Pan-cancer analysis of SETD2 mutation and its association with the efficacy of immunotherapy

Histone methyltransferase SETD2 plays a critical role in maintaining genomic integrity and stability. Here, we investigated the characteristics of SETD2 somatic mutation in the cancer genome atlas pan-cancer cohort. Our data revealed that, compared with SETD2 nonmutant patients, SETD2 mutant patients had higher tumor mutation burden and microsatellite instability. In addition, the transcriptions of most genes related to immune activities were upregulated in patients with SETD2 mutant tumors. Further examination of cancer patients treated with immune checkpoint inhibitors suggested SETD2 mutation was associated with favorable clinical outcomes. These results have implication for the personalization of cancer immunotherapy.

Immune checkpoint inhibitors (ICIs) targeting programmed cell death protein-1 (PD-1), programmed cell death ligand 1 (PD-L1), and cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) can significantly improve the overall survival (OS) in cancer patients 1 . However, most patients cannot benefit from immunotherapy and reliable biomarkers are warranted 2 . Although the US Food and Drug Administration (FDA) has approved the application of PD-L1, defective mismatch repair or microsatellite instability high (dMMR/ MSI-H), and tumor mutation burden (TMB) in clinical practice, we and others have shown these biomarkers are imperfect 2,3 .
Histone methyltransferase SETD2, the sole human gene responsible for the trimethylation of histone H3 at lysine 36 (H3K36me3), plays a critical role in maintaining genomic integrity and stability by several distinct pathways 4 . Pfister et al. found SETD2 was necessary for homologous recombination repair 5 , depletion of SETD2 shows MSI and an increased spontaneous mutation frequency, characteristic of dMMR cells 6 . SETD2 also provides an alternative mechanism that leads to DNA damage repair through interacting with p53 tumor suppressor 7 . Moreover, SETD2 can directly change the chromatin accessibility, which will generate RNA processing defects 8 . It is estimated that mRNA processing defects occur in 25% of expressed gene across the whole genome when SETD2 is mutant 8 . We speculate the mutation of SETD2 results in the enrichment of tumor mutationspecific neo-antigens in the cell surface, the immune system will recognize and attack these cells with the help of ICIs. The unique features of SETD2 mutation makes it a potential biomarker for cancer immunotherapy. Accordingly, with accumulated data that are publicly available, here we conducted a comprehensive analysis to examine the characteristics of SETD2 mutation and its association with the efficacy of immunotherapy.
MSIsensor is an effective and efficient tool for deriving MSI status 9 . MSIsensor scores in patients with SETD2 mutant cancer (0.12; 0.01-0.84) were significantly higher than the scores in patient with SETD2 nonmutant cancer (0.05, 0.00-0.31; P < 0.0001; Fig. 1f). There was no correlation between the frequency of SETD2 mutation and median MSIsensor scores (correlation coefficient, 0.10; P = 0.71). The associations between MSIsensor scores and SETD2 mutation in different tumors were presented in Supplemental Fig. 1b. To further validate the association between SETD2 mutation and MSI status, we also examined the MSI MANTIS 10 scores in patients with SETD2 mutant cancer (0.32, 0.30-0.34) and patients with SETD2 nonmutant cancer (0.31, 0.29-0.33; P < 0.0001). Of note, the scores showed no differences among various subtypes of SETD2 mutation (Fig. 1g). MSH2, MSH6, MLH1, and PMS2 played critical roles during the mismatch repair (MMR) process 11,12 , the mutation in any of these four MMR genes might cause MSI-H. Here, we investigated the co-occurrence patterns of these four MMR mutant genes and SETD2 mutation (Fig. 1h). Compared with patients with SETD2 nonmutant cancer, patients with SETD2 mutant cancer harbored more MMR mutant genes (MSH6, 1.24% vs.14.38%; MSH2, 0.98% vs.11.73%; MLH1, 1.06% vs.8.41%; PMS2, 0.98% vs.9.73%; P < 0.0001 for all four genes).  genes were upregulated in SETD2 mutant samples, and many showed statistically significant. These results suggested the immune system was more active in SETD2 mutant cancer, which might be recognized as immunologically "hot" tumor. Moreover, our data provided strong evidence that cancer epigenetic driver mutations could shape tumor immune phenotype.
To investigate whether these distinct characterisitics of SETD2 mutation could translate into cancer prognosis, we compared the OS (P = 0.38, Fig. 1i), disease-free survival (P = 0.53, Fig. 1j), disease-specific survival (P = 0.76, Fig. 1k), and progress-free survival (P = 0.96, Fig. 1l) between patients with SETD2 mutant cancer and patients with SETD2 nonmutant cancer. The prognosis and survival for cancer patients in TCGA cohort were independent of SETD2 mutant status.

