Quantification of tumour-infiltrating lymphocyte subsets: a practical immuno- histochemical method

Background A need for efficient quantification of tumour-infiltrating T and B lymphocytes (TIL) arose in designing a case-control study of breast cancer outcomes. Methods Ten adjacent x 40 digital images of 4μ sections immunostained for CD3, CD4, CD8, and CD20 were acquired for each of 16 randomly chosen carcinomas (640 images total). 'Color range' and 'Histogram' tools in Adobe Photoshop 7 were used to count the number of pixels in each image matching a partition of Lab colour representing the diaminobenzidine reaction product. This number was converted to a cell count per square mm calibrated from one, two, three or all 10 available images for each case and antibody. Results Variation in the number of pixels per immunostained cell necessitated individual calibration for each antibody / case combination. Calibration based on two fields containing the most labelled pixels gave a cell count minimally higher (+5.3%) than the count based on 10-field calibration, with 95% confidence limits 14.7 to +25.3%. Median and interquartile range for cell counts per mm2 were 6280 (490-14,200) for CD3; 5160 (1960-20,000) for CD4; 4280 (2997-9126) for CD8; and 1860 (54-10031) for CD20. As TIL density varied ∼100-fold between different antibodies and cases, this accuracy and precision are adequate for the purpose. Conclusion The methodology described offers sufficient accuracy, precision and efficiency to quantify the density of TIL sub-populations in breast cancer using commonly available software, and lends itself to batch processing. Abbreviations DAB Diaminobenzidine TIL Tumour infiltrating lymphocytes Pi Number of labelled pixels in image i (1≤i≤10) ΣPi Number of labelled pixels in a set of 10 images pi Number of labelled pixels in a calibration area in image i pm, pm', pm'' Number of labelled pixels in calibration areas of first three of a set of 10 images ranked by the number of labelled pixels ci Number of cells present in a calibration area (counted) in image i cm, cm’, cm’’ Number of labelled cells in calibration areas of first three of a set of 10 images ranked by the number of labelled pixels Ci Number of cells in image i (counted or estimated) ΣCi, Number of cells in a set of 10 images (estimated) Fn Calibration factor calculated from n images (n = 1,2,3 or 10) RGB Colour space defined by 8 bit red/green/blue colour axes Lab Colour space defined by luminance (L), green/red and yellow/blue colour axes