The disease Diabetes Mellitus was first described in an Egyptian papyrus, discovered by Ebers in the tomb of Thebes in Egypt in 1862, which is said to have been written between 3000 and 1500 BC. The first use of the term 'Diabetes Mellitus' is accredited to Aretaeus of Cappadocia and Apolonius of Memphis in the second century AD. 'Diabetes' stems from the Greek word for 'pipe-like' because nutrients begin to pass through the system rather than being utilised. 'Mellitus' is Latin for 'honey' or 'sweet', to distinguish the disease from 'Diabetes Insipidus', which is a pituitary disorder in which large volumes of sugar-free urine are passed.
In the middle of the nineteenth century, evidence from autopsies started to suggest a link between the pancreas and Diabetes Mellitus. Diabetics were sometimes seen to have pancreas damage, and patients with damaged pancreases almost always had diabetes. In 1869 Langerhans discovered the existence of two systems of cells in the pancreas: the acinar cells, secreting the pancreatic juice into the digestive system, and islets floating between the acini, with some as yet unknown function. In 1889 Minkowski and Von Mering depancreatised a dog, causing a state of polyuria indistinguishable from diabetes. This was the first direct evidence of the link between diabetes and the pancreas. They also showed that it was not the absence of the pancreatic juice that caused diabetes by studying the effect of ligating the pancreatic ducts rather than removing the whole pancreas. In most cases this caused minor digestive problems, but never diabetes.
The Frenchman Hédon proved in 1893 that a total pancreatectomy was necessary to cause Diabetes Mellitus. He blocked the flow of pancreatic juice, removed most of the pancreas, and grafted the small pancreatic remains just under the skin of his test subjects, for easy removal at a later stage of the experiment. The blood supply to this piece was left as normal as possible. At this stage no diabetes was established. After removing the graft, Diabetes Mellitus could be diagnosed immediately.
During the 1890s it was discovered that several diseases could be treated by feeding patients extracts of thyroid. In analogy, it was tried extensively, probably by more that 400 researchers worldwide [Bliss, 1982], to treat diabetic patients by feeding them pancreatic extracts, without success. Mildly positive effects could never be reproduced by others. Usually the toxic side-effects were far worse than the positive effects, although sometimes the side-effects were seen as positive, e.g. kidney failure could change the urine in such a way that diabetes was no longer diagnosed.
In 1901 Opie showed a direct link between Diabetes Mellitus and damage to the islets of Langerhans, generating a wide belief in an internal secretion in the islets, responsible for the prevention of diabetes. In view of the failed pancreas therapy experiments, scientists suggested that the exocrine pancreatic secretion might destroy the active component of the internal secretion. Two distinct types of pancreas seemed obvious choices for attempts to create pure internal secretion: that from foetuses, in which islets develop before acinar cells, and that from certain types of fish, which have anatomically distinct islet and acinar parts. There are no records to prove the first was tried in the early years, the second was tried between 1902 and 1904, but was not promising. Several scientists all over the world attempted to isolate and purify substances from the pancreas that were supposed to cure Diabetes Mellitus. Meanwhile, several diabetologists kept believing that a diet was the only good way to treat diabetics.
Until the 1910s opium was the only widely used medicine in the treatment of Diabetes Mellitus. However, this could only dull the patients' despair, but did nothing to cure or treat. Further treatment consisted of more or less trendy diets. In the late 1850s Piorry advised the use of extra sugar, to compensate for the loss of sugar into the urine. This 'eating a lot to compensate' was practised until the early 1900s. In Paris under German siege, in 1870, Bouchardat noticed that rationing of food caused the disappearance of glycosuria in diabetic patients, while exercise also seemed to have a positive effect. The idea settled that maybe the body should be put under as little metabolic strain as possible by limited eating.
At the time of the earliest tests of pancreatic extracts in the treatment of Diabetes Mellitus, America had two leading diabetologists who did not believe in pancreas therapy. They were Allen and Joslin. They both practised 'starvation treatment' where the patients are undernourished for a certain amount of time. They argued that apart from the carbohydrate metabolism the protein and fat metabolisms in diabetic patients were also affected. By cutting down on food until the patient's body was relieved of all metabolic strain, and then slowly building up again until a reasonably healthy diet was achieved, many diabetics could live years longer. Some patients, however, did not even tolerate the minimum amount of food (the 'living diet'), and succumbed quickly.
