Determining an immunohistochemical profile to predict response to intravesical bacillus Calmette–Guérin (BCG) in patients with high-grade non-muscle invasive bladder cancer.

Abstract

847 Background: Intravesical Bacillus Calmette–Guérin (BCG) remains the gold standard for the management of high-grade non-muscle invasive bladder cancer (NMIBC). However, up to 70% of patients fail BCG therapy. We aimed to define potential mechanisms for BCG resistance and develop an immunohistochemical (IHC) panel to help identify patients likely to respond to intravesical BCG. Methods: Patients with NMIBC undergoing induction intravesical BCG at a tertiary institution were identified. Twelve BCG-responders and 13 non-responders were matched for patient and tumour factors. RNA sequencing was performed with hierarchical clustering and Gene Set Enrichment Analysis to identify response resistance mechanisms. Immune cell subsets were measured using pre and post-BCG therapy for CD4, CD8, T-Bet, GATA-3 and PD-1 antibodies. GATA-3 and T-Bet stains were used as surrogates for Th-2 and Th-1 cells, respectively. T-tests were used to assess differences. An integrated IHC panel using CD4, CD8, T-Bet, GATA and PD1 was developed and correlated with BCG response using Receiver Operator Characteristic (ROC) curves. Results: On hierarchical clustering, BCG-responders and non-responders had distinct gene expression profiles pre- and post-BCG. Prior to exposure to BCG, non-responders had enrichment for a pro-inflammatory gene signature with a higher CD4:CD8 ratio (2.94 vs 1.71, p=0.0003) and a higher Th-2/Th-1 (GATA/T-Bet) ratio (5.95 vs 2.97, p=0.0026) compared to BCG-responders, on IHC. However, on exposure to BCG, non-responders had no changes to the CD4:CD8 or Th-2/Th-1 ratios but had a 78% increase in the PD-1 expression (MD 20.83/5hpf, p=0.016) with BCG, indicating T cell exhaustion. In contrast, upon exposure to BCG, responders had an increase in the Th-2/Th-1 ratio (mean difference 0.9323, p=0.0228) with enrichment of the natural killer cell pathway compared to non-responders (p<0.05). The area under the ROC curve using the integrated IHC panel to predict BCG response was 0.864 (95% CI: 0.663 to 1.000). Conclusions: BCG non-responders had a pro-inflammatory TME, which showed marked T cell exhaustion on exposure to BCG. In contrast, responders had a baseline TME with low CD4:CD8 and low Th-2/Th-1 ratios at baseline, which allowed activation of both humoral and adaptive responses to BCG. These immune changes can be utilised as an IHC panel to predict response to BCG therapy.