Chemokine Receptor
For survival analysis, a total of 2734 patients from eight studies were included (Table 1). SETD2 mutation was associated with significantly better OS (hazard ratio (HR), 0.55; 95% confidence interval (CI), 0.46-0.65; P < 0.0001; Fig. 3b). This association remained robust after adjusting for confounding factors, including age, sex, cancer type, treatment strategy, and TMB (Fig. 3c), suggesting SETD2 mutation was not a prognostic, but a predictive biomarker for cancer immunotherapy.
Due to the success of POPLAR and OAK, two multicenter randomized controlled trials conducted in patients with non-small cell lung cancer, FDA granted the application of atezolizumab in clinical practice 14 . Here, we specifically examined the association between SETD2 mutation and various clinicopathological characteristics in patients enrolled in POPLAR and OAK. As shown in Table 2, more PD-L1-positive tumors and higher TMBs were discovered in patients with SETD2 mutant cancer.
In summary, our data reveal that SETD2 mutation is correlated with higher tumor mutation burden and MSI, and more immune activities in cancer. Moreover, SETD2 mutation status is a potential biomarker in predicting the clinical outcomes in patients treated with ICIs.

Study design
Our study was deemed exempt from institutional board approval and patient informed consent because all data are deidentified and publicly available. The nonsynonymous mutations were defined as frameshift, missense, nonsense, splice site, nonstop, and translation start site changes. Truncating mutations were defined as nonsense, nonstop, frameshift deletion, frameshift insertion, and splice site. Inframe mutations included inframe deletion and inframe insertion.

TCGA data
TCGA database included sequencing and clinicopathological data from patients with over 30 types of tumors. All data included for prevalence analysis of SETD2 mutations and CNA, subtype analysis, 3D protein structure, mutation counts, MSIsensor score, MSI MANTIS score, and survival analysis were queried and downloaded from the cBioPortal for Cancer Genomics database (https://www.cbioportal.org) 15 . To study the association between SETD2 mutation and immune characteristics, KIRC, COAD, LUAD, BLCA, and UCEC data obtained from TCGA were analyzed using TISIDB (http://cis.hku.hk/TISIDB) 16 , a database integrated multiple types of data resources in onco-immunology.

Data analysis of patients with immunotherapy
We searched "immune checkpoint blockade clinical trials" across all tumor types on ClinicalTrials.gov for status as completed. The treatment strategies were classified as anti-PD-L1 (avelumab, atezolizumab, and durvalumab), anti-PD-1 (nivolumab, pembrolizumab, and cemiplimab), and anti-CTLA-4 (ipilimumab and tremelimumab), in each tumor type. Then, we conducted systematic search of PubMed database for potential trials in November 2020. Two investigators (M.L. and B.Z.) independently screened the full texts were checked for their eligibility. Any discrepancy was resolved by discussion. The selection criteria were prespecified. To be eligible, studies had to meet the following standards: (1) population: clinical trials including over 30 adult patients with solid tumor; (2) intervention: at least one arm in the trial was treated with ICIs irrespective the dosage and duration of the treatment; and (3) outcomes: reported information regarding SETD2 mutation status and OS. In addition, the reference lists of all trials fulfilling the eligibility criteria were also checked for possible relevant studies. When multiple publications of the same study appeared, only the most recent and/or most complete reporting study were included. We retrospectively collected clinical data of cancer patients samples from three melanoma studies [17][18][19] , two lung cancer trials [20][21][22] , one renal cancer datasets 23 , and two cohorts, including multiple tumors 24,25 . After removing patients samples without survival information, a total of 2734 patients treated with ICIs were included in this study.

Statistics
Survival analysis was analyzed by Kaplan-Meier method and compared using log-rank test. It was censored at the last date that the patient was not dead. HR was calculated by Cox proportional hazards model and 95% CI was reported. Median OS time and 95% CI were presented where relevant. Spearman's ρ correlation coefficient was calculated. The relations between various clinical characteristics and SETD2 mutation were evaluated with χ 2 test, Student's t test, or Fisher's exact test depending on the context. Two-sided P < 0.05 was considered statistically significant. All statistical analysis was conducted by MedCalc 18.2.1 (MedCalc Software, Belgium). The threshold for PD-L1 positivity and negativity was that PD-L1 stained cell accounted for 1% of tumor cells or immune cells.
The bold values mean P < 0.05. NA not available.
M Lu et al.