As early as the 1890s Paulesco, a Romanian scientist in Paris, developed an interest in the internal secretion of the pancreas. He regained his interest in 1916, when he did his first experiments with extracts in Bucharest. The First World War and Austrian occupation prevented serious experiments until 1919, and he published some successes with his 'pancréine' in 1920 and 1921.
In the early 1900s Zuelzer, in Berlin, developed a pancreatic extract he named 'acomatol' with which he managed to bring back a 50 year old patient from a coma in 1906. This extract, produced for the Schering company, was probably very contaminated and produced many side-effects. It was tested in Minkowski's clinic in 1909, where it was concluded that the positive and negative effects were due to the same component. This caused Zuelzer's funding to be withdrawn, and he stopped publishing. He persisted with his experiments, however, and produced a new extract for Hoffman-La Roche, for which he never published the extraction methods. With hindsight, this extract was probably much better than the first, if the convulsions that were reported were hypoglycaemic reactions (signs of a low blood-sugar level).
Many other researchers worked on the extraction of the internal secretion of the pancreas, both in Great Britain and the United States of America. Some, like Dewitt and Scott, used pancreatic duct-ligation to atrophy the acinar cells. This would remove the juice that would, as they believed, have destroyed the internal secretion. Others used alcohol (like Zuelzer) to remove the pancreatic juice. Among these were Knowlton and Starling, and Murlin and Kramer. Most of their results were irreproducible. Two Americans, Kleiner and Meltzer, did produce promising results of a decline in blood-sugar in depancreatised dogs, caused by administering extract of the dogs' own pancreases. Most of the control experiments were satisfactory. This work stopped abruptly in 1919 because Kleiner left the laboratory [Bliss, 1982].
There were several meetings among those scientists to discuss the prospects of the research. Especially the intervention of a Scotsman, Macleod, who worked in Ohio, and who had been in contact with the researchers in Britain and seen their disappointing results, caused some of the above to stop their experiments.
On October 31, 1920, a fairly inexperienced general surgeon in London near Toronto (Canada), Banting, read an article on pancreatic duct-blockage. This gave him the idea of duct-ligation in order to isolate the internal secretion of the pancreas. This, as seen above, had been tried before, but the article did not mention it and Banting did not know. This is what he wrote in his note-book [Bliss, 1982]:
"Diabetus
Ligate pancreatic ducts of dog. Keep dogs alive till acini degener-
ate leaving Islets.
Try to isolate the internal secretion of these to relieve glycosurea"
On November 8 he managed to arrange a meeting with Macleod, who worked in Toronto at that time. Although Macleod had discouraged several scientists in the field of pancreatic extracts, Banting's enthusiasm caused him to offer a laboratory and some animals over the next summer holiday period, and the help of a student, Best, as research assistant. In March 1921 Banting decided to take Macleod up on his offer, after which the first dogs were depancreatised on May 17. Macleod had advised to use Hédon's method of pancreatectomy leaving a small piece grafted under the skin to be removed later.
The first pancreatic extract was prepared and tested on July 30, with temporary success. The dog died a day later. The pancreas used had been removed seven weeks previously and left to degenerate. Another dog, brought back from a coma, also died within a day. In order to speed up, a full pancreatectomy was tried successfully on August 3, after which Hédon's procedure was not used again. Banting and Best named their preparation 'Isletin' in notes on the experiments on the dog that had the first total pancreatectomy.
Testing urine was sometimes difficult, because the volume of it decreased after injections of extract; in some cases urination ceased altogether. But during the second decade of the twentieth century methods of blood-sugar testing had been developed and improved, which were much more accurate than urine sampling. These tests made diabetes research more efficient and reliable.
Another time-saving idea was tried on August 17: extract of fresh, non-degenerated pancreas. Banting and Best failed to recognise the positive results and persisted with their faulty hypothesis that degeneration of the pancreas was necessary to obtain pure internal secretion. Boiling extract rendered it inactive, exhausting the pancreas' external secretion with secretin was too effective. Extracts prepared with secretin lowered blood-sugar quickly but caused profound shock. More and more control experiments were carried out, e.g. in vitro sugar-burning capacity and checking the activity of mixtures with trypsin. Macleod recognised it was this thorough testing that needed to be extended in order to make it impossible for critics and pessimists to deny the positive effects. The experiments would involve pre-injection blood tests, more frequent blood sampling after injection so as not to miss the effect, and establishing that it was a real blood-sugar lowering effect rather than dilution phenomena caused by the injections of reasonably large amounts of extracts.
