Immune checkpoint inhibitors (ICIs) targeting cytotoxic T lymphocyte antigen-4 (CTLA-4), or programmed cell death protein 1 (PD-1) and its ligand (PD-L1) interaction have shifted the treatment paradigms and significantly improve the overall survival (OS) in diverse cancers (1–4). Nevertheless, ICIs could only benefit a minority (~20%) of unselected population (5). Herein, there is an urgent need to develop novel predictive biomarkers for the majority of patients, who could not benefit from ICIs treatment. The mutational landscape of tumor cells is a direct reflection of tumor immunogenicity and could dictate the extent and phenotype of immune infiltrates (6–8). Understanding the relationship between tumor genomic alterations and response to ICIs could lay a foundation for the development of novel predictive biomarkers and therapeutic strategies to improve the clinical benefit (8).

Lysine (K)-specific demethylase 5C (KDM5C) is a histone demethylase that specifically removes methyl residues from tri-, di-, and monomethylated lysine 4 on histone H3 lysine 4 (H3K4), thus resulting in suppressing gene transcription by reducing H3K4 trimethylation levels (9–11). Previous studies reported that genetic alterations of KDM5C were common in various types of cancers including breast, colon, ovarian, prostate cancer and so on. It plays a significant role in the tumorigenesis, cancer cell proliferation, invasion, drug resistance and the regulation of tumor-related gene expression (12–14). Moreover, a recent elegant study analyzed the multi-omics data of 823 advanced renal cell carcinoma and found that somatic mutations in KDM5C correlate with high angiogenesis and AMPK/fatty acid oxidation gene expression, which was enriched in ICIs beneficial group. These findings revealed the contribution of KDM5C to antitumor immune response. Therefore, it is valuable to explore the predictive value of KDM5C alterations for ICIs treatment outcome in multiple cancers.

Here, we performed this pan-cancer analysis to investigate KDM5C alterations frequency and their predictive significance for ICIs treatment outcomes across cancer types. We also evaluated the relationship between KDM5C alteration and immune infiltrates and signatures by using online database to unravel the potential mechanism.

To delineate the immune microenvironment features of tumors with KDM5C alterations, we calculated and compared immune infiltrates, immune signatures and immune-related gene expression between KDM5C altered and wild type group by using RNA-seq data from TCGA. The correlation between KDM5C expression and immune checkpoints expression in different cancers was evaluated by using online database, named Tumor Immune Estimation Resource (TIMER). The statistical methods were listed in this website (https://cistrome.shinyapps.io/timer/) and their previous publications (17, 18). The abundance of tumor infiltrating leukocytes, including CD8⁺ T cells, CD4⁺ T cells, regulatory T cells (Tregs), dendritic cells, B cells, macrophage, myeloid-derived suppressor cells (MDSC), NK cells, mast cell, neutrophils, endothelial cells and cancer-associated fibroblasts (CAFs), was estimated by using different bioinformatic algorithm and compared between KDM5C altered and wild type group.

To our knowledge, this study firstly reported the frequency of KDM5C alterations and its pan-cancer predictive value to ICI treatment in various cancers. KDM5C alterations were a negative prognostic marker in whole group but it might be utilized to predict survival benefit from ICI treatment across diverse cancers. Although KDM5C altered tumors had significantly higher TMB level, multivariate analysis showed that KDM5C alterations was associated with significantly longer OS independent of TMB. Moreover, we did not observe the association between KDM5C alterations and prolonged OS in patients with MSS solid tumors, suggesting that it may not be suitable for predicting ICI treatment outcome in MSS solid tumors. Mechanistically, KDM5C altered tumors was found to be markedly correlated with enhanced tumor immunogenicity and immunosupportive features of anti-tumor microenvironment.

In this pan-cancer analysis, the frequency of KDM5C alterations was 2.1% in a cohort of 45614 cancer patients, with esophagogastric cancer, endometrial carcinoma and renal cell carcinoma having the highest levels of KDM5C alterations, which was similar to previous publications (21, 22). Interestingly, we found a positive association between co-occurrence of KDM5C alterations and some common epigenetic regulatory genes including PBRM1, KMT2C, SETD2, NSD1, etc. In spite of the unclear biological function of these co-mutations, several previous studies have shown that these diver genes are very important tumor suppressor genes in renal cell cancers and could contribute to the aggressive phenotype, therapeutic efficacy and/or prognostic value (23–29). Therefore, it would be valuable to unravel the biological and molecular mechanisms, and impact on clinical outcome of this co-occurrence for specific cancer types in future studies.

