Background Adoptive cell transfer (ACT) of neoantigen-reactive tumor-infiltrating lymphocytes (TILs) is an emerging therapeutic modality for solid cancers. A growing body of clinical data in the TIL-ACT field supports the potential for the identification, selection, and expansion of tumor-reactive T cells to drive objective response in patients. We believe improvement in the method for identifying neoantigens may further increase the breadth and number of tumor-reactive T cells. Tissue biopsy based neoantigen identification can be limiting due to inter- and intra-tumoral heterogeneity and tissue access. Here, we applied whole exome DNA and RNA sequencing on patient liquid biopsy samples to assess the sensitivity of tumor variant detection and prioritization of neoantigen peptides in comparison with tissue data and to potentially improve target yield.

Methods Matched solid tissue and blood samples were collected from 10 patients (CRC, breast, melanoma, or NSCLC). For solid tissue, whole exome sequencing (WES) and transcriptome libraries were prepared using standard tissue protocols. For blood samples, cell-free DNA (cfDNA) and circulating RNA (cRNA) exome libraries were prepared using Illumina RUO library prep kit reagents. Solid tumor variant calling and neoantigen identification were performed by the Turnstone BFX-4101 platform. Liquid biopsy tumor variants and neoantigens were identified using the DRAGEN^TM Bio-IT platform and pVACtools suite.

Results Ultra-deep WES (>15,000x) of cfDNA resulted in 100% identification of known small variants at 0.5% variant allele frequency (VAF) in control samples, and >80% sensitivity for 0.2% VAF small variant detection in patient samples. Both analytical pipelines achieved 100% sensitivity on a dataset comprising experimentally determined immunogenic peptides. Concordance of somatic variant calls made between the solid and liquid biopsies ranged from 22%-88% concordance in 4 samples, while 6 samples showed no concordance. Further analysis showed a positive correlation between variant concordance and the percent tumor fraction in the liquid biopsy (from 3.6% to 0% tumor fraction in high to low concordance samples, respectively). For those samples with variant level concordance, up to 40% concordance was observed on neoepitope peptide identification between solid and liquid biopsies. In addition, the liquid biopsy data resulted in up to 29x more peptide calls than the solid tissue, suggesting the blood samples may contain unique tumor fragments not detected in solid tissue biopsy.

Conclusions Minimally invasive liquid biopsy is viable for detection of somatic variants with the potential to broaden selection of tumor-reactive TILs and improve objective response.

Acknowledgements Authors U. Singh and C. Laing contributed equally

Ethics Approval This study was approved by the Advarra IRB; IRB#00000971; IRB#21.299. Informed consent was obtained from all participants.