In the first half of September 1921 different ways of injecting were tested. Rectal injections had no effect, while subsequent intravenous injection did prove effective. On September 17 injections were given subcutaneously for the first time, but the results were not satisfactory, and Banting and Best decided it was not worth trying again until they had trypsin-free extracts. By the end of September two more respected Toronto doctors/scientists, Starr and Henderson, had become involved in attempts to keep the promising research going by providing lab space and money.
Some time between October and December 1921 Best read a publication by Paulesco from July 1921. The blood-sugar data quoted by Paulesco were so different from the generally accepted values for hyperglycaemia (probably due to different techniques used by Paulesco), that Best was not impressed and chose to ignore the information in the article.
In their first paper, describing work done up to November 10, which was published in February 1922, Banting and Best concluded that, although they had "always observed a distinct improvement in the clinical condition of diabetic dogs after administration of extract of degenerated pancreas", it was still too early for clinical trials. By then, they had started a so-called longevity experiment, keeping a pancreatectomised dog alive for as long as possible.
A visiting biochemist, Collip, became actively involved in designing better experiments. Banting discovered information about foetal pancreases, from which active extracts could be prepared without ligation or degeneration because of the relatively low content of acinar tissue; this meant that extract could now easily be produced in abundance from fresh, whole foetal pancreas. Time had come to try to capture "the active principle". New bacterial filtering methods produced more sterile extracts. Subcutaneous injections became possible, spreading action over a longer period, thus preventing shock. A new blood-sugar test was introduced, probably by Collip.
On November 23, Banting was injected subcutaneously with 1 1/2 cc of Berkefeld filtered extract. The group had become impatient, wanted to get into clinical testing. This extract did not seem to have any harmful effect, but blood-sugar was not measured.
One longevity experiment ended on December 2; the dog died after convulsions, due to anaphylactic shock. In retrospect, it could have been hypoglycaemic shock, since more extract seemed to make it worse. The next longevity experiment started four days later, in which the extraction was performed with alcohol instead of aqueous saline, which was easier to evaporate in order to concentrate the extract. This lead to the idea that the active principle could be extracted from adult pancreases with alcohol too. Adult pancreases were available more cheaply than foetal pancreas. Although Banting and Best had started using alcohol and adult pancreases before Collip joined them, his experience and expertise were much needed assets in the group. He started using rabbits, and discovered that even healthy animals had a blood-sugar lowering response to extract. He also found that the residue and not the filtrate of a final filtering step contained the really powerful active principle. On December 20, 1921, Joe Gilchrist received tested, potent extract by mouth, with no benefit after a day. At that time it still was not firmly established that only injections would work.
Apart from testing urine and blood for sugar, Collip started testing the urine for ketones and measuring the liver's glycogen, which show whether liver function could be restored with the extracts. He also discovered hypoglycaemic shock, a state of apparent toxic reaction which could be relieved by administering glucose solution.
Because Collip was a much more self-sufficient, experienced and thorough researcher, his results were much better received than those produced by Banting and Best. Also, more and more people became involved in the experiments. This made Banting feel he had been overtaken, and that his idea had been taken out of his hands. Relationships within the group deteriorated quickly, and even became violent at times. Because of this animosity, Banting and Best decided to have a first official clinical test on January 11, 1922, on Leonard Thompson. The timing was wrong. It was still too early, and the extract did not perform as well as expected.
Meanwhile, Collip discovered the active principle could be purified to a certain extent by gradually precipitating other protein with increasing amounts of alcohol. Only at around 90% alcohol the active principle would precipitate, leaving it pure enough not to cause abscesses at injection sites. His extract underwent a first clinical test on January 23, also on Leonard Thompson. Two days later the group signed an agreement to work together as a group again rather than trying to compete with each other.
During February more clinical tests were carried out, all with favourable results. At the same time, Paulesco started clinical trials, independently and without knowledge of the work in Toronto. In April, the Latin-rooted name 'insulin' was proposed. The name 'insuline' had been suggested for the hypothetical internal pancreatic secretion twice before, in 1909 and again in 1916, by two independent scientists, and unknown to the Toronto group.
On May 3, 1922, the discovery of insulin was officially announced to the medical world by Macleod.
After the actual isolation of insulin in 1922, it took another six years for Wintersteiner to establish that insulin is a protein. And it was not until 1955 that the primary structure of insulin was elucidated by Sanger and co-workers. An account of this process can be found in the transcription of Sanger's Nobel Prize lecture.
