Morphometric analysis Beta cell area (defined as the area of cells stained by insulin) was quantified using computer-assisted morphometric analysis of slides digitally stored in the Aperio System. We obtained the ratio between the insulin stained areas and the total pancreatic section area, i.e. the relative volume of beta cells in patients with diabetes and in normal controls. For each diabetic pancreas we then determined the beta cell index defined as the relative volume divided by the mean relative volume of beta cells of controls multiplied by 100. The beta cell index thus expressed the beta cell area of each diabetic pancreas as a percentage of normal control pancreases. Double immunofluorescent staining for insulin and glucagon was used to determine the percentage of islets devoid of beta cells in pancreases with residual beta cells (i.e. the percentage of insulin-deficient islets). In each section we determined the percentage of insulin-deficient islets (i.e. islets composed of non-beta cell vs the total number of islets). Only islets completely devoid of beta cells were considered insulin-deficient islets.

For each childhood-onset case, the pancreas was processed in three regions (head, body and tail) and fixed overnight in 10% (vol./vol.) neutral buffered formalin (ThermoFisher, Waltham, MA, USA) prior to processing to paraffin blocks. Tissue sections (5 μm) were obtained stained with haematoxylin and eosin, immunohistochemistry and immunofluorescence. Pancreatic tissue sections were stained: (1) with antibodies to insulin or glucagon, and a panleucocyte marker CD45 and/or a T cell marker (CD3; and in a subset of diabetic pancreases with CD4, CD8, CD20 and CD68 antibodies); and (2) with a cocktail of antibodies to endocrine non-beta cell hormones composed of glucagon, pancreatic polypeptide and somatostatin. In diabetic pancreases with insulitis (as defined below), the infiltrating T cells were further characterised with antibodies to CD4 and CD8 (as well antibodies to CD20 and CD68 to detect B lymphocytes and macrophages). Antibody binding was detected with appropriate secondary antibodies to guinea pig, rabbit and mouse immunoglobulins conjugated with alkaline phosphatase and peroxidase (MACH2 Polymer Systems; Biocare, Concord, CA, USA), followed by chromagen development with fast red (Vector, Burlingame, CA, USA) or diaminobenzidine (Vector). A scanner (CS ScanScope; Aperio, Vista, CA, USA) was used to produce whole slide images for both haematoxylin and eosin and immunohistochemistry-stained slides. In a subset of diabetic pancreases (pancreases containing residual beta cells), the pancreas was also double-stained with antibodies to insulin and class I HLA molecules. Immunofluorescence was performed by incubating the tissue sections with antibody to insulin, glucagon and survivin (Abcam, Cambridge, MA, USA), and with appropriate anti-rabbit and anti-guinea pig immunoglobulins conjugated with aminomethylcoumarin acetate, Cy3 or Cy5. The sections were photographed at ×20 magnification using a microscope B651 (Olympus America, Center Valley, PA, USA) connected with a digital imaging system (Image pro plus, version 6.2; Media Cybernetics, Bethesda, MD, USA) with a camera (Pro 150ES; Pixera, San Jose, CA, USA). Additional images were recorded on an epifluorescence microscope (Microphot FXA; Nikon Instruments, Melville, NY, USA) with a monochrome digital camera (Roper Micromax; PerkinElmer, Waltham, MA, USA) and appropriate software (Intelligent Imaging Innovations, Denver, CO, USA). The photos are displayed in pseudocolour. Immunoperoxidase staining of survivin was performed by incubating the sections with antibody to survivin and visualising with Cytomation Envision+System-HRP (DAB; Dako, Carpenteria, CA, USA). Class I HLA immunostaining was performed with mouse monoclonal anti-human HLA-ABC, Clone W6/32 (Dako), directed against a monomorphic epitope on the 45 kDa polypeptide products of the HLA-A, -B and -C loci. Insulin was detected with polyclonal guinea pig anti-swine insulin (Dako). All secondary goat antibodies (Life Technologies, Carlsbad, CA, USA) were highly cross-adsorbed against the host species of the other primary antibodies and conjugated with Alexa 488 (green) or 594 (red). Anti-fade reagent (Prolong Gold) with DAPI (Life Technologies) was applied upon staining and sections were imaged using an epifluorescence microscope (Eclipse 80i; Nikon) equipped with a mercury arc lamp (X-Cite, Mississauga, ON, Canada) and a digital camera (DXM1200C; Nikon). An air objective (×20; Nikon) was used with a 0.75 N.A. Positive and negative controls for HLA-ABC staining included human spleen and isotope-matched primaries, respectively. Cross-reactivity with the insulin primary was also experimentally excluded. In additional experiments, HLA immunofluorescence staining was performed with rabbit polyclonal antibody (Abcam). An irrelevant rabbit polyclonal antibody served as a negative control.