The APOBEC enzyme family drives a distinct mutational pattern found across many cancers. APOBECs (apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like) are cytosine deaminases that convert cytosine to uracil in DNA, leading to characteristic C-to-T and C-to-G mutations at TCW motifs (where W is adenine or thymine). This APOBEC mutagenesis signature has been identified as one of the most prevalent mutational processes in human cancers.
Among all solid tumors, bladder cancer (BLCA) carries one of the highest APOBEC mutation burdens. APOBEC signatures are dominant in BLCA mutational patterns across multiple whole-exome sequencing cohorts, making BLCA a primary model for studying APOBEC-driven tumorigenesis and its clinical implications.
The APOBEC family comprises 11 members with diverse tissue expression patterns and cellular functions. While APOBECs serve antiviral roles by mutating viral genomes, their activity can also cause collateral genomic damage in host cells. Whether APOBEC mutagenesis promotes or inhibits cancer progression has been context-dependent and inconsistent across studies.
A key open question was whether APOBEC mutagenesis in BLCA correlates with immune activation and treatment response. APOBEC-driven mutations may generate immunogenic neoantigens that sensitize tumors to immune checkpoint blockade (ICB), but no comprehensive analysis had established APOBEC mutagenesis enrichment score (AMES) as a standalone biomarker for prognosis and ICB benefit across multiple independent BLCA cohorts.
Pan-cancer APOBEC expression and mutagenesis were profiled across 9,765 tumor samples from 28 solid cancer types in TCGA. Three whole-exome sequencing cohorts were used to characterize mutational signatures in BLCA specifically: TCGA-BLCA, BGI-BLCA, and DFCI/MSKCC-BLCA. Four independent GEO cohorts (GSE13507, GSE32894, GSE48075, and E-MTAB-4321) plus a clinical immunotherapy cohort (Samstein's MSK-IMPACT cohort of 140 advanced BLCA patients who received ICB) were used for validation.
The APOBEC mutagenesis enrichment score (AMES) was calculated for each tumor by normalizing TCW mutation count against background mutation rate, adjusting for total mutation burden to isolate the relative contribution of APOBEC-mediated mutagenesis from overall mutation load. This background correction addresses a key limitation of raw TCW counting, where high-TMB cancers like melanoma appear APOBEC-rich solely due to mutation volume.
Random forest algorithm (1,000 trees) identified the most important APOBEC family members contributing to TCW mutations and AMES. Single-cell RNA sequencing datasets (GSE130001 and GSE145281) were analyzed using Seurat, with pseudotime trajectory analysis by Monocle2 to trace APOBEC expression along cell differentiation paths in both tumor epithelial cells and peripheral blood immune cells.
A prognostic APOBEC mutagenesis-related risk score (AMrs) was constructed using LASSO Cox regression. Starting from 401 differentially expressed genes between AMES-high (greater than 4) and AMES-low (less than 1) groups, univariate Cox regression filtered 45 candidates, and LASSO regularization with 10-fold cross-validation (optimal lambda = 0.022) retained 21 genes as the final AMrs signature. Drug sensitivity was analyzed using 1,837 compounds across GDSC, CTRP, and PRISM databases.
Among all 28 solid cancer types, bladder cancer showed the highest AMES. While melanoma (SKCM) and endometrial cancer (UCEC) ranked highest in raw tumor mutation burden, BLCA led all cancer types in background-adjusted APOBEC mutagenesis enrichment -- demonstrating that APOBEC-specific mutation patterns are proportionally more dominant in BLCA than in any other tumor type analyzed.
Random forest analysis identified APOBEC3B as the most important contributor to TCW mutations pan-cancer, while both APOBEC3A and APOBEC3B were the greatest contributors to AMES specifically. APOBEC3B was uniquely dysregulated across all 16 cancer types with matched adjacent normal tissue data, being upregulated in 14 types while downregulated only in colon and thyroid cancers.
Across 9,550 pan-cancer samples, AMES divided into three prognostic levels: AMES-low (42.26%), AMES-moderate (37.74%), and AMES-high (20.00%). In BLCA specifically, AMES showed the highest correlation with tumor mutation burden among all cancer types (r = 0.514, p less than 2.2e-16), and TMB, tumor neoantigen burden, and intra-tumor heterogeneity all increased stepwise with AMES level.
FGFR3 mutation -- a hallmark of low-grade bladder cancer -- was inversely associated with AMES level. FGFR3 was identified as a driver gene in AMES-low and AMES-moderate groups but absent from AMES-high samples, with FGFR3 mutation frequency declining as AMES increased. The AMES-high group instead showed enrichment for DNA damage response (DDR) pathway mutations, particularly in nucleotide excision repair and homologous recombination pathways.
Single-cell RNA sequencing revealed that APOBEC3B is specifically expressed in malignant epithelial cells within BLCA tumors. Analysis of two BLCA scRNA-seq datasets confirmed that APOBEC3B was highly expressed in malignant epithelial subclusters (M-C2 and M-C3) but barely detectable in normal epithelial cells, fibroblasts, endothelial cells, or myofibroblasts. Pseudotime trajectory analysis showed APOBEC3B expression rising progressively along the malignant evolution pathway.
APOBEC3B expression correlated significantly with cell cycle progression (CCP) score in malignant epithelial cells (r = 0.437, p less than 0.001), and was stepwisely elevated from normal bladder tissue to adjacent normal tissue to primary BLCA tumor samples -- establishing a consistent pattern of APOBEC3B upregulation during malignant transformation.
