Recent Progress in Hormone Research: Proceedings of the 1970 Laurentian Hormone Conference

By E. B. Astwood

Recent Progress in Hormone Research, Volume 27 covers the proceedings of the 1970 Laurentian Hormone Conference held in Mont Tremblant, Quebec, Canada on August 29-September 4, 1970. The book discusses the x-ray analysis and the structure of insulin; spontaneous hyperglycemia and/or obesity in laboratory rodents; and the biological properties of the growth hormone-like factor from the plerocercoid of Spirometra mansonoides. The text also describes studies on human chorionic gonadotropin; studies on the structure of thyrotropin and its relationship to luteinizing hormone; and ultimobranchial follicles in the thyroid glands of rats and mice. The use of antibodies for characterization of gonadotropins and steroids; the biosynthesis of pregnenolone; and the metabolism and protein binding of sex steroids in target organs are also considered. The book further tackles the regulation of gene expression in Escherichia coli by cyclic AMP; the mechanism of action of ACTH; and the role of vitamin D and its relationship to parathyroid hormone and calcitonin. The text then encompasses the production and secretion of testicular steroids; the factors affecting the secretion of steroids from the transplanted ovary in the sheep; and the pilot gland approach to the study of insulin secretory dynamics. The analysis of the response to ACTH by rat adrenal in a flowing system is also looked into. Biochemists, physiologists, pathologists, endocrinologists, people working in laboratories of cancer research, chemical crystallography, and molecular biophysics will find the book invaluable.

• Language: English
• Publisher: Academic Press
• Release date: Oct 22, 2013
• ISBN: 9781483219486

Book preview

AMERICA

PREFACE

The speakers on the program of the 1970 Laurentian Hormone Conference set a new standard of excellence, and the favorable comments from the membership gladdened the hearts of the Committee on Arrangements. One innovation contributing to this performance was the introduction by the first speaker, Dr. Blundell, of the use of two projectors for slides and two screens. This added greatly to the clarity of exposition in his difficult task of depicting the tertiary structure of insulin—the overall effect was most pleasing to the audience. This manner of presenting his material was Dr. Blundell’s third choice; he would have preferred to use fully stereoscopic projection or, failing this, three screens, but suitable equipment was not available. Other speakers were quick to recognize how the new double projection could be used to advantage in their presentations and promptly reshuffled their slides accordingly.

The article on growth in mammals induced by a tapeworm was enhanced by a presentation shared by a classical helminthologist and an endocrinologist, and this glimpse at comparative physiology was extended by the paper dealing with diabetes in a variety of animals including some exotic ones. And so it went, with novelties and new scientific breakthroughs interspersed throughout the week.

The meeting was held as usual at Mont Tremblant Lodge, Mont Tremblant, Quebec, Canada, August 29 to September 4. Still further improvements in the amenities of the Lodge and the delicious food have subdued the voices of those of the membership favoring moving the meeting about from place to place.

Any favorable comments that might be made about the presentations could equally well be applied to the manuscripts submitted for this volume. Most of them were turned in promptly and all of them were in finished, impeccable order—an editor could not ask for more. The particular interests of the topics covered this year should make this a popular volume. It will be possible for the interested reader to find in one place the most recent work on the glycoprotein hormones and on their unique bimolecular makeup. The extensive reports on the radioimmunoassays of steroid and protein hormones and on the binding of steroid hormones to target tissues will also be of general interest.

The fascinating story growing out of recent work on vitamin D, which makes it seem more like a hormone than an accessory food factor, is told in a reserved but graceful manner by the discoverer of much of the new information. If any one compound were to be singled out as the parent of the steroid hormones, the choice would be pregnenolone, and the authoritative account of the biosynthesis of this important substance will be a valuable source of reference.

The mediation of the action of many hormones through activation of the enzyme adenylcyclase is a topic of wide popular interest at present. The two scholarly papers, one by Dr. Pastan and one by Dr. Garren and their respective collaborators, provide the basis for careful thought and deliberation on the part of those who would understand and evaluate the vast literature in this area.

