Abstract 2717: Exploring immune contexture using computational approaches in HER2-positive breast cancer

Abstract

Abstract Introduction: Tumor infiltrating lymphocytes (TILs) have a significant effect on tumor progression and survival outcome, and serve as a potential biomarker for pathologic complete response (pCR) in neoadjuvant HER2-positive breast cancer therapy. The assessment of TIL abundance and organization prior to treatment initiation has been associated with improved survival and response in breast cancer. This study aims to assess immune contexture using in silico approaches in a meta-analysis of neoadjuvant trastuzumab therapy. Methods: A literature search was conducted on baseline and on-treatment mRNA data from neoadjuvant studies that enrolled patients who received HER2-targeted therapies. A total of 322 baseline samples from eight neoadjuvant datasets were retrieved from Gene Expression Omnibus (GEO) repository (0. Of these studies, two had available on-treatment samples (03-311; n=50, TRIO-US B07 (n=64). The raw data were pre-processed by normalization and background correction using ‘limma’ and ‘oligo’ packages in R. mRNA profiles of baseline and paired samples were analyzed using in silico approaches CIBERSORT and ESTIMATE. Response information was assigned to each sample to observe the proportion of immune cell contexture in pCR and non-responder (NR) groups in baseline analysis. Generalized linear mixed model (GLMM) analyses on baseline data from CIBERSORT was conducted using the ‘lme4’ package in R to investigate whether a systemic relationship of appropriate immune cell signatures occur across the pooled data. Statistical hypothesis on baseline data was applied to observe differences between pCR and NR groups. Testing on paired data was conducted to observe the differences in immune contexture between baseline and on-treatment. Results: Preliminary results of baseline analysis using CIBERSORT revealed a non-significant trend in reduced follicular helper T-cells and plasma cells in the pCR group amongst three neoadjuvant studies (Brodsky, CHER-LOB and TransNOAH). ESTIMATE analysis observed a trend in elevated immune infiltrate in pCR groups from two studies (Brodsky and CHER-LOB), but was not significant after multiple correction. Paired data analysis using CIBERSORT revealed that single-agent trastuzumab reduced activated natural killer cells (p<0.05) in the 03-311 study, whilst single-agent trastuzumab did not lead to changes in immune contexture in TRIO-US B07. Furthermore, trastuzumab in dual combination with lapatinib increased resting dendritic cells, memory B-cells and stromal cells, and reduced naive B-cells and activated dendritic cells (p<0.05) in TRIO-US. ESTIMATE analysis on paired data observed immune infiltration following trastuzumab alone in 03-311 (p=0.01) and stromal infiltration following dual combination in TRIO-US B07 (p=0.03). Neither immune subset was associated to pCR in all treatment arms of the on-treatment analysis. Citation Format: Dalal AlSultan, Alex J. Eustace, Stephen F. Madden, John Crown. Exploring immune contexture using computational approaches in HER2-positive breast cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2717.