The Evolution of Insulin Glargine and its Continuing Contribution to Diabetes Care

Prior to the 1920s, the prognosis for people with insulin-requiring diabetes mellitus was very poor, with limited treatment options being available and high resultant morbidity and mortality, particularly in children and young adults [1]. Generally, confirmation of the diagnosis of diabetes in this ‘pre-insulin’ era meant eventual coma and subsequent death, often within 2 years of diagnosis [2]. At this time, physicians had to manage the disease through dietary modification alone, with some affected individuals being restricted to a diet with an almost negligible carbohydrate intake in a bid to control blood glucose levels [3‐5]. In such circumstances, the benefit of such ‘starvation’ diets, involving repeated fasting and prolonged undernourishment, was relatively short-lived, providing only a modest extension of life [4]. Furthermore, there was little or no evidence to support longer-term efficacy benefits of undernourishment therapy, which was accompanied by a risk of infection, inanition, and poor quality of life [4]. However, such an approach would have improved the metabolic status of those with ‘non-insulin-dependent’ diabetes, which was not recognized as a separate entity until the mid-1930s and so this differentiation between diabetes types could not be used in treatment decisions at the time [6]. Indeed, dietary intake continues to be a mainstay of the management of type 2 diabetes mellitus (T2DM), taking on increased importance as obesity has become an increasing problem worldwide.

The magnitude of the discovery of insulin must therefore be viewed in the context of a disease that resulted in rapid deterioration and death. The eventual isolation and purification of insulin in a series of relatively crude experiments in the early 1920s heralded a new dawn for the management of diabetes, providing hope for the first time to all those people suffering from this debilitating disease. The pivotal discovery of insulin at the University of Toronto built on the work of earlier scientists, culminating in the development of an effective pancreatic extract (Fig. 1).

Diabetes has been known about for millennia, with the first known description of symptoms allied to diabetes and suggested treatment written circa 1500 BC in the Ebers papyrus from ancient Egypt [1, 7]. Circa 230 BC, Apollonius of Memphis first used the term ‘diabetes’; however, its cause and the organ responsible for this condition were not elucidated until more than two millennia later (in the late nineteenth century) [1, 7].

In 1889, in the laboratory of Oscar Minkowski and Joseph von Mering, it was observed that a dog developed diabetes, with all of the characteristic symptoms of the disease, after total pancreatectomy [8]. This finally confirmed the central role of the pancreas in the etiopathogenesis of diabetes. Later in 1893, Eduoard Hédon further demonstrated that, in animals, complete removal of the pancreas was required for the development of diabetes [9]. Hédon observed that grafting a small piece of pancreatic tissue under the skin after total pancreatectomy alleviated diabetes, but it promptly returned on removal of the tissue. This was also demonstrated independently by Minkowski and, in 1893, led Gustave-Eduoard Laguesse to suggest that the small clusters of ductless cells within the pancreas—that had been described by Paul Langerhans in his doctoral thesis, and that he named the Islets of Langerhans—could be the source of the substance involved in glucose control [5, 10]. This association between the islet cells and diabetes was confirmed in 1901 by Eugene Opie who connected the degeneration of the islet cells to the appearance of diabetes [1, 11, 12]. Subsequently, numerous attempts were made to isolate the glucose-lowering substance, with varying success, by a number of investigators, including Georg Ludwig Züzler (alcohol and saline extraction) [13], Nicolas Paulesco (ice-cold water) [14], Ernest Lyman Scott (acid and ethanol extraction) [15], Israel Kleiner and Samuel James Meltzer (water and saline dilution) [16], John Rennie and Thomas Fraser (boiling water and weak acetic acid) [17], and CP Kimball and John Murlin (acid and alcohol extraction) [18], among others [1, 5, 19, 20].

In 1907, Züzler removed the pancreas from a dog, extracted the pancreas with alcohol and injected the same dog with this extract, which he called ‘acomatol’. He observed that it reduced the amount of glycosuria and raised the pH of the blood [1]. Soon after, in 1908, he used the extract to revive a subject who was in diabetic coma. However, the treatment produced severe complications, which were ascribed to toxic impurities in the extract. The person later died when the supply of the extract ran out; nevertheless, Züzler continued his attempts to produce a pure extract for a further few years and, in 1911, Hoffman-La Roche assisted Züzler’s creation of an experimental laboratory [5, 13, 19, 21]. On 28 May 1912, acomatol was granted a patent (Patent 1027790) [21].

