Phenoimager ht automated quantitative pathology imaging system

The validation pipeline and details of the development of the multiplex immunolabeling protocol have been described previously by our group and are shown in supplementary material, Figure S1 and Supplementary materials and methods [15 (link), 16 (link)]. Multiplex immunofluorescence staining was performed using the LabSat® Research platform (Lunaphore Technologies, Tolochenaz, Switzerland), a fully automated tissue‐staining instrument for rapid immunostaining which utilizes a microfluidic technology for the rapid and uniform delivery of reagents to tissue samples [17 (link)]. In brief, each TMA section was subjected to six successive, automated rounds of antibody staining, including pan‐cytokeratin, CD8, CD68, FOXP3, PD‐1, and PD‐L1. Nuclei were counterstained with spectral DAPI (Akoya Biosciences, Marlborough, MA, USA). A full protocol is provided in Supplementary materials and methods. TMAs were scanned using a PhenoImager HT Automated Quantitative Pathology Imaging System (Akoya Biosciences). Image analysis was performed using the inForm software framework (version 2.4.8, Akoya Biosciences), shown in Supplementary materials and methods.

Multiplexed immunofluorescence TMA slides were scanned on a PhenoImager HT Automated Quantitative Pathology Imaging System (Akoya Biosciences). Briefly, a spectral library containing the spectral peaks emitted by each fluorophore from single stained slides was created using inForm software (version 2.4.8, Akoya Biosciences). This spectral library was used for spectral unmixing of the images, allowing color-based identification of the markers of interest. Autofluorescence was determined on an unstained endometrial carcinoma tissue. Each TMA core image was spectrally unmixed and exported as a component TIF image (2656 × 2656 × 7 pixels) using Akoya Biosciences’ inForm software. Component TIF images were then imported into the NaroNet deep learning framework. Due to tissue loss during multiplexed staining, each patient contributed one or two TMA tumor cores.

For IHC labelling of FFPE tissue sections, antigens were unmasked by heat-induced epitope retrieval (HIER) using either 10 mM citrate solution (pH 6.0) and/ or Tris ethylenediaminetetraacetic acid (TE) solution (pH 9.0). Tissue sections were blocked with 5% normal goat serum in phosphate-buffered saline (PBS; blocking buffer) before incubating with primary antibodies either overnight at 4 °C or for 2 h at RT (for primary antibody details, see Supp. Table S1). The resulting antigen–antibody complexes were detected using Alexa Fluor-conjugated secondary antibodies (1/400; AlexaFluor 488, 555, 647—secondary antibodies are listed in Supp. Table S1) and cell nuclei were labelled with DAPI (Invitrogen, UK). If applicable, tissue sections were then counterstained with 0.5% Thioflavin S (ThS; diluted in water; Sigma Aldrich, cat. no. T1892). Sections were mounted in fluorescence mounting media (Agilent, UK, cat. no. S302380-2) before imaging. All slides were imaged on the Akoya PhenoImager HT™ Automated Quantitative Pathology Imaging System (CLS143455) and images were processed using the InForm™ image analysis platform (Akoya Biosciences, US). Quantification analysis of the immunofluorescent signal was performed using HALO® (Indica Labs), a gold standard image analysis platform for quantitative analysis of IHC data. See Supplementary Materials and Methods for further details.