging molecular classification of clear cell renal cell carcinoma. Gerlinger et al[22] have shown that intragenic VHL mutations and loss of 3p are the only uniformly truncal events in ccRCC. We have shown that loss of PBRM1 and BAP1 are mutually exclusive and that loss of PBRM1 is associated with better outcomes than loss of BAP1.[10,13,23] Herein we validate previous results from a meta-analysis showing that PBRM1 and SETD2 mutations cooperate in ccRCC.[24] Specifically, we find that H3K36me3 loss is 3 times more likely in PBRM1-negative tumors than in those that express PBRM1. These data further sub-classifies PBRM1-negative tumors into those with and without SETD2 loss.