All imaging was performed using the Vectra 3.0.5 Automated Quantitative Pathology Imaging system (Perkin Elmer). 20X multispectral images covering the entire tumour for each tissue were acquired using the DAPI, FITC, Cy3, Texas Red and Cy5 fluorescent channels. Spectral unmixing of multispectral images was performed in inForm v2.4.2 (Akoya Biosciences) based on signals acquired from single colour controls.
Multispectral images were exported for image analysis in HALO v2.3 (Akoya Biosciences) and stitched together to create a single high-resolution multispectral image for each tumour, upon which image analysis was performed. An algorithm was developed to detect individual cells based on nuclear DAPI staining. Positivity thresholds were set for each marker based on staining intensity. Data for each cell, including marker expression, was exported for secondary analysis in TIBCO Spotfire v7.8. Cell phenotyping was performed based on cell marker expression as outlined inSupplement Table 5 , and total cell counts as well as cell counts for each phenotype were exported from TIBCO Spotfire.
Multispectral images were exported for image analysis in HALO v2.3 (Akoya Biosciences) and stitched together to create a single high-resolution multispectral image for each tumour, upon which image analysis was performed. An algorithm was developed to detect individual cells based on nuclear DAPI staining. Positivity thresholds were set for each marker based on staining intensity. Data for each cell, including marker expression, was exported for secondary analysis in TIBCO Spotfire v7.8. Cell phenotyping was performed based on cell marker expression as outlined in