In contrast, APOBEC3A showed a complementary expression pattern in immune rather than tumor cells. In peripheral blood mononuclear cell (PBMC) scRNA-seq from BLCA patients, APOBEC3A was specifically expressed in FCGR3A+ monocytes and became highly expressed at the terminal differentiation stage of monocyte pseudotime trajectory, suggesting a role in monocyte-to-M1 macrophage differentiation.
Bulk RNA-seq immune infiltration analysis confirmed the APOBEC3A-M1 connection. APOBEC3A showed the highest correlation with M1 macrophage abundance among all APOBEC family members (r = 0.334, p = 5.09e-12), and BLCA samples with higher APOBEC3A expression had significantly greater M1 infiltration -- providing mechanistic evidence linking APOBEC3A to anti-tumor immune activation.
Higher AMES correlated with activated anti-tumor immunity across multiple functional readouts. Gene set enrichment analysis showed "adaptive immune response" as the most significantly upregulated pathway in AMES-high BLCA samples (NES = 2.16, p = 1.8e-28). CD8+ T cell abundance, immune infiltration score, T cell-inflamed gene expression profile, IFN-gamma response, and CYT cytolytic activity score all increased stepwise with AMES level.
Immune checkpoint molecules including PD-1, PD-L1, CTLA-4, and TIGIT were significantly elevated in AMES-high samples, and SubMap transcriptional similarity analysis showed that AMES-high samples aligned closely with ICB responders in two immunotherapy cohorts (IMvigor210 and Roh's cohort), suggesting these tumors share transcriptional features with tumors that benefit from checkpoint blockade.
Kaplan-Meier analysis demonstrated that higher AMES predicted better overall survival (HR = 0.5954, 95% CI = 0.4436 to 0.7992, p = 0.0005) and cancer-specific survival (HR = 0.5643, 95% CI = 0.3956 to 0.8049, p = 0.0015) in the TCGA-BLCA cohort. In multivariate Cox regression, AMES was the only independent protective factor for survival, with advanced stage and older age as the only risk factors.
In Samstein's ICB-treated cohort of 140 advanced BLCA patients, AMES outperformed TMB and DDR status as a survival predictor. Higher AMES correlated with better OS (HR = 0.4863, 95% CI = 0.2931 to 0.8068, p = 0.0124), and in multivariate analysis, AMES was the only significant protective factor, surpassing both tumor mutation burden and DNA damage response pathway mutation status.
A 21-gene APOBEC mutagenesis-related risk score (AMrs) was constructed and validated across five BLCA cohorts. After LASSO Cox regression on 401 differentially expressed genes between AMES extremes, 21 genes retained Cox coefficients. AMrs was calculated as a linear combination of these gene expression values weighted by their regression coefficients, enabling individual patient risk stratification.
In the TCGA-BLCA training set, patients with higher AMrs had substantially worse cancer-specific survival (HR = 3.570, 95% CI = 2.511 to 5.076, p less than 0.0001). This was independently validated in three GEO cohorts with consistently strong hazard ratios: GSE13507 (HR = 3.916, p = 0.0003), GSE32894 (HR = 8.242, p less than 0.0001), and GSE48075 (HR = 3.751, p less than 0.0001).
In a cohort of 460 non-muscle-invasive bladder cancer (NMIBC) patients, AMrs also predicted progression-free survival (HR = 5.509, 95% CI = 2.624 to 11.56, p less than 0.0001), extending the model's clinical utility to early-stage disease where treatment decisions hinge on risk of progression to muscle-invasive disease.
Drug sensitivity screening identified cell cycle inhibitors as promising therapeutic options for high-AMrs patients. From 1,837 screened compounds across GDSC, CTRP, and PRISM databases, AMrs negatively correlated with estimated sensitivity to several compounds. The two most promising from GDSC were BI-2536 and RO-3306, both cell cycle inhibitors -- consistent with the finding that APOBEC3B correlates with cell cycle progression in BLCA malignant cells.
The study positions AMES as a step from quantifying mutation burden to assessing mutation quality. TMB counts all non-silent mutations indiscriminately, but not all mutations contribute equally to immune activation and ICB sensitivity. APOBEC mutagenesis generates mutations at specific sequence contexts that are particularly effective at creating neoantigens and triggering immune responses, making AMES a more mechanistically grounded ICB predictor than raw mutation counting.
The superiority of AMES over DDR mutation status in the ICB-treated cohort has an important implication: APOBEC activity contributes to DDR gene mutations as part of its broader mutational impact. AMES therefore partially captures DDR-related immunogenicity while also reflecting additional neoantigen-generating mutations beyond DDR pathways, explaining its broader predictive power.
Single-cell data revealed distinct biological roles for the two most clinically relevant APOBEC family members in BLCA: APOBEC3B promotes malignant epithelial cell cycle progression while APOBEC3A drives pro-inflammatory monocyte differentiation toward M1 macrophages. This functional separation means APOBEC mutagenesis influences BLCA biology through both tumor-intrinsic (genomic instability) and tumor-extrinsic (immune activation) mechanisms simultaneously.
The dual nature of APOBEC activity -- driving both mutagenesis and immune activation -- creates a paradox where the same mutational process that causes genetic instability also generates the neoantigens that make tumors vulnerable to immune attack. In BLCA, the net effect appears favorable: higher APOBEC activity leads to more immunogenic tumors that respond better to ICB, suggesting that targeting APOBEC pathways could be counterproductive in this disease.