Based on the knowledge of protein chemistry in general and the composition of insulin in particular, by 1943 Sanger started investigating the sequence of amino acids of insulin. Chibnall and his colleagues had shown a high content of free a-amino groups, i.e. a relatively high number of N-terminal residues, one of which had been determined by Jensen and Evans (1935) to be phenylalanine. Until 1952 Sanger believed the molecular weight of insulin to be 12,000. In that year, Harfenist and Craig showed it to be around 6,000, using the method of partial substitution by 1-fluoro-2,4-dinitrobenzene (FDNB), separation of the reaction products and colorimetric analysis of the monosubstituted derivative for the dinitrophenyl group.
For further study of the N-termini, Sanger developed the DNP-method [1945], which was later also used for many other proteins. The result was the discovery of two phenylalanine and two glycine termini, along with non-terminal e-DNP-lysine. The assumed four polypeptide chains were thought to be linked by disulfide bridges due to the relative abundance of cysteines [du Vigneaud et al., 1939]. By oxidation with performic acid [Sanger, 1949], the crosslinks were broken, resulting in two fractions, A and B. Fortuitously, the two types of amino acids that could have confounded this experiment by reacting with performic acid, methionine and tryptophan, are not present in insulin. Fraction A contained the smaller number (around 20) of residues, only 12 unique ones, of which none were basic. Four of them were cysteine. Fraction B had 30 residues, two of which were cysteine. It seemed there was only one type of glycine chain and one type of phenylalanine chain, which was confirmed in 1949. Mild acid hydrolysis of the DNP-derivatives of the fractions made it possible to study the N-terminal sequences, resulting in Phe-Val-Asp-Glu (fraction B) and Gly-Ile-Val-Glu-Glu (fraction A).
This meant there were only two types of chain, and not four different ones, so the 12,000 molecular weight insulin was built up of two identical halves. Or, alternatively, the actual molecular weight was 6,000. During 1950 Sanger and Tuppy managed to sequence the whole of fraction B. Separation of the complex partial hydrolyzate by various means resulted in simpler mixtures of which direct analysis was possible. They ended up with a puzzle of about 45 peptides of varying lengths from which 5 sequences could be deduced:
| Phe-Val-Asp-Glu-His-Leu-CySO3H-Gly (N-terminal sequence) | (1) |
| Gly-Glu-Arg-Gly | (2) |
| Thr-Pro-Lys-Ala | (3) |
| Tyr-Leu-Val-CySO3H-Gly | (4) |
| Ser-His-Leu-Val-Glu-Ala | (5) |
Four amino acids were still missing, and it was impossible to establish how the sequences would be joined together. The use of proteolytic enzymes (pepsin, trypsin and chymotrypsin) [Sanger and Tuppy, 1951] as hydrolytic agents instead of acid, solved this problem by producing different peptides, which also included the missing residues. Determination of the sequence of fraction A was completed by 1953. This was more difficult because of the specific amino acid content and the lower susceptibility to enzymatic hydrolysis. Ionophoresis at around pH 3 was needed to separate the problematic cysteic acid peptides.
By that time, it was firmly established that the molecular weight of insulin was, in fact, 6,000. So the only remaining question, that of the number and nature of the disulfide bridges, was reduced to finding two between chains A and B, and one intrachain bridge in chain A. In order to do this, unoxidised insulin was subjected to hydrolysis to isolate peptides with intact cystines. Acid hydrolysis was not suitable because of disulfide rearrangement reactions, but enzymatic, neutral and alkaline hydrolysis all proved useful. Thus the complete sequence of insulin was deduced (see figure 1.1).
Figure 1.1: Sequence of insulin.
Sanger used bovine insulin
in his studies. Note the specific way of depicting amides. During strong acid
hydrolysis the amide groups of some glutamines and asparagines are also
hydrolysed. The real identity of these residues could be more appropriately
determined by enzymatic hydrolysis [Sanger
et al., 1955].
Human insulin differs from bovine insulin at positions A8, A10 and B30, where it has threonine, isoleucine and threonine, respectively. The molecular weight of human insulin is 5,808 [Hansen and Brange, 1987]. The net charge of insulin is zero at pH 5.5, in good agreement with the isoelectric pH of 5.3-5.35 as originally determined by Wintersteiner and Abramson in 1933.
Insulin had been crystallised for the first time by Abel (1926). Over the years, the crystallisation method for these rhombohedral crystals was standardised. The discovery of Scott (1934) of the need for zinc in the crystals, was a big step forward in this. In 1935 Crowfoot received a first sample of finely crystalline insulin. She recrystallised the material according to Scott's method and took the first X-ray photographs, which were published in the same year [Crowfoot]. They were not the first X-ray photographs of protein crystals ever, for pepsin had been used before [Bernal and Crowfoot, 1934]. The first crystallisations were approached as an organic chemist would, so the crystals were dried with alcohol. Soon, however, the positive influence of mother liquor was an established fact, and X-ray measurements of wet insulin crystals were published [Crowfoot and Riley, 1939].