Previous studies reported KDM5C is required for proper DNA replication at early origins and its alterations could lead to genomic instability in sporadic renal cancer (30, 31). We thus evaluated the association between KDM5C alteration and TMB level. As expected, our results showed that KDM5C altered tumors had significantly higher TMB level than wild type ones in two independent cohorts, indicating that KDM5C alterations could be considered as predictive biomarkers for ICI treatment. Having noticed this relationship, we then investigated both predictive and prognostic significance of KDM5C alterations. In whole group, patients with KDM5C alterations had a significantly shorter OS than those with wild type, suggesting that KDM5C alterations could not confer an intrinsic survival benefit to treatment-naïve patients receiving ICI treatment. In ICI-treated cohort, patients with KDM5C alterations had a substantially prolonged OS. Moreover, subgroup analyses showed the association between KDM5C alterations and OS was independent of TMB in patients receiving ICI. More interestingly, in NSCLC treated with ICI, we found patients with KDM5C alterations had the significantly longer PFS than wild type groups. Collectively, KDM5C alterations could be considered as a potential pan-cancer predictive biomarker for ICI treatment, especially for NSCLC.

As a histone demethylase, KDM5C could suppress gene transcription by reducing H3K4 trimethylation levels (9–11). KDM5C plays a significant role in the tumorigenesis, cancer cell proliferation, invasion, metastasis and drug resistance (12–14). Recently, an elegant study analyzed the multi-omics data of 823 advanced renal cell carcinoma and found that somatic mutations in KDM5C correlate with high angiogenesis and AMPK/fatty acid oxidation gene expression, which was enriched in ICIs beneficial group (32). These findings suggested that KDM5C altered tumor would have specific immune microenvironment features. In this study, we observed that KDM5C altered tumors had markedly higher TMB and were associated with anti-antitumor immune signatures, indicating that KDM5C altered tumors would possess the enhanced tumor immunogenicity and relatively immunosupportive microenvironment, supporting its predictive value to ICI treatment.

Pan-cancer universality of immunotherapy targeting PD-1 and PD-L1 interaction challenges us to rethink the investigation and development of predictive biomarkers. To date, MSI-high (MSI-H) is the only pan-cancer biomarker approved by the FDA with a relatively low frequency (~4%) (33, 34). MSI-H is common in digestive cancer including colorectal cancer and gastric cancer, while KDM5C alterations were more common in endometrial and renal cell carcinoma, indicating the predictive value of MSI-H and KDM5C alterations is not overlapped. Notably, KDM5C alterations could not predict the clinical outcome in patients with MSS solid tumors receiving ICI, which need future investigation. Collectively, the pan-cancer predictive significance of KDM5C alterations and its complementation to MSI-H in ICI therapy are anticipated.

There are several limitations that should be acknowledged. First, the origin of included cohorts was diverse, which could result in the selection bias and inconsistency of data quality. Combining different groups of patients with distinct histologies without meta-analysis could lead to the methodological pitfalls. Second, the KDM5C altered cohort included both gain (e.g., amplifications) and loss (e.g., deletions) of function alterations whether they could cause the same survival or ICI response difference compared to WT remained undetermined. Without adjustment per histology and type of KDM5C alterations, these results should be interpreted with caution. Third, due to the unavailable PD-L1 expression results from online database, we could not evaluate the relationship between KDM5C alterations and PD-L1 expression. Last but not least, in patients with MSS tumors, only six patients had KDM5C alterations. The association between KDM5C alterations and prolonged OS in MSS tumors needs further exploration.

In summary, the present study firstly provides the evidence that KDM5C alterations were associated with enhanced tumor immunogenicity and inflamed anti-tumor immunity, which result in prolonged OS in cancer patients treated with ICIs. The predictive value of KDM5C alterations were independent of tumor mutational burden and microsatellite status, suggesting that KDM5C alterations could be considered as a potential pan-cancer predictive biomarker for ICI treatment. In the future, we still need to investigate the exact molecular mechanism and large-scale, prospective studies are also warranted.

Publicly available datasets were analyzed in this study. This data can be found here: https://www.cbioportal.org.

Ethical approval was waived since we used only publicly available data and materials in this study. The patients/participants provided their written informed consent to participate in this study.

X-JC and LZ designed this study. X-JC and AR collected the clinical and sequenced data. X-JC performed the statistical analyses. X-JC, LZ and E-DZ drafted the manuscript. LZ and E-DZ provided critical comments, suggestions and revised the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.