The last four chapters in this volume derive from a symposium on in vitro methods for studies in endocrinology that was held on the last day of the conference. Each speaker was allowed only 20 minutes for his presentation; however, the published material is much fuller and more complete.

The Committee on Arrangements wishes to express its gratitude to the management of the Lodge for their splendid cooperation during the conference and to the members of the staff of Academic Press for the efficient and skillful manner in which they have brought out this hopefully faultless and certainly handsome volume. The Committee is also most grateful to its executive secretary Miss Joanne Sanford and to her helpers Miss Lucy Passalapi and Mrs. Mina Rano for their diligence and skill in making most of the arrangements and for recording and collating the discussion. We also wish to thank the members who served as Chairmen of the sessions, Drs. W. H. Daughaday, A. E. Wilhelmi, D. H. Solomon, S. Lieberman, L. T. Samuels, R. W. Butcher, G. Nichols, and K. Savard.

If the spirit and interest of the last few meetings and the welcome acceptance of the last few volumes can be used as a guide, we can look to the future and confidently anticipate that there will be many more good conferences and a continued demand for annual additions to Recent Progress in Hormone Research.

E.B. ASTWOOD,     Boston, Massachusetts

May, 1971

X-Ray Analysis and the Structure of Insulin¹

T.L. BLUNDELL, G.G. DODSON, E. DODSON, D.C. HODGKIN and M. VIJAYAN²,     The Chemical Crystallography Laboratory and Molecular Biophysics Laboratory, University of Oxford, Oxford, England

Publisher Summary

This chapter focuses on X-ray analysis and the structure of insulin. Insulin is one of the most widely studied of all hormones. The fact that insulin will also crystallize in other forms helps in understanding the symmetry of the insulin dimers and hexamers in the rhombohedral crystals. The monoclinic and rhombohedral forms contain a hexamer of insulin that has both threefold and twofold symmetry. In the rhombohedral crystals of insulin, there is a threefold axis only; there is no centrosymmetric projection. This means that an alternative approach is required to analyze the heavy atom derivatives. The chapter discusses three experimental stages required for a successful analysis of the crystal structure of a protein: (1) crystallization of the protein, (2) preparing crystals of the protein modified by the addition of atoms of high atomic weight, and (3) collection of X-ray data and analysis of the X-ray intensities from the crystals of the native protein and heavy atom derivatives to derive the heavy atom positions and the electron density map of the protein. This analysis should be done in the highest resolution possible, and X-ray intensity data should be extended to well beyond 2 Å spacing. This will be particularly valuable in trying to understand the intricate packing of the molecules in the dimer and the hexamer.

I Introduction

Insulin is one of the most widely studied of all hormone molecules. It is not only central to research on diabetes. It is also one of the smallest of those complex molecules, the proteins, and, therefore, has been the subject of many studies by organic and physical chemists. An important theme in the work of both biologists and chemists has been the effort to understand its three-dimensional structure, and the literature contains many speculative models. Now we are pleased to be able to describe the first detailed results on the structure of insulin—indeed of any protein hormone. These have been obtained by X-ray analysis of insulin crystals.

In this discussion we will not attempt to give a full description of developments in experimental technique and mathematical analysis that made this advance possible. Instead we will attempt only to outline the problems that made the X-ray analysis of insulin less straightforward than that of many other proteins. This will allow greater discussion of the crystal structure itself. Also we will describe some of the features of the structure that appear relevant to an understanding of the chemistry and biology of insulin.

II Preliminary Crystallographic Studies

Insulin was first crystallized in a rhombohedral form in 1926 by Abel. However, this could not be reliably repeated with purified insulin until D. A. Scott discovered that the presence of zinc ions was necessary for crystallization. He later experimented on the replacement of zinc by other metals and showed that insulin crystals could grow in the presence of iron, cobalt, nickel, and cadmium as well as zinc.

