Highly purified regular porcine insulin was given by portable insulin pumps through indwelling vena cavai catheters to 17 (13 normal, and 4 pancreatectomized) dogs initially weighing 15 ± 2 kg at rates ranging from 2 to 10 mU/min (total 17–250 mg) over time periods ranging from 37 to 252 days. During the course of the study, many of the animals lost weight and became anemic. Since these conditions persisted and weight loss progressed even after cessation of insulin infusion, as many of the dogs as possible (15 of 17) were autopsied for microscopic studies. Large amounts of amyloid were demonstrated in the liver, kidney, spleen, and/or pancreas in 55% (6/11) of normal, and in 75% (3/4) of pancreatectomized dogs. The amyloid deposits were Congo red positive, exhibited classical apple green fluorescence under polarized light, and possessed the characteristic ultrastructural features of amyloid. Massive deposits of amyloid were observed in animals receiving as little as 17 mg of insulin over a time span of 52 days. In those animals with hepatic amyloid, marked hepatomegaly was present (i.e., 1200 ± 250, X ± SD, versus 300 ± 25 g for normal animals) and preterminal serum alkaline phosphatase levels were markedly elevated (434 ± 285 versus 30 ± 14 IU/L for animals without hepatic amyloid). The magnitude of the hepatic amyloid deposits precludes the possibility that they represent insulin aggregates or insulin-derived products per se. No evidence of amyloid was present in any of the tissue biopsy specimens obtained prior to insulin infusion.
Moreover, the possibility that this represents an immune response to the injected porcine insulin has to be viewed in light of the fact that the amino acid sequences of dog and porcine insulins are identical. It is of particular interest that the affinity of the amyloid deposits for Congo red stain was totally abolished by prior permanganate treatment, suggesting that the amyloid was derived from serum amyloid A protein rather than from immunoglobulin light chains or insulin aggregates per se. Further evidence that the protein was of the AA-type came from the initial biochemical characterization. Gel filtration on Sephadex G100 in 6 M guanidine hydrochloride identified two small molecular weight peaks of about 13,000 and 25,000 daltons, both of which inhibited the radioimmunoassay for human AA protein.
The consistent finding that the insulin solutions recovered from the pump reservoirs were in an aggregated state suggests that under the conditions of these experiments insulin was physically altered in such a way as to render it amyloidogenic when administered to the dogs. Filters to trap such aggregates were not present. The relative importance of the route of administration, of the physicochemical state of the insulin, and of the status of the recipient's immune system in initiating amyloid deposition remain to be determined.