Internship Presentations

Using multiplex imaging to understand spatial heterogeneity in breast cancer

Yingqi Li

Mentors: Dr. Simina Boca, Innovation Center for Biomedical Informatics, Department of Oncology and Department of Biostatistics, Bioinformatics and Biomathematics, Georgetown University; Dr. Rebecca Riggins, Lombardi Comprehensive Cancer Center, Georgetown University

Date/Time: August 27, 2019 at 2pm

Location: Room 1300, Harris Building

Breast cancer is commonly classified by immunohistochemistry (IHC) detection of human epidermal growth factor 2 (HER2), estrogen receptor (ER) and progesterone receptor (PR) protein expression. Expression of any one of these biomarkers drives clinical management and treatment decisions, since there is a range of drugs available to treat either ER- or HER2-positive breast cancer. However, 50-60% of HER2+ breast cancers also co-express ER. Previous studies have shown that co-expression of ER is associated with worse response to HER2-targeted treatment prior to surgery, and unfortunately, combined treatment with HER2- and ER-targeted treatment together does not reliably improve this response. Manipulating single positive breast cancer cell lines (ER+ or HER2+) to express the other protein (and vice versa) has revealed an inhibitory relationship between HER2 and ER. In single positive breast cancer cell lines selected for resistance to a receptor-targeted drug, increased expression and activity of the other target is often used by the tumor cell as an escape mechanism. This antagonistic relationship is hypothesized to exist in HER2+/ER+ disease, but there is a critical gap in knowledge of how ER expression mechanistically influences the biology of these tumors.

The goal of this project is to quantify the heterogeneity of ER and its spatial relationship to the proliferative marker Ki67 at the protein level in primary breast tumors using multispectral quantitative IHC data, using the phenoptr and phenoptrReports R packages. We considered imaging data from 72 samples of patients who were diagnosed with ER+ breast cancer based on at least 10% of their cells showing ER expression via IHC. We generated R/R markdown scripts to calculate and visualize distances between cells expressing different proteins. We first used the phenoptrReport package to consolidate the cell segmentation data and generated a summary report containing description of the fields, tissue category and phenotype counts for each tissue microarray slide. Next, the consolidated data were used to calculate the distances between each phenotype and the cell counts within a desired radius using the phenoptr package. The nearest distances between ER+ and Ki67+ cells were then used together with the composite and component image data to visualize the spatial relationship. Preliminary findings show that ER+ cells are closer to their nearest Ki67+ cells than either the HER2+ or PR+ cells. This work is an important preliminary step for defining the relationship between ER and other biomarkers in ER+ vs. HER2+/ER+ breast cancer, and its future clinical implications. 

Summer 2019