The interpretation of the X-ray patterns had started with the use of Patterson's ideas on the determination of the components of interatomic distances in crystals. With the simplifications introduced by calculating Harker sections only, the maps showed, in the case of insulin, strong features at 10 and 22 Å interatomic distances. When the first amino acid crystal structure, glycine, was solved in 1939 through the application of the Patterson synthesis, solving protein structures with the same methods still seemed "madness", according to Patterson himself, as mentioned by Crowfoot Hodgkin in 1968. The idea of replacing zinc isomorphously with heavier ions had taken root almost as soon as the first photos had been taken. However, the intensity changes introduced by cadmium were too small to be of any help. Iodination experiments were not followed through, probably partly because of the war, and because of the lack of experience with protein structure analysis in general. Until around 1950 most of the interpretation consisted in building models, almost by guessing and then checking whether they could produce the diffraction patterns observed. All the proposed models exhibited the correct symmetry, namely 32 [Hodgkin and Riley, 1968], but nobody recognised the possibility of calculating the correct size of the insulin molecule and the contents of the asymmetric unit then. The idea of utilising multiple crystal forms in structure determination was described by Crowfoot in 1938. Unfortunately, the many different shapes of crystals observed all turned out to be the same crystal form.
It was not until the primary structure of insulin became available in 1955 that Hodgkin's group in Oxford started devoting most of their time to the problem of solving the three-dimensional structure of insulin. Schlichtkrull's methods of crystallisation and his investigations into the exact amount of zinc in those crystals, along with earlier experiments with cadmium insulin crystallisation according to Scott, paved the way for the production of lead insulin. It proved possible to remove zinc by soaking in EDTA and then replace the zinc with lead. X-ray measurements were taken of the isomorphous series of zinc-free, zinc, cadmium and lead insulin [Hodgkin and Riley, 1968]. Especially the differences between metal-free insulin and lead insulin were large, but the feasibility of solving the structure with those alone, was still in doubt at that time.
However, it was not long at all until enough isomorphous heavy atom derivatives were found to solve the structure to a resolution of 2.8Å. Five derivatives were used for the phase determination:
Anomalous dispersion measurements were also taken. So late on Sunday night, August 3, 1969, insulin was solved (for a facsimile of the original note announcing the solution of the insulin structure, see Dodson, Glusker and Sayre (1981)), and the structure of rhombohedral 2-zinc insulin to a resolution of 2.8Å was published on November 1, 1969 [Adams et al.]. The peptide chain, the disulfides and the aromatic residues were well defined. A few regions, especially on the outside of the molecule, were not completely clear.
After the determination of the primary structure of insulin by Sanger et al. in 1955, several groups in China started the chemical synthesis of insulin in 1958. They succeeded in 1965, after which some pressure was applied to determine the three-dimensional structure. The Cultural Revolution delayed the start until the beginning of 1967 [Tang, 1981]. The structure of insulin to 2.5Å resolution was published in 1971. In 1972, when the Oxford group had extended their calculations to include data to 1.9Å resolution, Hodgkin visited China and it was decided the two groups would carry out further refinement separately and simultaneously. This resulted in publications by the Beijing group of the structure at 1.8Å resolution and at 1.2Å , and at 1.5Å resolution by the Oxford group. Meanwhile, both groups started studying different insulin species, for example insulin shortened at the B chain C-terminus [Peking Insulin Structure Group, 1976], 4-zinc insulin [Bentley et al., 1976], insulin crystallised in cubic form [Dodson et al., 1978] and insulin from hagfish [Cutfield et al., 1979].
The biosynthesis of insulin starts in the nucleus of the B-cells of the pancreas. Most species have only a single insulin gene [Bell et al., 1980]. The DNA contains two intervening sequences, one in the 5' part of the mRNA that remains untranslated, and one almost in the middle of the sequence for the C-peptide region of proinsulin, 179 and 786 base pairs respectively. After transcription the mRNA ends up in the cytoplasm of the cell. It is thought [Steiner, 1983] that translation of the mRNA into the 110 amino acid preproinsulin (species: rat) starts while the ribosome is free in the cytosol. The signal sequence of the preproinsulin anchors the ribosome to the membrane of the rough endoplasmic reticulum (RER), after which the protein is translocated into the ER lumen. The 59 "extra" amino acids in the preproinsulin appear to contain all the information for accurate processing and secretion of the mature hormone. The prepeptide has many hydrophobic side chains and is exactly long enough to span the RER membrane so that the prohormone can start folding as soon as its N-terminus enters the ER lumen. There is evidence that this anchoring function of the prepeptide enhances the efficiency of protein folding. However, the prepeptide will be cleaved off by signal peptidase on the inner surface of the membrane before the whole of the peptide chain has entered the lumen. This is necessary for the correct bridging of the last two cysteines, one of which is the last-but-one in the nascent chain.
