Abstract 209P
Background
Immune-cell indices of immunohistochemistry (IHC) are calculated in countless research studies; often to identify prognostic or predictive biomarkers in cancer patients. The study aimed to investigate how the location of immune cells and the amount of tumor epithelia may affect their calculation in breast cancer (BC) patients whilst exploring multiple sample techniques for IHC immune-cell quantification.
Methods
Formalin-fixed, paraffine-embedded pretreatment tumor tissue (N=227; full-cut slides) from BC patients was stained with two multiplexed chromogenic IHC using CK7/19, CD8, CD4, FOXP3, CD20, CD66b, and CD68. AI-powered image analysis (U-net) determined their area-based indices in the reference spaces: total tumor area, tumor stroma, tumor epithelia, a stromal tumor brim, and a fixed area for each lesion using multiple subsamples (mean no., 100) within tumor stroma (in total, 1 mm2) and epithelia (in total, 1 mm2).
Results
Immune cells were predominantly situated in tumor stroma and rarely epithelia (P<0.01). The most infiltrative cell type was CD68 with a mean index of 3% for epithelia, yet 32% for stroma. In remainder cells, indices were <0.7% for epithelia. The amount of epithelia ranged from 1-60%, which strongly affected index calculations that included the total area of epithelia (P<0.01). In addition, solidity of tumor islands was highly associated with the number of infiltrative CD8, CD66b, and CD68 cells (P<0.01).
Conclusions
The location (stroma vs epithelia) of each immune cell of interest should be considered in connection to its index calculation because the amount of tumor epithelia may skew the result. Alternatively, the novel, unbiased subsampling developed in this study could be used. High solidity meant fewer infiltrating cells, which, conceivably, may affect both prognostic and predictive utility of immune cells.
Legal entity responsible for the study
The authors.
Funding
Danish Cancer Society.
Disclosure
All authors have declared no conflicts of interest.
Resources from the same session
43P - Machine learning radiomics based on CT to predict response to lenvatinib plus tislelizumab based therapy for unresectable hepatocellular carcinoma
Presenter: Gang Chen
Session: Poster Display session
44P - Machine learning-based prediction of survival in patients with metastatic renal cell carcinoma receiving first-line immunotherapy
Presenter: Ahmed Elgebaly
Session: Poster Display session
45P - Gut microbiome signatures for exploring the correlation between gut microbiome and immune therapy response using machine learning approach
Presenter: Han Li
Session: Poster Display session
Resources:
Abstract
46P - Abnormal gut microbiota may cause PD-1 inhibitor-related cardiotoxicity via suppressing regulatory T cells
Presenter: Zeeshan Afzal
Session: Poster Display session
47P - Correlation of clinical, genetic and transcriptomic traits with PD-L1 positivity in TNBC patients
Presenter: Anita Semertzidou
Session: Poster Display session
48P - The A2AR antagonist inupadenant promotes humoral responses in preclinical models
Presenter: Paola Tieppo
Session: Poster Display session
49P - Highly potent novel armoured IL13Ra2 CAR T cell targeting glioblastoma
Presenter: Maurizio Mangolini
Session: Poster Display session
50P - Phase I trial of P-MUC1C-ALLO1 allogeneic CAR-T cells in advanced epithelial malignancies
Presenter: David Oh
Session: Poster Display session
51P - Unlocking CAR-T cell potential: Lipid metabolites in overcoming exhaustion in ovarian cancer
Presenter: Xiangyu Chang
Session: Poster Display session
Resources:
Abstract
52P - Tumor-targeted cytokine release by genetically-engineered myeloid cells rescues CAR-T activity and engages endogenous T cells against high-grade glioma in mouse models
Presenter: Federico Rossari
Session: Poster Display session