In 1916, Paulesco injected a diabetic dog with a pancreatic extract, extracted with ice-cold water, and observed that this led to the death of the dog from hypoglycemia, with its blood glucose levels falling from 140 to 26 mg% [1, 5, 14]. In 1921, Paulesco presented papers at meetings of the Romanian Society of Biology on his experiments in dogs (21 April, 19 May, and 23 June), and these results were published in France on 23 July 1921 [14, 21]. These demonstrated that his pancreatic extract, ‘pancreine’, reduced blood sugar, ketones, urea, and urine in both normal and depancreatized dogs [22‐25]. Further details of this work were published in France on 31 August 1921 and, on 10 April 1922, he filed a patent application with the Romanian Government for pancreine [21, 23]. The publication of these results was delayed owing to the First World War, and this also meant that there was a hiatus during which his research was effectively put on hold, delaying his progress. In 1919, Kleiner reported on his work, whereby he extracted freshly ground dog pancreas with salted distilled water; in all 16 reported experiments, the extract caused a temporary decrease in the blood sugar levels of depancreatized dogs [16, 20]. However, these decreases in blood sugar levels were accompanied by mild toxic symptoms—most commonly, elevated temperature.

On 17 May 1921, Banting and Best began working together in Professor JR Macleod’s laboratory at the University of Toronto and, within 6 months, had succeeded in both extracting insulin and demonstrating that their crude extract reduced blood glucose in pancreatectomized dogs [26]. Over the course of the next 2 years (1921–1922), Frederick Banting, Charles Best, and John James Rickard Macleod, with the invaluable assistance of the chemist James Collip, achieved an improved extract of insulin from animal pancreata and successfully administered the extract to individuals with diabetes mellitus [26, 27]. Their original process involved ligating the pancreatic duct, therefore destroying the exocrine pancreas, and isolating the endocrine-producing islet of Langerhans from whole pancreas followed by acid-ethanol extraction [28]. This approach was adopted as Banting thought that this would yield a purer extract, free from trypsin, which would degrade the active principle. However, owing to the labor-intensive surgery needed to ligate the pancreas, production of the extract was slow and other approaches were attempted. Initially, they used secretin-exhausted glands for the extraction, which had been produced by slow injection of secretin over 4 h until the flow of pancreatic fluid through a cannula placed in the pancreatic duct stopped [20]. This was also extremely labor intensive and, on 6 December 1921, they extracted fetal-calf pancreas with slightly acidic 95 % alcohol, and the extract successfully lowered blood glucose levels. Finally, on 11 December 1921, they performed the extraction on whole adult cow pancreas, and this extract reduced a depancreatized dog’s blood sugar from 0.460 to 0.180 % in 3 h [20, 29]. This discovery that insulin could be extracted from whole pancreas was a major step towards its successful use in treating diabetes, as the extract was available from a cheap and readily available source material.

Having demonstrated that their extract reduced blood glucose levels in dogs, they moved on to human trials. The first administration to a person with diabetes, 14-year-old Leonard Thompson, occurred in January 1922 [20, 27, 30]. This resulted in a reduction of blood sugar levels from 0.440 to 0.320 %, as well as a drop in the 24-h excretion of glucose from 91.5 to 84 g [20, 27]. However, no clinical benefit was observed and severe local reactions, including abscesses, were observed. This extract was described as a murky, light-brown liquid, and Collip subsequently provided an improved extract of greater purity, which was tested on 23 January 1922 on Leonard Thompson [20, 27]. Frequent injections over the first 24 h of treatment resulted in immediate improvement, with blood sugar levels dropping from 0.520 to 0.120, and glucose excretion from 71.1 to 8.7 g, and the elimination of ketonuria, accompanied by an associated symptomatic improvement [1, 20, 27, 30]. This purer and more consistent extract did not result in such severe injection-site reactions, highlighting the importance of obtaining as pure an extract as possible. Subsequently, the patent for the production of insulin was given to the University of Toronto by Banting, Best, and Collip.