The length and not the amino acid content of the connecting peptide in proinsulin (26-35 residues of greatly varying sequence for all known species) appears its most important feature. Its function in folding and disulfide formation, i.e. maintaining the correct spatial separation of B30 and A1, could just as easily be exerted by a much shorter peptide. The reason for its excessive length may be that the prohormone chain needs to span the distance from the interface between the small and large ribosomal subunits (where translation takes place) to the inside of the ER, which is longer than 51 amino acids plus the necessary distance between B30 and A1 in the mature hormone [Steiner, 1983]. This is an example of the 'minimum length hypothesis' [Steiner, 1981]. After folding and disulfide bridging, proinsulin is transferred to the Golgi apparatus. There the proteolysis of the prohormone starts and the mature hormone is concentrated, sorted and packed into secretory granules, ready for extracellular release.
The actions of insulin have been known for quite some time [Steiner, 1977]:
Three other peptide hormones are produced in the islets of Langerhans in the pancreas:
Glucagon antagonises most of insulin's actions, while stimulating insulin secretion. Somatostatin inhibits the three other islet hormones and a range of hormones from different origins. Pancreatic polypeptide inhibits pancreatic secretion altogether [Johnston et al., 1988].
Once in the blood, insulin controls glucose homeostasis by stimulating the uptake of glucose into skeletal muscle and, to a lesser extent, into liver and adipose tissue. In muscle and adipocytes this uptake is mediated by the so-called insulin-sensitive glucose transporter GLUT-4, a process that is not yet understood. Other processes in the regulation of glucose homeostasis are: alterations in glycogen metabolism in muscle and liver and decreased gluconeogenesis in the liver. The enzymes involved in the insulin-regulated processes of glucose metabolism appear to be regulated by (de)phosphorylation of serine and/or threonine residues [Lee and Pilch, 1994].
All known actions of insulin are initiated at the plasma membrane by insulin receptors responding to ligand binding. A schematic view of the insulin receptor can be seen in figure 1.2.
Figure 1.2: Schematic view of the insulin
receptor.
Reproduced from Lee
and Pilch (1994).
The a-subunit of 723 amino acids contains the site or sites for insulin binding. The 620 amino acid b-subunit is built up of three regions: the extracellular, transmembrane and cytosolic domains. Both subunits are glycosylated, resulting in approximate molecular masses of 130 kDa and 95kDa respectively [Lee and Pilch, 1994]. The insulin holoreceptor is composed of two a-subunits and two b-subunits covalently linked by disulfide bridges to form a functional dimeric protein complex. The receptor gene is translated into a single ab-product of which two are linked together via disulfide bridges before being processed into separate a- and b-subunits. The covalent linkage of the two ab-heterodimers is unusual among the receptor/tyrosine kinase family of which the insulin receptor is a member. The insulin holoreceptor can, however, under mild conditions be reduced to ab-heterodimers which are functional monomeric receptor species with a lower ligand binding affinity than the holoreceptor. The exact position of the disulfide bridges both between a- and b-subunits and between ab-heterodimers are not known; different studies are contradictory [Lee and Pilch, 1994].
Ligand-receptor contact occurs within the a-subunit. The exact regions of contact remain incompletely defined despite a lot of research effort. Various truncated versions of the receptor have been studied for insulin binding and it seems that only those with an intact a-subunit are capable of binding the ligand, leading to the conclusion that the entire a-subunit is important in some way for insulin binding. At physiologically relevant hormone levels there seems to be a stoichiometry of one insulin molecule per holoreceptor, with negative cooperativity for binding of a second ligand [Lee and Pilch, 1994]. Thus insulin somehow breaks the symmetry of the receptor upon binding. Lee and Pilch propose a model for insulin binding to the receptor (see figure 1.3a) where at physiological concentrations insulin makes contact with two partial binding sites on separate halves of the receptor, resulting in high-affinity binding. It can be seen from the model that at high concentrations a second insulin molecule could potentially make contact with only one of the partial binding sites, resulting in low-affinity binding (figure 1.3b).