The first X-ray diffraction photographs of single rhombohedral crystals were taken by Crowfoot (1935) in Oxford. These showed that the crystals contained three equivalent units of weight about 12,000. When Sanger later determined the primary sequence of beef insulin (Ryle et al., 1955), this was found to correspond with two insulin molecules. We now know that insulin crystallizes from aqueous solution in two rhombohedral forms which contain a minimum of two zinc ions and four zinc ions, respectively, per six insulin molecules; in the basic repeating unit, the rhombohedral cell, each form contains three equivalent insulin dimers related by a 3-fold axis. Figure 1 shows the rhombohedral 2 Zn insulin crystals (Schlichtkrull, 1956), and Fig. 2 an X–ray diffraction photograph. The 3-fold symmetry is apparent in both. Table I records the X-ray data on 2 Zn insulin crystals (Harding et al., 1966).

TABLE I

Unit Cell Dimensions and Solvent Contents of 2 Zn Insulin and 4 Zn Insulin

Fig. 1 Rhombohedral 2 Zn Insulin Crystals.

FIG. 2 X-Ray diffraction photograph (precession of h k i 0 zone) of rhombohedral lead insulin crystals.

The fact that insulin will also crystallize in other forms helped us to understand the symmetry of the packing of the insulin dimers and hexamers in the rhombohedral crystals. Barbara Low and her colleagues have worked on zinc-free orthorhombic crystals obtained at low pH. They recognized that the asymmetric unit of these crystals contained two equivalent insulin monomers related by a 2-fold axis (Low and Einstein, 1960), and guessed that this might be so in the rhombohedral crystals as well. Later, using Rossmann and Blow’s functions, Dodson et al. (1966) were able to determine the position of this axis in the rhombohedral crystals. They also showed that a monoclinic form of zinc insulin, which crystallizes in the presence of phenol, contains six molecules in the asymmetric unit which has local 3-fold and 2-fold axes.

from this axis. In the monoclinic and 2 Zn rhombohedral forms, the 2-fold axes not only are perpendicular to the 3-fold axis, but also intersect it. This arrangement has 32 symmetry and is shown in Fig. 3. The 2-fold axes are not crystallographic axes. They relate only those molecules within the hexamer, that is within the unit cell, and they are, therefore, local axes. It was an early and correct guess that the two zinc atoms lie on the 3-fold axis related by the 2-fold axes.

FIG. 3 The arrangement of insulin molecules in the hexamer. The 3-fold axis is perpendicular to the plane of the paper, and the 2-fold axes lie in this plane.

III The Crystal Structure Analysis of 2 Zn Rhombohedral Insulin

The successful analysis of the crystal structure of a protein depends on three experimental stages.

1. The first of these concerns the crystallization of the protein. Large crystals, about 1 mm across, are required for an accurate analysis at high resolution. The very few accounts of hormone isolation and purification which describe good crystals indicate that this may be the main difficulty in the X-ray analysis of other protein hormones, but in the case of insulin, beautiful 2 Zn rhombohedral crystals of a suitable size were available. Most of our experiments have been carried out on pig insulin given to us by Dr. J. Schlichtkrull (Novo Terapentisk Laboratorium, Copenhagen) and recrystallized according to methods described by him. We are very grateful to Dr. Schlichtkrull for the careful work that made the production of such crystals possible.

2. The second stage involves preparing crystals of the protein modified by the addition of atoms of high atomic weight. The conformation and packing of the protein molecules in the crystals of these heavy atom derivatives should be the same as those in the native crystals; in other words, they must be isomorphous. It was the demonstration of the use of the method of isomorphous replacement by Max Perutz nearly twenty years ago that ensured the success of protein X-ray analyses. With insulin, this procedure involved many difficulties.

Bernal realized that isomorphous replacement might be useful in the X–ray analysis of insulin in the very early days of protein crystallography, and he suggested using Scott’s cadmium crystals. However, the differences in the diffraction patterns of cadmium and zinc insulin crystals were small, and further work was not carried out at that time. More recently, Marjorie Harding and others made a careful study of the effect of heavy atom salts on insulin crystals and also attempted many cocrystallizations in the presence of these salts. The first breakthrough came in 1965 when it was discovered that the zinc ions could be removed from the rhombohedral crystals by using the chelating agent, EDTA. These zinc free crystals were stable enough to be removed for soaking in lead ions which became bound in a regular way. This derivative proved to be sufficiently isomorphous for a detailed analysis. However, the removal of zinc ions appeared to lead to systematic errors in the analysis, and in addition at least one other derivative was required for a complete determination by this technique.

