Mechanisms of augmented tumor immunogenicity via ATR inhibition in Merkel cell carcinoma

There is an urgent unmet need to develop novel therapeutic strategies for tumors that do not respond to immune checkpoint inhibition (ICI) via PD-1 pathway blockade. The ATR-mediated DNA replication checkpoint has been reported to have immune-augmenting properties; however, the mechanisms underlying these properties are not well characterized. Here we explore the potential immunogenic effects of ATR inhibition in Merkel cell carcinoma (MCC), a cancer that is particularly relevant due to its high proliferative index and frequent response to anti-PD-(L)1 therapy. ATR inhibition induced tumor cell cytotoxicity in both Merkel cell polyomavirus-positive and UV-induced MCC cell lines in the absence of exogenous DNA damage. ATR inhibition alone or in combination with low-dose radiation induced numerous proinflammatory TNF-NF-?B signals as assessed via bulk transcriptomic profiling. These included increased expression of MHC class-I alleles, antigen processing machinery, interleukins, chemokines and interferon genes associated with anti-tumor immune responses in diverse tumor types. In parallel, we observed enhanced surface exposure of the “eat-me” signal calreticulin on MCC cells and subsequent phagocytosis by human monocyte-derived macrophages. Given that MCC tumors are often cGAS-STING-deficient (including two cell lines examined here), these ATRi-induced mechanisms are significant as they were activated regardless of cGAS-STING functional status. These data provide a mechanistic basis for the clinical evaluation of ATRi in advanced ICI-refractory MCC (NCT05947500), and suggest biomarkers that may be associated with response in human MCC tumors treated with ATRi. Overall design: Early passaged WaGa, MKL2, MCC26 and MCC13-HL cell lines were used (sub-cultured less than 25 times). 72 hours of ATRN-119 (100 nM for MCPyV-positive and 250 nM for MCPyV-negative) ± 4 Gy radiation treatment was initiated in triplicate for cells seeded in 10 cm tissue-culture dishes. Untreated cells were used as negative control. Harvested cells were submitted for RNA extraction, library preparation and sequencing (DNBseq platform, BGI Genomics, Cambridge, MA, USA). Raw paired-end reads (~ 30 million per sample; 100 base pair read length) were quality-checked, adapter-trimmed, and aligned to GRCh38 with Bowtie236. Genes with counts less than 10 across all samples were filtered out. The subsequent raw RNA-Seq count data were normalized using FPKM (fragments per exonic kilobase per million) with effective library sizes estimated using TPM (transcripts per million), followed by log2 transformation with a pseudocount of 2, as implemented in the edgeR R/Bioconductor package37 and used for further analyses. Differential analysis at the gene level was performed using the Limma package38, with false discovery rate method used to adjust the p-values for multiple testing. Genes with an absolute log2 fold change of at least 0.5, normalized against mock-treated cells, were considered either up or downregulated and used for relative quantification across different treatment groups. Statistical significance was defined by a Q value of = 0.05. Dot plots were generated using the R ggplot2 package39. Transformed expression data matrices were uploaded to an integrative cloud-based multi-omics RNA-seq data mining browser from BGI Genomics (“Dr. Tom”) and then used as input for computation of treatment-induced transcriptomic enrichment via the Molecular Signatures Database (MSigDB) hallmark gene set analysis. Hallmark interpretations were queried using a minimum of 15 and a maximum of 500 genes per set for statistically significant signatures. Distribution of normalized enrichment scores (NES) and gene set enrichment analysis (GSEA)-defined false discovery rate (FDR) significance threshold ? 0.05, along with leading edge (LE) gene set sizes for a characteristic biological phenotype, can be found in Supplemental Figure 3. Variable clustering of gene expression data across all samples was performed and visualized through principal component analysis.