Figure 1.3: Schematic view of insulin binding as proposed by
Lee and Pilch.
a) One insulin molecule per holoreceptor at
physiological concentrations;
b) A second insulin molecule may bind,
less strongly, at high concentrations.
In essence, the model proposed by Lee and Pilch is the same as that proposed by Schäffer (1994), with high-affinity binding of the first insulin molecule to two different partial binding sites on the two a-subunits of the receptor, after which low-affinity binding of a second molecule is possible to only one partial binding site. Schäffer implicates specific insulin residues in the binding to both partial binding sites. The dimer-forming surface of insulin (mainly residues B24, B25, A21 and B12, and potentially some residues which are buried beneath the B chain C-terminus), named the 'classical binding site', is best re-named 'binding site 1', and binds to receptor binding site 1. It appears that the opposite side of the insulin molecule, known to be involved in hexamer formation, forms 'binding site 2'. The two most important residues in this site are leucines A13 and B17.
Following binding of the ligand, the receptor is rapidly autophosphorylated on some or all of seven tyrosine residues in the cytosolic domain of the b-subunit. Exactly how the signal for autophosphorylation is propagated from the site of ligand binding on the a-subunit to the intracellular autophosphorylation sites on the b-subunit is unclear. The most likely explanation seems a series of conformational changes, first between the a-a-subunits, then the transmembrane domains of the b-subunits, followed by the b-b-subunits, after which ATP-binding and autophosphorylation can take place [Lee and Pilch, 1994].
The insulin receptor family primarily regulates nutritional metabolic pathways, whereas all other receptor/tyrosine kinases mainly regulate cell growth and differentiation. The same subdivision exists in that the insulin receptor family has covalent links between the ab-heterodimers, and again in that these receptors do not form direct complexes with substrates and/or effector molecules after autophosphorylation. Instead, the insulin receptor, upon autophosphorylation of at least the tri-tyrosine subdomain, acquires exogenous kinase activity, phosphorylating its principal substrate: insulin receptor substrate 1 (IRS1). IRS1, in turn, binds effector molecules which are responsible for the actual processes of glucose transport and metabolism.
Since 1922, a lot of research has gone into the improvement of insulin therapy, both in terms of improved insulin preparations and ease of use.
Achieving purity was the first challenge. Abel discovered insulin could be crystallised, which became standard procedure in insulin purification only after Scott (1934) established that zinc was needed in order to crystallise insulin in rhombohedral form, a discovery inspired by his observation of zinc in the pancreas. Another crystallisation step reduced allergic reactions [Jorpes, 1949]. Chromatographic techniques started to play a role in the 1960s, leading to the first chromatographically purified insulin, Monocomponent insulin [Schlichtkrull et al., 1970].
The search for new insulin preparations with various desired properties was approached in many different ways. At first it was thought the number of injections needed every day could be reduced by using insulins with a retarded uptake from the injection site, which could be achieved by introducing basic additives as in protamine and isophane insulin [Hagedorn et al., 1936 and Krayenbühl and Rosenberg, 1946]. Addition of compounds like surfen or globin to acid insulin solutions, which produce heavily insoluble complexes upon neutralisation in tissue fluids, and complex formation of zinc with neutral insulin suspensions, had the same effect. However, the need for strict metabolic control in prevention of long-term complications called for the reinstatement of multiple injections, along with the development of rapid-acting insulins for the relief of the glucose-surge just after meal times, and mixtures of rapid-acting and intermediate-to-long-acting insulins. Developments in this field include Rapitard and Actrapid [Schlichtkrull, 1959 and Schlichtkrull et al., 1961].
The duration of action could also be influenced by the physical state and size of the insulin particles. Ultralente [Hallas-Möller, 1956] is an example of a long-acting crystalline insulin preparation. Initially, only bovine insulin was used for crystalline preparations, having a slightly longer action than porcine insulin in crystalline form. Gradually, longer acting porcine preparations became available. Then, in 1979, recombinant DNA techniques made it possible to produce human insulin in Escherichia coli [Goeddel et al., 1979]. Amazingly, this was done without knowledge of the nucleotide sequence of the human insulin gene. Crea et al. (1978) had chemically synthesised two separate genes for chain A and B (77 and 104 base pairs for 21 and 30 amino acids, respectively, plus start and stop codons and restriction site bases) following nucleotide sequences that had been designed from the amino acid sequences. By designing the nucleotide sequences the way they did, there was no need to produce every possible trinucleotide separately. They produced 29 oligonucleotides, made from carefully chosen di-, tri- and tetranucleotide building blocks. Goeddel et al. (1979) describe the actual assembly of the genes, the subsequent construction of the plasmid, and the expression and characterisation of the product. They show that the amino acid content of the product is indistinguishable from that of porcine insulin. It took another year before the actual nucleotide sequence of the human insulin gene was published by Bell and co-workers (1980). Only 21 out of 51 of the codons used by Crea et al. turned out to be the same in the correct DNA-sequence.
