1928O - Meta-analysis of tumour and T cell intrinsic mechanisms of sensitization to checkpoint inhibition

Background

A wide number of biomarkers have been associated with immune checkpoint inhibitor (CPI) response to date, however clarity on their reproducibility across larger patient cohorts with defined response criteria is lacking. In addition, systematic pan-tumor analyses may reveal the relative importance of tumour cell intrinsic and microenvironmental features underpinning CPI sensitization.

Methods

Here we collated raw whole exome sequencing and transcriptomic data on >1000 patients treated with checkpoint inhibitor (CPI) treatment across eight tumor types (CPI1000+ cohort), utilizing a uniform bioinformatics pipeline. In addition harmonized RECIST response measures were collected from each study to allow standardized clinical outcome analysis. A systematic literature search was conducted to identify previously published biomarkers, which were then tested across the >1000 patient cohort via meta-analysis. Finally, single cell sequencing of tetramer positive T cells from patient tumour tissue was conducted to further validate results.

Results

Clonal-TMB was the strongest predictor of CPI response, followed by TMB and CXCL9 expression. Subclonal-TMB, somatic copy alteration burden and HLA-evolutionary divergence failed to attain significance. Mutation signature analysis revealed apobec, UV and tobacco associated mutations predictive of CPI response even after correction for TMB. scRNA sequencing of clonal neoantigen-reactive CD8-TILs, combined with bulk RNAseq analysis of CPI responding tumors, identified CCR5 and CXCL13 as T cell-intrinsic mediators of CPI-sensitisation. Finally, combination of biomarkers together into a multi-variate predictive algorithm was shown to attain siginificantly higher AUC scores than TMB alone, in both validation (n=406) and independent test set (n=383).

Conclusions

We find that high clonal mutation burden, apobec/UV/tobacco mutation signatures, together with elevated CXCL9 and CXCL13 expression, as core features marking a tumor as likely to respond to CPI therapy. As biomarker datasets continue to grow in size there is tangible opportunity to build a more complete understanding of CPI response, and identify molecularly defined patient cohorts with high chance of response.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

CRUK, MRC.

Disclosure

K.R. Litchfield: Honoraria (self): Roche. S. Quezada: Full/Part-time employment: Achilles Tx. C. Swanton: Advisory/Consultancy: Pfizer; Advisory/Consultancy: Novartis; Advisory/Consultancy: GlaxoSmithKline; Advisory/Consultancy: MSD; Advisory/Consultancy: BMS; Advisory/Consultancy: Celgene; Advisory/Consultancy: AstraZeneca; Advisory/Consultancy: Illumina; Advisory/Consultancy: Genentech; Advisory/Consultancy: Roche-Ventana; Advisory/Consultancy: GRAIL; Advisory/Consultancy: Medicxi; Advisory/Consultancy: Sarah Cannon Research Institute; Shareholder/Stockholder/Stock options: Apogen Biotechnologies; Shareholder/Stockholder/Stock options: Epic Bioscience; Shareholder/Stockholder/Stock options: GRAIL; Leadership role, Shareholder/Stockholder/Stock options: Achilles Therapeutics. All other authors have declared no conflicts of interest.