On page 483 of this issue³, Lee and colleagues describe an advance that may pave the way to the use of insulin gene therapy in patients with type I diabetes. Using genetic engineering, the authors have produced an insulin analogue that retains an impressive 20–40% of the activity of native insulin. They inserted the DNA encoding this insulin analogue into a modified virus for delivery into diabetic animals. They then placed the entire DNA construct under the control of the promoter region of the L-type pyruvate kinase gene found in liver cells. This promoter is a regulatory element that normally controls the expression of the L-type pyruvate kinase gene in response to glucose.

The authors injected the resulting virus intravenously into rats and mice, and found that the viral genome became incorporated exclusively in liver cells. The animals were able to secrete the biologically active insulin analogue from their livers into their bloodstreams for up to 8 months. When rats were made diabetic by administration of the -cell toxin streptozotocin, and then treated with the viral construct, they maintained normal blood glucose levels throughout the entire 8-month study period. Similar results were obtained³ when the virus was administered to autoimmune diabetic 'NOD' mice⁴.

There are several clever features to this approach that contributed to the authors' success. The native insulin gene encodes proinsulin, which is a protein consisting of two chains, the A and B chains, linked by a 35-amino-acid sequence, the 'C peptide'. To produce insulin, proinsulin must be processed by proteolytic (protein-cleaving) enzymes in the pancreatic -cells to remove the C peptide⁵. This releases the A and B chains, which are then connected and folded to create native insulin⁵.

The proteolytic processing steps are complex, and rely on cleavage enzymes and other machinery⁶. This has been a significant obstacle in the past, because most tissues do not express the proper cleavage enzymes and other factors needed to process proinsulin. Attempts to overcome this problem by inserting alternative cleavage sites in the proinsulin molecule, which can be recognized by liver-specific proteolytic enzymes, have been only partially successful. Lee et al.³ have addressed this problem by replacing the 35 C-peptide residues with a short, seven-amino-acid peptide, avoiding the need for proteolytic processing. Whereas proinsulin has only 1–2% of native insulin's biological activity, Lee et al .'s single-chain insulin analogue was 20–40% as active as native insulin.

Normally, glucose is the main molecule that induces the secretion of insulin from the pancreas. An ideal gene-therapy system would require insulin secretion to be similarly responsive to changes in blood glucose levels. Lee et al .³ achieved this by placing the gene encoding the single-chain insulin analogue under the control of the glucose-responsive L-type pyruvate kinase promoter. Surprisingly, this seems to have worked reasonably well. Ready-made insulin is normally stored in pancreatic -cells and is secreted by a process of stimulated 'exocytosis', which allows rapid responses to molecules such as glucose. With the viral construct, changes in glucose levels modulate expression of the single-chain-insulin gene. By definition, this is a sluggish process compared with stimulated exocytosis, and one would expect that secretion of the insulin analogue in response to glucose would be substantially delayed. This was in fact seen in Lee et al.'s study. Remarkably, however, the changes in blood glucose levels were not that different between healthy control and virus-treated animals.

The virus-treated animals were able to maintain reasonably normal fasting (between-meal) blood glucose levels over the 8-month study period. Although important, this degree of 'normoglycaemia' might reflect unregulated secretion of the single-chain insulin. It is well known that basal levels of insulin secretion have a major impact on fasting glucose levels⁷. However, constant basal insulin levels are inadequate to control glucose concentrations after eating⁸, and some method for timing insulin delivery to glucose ingestion is needed. This is why it is interesting that the virus-treated animals controlled blood glucose levels so well after eating, despite the delayed kinetics of insulin secretion.

But rodents are quite different from humans with respect to maintaining glucose levels, and extending these results to human physiology may prove a challenge. Rodent livers have much higher basal rates of glucose production than human livers^{9, 10}. So small effects of insulin on the liver may be able to control post-eating glucose levels in rodents, but may be less effective in humans. In addition, given that the viral construct integrated directly into the chromosomes of the liver cells, these cells may be exposed to abnormally high local levels of insulin, which are not reflected in the measurements of single-chain-insulin concentrations in the blood. One would also expect that the delayed and prolonged insulin secretion in the virus-treated animals would lead to hypoglycaemia (abnormally low glucose levels) several hours after glucose ingestion. Indeed, post-digestion blood glucose levels were lower in the virus-treated animals than in controls.

Whether Lee et al.'s approach³ can be applied to humans remains to be determined. One of the biggest obstacles to anti-diabetes therapy is the hypoglycaemia that often accompanies efforts to maintain tight glucose control. Administration of too much insulin through an irreversible gene-delivery system is of concern. And an individual's overall insulin secretion changes with diet, body weight, age, growth status, and so on. So a gene-therapy approach must be able to modulate insulin delivery in response to changing needs over time. Placing insulin secretion under the control of glucose is nature's way of overcoming this problem, and Lee et al.'s use of a glucose-responsive promoter in the viral vector is a definite step forward. Despite these issues, the paper represents a good example of how basic research can be applied to problems of clinical significance.