The final success was the result of a further, very systematic study of the soaking of insulin crystals in heavy atom-containing solutions. Previously unsuccessful experiments were repeated with uranyl and uranyl-fluoride ions, and by careful control of temperature, concentration, and buffer, usable derivatives were prepared. It was also discovered that a different lead derivative could be produced by increasing the concentration of lead ions. A series of heavy atom-substituted aldehydes were also reacted with insulin in the hope that they would form Schiff’s bases with the insulin α- and ε-amino groups. As a result of a great number of studies, it was found that mercurated metahydroxybenzaldehyde gave a reproducible, isomorphous derivative.

Thus, in 1969 we found ourselves with five isomorphous heavy atom derivatives. Each of them seemed usable, but we had serious reservations about all. Of the new derivatives, high lead concentrations damaged large crystals, the uranyl derivatives had disordered heavy atom positions, and the mercurated aldehyde preparations gave very low substitution. We can now see that these difficulties in forming good heavy atom derivatives mainly arose from the very close packing of the insulin molecules in the hexamer and of the hexamer in the rhombohedral lattice; the amount of solvent in the crystals of the rhombohedral form, 30–34% (Table I), is rather lower than for most protein crystals. Also the absence of reactive sulfhydryl groups and the presence of an exposed disulfide bridge precluded the use of many of the reagents successful with other proteins.

3. The third stage involves the collection of X-ray data and the analysis of the X-ray intensities from the crystals of the native protein and heavy atom derivatives in order to derive the heavy atom positions and finally the electron density map of the protein.

The rhombohedral space group, R3, made this part of the analysis complicated. All crystals containing protein molecules formed from L-amino acids must be noncentrosymmetric. In fact, centrosymmetric crystals offer great advantages in simplification of structure analysis, and it is fortunate that many protein crystal structures contain an evenfold axis of rotation which leads to an apparent centrosymmetric arrangement in projection. This is true for hemoglobin, lysozyme, ribonuclease, and other enzyme crystals so far studied, and it allowed straightforward exploratory studies when derivatives were being evaluated.

, and this involved recording about 60,000 intensities.

The interpretation of our X-ray data showed that the binding patterns in the lead and uranyl derivatives were very complicated. Nowhere did the ions occupy all the available equivalent positions in the crystal. The binding was characterized by a statistical distribution among five or more sites giving an average of about four metal ions per hexamer. The two lead derivatives had similar sites, but the increased concentration of lead ions in one led to an increased occupancy. The amino acid side-chains involved in the heavy atom binding sites can now be identified. In all cases, the heavy atoms are on the surface of the dimer; none penetrates the nonpolar core. Most of the metal ions are close to the carboxylate groups of one or more glutamate residues.

resolution. A measure of the electron density is given by the average figure of merit, which was 0.8, and this led us to feel that the map was of good quality.

IV The Electron Density Map

Parts of the electron density map are shown in . Several sections perpendicular to the 3-fold axis are shown in a diagrammatic form in these figures.

FIG. 4 resolution showing the appearances of residues B9, B10, and B11 of the insulin molecule. The section is taken perpendicular to the 3-fold axis. The atomic positions of one of the three equivalent groups are shown.

FIG. 5 resolution showing the region of the 2-fold axes of symmetry (OP and OQ). The atomic positions of some selected residues are shown. The sections are computed perpendicular to the 3-fold axis.

apart, and the two zinc ions per hexamer of insulin are in agreement with the finding of Schlichtkrull.

