Tumor Heterogeneity Confounds Lymphocyte Metrics in Diagnostic Lung Cancer Biopsies

Abstract

Context.— The immune microenvironment is involved in fundamental aspects of tumorigenesis, and immune scores are now being developed for clinical diagnostics. Objective.— To evaluate how well small diagnostic biopsies and tissue microarrays (TMAs) reflect immune cell infiltration compared to the whole tumor slide, in tissue from patients with non–small cell lung cancer. Design.— A TMA was constructed comprising tissue from surgical resection specimens of 58 patients with non–small cell lung cancer, with available preoperative biopsy material. Whole sections, biopsies, and TMA were stained for the pan-T lymphocyte marker CD3 to determine densities of tumor-infiltrating lymphocytes. Immune cell infiltration was assessed semiquantitatively as well as objectively with a microscopic grid count. For 19 of the cases, RNA sequencing data were available. Results.— The semiquantitative comparison of immune cell infiltration between the whole section and the biopsy displayed fair agreement (intraclass correlation coefficient [ICC], 0.29; P = .01; CI, 0.03–0.51). In contrast, the TMA showed substantial agreement compared with the whole slide (ICC, 0.64; P < .001; CI, 0.39–0.79). The grid-based method did not enhance the agreement between the different tissue materials. The comparison of CD3 RNA sequencing data with CD3 cell annotations confirmed the poor representativity of biopsies as well as the stronger correlation for the TMA cores. Conclusions.— Although overall lymphocyte infiltration is relatively well represented on TMAs, the representativity in diagnostic lung cancer biopsies is poor. This finding challenges the concept of using biopsies to establish immune scores as prognostic or predictive biomarkers for diagnostic applications.