Enhancer RNA-based modeling of adverse events and objective responses of cancer immunotherapy reveals associated key enhancers and target genes

Abstract

Immune checkpoint inhibitors (ICI) targeting PD-1/PD-L1 or CTLA-4 are emerging and effective immunotherapy strategies. However, ICI-treated patients present heterogeneous responses and adverse events, thus demanding effective ways to assess benefit over risk before treatment. Here, by integrating pan-cancer clinical and molecular data, we tried to predict immune-related adverse events (irAEs, risk) and objective response rates (ORRs, benefit) based on enhancer RNAs (eRNAs) expression among patients receiving anti-PD-1/PD-L1 therapies. We built two tri-variate (eRNAs) regression models, one (with ENSR00000326714, ENSR00000148786, and ENSR00000005553) explaining 71% variance (R=0.84) of irAEs and the other (with ENSR00000164478, ENSR00000035913, and ENSR00000167231) explaining 79% (R=0.89) of ORRs. Interestingly, target genes of irAE-related enhancers, including upstream regulators of MYC, were involved in metabolism, inflammation, and immune activation, while ORR-related enhancers target PAK2 and DLG1 which participate in T cell activation. More importantly, we found that ENSR00000148786 probably enhanced TMEM43/LUMA expression mainly in B cells to induce irAEs in ICI-treated patients. Our study provides references for the identification of immunotherapy-related biomarkers and potential therapeutic targets during immunotherapy.

Keywords: TCGA; TMEM43/LUMA; adverse effect; drug responses; enhancer RNA (eRNA); immune checkpoint block therapy; pan-cancer analysis.

Figure 1

Construction of eRNA-based prediction models for irAE ROR (risk) and ORR (benefit) of immunotherapy. (A) The step-by-step workflow of this study. (B) Multicollinearity (VIF) analysis for top ten eRNA expression in predicting irAEs. Six eRNAs showed no multicollinearity, while 4 eRNAs showed strong multicollinearity. (C) Spearman correlation (Rs) between irAE-correlated eRNAs. The shade of the square indicates the Rs, and the size indicates significance (* indicates statistical significance P < 0.05). (D) Combined effects of the final trivariate model of predicting irAEs (R=0.84, P=2.1e-6). The model is 0.1912*ENSR00000005553+0.4097*ENSR00000326714+0.1953*ENSR00000148786+0.2942. (E) Multicollinearity analysis for top ten eRNA expression in predicting ORR. Two eRNAs showed no multicollinearity, while 8 eRNAs showed strong multicollinearity. (F) Spearman correlation between ORR-correlated eRNAs. The shade of the square indicates the Rs, and the size indicates P-value (* indicates statistical significance P< 0.05). (G) Combined effects of the final model of predicting ORR (R=0.89, P=3.3e-7). The model is 0.0953 + 0.0649*ENSR00000164478+0.0032*ENSR00000035913+0.1687*ENSR00000167231. irAE, immune-related adverse events; ROR, reporting odds ratio; ORR, objective response rates; LUAD, lung adenocarcinoma; SKCM, skin cutaneous melanoma; LUSC, lung squamous cell carcinoma; KIRC, kidney renal clear cell carcinoma; PRAD, prostate adenocarcinoma; BLCA, bladder urothelial carcinoma; MESO, mesothelioma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; UCEC, uterine corpus endometrial carcinoma; SARC, sarcoma; ESCA, esophageal carcinoma; PAAD, pancreatic adenocarcinoma; OV, ovarian serous cystadenocarcinoma; HNSC, head and neck squamous cell carcinoma; STAD, stomach adenocarcinoma; THCA, thyroid carcinoma; CHOL, cholangiocarcinoma; ACC, adrenocortical carcinoma; READ, rectum adenocarcinoma; COAD, colon adenocarcinoma; LIHC, liver hepatocellular carcinoma; LGG, brain lower-grade glioma; GBM, glioblastoma multiforme; UVM, uveal melanoma; UCS, uterine carcinosarcoma.

Figure 2

Visualization of enhancer-target interaction network and functional enrichment. (A) Target genes of irAE-related enhancers ENSR00000005553, ENSR00000326714, and ENSR00000148786. (B) Target genes of ORR-related enhancers ENSR00000164478, ENSR00000164478, and ENSR00000035913. (C, D) Protein-Protein Interaction (PPI) networks of target genes of enhancers in the prediction models of irAE (C) or ORR (D).

Figure 3

The ENSR00000148786 putative target TMEM43/LUMA potentially induced irAE from B cells. (A, B) Significant correlation between irAE and expression levels of eRNA ENSR00000148786 (A) or TMEM43/LUMA (B, C) Sorted mean expression levels of eRNA ENSR00000148786 across cancer types, with the arrow indicating the highest expression in DLBC (diffuse large B-cell lymphoma). (D) Significant eRNA-target correlation between ENSR00000148786 and TMEM43/LUMA was the highest in DLBC among the three significant cancer types.

Figure 4

irAE-related genes were involved in oncogenic and other pathways. (A) The RAF1 and RTK-RAS signaling pathway. Pathway graphs were generated by using the cBioPortal website. (B) The MLXIPL and MYC signaling pathway. (C) GO enrichment of genes regulated by irAE-correlated enhancer ENSR00000326714. (D) GO enrichment of genes regulated by ORR-correlated enhancer ENSR00000164478. (E) KEGG pathway enrichment of genes regulated by ORR-correlated enhancer ENSR00000164478.