The discovery of insulin represented a significant moment in medical history, with Banting and Macleod being awarded the 1923 Nobel Prize in Physiology or Medicine for the discovery of insulin. Banting shared his prize money with Best, and Macleod shared his with Collip; however, there is still controversy over the awarding of this Nobel Prize [20, 21]. The availability of insulin meant that people with insulin-requiring diabetes could now survive and successfully manage their disease.

The goal of exogenous insulin therapy is to mimic normal endogenous insulin secretion, which adapts to fasting and prandial conditions. When insulin therapy was first introduced for clinical use, it was available only as a short-acting formulation requiring multiple daily injections. This meant that, from the outset, there was a drive to develop new formulations of insulin and explore different routes of administration to make insulin treatment easier for people to manage and tolerate. Early attempts were made to administer insulin by alternative routes. Enteral administration met with very little success, as insulin is destroyed by enzymes in the stomach and small intestine, and there is no appreciable absorption from the large intestine [52]. However, this approach has seen resurgence, with several phase II trials of oral insulin therapy ongoing. Other routes of administration—such as inhalation—were also deemed to be unsuccessful due to poor bioavailability [52]. Consequently, attention was redirected to protracting the absorption of subcutaneously administered insulin in order to reduce the number of daily injections required. This led to the development of insulin preparations with prolonged effect, described as ‘intermediate-’ and ‘long-acting’ insulin formulations (Fig. 1) [53]. The pharmacokinetics of currently available insulin preparations are summarized in Table 1.

The early attempts at developing protracted forms of insulin therapy made use of additives such as gum arabic solutions, oil suspensions, and lecithin emulsions to delay subcutaneous absorption [52]. Attempts were also made to prolong the action of insulin by administering the insulin solution with a vasoconstrictor, such as pituitrin or epinephrine; however, all of these met with little success [52]. The next generation involved combining neutral suspensions of insulin with zinc ions and/or highly basic proteins such as protamines [52, 54]. In the 1930s, globin and surfen insulin were developed as potential prolonged-effect insulin preparations, the latter being formulated using a synthetic urea as an alternative to protamine [54‐57]. At a similar time, protamine zinc insulin was developed [54, 58, 59]. This was a preparation of insulin with excess protamine and a small amount of zinc, which prolonged the hypoglycemic effect of the insulin beyond 24 h [54, 58]. Despite this protracted period of action, the use of protamine zinc insulin was limited by a greater risk of hypoglycemia, as well as a slow onset of action, which necessitated the addition of soluble insulin for immediate action [54, 60‐62]. Unfortunately, the admixture of protamine zinc insulin and soluble insulin was unstable and the two types of insulin had to be given as two separate injections [63]. In 1946, neutral protamine Hagedorn (NPH) insulin, an intermediate-acting insulin developed by Hans Christian Hagedorn, became available [64]. This is a stable ‘protamine zinc insulin’ modification that combines insulin and protamine in ‘isophane’ proportions (i.e. no excess of insulin or protamine) at neutral pH in the presence of a small amount of zinc and phenol or phenol derivatives [65]. This insulin preparation has continued to be used as a ‘basal’ insulin up to the present time, recommended as a once- or twice-daily insulin, used either alone or in combination with a soluble insulin, as required. The timeline for the development of long-acting insulin preparations is shown in Fig. 2.

Early protracted animal insulins developed by Hallas-Møller and Schlichtkrull capitalized on the varying solubilities of their components (i.e. porcine and/or bovine insulins) at physiological pH. The lente family of insulins (semilente, lente, and ultralente) were created by complexing neutral insulin suspensions with small amounts of zinc ions, in the absence of any added foreign proteins or synthetic compounds [54, 66]. This provided a spectrum of time–action characteristics. The original lente insulin, which had an intermediate timing of action similar to that of NPH insulin, comprised a 30:70 mixture of amorphous porcine insulin and crystalline bovine insulin particles [67, 68]. Bovine ultralente insulin formed fairly large crystals (30 μm) that remained in a subcutaneous depot for a number of days, resulting in a duration of action similar to that of protamine zinc insulin, enabling once-daily administration [54].