The electron density contains continuous chains, which are the polypeptide backbone of the insulin molecule. Most carbonyl oxygens are shown by small but well defined peaks of density in the chain, and the amino acid side chains also have density continuous with the backbone density. Figure 4 shows part of an α-helix. The side chains are those of B9 serine, B10 histidine, and B11 leucine, and the atomic positions are indicated for one of the equivalent positions. Figure 5 illustrates the appearance of other residues in the electron density map. In particular, the rings of B24 phenylalanines are very well defined, as are the carboxylate ions of B13 glutamates.

Figure 5 also demonstrates another feature of the electron density map, and this is its symmetry. The 3-fold symmetry is evident. Also there are approximate 2-fold axes perpendicular to the 3-fold axis, marked OP and OQ. This arrangement of electron density agrees well with the predictions using the functions of Rossmann and Blow that have been discussed in Section II.

V The Structure of the Insulin Monomer

The electron density map was interpreted initially from the hypothesis that the residue coordinated to the largest peak, the zinc atom, would be a histidine linked through an imidazole nitrogen. This had been predicted by Tanford and Epstein (1954). The choice of this histidine as B10 rather than B5 was decided by examining the fit of the adjacent residues in the B–chain to the electron density. In particular, the disulfide bridges were easily identifiable in the electron density. The interpretation of the density close to B10 is shown in Fig. 4. Once this part of the molecule had been correctly identified, the interpretation of the electron density on the basis of the known primary structure of pig insulin was reasonably straightforward. However, parts of the molecule such as A12 to A18 and also the C-terminal of the B-chain presented some difficulties.

There was some ambiguity in the interpretation arising from solvent bound to carbonyl oxygens on the outside of the molecule. Also the density was rather more diffuse for side chains that were free to move in the solvent region. Therefore, our initial interpretation was followed by a careful model building study using an optical superposition device designed by Richards (1968). This device allowed us to superimpose our model directly onto the electron density maps by means of a half-silvered mirror. This technique ensures that the final model is a best fit to the electron density.

The conformation of the B chain interpreted by this method is shown in Figs. 6 and 7. These projections of the structure down the 3-fold axis show the backbone and all the atomic positions, respectively, of the B chain. The 3-fold and approximate 2-fold axis positions, OP and OQ, are indicated in the diagrams for reference. The conformation of the B chain of the other independent insulin molecule related by these 2-fold axes is very similar, but not identical. The differences will be discussed later.

FIG. 6 The main chain conformation of the B chain as viewed along the 3-fold axis.

FIG. 7 The main chain conformation and the side chain distribution of the B chain as viewed along the 3-fold axis.

The first seven residues in the B chain have an extended configuration but the chain undergoes a sharp turn at the glycine at B8. The residues B9 to B19 form part of a well defined right-handed helix which seems to have classical α-helix geometry in the middle but is rather opened out at the ends. The helix separates the cystine disulfide bridges at B7 and B19 and it appears to be an important structural feature in the insulin molecule. The residues following the helix B20 to B23 constitute a U turn in the polypeptide chain so that the remaining residues B24 to B30 can form an extended chain lying antiparallel and against the helix B9 to B19. The chain folding brings residues quite apart in the primary sequence, such as B11 leucine and B26 tyrosine, into contact with each other.

The A chain has a less extended tertiary structure, and this is shown in Figs. 8 and 9. The structure is projected down the 3-fold axis with the axes of symmetry marked for reference. The conformation of the A chain of the other independent molecule is again very similar.

FIG. 8 The main chain conformation of the A chain as viewed along the 3-fold axis.

FIG. 9 The main chain conformation and the side chain distribution of the A chain as viewed along the 3-fold axis.

Residues A2 to A8 make a short piece of distorted α-helix, and the right-handed helical sense is retained for the residues A9 to A11 although the pitch of the helix is quite different from that of an α-helix. This arrangement accommodates the A6–A11 cystine disulfide bridge easily. The residue A12 has an extended form so that the following amino acids A13 to A19 can form a distorted right-handed helix with its axis nearly antiparallel with that of the helix A2 to A8. This region of the molecule is most difficult to interpret, and very careful model building was necessary to eliminate some ambiguities. This is because this helix is on the outside of the molecule with many side chains free to move. There are also ordered solvent molecules bound to many of the carbonyl oxygens in such a way that they may easily be mistaken for small side chains. We now think that the pitch of the helix is intermediate between a 310 and 313 or α-helix, but the carbonyl groups make an angle to and are not parallel with the helix axis. The peptides from A19 to A21 have an extended configuration.