Production of human insulin could also be achieved by conversion of porcine insulin [Markussen, 1982] or biosynthesis in Saccharomyces cerevisiae rather than E. coli [Markussen et al., 1986]. Over the years, the production of many therapeutic insulins has involved chemical alteration of the molecule. Nowadays, modified insulins can be synthesised by mutation of the genes used in E. coli or S. cerevisiae, thus facilitating structural and functional studies. But apart from the scientific possibilities opened up by the availability of recombinant insulin, there was a pressing need for a new source of the protein, because the demand for insulin for therapy was outgrowing the supply of slaughter pancreases for the isolation of insulin. Even now [Kott, 1996] the major insulin manufacturers are not able to provide insulin for every area of the world, and cheaper beef and pork insulin is still produced, especially for third world countries.
Because of the varying needs of diabetic patients, insulin has to be available in multidose quantities. This requires the addition of antimicrobial preservatives. Banting and Best used tricresol to that effect, and to this day phenol and derivatives like m-cresol and methylparaben are used throughout the range of therapeutic agents. Other additives include sodium chloride or glycerol as isotonic agents, and certain buffers [Brange, 1987].
Alongside research into better and purer preparations, came the development of techniques of administering them. At the time of Banting and Best it was already established that oral therapy did not have any desired effect, and that subcutaneous or intravenous injections were the only option. Many other routes of absorption were tested, including rectal administration and absorption by mucosae. Success was limited. Even with more modern technology of aerosol powder [Wigley et al., 1971], surfactants [Hirai et al., 1981], liposome-enclosure [Dapergolas and Gregoriadis, 1976] or polymer-crosslinking [Saffran et al., 1986] the desired efficiency and bioreactivity has not been achieved. The subcutaneous implantation of vinyl-ethylene copolymer pellets [Creque et al., 1980 and Brown et al., 1986] or biodegradable insulin-albumin microbeads [Goosen et al., 1983], releasing insulin slowly and constantly over a longer period of time, seems more promising. Most recently, reports have appeared on glucose-responsive insulin release from certain polymeric systems [Shiino et al., 1995 and Valuev, 1995]. However, the only techniques in actual clinical use are based on injections. Hospitals operate systems of continuous infusion, either according to continuously measured glucose concentrations [Albisser et al., 1974 and Pfeiffer et al., 1974] or a pre-programmed schedule [Slama et al., 1974]. These insulin pumps are not yet available to the general public, although additional research is done on implantable pumps [Buchwald et al., 1981]. Today's most patient-friendly portable insulin delivery is the NovoPen, which has reduced injections to just pressing a button. Most diabetics seem to prefer this to other available options [Jefferson et al., 1985 and Walters et al., 1985].
In 1979, the National Diabetes Data Group formally classified Diabetes Mellitus and other categories of glucose intolerance as follows:
Apart from the clinical classes mentioned above, there are two so-called statistical risk classes [NDDG, 1979]:
As described above, human insulin consists of 51 amino acids, divided into two chains, commonly labelled A and B, with 21 and 30 amino acids respectively. The chains are linked by three disulfide bridges, two forming interchain cystines at A7-B7 and A20-B19, and one forming an intrachain cystine at A6-A11. A piece of antiparallel b-sheet is formed upon dimerisation: residues B23 to B28 of one monomer lie antiparallel to the same stretch in the other monomer. There are two very small a-helices in the A chain, and a three turn a-helix running from residues B9 to B19 is found in every insulin structure known so far.
Although insulin is only a small protein hormone with a characteristic three-dimensional structure, the sequence varieties in nature and the biological activities of these different forms in vitro, are quite diverse. This diversity provides a tool for research into more effective and efficient insulins for the treatment of Diabetes Mellitus. Solving modified insulin structures will help understand the differences in structure-function relations of all these insulins. It is also thought that understanding properties (e.g. conformational changes) in a small protein like insulin will contribute to the understanding of proteins in general.