The remainder of this review discusses insulin analog preparations that are designed to possess a protracted action, with a focus on insulin glargine, which was the first ‘long-acting’ insulin and was approved for clinical use in 2000.

The main role of basal insulin secretion is to limit hepatic glucose production and lipolysis in the fasting state, particularly overnight, without impairing glucose availability for brain function [69]. However, older basal insulin preparations, e.g. NPH and lente insulins, are acknowledged to be associated with a number of limitations, such as variable absorption with notable inter- and intra-individual variation, and discernible peak plasma concentrations after subcutaneous injection, thus increasing the risk of hypoglycemia (in particular, nocturnal hypoglycemia). Therefore, individuals treated with NPH insulin before the evening meal or before bed may be at an increased risk of fasting hyperglycemia. In addition, due its activity of less than 24 h duration, a second dose in the morning is often required. For example, even the longest-acting preparation, human ultralente, with a peak insulin level at 10–14 h post-injection, did not always provide adequate basal coverage with once-daily administration at the lower dose levels [51, 58].

Therefore, in an attempt to avoid the shortcomings of conventional basal insulin therapies, long-acting basal insulin analogs were developed. To date, there have been two main protraction strategies used: (1) modification of the insulin molecule to achieve a low solubility at physiological pH, e.g. insulin glargine; (2) the addition of a fatty-acid chain of variable length to the insulin molecule, which can bind to albumin, forming a circulating depot from which the insulin analog is slowly released, e.g. the insulins detemir and degludec. More recently, a third strategy is being explored that involves the pegylation of insulin, e.g. LY2605541 (insulin peglispro), which is currently undergoing extensive clinical evaluation [70, 71].

The elucidation of the chemical structure of animal insulin by Frederick Sanger and his group [72, 73] and, subsequently, human insulin by Nicol and Smith [35], and the determination of its three-dimensional structure by means of X-ray crystallography by Dorothy Hodgkin and colleagues [74], as well as by the Chinese Insulin Group, helped to reveal the relationship between proinsulin and insulin and the spatial arrangement of the insulin molecules within the hexamers (Fig. 3). These discoveries paved the way for the synthesis of insulin and the eventual development of new forms of rapid- and protracted-acting insulin preparations based on alterations to the structure of the insulin molecule itself (insulin analogs).

The human insulin molecule is a polypeptide with a molecular mass of 5,808 Daltons, comprising an A and a B chain connected by two disulphide bridges (Fig. 4a). By changing the amino-acid sequence in such a way that it does not prevent the interaction with either the insulin receptor or insulin-like growth-factor receptor (i.e. protein engineering), the ‘absorption kinetics’ of the insulin molecule can be altered [75]. This process has been widely and successfully used in the creation of short-acting analogs, working on the principle that hexamer stability in the subcutaneous depot could be decreased by alterations to structure or charge, leading to an increased dissociation rate of the hexamers into dimers and monomers at the site of injection, thereby enhancing the absorption of insulin into the systemic circulation [76].

During the 1980s, initial attempts at creating long-acting insulin analogs involved the addition of positive charges to the insulin molecule, either by removing carboxylates (Glu, Asp), or by the introduction of lysine or arginine using single-chain insulin precursors [54, 77‐80]. Early efforts by Novo Nordisk involved changing Glu^B27 to arginine and replacing the terminal carboxylate of the B chain by an amide (Thr^B30–NH₂) [77‐79]; further structural modifications created NovoSol Basal, a Gly^A21Arg^B27Thr^B30 insulin amide [81]. Although NovoSol Basal achieved prolonged absorption compared with ultralente, it required double the dose for comparable glycemic control. NovoSol Basal also had low intra-individual variability, but high inter-individual variability. This agent failed in clinical testing in 1989, and this was thought to be due to subcutaneous crystal formation and degradation of the drug in the subcutaneous depot by significant macrophage infiltration, leading to reduced bioavailability [75].

