Pan-cancer Analysis of Convergent Immunoglobulins from Tumor-infiltrating B Cells

University Of Tx Md Anderson Can Ctr, Houston TX

Investigators

Abstract

Project Summary Tumor infiltrating B cells (TIL-B) express B cell receptors (BCR) and antibodies (Abs) that can have a role in antitumor function. What tumor antigens they target can give insight into anti-tumor immune function and potentially leveraged for therapeutic strategies. Public tumor RNA-seq databases such as the TCGA, represent a wide and deep data source that include TIL-B BCRs. Our analysis of these datasets reveals the surprising finding of BCRs with the same V, J and complementarity-determining region 3 (CDR3) shared across patients. The occurrence of this phenomena by random chance is extremely rare (as low as one in 10 trillion) and therefore suggests a convergent response shared in patient tumor types. Although BCR repertoire has been analyzed in small groups of patients, a wide BCR analysis within and across tumor types has not been done. Analyzing convergent BCRs could inform what common antigen targets the immune system is focusing on. The bulk of tumor sequencing data exists as NGS RNA-seq. A challenge for reconstructing a BCR is matching the immunoglobulin heavy and light chains. To accomplish this, we have developed a comprehensive algorithm based on our Frequency Inference Model (FIM) to predict the paired heavy and light chains from sequencing datasets. We propose to refine our sequence analysis by leveraging machine learning algorithms to inform our ability to clone, produce and characterize Abs shared across patient tumor samples. This proposal will characterize antibodies reconstructed from RNA-seq datasets and test their tumor and antigen recognition characteristics. Abs will be tested for anti-tumor potential and antigen identification will be done on a small cohort with the following aims. Aim 1: Analyze immunoglobulins from different cancer types in TCGA. Convergent immunoglobulins will be identified, and matching chains identified by leveraging the deep learning algorithms. Aim 2: Produce shared Abs using the MDACC Recombinant Antibody Production Platform based on the computationally derived immunoglobulins. The reactivity to cell lines and tumor samples will be analyzed. Aim 3: Elucidate targets of TIL-Bs and their cognate BCRs and evaluate the anti-tumor function the antibodies. A pipeline to identify antigen epitopes will be used. The anti-tumor function of the antibodies will be assessed. With the completion of these aims we will have a proof-of-concept that our pipeline can recapitulate shared BCRs and assess tumor targeting potential. This platform would enable identifying shared patient antigen response that could be potentially leveraged to develop therapeutic targeting strategies.