Distinct gene expression patterns identify patients who relapse after neoadjuvant pembrolizumab and radical cystectomy in the PURE-01 study.

Abstract

549 Background: The PURE-01 clinical trial reported the use of neoadjuvant treatment with pembrolizumab prior to radical cystectomy (RC) in patients with muscle-invasive bladder cancer (MIBC; cT2–3N0M0; PMID 30343614). We have previously reported that specific molecular subtypes (i.e. basal or claudin-low) and immune signatures are associated with more favorable survival. However, reports on the detailed tumor biology of patients (pts) who relapsed after neoadjuvant pembrolizumab are lacking. Herein, we provide a detailed characterization of relapsing pts and identify distinct gene expression patterns with potential utility as novel biomarkers. Methods: Microarray data from transurethral resection of the bladder tumor (TURBT; N=102) and matched RC (N=25) tissue from the PURE-01 trial were analyzed. Gene expression signatures and molecular subtypes were assigned as previously described (PMID 32165065). The Kaplan-Meier method was used to estimate differences in patient outcomes. Immune-signatures were split by median for survival analysis. All significance testing used a two-sided t-test at a threshold of 0.05. Results: Differential gene expression analysis identified KRT20 and H19 to be enriched in relapsing patients (both p<0.001). Expression below the median was predictive of favourable 3-year overall (OS) and recurrence-free (RFS) survival (OS, 37.3% vs. 21.6%, p=0.118; RFS, 35.3% vs. 21.6%, p<0.01 and OS, 41.2% vs. 17.6%, p=0.03; RFS, 39.2% vs. 17.6%, p=0.012, respectively). Distribution of molecular subtypes showed tumors with neuronal character (NE-like) were enriched in pts who relapsed. Low immune-signatures, including Interferon-gamma and -alpha, were predictive of relapse (3-year RFS, 31.4% vs. 25.5%, p<0.001 and 31.4% vs. 25.5%, p<0.01, respectively), whereas Immunophenoscore for CD4 was not. None of these markers were predictive for relapse in matched RC-samples. Conclusions: Relapsing patients demonstrated distinct biological patterns in pre-therapy tumor samples that show prognostic value, potentially making an earlier identification and therefore adapted treatment possible. Low expression of immune-related signatures was associated with relapse.