CD3, CD68, and glucagon stainingThe coimmunostaining for CD3, CD68, and glucagon was performed on the automated platform Discovery XT-VENTANA. After deparaffinization and antigen retrieval using CC1 solution, slides were incubated 1 hour at 37°C with a mix of anti-CD3 (A0452, Dako) and anti-CD68 (M0876, Dako) antibodies, before incubating anti-mouse (FP-SC 4110, Interchim) and anti-rabbit (FP-SB 5110, Interchim) secondary antibodies for 45 minutes at 37°C. Anti–glucagon-FITC antibody (BS-3796-A488, Interchim) was subsequently incubated for 45 minutes at 37°C. Slides were digitalized with the slide scanner NANOZOOMER 2.0RS/C10730-12 (Hamamatsu; objective ×20, resolution 0.46 μm/pixel). Image analysis was performed on the whole pancreas slide by using an algorithm from HALO platform combined with a tissue classifier (Indica Labs). Segmentation between exocrine/endocrine pancreas and other tissue types (connective tissues) were processed using Tissue Classifier based on color, texture, and contextual features. Islet regions were identified as a contiguous area of glucagon. CD3 and CD68 cells were detected according to thresholds of intensity within the endocrine and exocrine pancreas. The algorithm calculates the total number of cells within each pancreas part, the number of CD3+ cells, the number of CD68+ cells, and the areas of endocrine and exocrine parts of the pancreas. Quantification was performed under blinded conditions, with anonymized slides from control and T1D donors.

The HERV-W-Env IHC was performed on the automated platform Benchmark (Ventana, Roche) with the detection kit UltraView DAB (brown chromogen), without pretreatment. GN_mAb_Env03 monoclonal antibody was developed by GeNeuro and has already been validated in several publications (24, 25, 30). GN_mAb_Env03 monoclonal antibody was used at a concentration of 5 μg/ml for the 77 slides. Eight additional slides (4 T1D, 4 non-T1D) were used as controls, with mouse IgG2a isotype. Counterstaining was applied with hematoxylin II and bluing reagent. Slides were digitized with the slide scanner (Hamamatsu), objective ×20, and quantification was made using Indica Labs HALO platform. An algorithm was designed based on pattern recognition that discriminates pancreas tissue (analyzed areas) from fatty inclusions, vasculo-nervous structures, and surrounding connective tissue (excluded areas). Image analysis based on red, green, and blue (RGB) spectra was used to detect brown staining (DAB) within the positive areas (pancreas). The algorithm was designed to allow the detection of specific brown staining according to a threshold of intensity, and nonspecific edge staining of sections was not taken into account. It calculates pancreas area (mm²), staining area (mm²), and percentage (% stained area/pancreas area). Quantification was performed under blinded conditions, with anonymized slides from control and T1D donors.