Loss of sympathetic innervation to islets of Langerhans in canine diabetes and pancreatitis is not associated with insulitis

Canine diabetes mellitus (DM) affects 0.6% of the canine population and yet, its etiology is poorly understood. Most affected dogs are diagnosed as adults and are insulin-dependent. We compared pan-leukocyte and sympathetic innervation markers in pancreatic islets of adult dogs with spontaneous DM (sDM), spontaneous pancreatitis (sPanc), both (sDMPanc), toxin-induced DM (iDM) and controls. We found evidence of decreased islet sympathetic innervation but no significant infiltration of islets with leukocytes in all disease groups. We show that loss of sympathetic innervation is ongoing in canine DM and does not necessarily precede it. We further found selective loss of islet-associated beta cells in dogs with sDM and sDMPanc, suggesting that collateral damage from inflammation in the exocrine pancreas is not a likely cause of DM in these dogs. The cause of this selective loss of beta cells needs to be further elucidated but overall, our findings are not supportive of an autoimmune process as a cause of sDM in adult dogs. The loss of sympathetic innervation in sPanc in dogs that do not suffer from DM links the disease in the exocrine pancreas to a pathological process in the endocrine pancreas, suggesting pancreatitis might be a potential precursor to DM.

after DM induction, dogs were treated with various insulin formulations with no other intervention (other than routine healthcare preventatives including anthelmintic medications and standard anti-viral vaccines; no antibiotics were administered). Body weight and general health assessment were performed weekly and throughout the 6 month period prior to sampling. All animals maintained their body weight and ideal body condition score with no clinical evidence of gastrointestinal disease or pancreatic disease other than DM. Dogs were monitored daily and fed a standard canine laboratory chow of 300g/day throughout the course of the study to meet their energy requirement. Three weeks prior to tissue harvesting, dogs were treated with fenofibrate (TricorÒ) at a dose of 10mg/kg orally once daily for 3 weeks, as part of a different study. They were then euthanized and pancreata were collected within 30 minutes.

Tissue processing
Formalin fixed and paraffin embedded tissue sections were routinely deparaffinized in xylene and serial ethanol dilutions, following heat-induced antigen retrieval (antigen retrieval buffer; Dako). Samples were further blocked with normal donkey serum and FcR Blocking Reagent (Miltenyi) and incubated overnight with the appropriate primary antibodies (Table S2).
Two multiplex panels were used: 1) insulin-glucagon and 2) chromogranin A (CgA)-tyrosine hydroxylase (TH)-CD45. Twenty-four hours later, slides were extensively washed and treated with the appropriate set of secondary antibodies (Table S32). Finally, nuclei were labeled with DAPI (ThermoFischer Scientific). All immunohistochemistry and immunofluorescence studies had several layers of negative and positive controls to ensure appropriate interpretation of our findings. Specifically, we used a 'No antibody' control to determine the level of tissue autofluorescence and 'No primary antibody control' to determine the level of non-specific secondary antibody binding. For CD45 staining, canine gut and lymph node tissues were used as positive controls. All pancreatic tissue sections had TH staining within the exocrine pancreas serving as positive control.

Image acquisition
Each group (Controls, sDM, sDMPanc, sPanc and iDM) had optimized image settings applied across the entire group. To mitigate any bias in field selection for TH and CD45 analysis, prior to looking at the 20x stitched image, the Alexa Fluor-488 and Cy3 channels were turned off. DAPI and Alexa Fluor-647 (CgA) remained visible as the 20x stitched image was scanned and acquired. Once the 20x field of view was set, the tissue was divided into 16 quadrants. With only DAPI and CgA visible, a representative islet was selected for high power Z-stach imaging in each quadrant: In a quadrant with several small islets, a small islet would be selected while in a quadrant with mostly large islets, a large islet would be selected. For the purpose of image analysis, we defined islets as a cluster of 3 or more CgA+ cells. If there were no islets available to image, areas with less than 3 CgA+ cells in a cluster were imaged and labeled as 'endocrine cells' instead of 'islets'. Given the small size of some tissue samples, some of the quadrants displayed only empty spaces. To be able to exclude empty spaces from analysis, only 12 of the 16 quadrants per tissue were selected for imaging to set consistency across all groups of tissues.
Only after 12 imaging areas were selected, the TH channel was made visible to avoid cutting off TH signal associated with the pre-selected islets. Subsequently, selected islets in quadrants were imaged at 63x and z-stacked according to tissue thickness.

Image analysis
Quantitative image analysis was performed on 63x images using ImageJ (NIH, Bethesda, MA). There were 4 total channels involved in the analysis: Nuclei (DAPI), TH (Cy3), CD45 (Alexa Flour 488), and CgA (Alexa Flour 647). Macro scripts were applied across all images to convert them into TIFF files, convert the z-stacks into z-projects, automate background noise reduction and remove autofluorescence. Thresholding the TH and CgA signal was performed in the same manner. After a composite image was created using the optimized channel settings, CD45+ cells were identified and counted.