A di-arginine (Arg^B31Arg^B32) preparation (Fig. 4b) [82, 83] was obtained by trypsin cleavage of the biosynthetic precursor, proinsulin [84, 85]. As these two arginines are present in proinsulin, linking the B chain to C peptide, and are cleaved off during its metabolism, retaining them was seen as a logical approach to the development of a long-acting insulin [86]. Patents on this structural innovation were filed in 1983 and 1984 by Hoechst AG, a predecessor company of Sanofi. Initially, the simple concept behind these structural modifications was that they would cause a shift in the isoelectric point of the insulin analog from 5.4 towards a neutral pH, consequently lowering solubility at physiological pH. The injection of such a preparation would then result in amorphous precipitation and possibly crystallization in the subcutaneous tissue, leading to delayed absorption into the circulation [75]. Early results confirmed the efficacy of Arg^B31Arg^B32 insulin in rabbit models; although, in canine models after subcutaneous injection, this effect did not show any benefit over NPH insulin (Fig. 5a) and, consequently, development of this formulation was discontinued.

Subsequent investigation, in animal models, of subcutaneous injection of acidic solutions with varying zinc concentrations revealed that the lowest total potency of 20 µg/mL was the ideal concentration for Arg^B31Arg^B32 insulin crystallization in vitro, and may indicate a change in the morphology of the subcutaneous precipitate from amorphous to crystalline (Fig. 5b).

Findings from studies of long-acting insulin analogs demonstrate that, even in cases such as NovoSol Basal and Arg^B31Arg^B32 insulin, which have similar solubility profiles, variations in chemical structure can produce markedly different pharmacokinetic profiles and pharmacodynamic outcomes in vivo. This is highlighted by a study of NovoSol Basal in dogs, which reported lower total blood glucose-lowering properties than for insulin glargine (Fig. 5c). In addition, the failure of analogs such as Arg^B0 to exhibit prolonged glucose-lowering activity in spite of an increased isoelectric point suggested that merely increasing the isoelectric point was too simple a concept to achieve prolonged activity; rather, the impact of the three-dimensional structure seemed to play a major role.

It was at this point that structural biologists entered the stage. Initially, attention was focused on the role of phenol in stabilization (dimer-to-dimer interaction) of the insulin hexamer and monoclinic insulin crystals [87, 88]. Phenol or its derivatives were introduced as a bacteriostatic agent and preservative in insulin preparations; however, it was subsequently realized that phenol was critical in stabilizing the dimer–dimer interactions within the hexamer, thereby protracting the action of insulin [76, 88, 89]. Subsequent crystallographic analyses of the long-acting insulins focused on the contact areas between insulin hexamers in phenol-containing monoclinic crystals. It could be shown that the addition of extra arginine residues at the C-terminal of the B-chain leads to an increase in the number of polar interactions between hexamers in the crystals, concomitant with a higher packing density of the crystals and a reduced water content. Thus, phenol-containing, monoclinic crystals of human insulin comprise 49.8 % solvent (Fig. 6a). Attachment of two arginine residues at the C-terminus of the B chain, as in Arg^B31Arg^B32 insulin, introduces many additional hydrogen-bonding and salt-bridge interactions between neighboring hexamers, leading to a shrinkage of the unit cell of the monoclinic crystals with a lower water content (43 %), and a higher packing density (Table 2; Fig. 6b). The addition of a single arginine residue at the C-terminus of the B chain (Arg^B31 insulin) resulted in a water content and packing density between that seen with human insulin and Arg^B31Arg^B32 insulin (Table 2; Fig. 6c). The activity of Arg^B31 insulin is protracted by about 2 h compared with human insulin. A correlation between crystal packing density and duration of activity was observed, suggesting that stable crystal formation at physiological pH was important for the protraction of time–action. In the case of Arg^B31Arg^B32 insulin, pre-formed crystals with their tight packing are subject to very slow solubilization following subcutaneous injection, and this may ultimately explain the loss of activity following injection of solutions with high crystallization tendencies. Thus, although low solubility at physiological pH is necessary for prolonged duration of insulin action, it is not sufficient alone: attention must also be paid to the degree of inter-hexamer interaction and crystal stability against dissolution. It is also of interest that even uncharged residues that are capable of making extra interactions and/or reducing the solubility of the insulin derivative (such as phenylalanine in both positions B31 and B32) were found to lead to a prolonged activity profile; thus, the effect did not only depend on a shift of the isoelectric point. Even in the case of Phe^B31Phe^B32 insulin, the correlation between reduced water content of the crystals (46.9 %) and the duration of activity was found to hold true.