AIMS: This study was designed to test programmed cell death 1 (PD-1) expression of T cells, the hallmark of T cell exhaustion, in different 'immune-classes' of colorectal carcinoma microenvironments as delineated by unsupervised hierarchical cluster analysis. METHODS AND RESULTS: A tissue microarray was made with punches from the invasive margins of 40 microsatellite-unstable and 34 microsatellite-stable colorectal carcinomas. Immune cells were phenotyped by CD8, granzyme B, CD4, FoxP3, CD68, S-100, PD-1 and programmed cell death ligand 1 (PD-L1) immunohistochemistry; tumour area per tissue spot was quantified by cytokeratin (CK)18 immunohistochemistry. For each tissue spot, intra-epithelial immune cells were counted and densities of the various immune cells were calculated. Unsupervised hierarchical cluster analysis with these data yielded a group of 'anergic/immune-naive' microenvironments (47.3%), a group of 'intermediates' (27.0%) and a group of 'immunoreactives' (25.7%) in which PD-1 expressing T cells were prominent. Sixteen of 19 tissue spots representing immunoreactive microenvironments derived from microsatellite-unstable tumours and three were from microsatellite-stable tumours. Further phenotyping of intra-epithelial T cells by sequential immunohistochemistry showed frequent granzyme B/CD8 co-expression, whereas PD-1/CD8 co-expression was more variable. Using receiver operating curve (ROC) analysis, assignment to immune classes was seen to be feasible with good sensitivity and specificity by CD8 counts only. CONCLUSION: A subset of colorectal carcinoma microenvironments is distinguished from the rest by an immune cell composition suggestive of active host anti-tumour immune defence, but this appears to be antagonized by a brisk undercurrent of T cell exhaustion. This observation may have implications for selecting colorectal carcinoma patients for immune checkpoint therapy.
AIMS: This study was designed to test programmed cell death 1 (PD-1) expression of T cells, the hallmark of T cell exhaustion, in different 'immune-classes' of colorectal carcinoma microenvironments as delineated by unsupervised hierarchical cluster analysis. METHODS AND RESULTS: A tissue microarray was made with punches from the invasive margins of 40 microsatellite-unstable and 34 microsatellite-stable colorectal carcinomas. Immune cells were phenotyped by CD8, granzyme B, CD4, FoxP3, CD68, S-100, PD-1 and programmed cell death ligand 1 (PD-L1) immunohistochemistry; tumour area per tissue spot was quantified by cytokeratin (CK)18 immunohistochemistry. For each tissue spot, intra-epithelial immune cells were counted and densities of the various immune cells were calculated. Unsupervised hierarchical cluster analysis with these data yielded a group of 'anergic/immune-naive' microenvironments (47.3%), a group of 'intermediates' (27.0%) and a group of 'immunoreactives' (25.7%) in which PD-1 expressing T cells were prominent. Sixteen of 19 tissue spots representing immunoreactive microenvironments derived from microsatellite-unstable tumours and three were from microsatellite-stable tumours. Further phenotyping of intra-epithelial T cells by sequential immunohistochemistry showed frequent granzyme B/CD8 co-expression, whereas PD-1/CD8 co-expression was more variable. Using receiver operating curve (ROC) analysis, assignment to immune classes was seen to be feasible with good sensitivity and specificity by CD8 counts only. CONCLUSION: A subset of colorectal carcinoma microenvironments is distinguished from the rest by an immune cell composition suggestive of active host anti-tumour immune defence, but this appears to be antagonized by a brisk undercurrent of T cell exhaustion. This observation may have implications for selecting colorectal carcinomapatients for immune checkpoint therapy.