2760 - Comparative analysis of tumor mutational burden (TMB) prediction methods and its association with determinants of the tumor immune microenvironment of urothelial bladder cancer (UBC)

Background

TMB is an emerging biomarker for immune-therapy including metastasized UBC. The evaluation of reliable methods to determine the TMB value is a prerequisite for successful application of IO therapies.

Methods

TMB was assessed in genomic DNA from 39 muscle-invasive UBC by whole genome sequencing (GPS Cancer^©, NantHealth). In 45 cases sequencing with the TruSight Oncology 500, (TSO500, Illumina) was performed. Immune cell (IC) infiltrates were analyzed by CD3, CD8, CD56, PD-1 and CD68 immunohistochemistry. PD-L1 status was evaluated by the SP263 assay (Ventana). Intrinsic subtypes (MDACC-approach) were assessed via a Nanostring-assay. Immune phenotypes were assessed by (I) spatial distribution of IC and (II) a T-cell inflammation related gene expression signature consisting of 29 genes (Nanostring). Results were validated in the TCGA MIBC cohort.

Results

Correlation between TSO500 and WGS revealed high concordance (spearman-correlation: r = 0.68, p < 0.0001). Overall percentage agreement was 76.9% for a cut-off of 10 mut/mb and 84.6% for 15 mut/mb. We found no significant association of high TMB at different cut-offs (10 mut/mb, 15 mut/mb) with inflammation status and different immune cell populations in both cohorts. TMB was not associated with intrinsic subtypes. Spatially distributed IC phenotypes or PD-L1 status was not associated with TMB. Hierarchical clustering of all factors revealed four distinct subgroups: Cluster A Inflamed, PD-L1 “high”, TMB high (n = 5); Cluster B Inflamed, PD-L1 “high”, TMB low-intermediate (n = 14); Cluster C: Uninflamed, PD-L1 “low”, TMB intermediate-high (n = 14); Cluster D Uninflamed, PD-L1 “low”, TMB low (n = 12).

Conclusions

This study provides insights into the reliability of a NGS panel for the prediction of TMB in MIBC which shows comparable performance to WGS. High TMB seems to be a characteristic of MIBC which occurs regardless of PD-L1 and inflammation status. TMB could possibly identify additional patients who do not fulfill the PD-L1 assessment based criteria for checkpoint inhibition.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Illumina, California, United States of America NantHealth, Culver City, California, United States of America.

Disclosure

M. Eckstein: Honoraria (self), Advisory / Consultancy, Speaker Bureau / Expert testimony: AstraZeneca; Advisory / Consultancy: Janssen-Cilag. All other authors have declared no conflicts of interest.