Elizabeth Phillips

Background: IrAEs are prevalent in patients treated with immune checkpoint inhibitors (ICI) and mimic classical autoimmune diseases. IrAE risk prediction could improve clinical decision making and patient safety for ICI-treated patients. Classical autoimmune diseases show strong associations with germline polymorphisms and with Human Leukocyte Antigen (HLA) variants, though it is unclear if these same genetic risk factors predispose to irAEs. To identify both tolerant and irAE patients, we mined Vanderbilt University Medical Center (VUMC) BioVU, a biorepository of >300K DNA samples linked to de-identified medical records and evaluated genome-wide and imputed HLA class I and II associations with irAE risk. Methods: We identified 1,472 ICI-treated BioVU subjects: 428 with ≥ 1 irAE, and 1044 with no irAEs ≤ 6 months from last ICI treatment. Genotyping was performed by Illumina MEGAEX array. Records were screened for use of irAE therapies (e.g. systemic corticosteroids, infliximab, levothyroxine) after ICI treatment, with chart adjudication. Association tests were performed using Firth’s logistic regression adjusting for age, sex, ICI regimen, and 3 principal components to assess 1) HLA Class I and II alleles; 2) HLA evolutionary divergence; and 3) GWAS to evaluate total irAEs, organ-specific irAEs, or other specific irAEs (sample size permitting). Endocrine irAEs were excluded to be presented in a separate focused study. Results: In GWAS analyses, 53 SNPs and 23 SNPs were identified significantly (nominal p ≤ 5e-8) associated with GI and neuro-toxicity, respectively. We did not identify any significant HLA class I or II associations with irAE or organ-grouped irAEs, including most evolutionary divergence association tests. We noted marginal association (FDR=0.12, OR=6.85) of rheumatic irAEs with HLA-DRB1*10:01, an HLA class II allele with RA-specific epitope motif. Patients experiencing lung irAEs were more likely to have a high HLA class I and II evolutionary divergence sum (p = 0.013 and p=0.046, respectively). Assessment of associations with specific irAEs, validation of known associations with irAEs or autoimmune disorders, including polygenic risk scores, are underway. Conclusions: We had several SNP-irAE associations, though appropriate classification schemas to maximize power and detect genetic variation remain challenging. We will soon expand with 1,800 additional ICI-treated patients (n=~3,300 in toto), creating a robust resource for irAE genetic associations.