This arrangement means that the A chain is also folded upon itself. Thus the phenyl ring of A19 tyrosine is in van der Waals contact with the side chain of A2 isoleucine. Other results are the close proximity of the amide functions of A5 and A15 glutamines and also the van der Waals contact between A16 leucine side chain and the sulfur atoms of A6–A11 cystine. The packing of the A and B chains together in the insulin molecule is shown in Figs. 10 and 11. These are projections of the backbone and full molecule, respectively, down the 3-fold axis in a similar way to the diagrams of the separate chains. In order to give a 3-dimensional picture of the insulin molecule, the molecule viewed perpendicular to the 3-fold axis is included in Fig. 12.

FIG. 10 The main chain conformation of an insulin monomer as viewed along the 3-fold axis.

FIG. 11 A view of the complete molecule, including side chains, along the 3-fold axis.

FIG. 12 A view of the molecule along a direction perpendicular to the 3-fold axis.

The most obvious stabilizing forces result from the disulfide bonds. One of these, B7–A7, is on the outside of the molecule while B19–A20 is more concealed but, nevertheless, still accessible to the solvent. The A6–A11 disulfide is completely buried and forms part of the nonpolar core of the insulin molecule. The general disposition of these groups is consistent with the ease with which the interchain bridges are reduced electrolytically (Markus, 1964) or by chemical reagents (Zahn and Gattner, 1968). The unreactive nature of the intrachain disulfide had previously led Cecil and Wake (1962) to predict that it would be buried.

[See figures on the following three pages.]

The second important feature of the structure is the existence of a completely nonpolar core. The regions of nonpolar intrachain contacts are brought together in the complete molecule and define a hydrophobic center comprising the residues: B6, B11, B15, and A16 leucines; B18 valine; B24 phenylalanine and the phenyl ring of B26 tyrosine; the A6–A11 and part of the A20–B19 cystines. Some of these residues are partly exposed in the monomer, but all must contribute to a positive entropy term in aqueous solution. The location of residues in a way that gives a nonpolar core is very nicely illustrated by the Schiffer and Edmundson (1967) wheel shown in Fig. 13. This is a diagrammatic representation of the α-helix between B9 and B19, projected down the helix axis. One arc of the wheel is comprised of nonpolar residues, contributing to the nonpolar core. The other side has polar residues which are on the outside of the insulin molecule.

FIG. 13 Edmundsen-Schiffer wheel for residues B9 to B19.

resolution as hydrogen atoms cannot be seen and the bonds must be inferred from model building. However, we suspect that ion pairs are formed between the C-terminal carboxylate group of A21 and the guanidinium group of B22 arginine, and possibly between B29 lysine α-amino group and the carboxylate group of A4 glutamate. The latter region is difficult to interpret unambiguously as the density is diffuse and there may be some local distortions caused by close contacts between hexamers. There are a number of possible hydrogen bonds, some of the more obvious ones of which are between A19 carbonyl and B25 α-nitrogen, A11 carbonyl and B4 α-nitrogen, B4 carbonyl and A11 α-nitrogen, and A7 carbonyl and B5 imidazole nitrogen.

The complete insulin molecule is a compact 3-dimensional unit with only the terminals of the B chain free from the main structure. The helix content is difficult to estimate. The crystal structure shows only about 20% of the residues in good α-helices, but nearly 50% contribute to right-handed helices, which are often short and distorted but have a pitch close to that of an α-helix. There is no left-handed helix.

Two faces of the molecule adjacent to the axes OP and OQ are mainly nonpolar, while the remainder of the surface residues are mainly polar. Some of the polar residues, such as A5, A15, and A21 backbone, may be involved in stabilizing interactions; others like A8 and A9 are free in the solvent.

