Vaccination targeting azacytidine-induced neoantigens in the C1498 murine AML model

Abstract Hypomethylating agents, including azacytidine (AZA), are a cornerstone of frontline acute myeloid leukemia (AML) therapy. AZA inhibits DNA methyltransferase and exerts its therapeutic effect through both direct cytotoxicity and epigenetic re-expression of silenced genes, such as tumor suppressor genes. Growing evidence suggests that AZA also upregulates tumor neoantigens, including those derived from cancer-testis antigens (CTAs) and endogenous retroviral elements (ERVs), and the resulting anti-tumor immune activation may contribute to clinical benefit. We therefore tested whether neoantigen vaccination could potentiate this effect and improve survival in an AML mouse model. We first confirmed that the C1498 AML murine model is sensitive to AZA in vivo; C57Bl/6J mice treated with AZA (5 mg/kg, i.p., Days 3-7) had significantly greater survival compared to controls (Log-rank test p=0.001, HR=0.22, 95% CI 0.05,0.97). We next aimed to determine whether AZA enhanced anti-tumor immunity, a mechanism which may contribute to this in vivo therapeutic effect. Luciferase-expressing C1498 cells were pre-treated with AZA (1.0 µM, 72 h) or DMSO, washed, and co-cultured for 24h with splenic CD3⁺ T cells isolated from C1498-bearing C57Bl/6 mice (E:T = 10:1). After normalizing to drug-only controls, Aza-pretreated C1498 cells showed significant lower viability measured by relative luminescence units (RLU) compared to DMSO-treated cells (ΔRLU = –0.08, 95 % CI –0.12 to –0.04, p<0.01), confirming that azacytidine sensitizes C1498 cells to T-cell killing independently of its direct cytotoxic effect. To profile changes in the neoantigen landscape induced by azacytidine, we treated C1498 cells with AZA (1 µM, n=3) or DMSO (n=3) for 72h and performed RNA-seq and whole-exome sequencing. These sequences were analyzed using the Landscape of Effective Neoantigens Software (LENS) to predict neoantigen expression in AZA-treated and DMSO control cells (Vensko et al. 2023). LENS identified single nucleotide variants (SNVs), insertions/deletions (InDels), fusions, splice variants, cancer testis antigens (CTAs), and endogenous retroviruses (ERVs) in the transcriptomes and translates these into 8-11-mer peptides. Peptides were annotated with NetMHCpan-4.1 to retain those in top 2 percentile predicted binding affinity as these are most likely to be presented by tumor cells. Analysis of these predicted neoantigens identified 189 peptides with increased expression measured by TPM in AZA-treated cells compared to controls, and 32 peptides with exclusive expression in AZA-treated cells. We designed two pooled synthetic-long-peptide vaccines targeting these antigens upregulated by AZA. Vaccine 1 contained 6 peptides with ≥1.5-fold increased expression but detectable baseline levels in untreated cells, while Vaccine 2 included 5 peptides expressed exclusively after AZA treatment. Each 25–30 amino acid SLP was generated by extending the 8–11-mer peptides with their native flanking sequences to enhance uptake and processing by antigen-presenting cells. Peptides (100 µg each) were pooled and mixed with CpG (50 µg) for subcutaneous administration. C57BL/6J mice (n = 3 per group) were assigned to six treatment arms: Vaccine 1 + AZA, Vaccine 2 + AZA, sham vaccine + AZA, Vaccine 1 + DMSO, Vaccine 2 + DMSO, sham vaccine + DMSO. Vaccines were administered on Days 0 and 7, followed by intravenous injection of C1498 cells on Day 14. AZA (5 mg/kg, i.p.) was administered daily on Days 18–22, and anti–PD-1 antibody (200 µg, i.p.) was given on Days 18 and 21 to all groups. Survival was monitored and Kaplan–Meier curves were compared using log-rank tests. Neither Vaccine 1 nor Vaccine 2 further prolonged survival when combined with AZA (median 31 days vs AZA + sham 31 days). Vaccines alone did not improve survival compared to sham controls (median 19.5 vs 19 days). These findings indicate that, under current conditions, vaccination with neoantigens does not enhance the therapeutic benefit of AZA, suggesting that further optimization of vaccine immunogenicity or antigen selection may be necessary.