VI The Insulin Dimer

The asymmetric unit of the rhombohedral cell contains two insulin molecules, related by the approximate 2-fold axes OP and OQ. The contacts in the region of the OP axis are close and more intricate than those in the region of the OQ axis. We believe that the dimer in solution retains these contacts as shown in Fig. 14.

FIG. 14 Projection of the insulin dimer down the 3-fold axis.

As described above, the insulin molecule has a nonpolar surface adjacent to the axis, OP. This is also true for the second molecule. Thus, the contacts between monomers in the dimer are predominantly between nonpolar groups. The side chains of the residues close to the 2-fold axis OP, are shown in Fig. 15. The two B chain helices come close together so that the B12 valines are in van der Waals contact. Also the extended chains of residues, B23 to B27, of the two molecules are antiparallel so that the phenyl rings of B24 and B25 phenylalanines are brought into van der Waals contact with the equivalent groups on the other molecule. Thus, the nonpolar core of the monomer is extended through the dimer. Residues B26 tyrosine and B11 leucine as well as B24 and B12 mentioned above are no longer on the surface but are buried in the dimer. This type of interaction is consistent with the independence of the dimer dissociation constant of the ionic strength of the medium and with the easy dissociation of insulin in organic solvents (Harfenist and Craig, 1952; Fredericq, 1957).

FIG. 15 A view of the side chains close to OP viewed along the 3-fold axis.

The antiparallel polypeptide chains of the two molecules give rise to an antiparallel, hydrogen-bonded pleated sheet, and this is apparent in Fig. 16, which shows the polypeptide backbone for the dimer viewed down the OP axis. Four hydrogen bonds involving the carbonyl and α-nitrogen groups of B24 and B26 of both molecules can be accommodated in this arrangement.

FIG. 16 The main chain conformation of the dimer viewed along OP.

The importance to the dimer of interactions involving the C-terminal octapeptide of the B chain is consistent with the fact that the desoctapeptide insulin sediments as a monomer (Arquilla et al., 1969).

The very tight and intricate nature of the association of the insulin molecules in the dimer explains why the dimer is so stable and why for a time it was considered to be the basic unit of insulin.

The two insulin molecules are related by only an approximate 2-fold axis. While it has been emphasized that the two molecules have generally the same structure, they differ in several respects. One of the most striking of these differences can be seen from Figs. 5 and 15, which show that the B25 phenylalanines are not related by 2-fold axis symmetry. The phenyl ring of one molecule is in fact lying on the 2-fold axis. There are other marked, although perhaps less striking, divergences from 2-fold axis symmetry in other parts of the molecule.

The reason for this lack of symmetry cannot be identified at this stage. It may result from the fact that insulin molecules in the dimer have different environments in the crystal. On the other hand, the lack of symmetry may be characteristic of the isolated dimer. We suspect that the close packing of equivalent residues around the axis may not be easily accommodated with retention of the 2-fold symmetry, and further that the movement from a symmetrical relation might increase van der Waals contacts. These observations are relevant to the basic assumptions of the Monod symmetry model.

VII The Structure of the Insulin Hexamer

The insulin dimers are related by the crystallographic 3-fold axis to give a hexamer. The structure of the hexamer is shown in Fig. 17.

FIG. 17 A projection of the hexamer down the 3-fold axis.

apart and related by the 2-fold axes. Figure 4 shows that each zinc is bound to an imidazole nitrogen of one of the B10 histidines of each of the dimers; each zinc has bound to it three imidazole rings. The coordination of the zinc is completed by solvent molecules, and these molecules appear to be part of a complex arrangement of ordered solvent bound to the protein through B9 serine, B5 histidine, and B16 tyrosine.

In addition to the coordination of the imidazole rings to the zinc, interactions between the insulin dimers seem to be important to the stability of the hexamer. The insulin dimers are related by the approximate 2-fold axis, OQ, and contacts in this region are shown in Fig. 18. There are fewer van der Waals contacts between dimers than between monomers although the groups involved are mainly nonpolar. They include B1 phenylalanine, B6 leucine, B14 alanine, B17 leucine, A13 leucine, and A14 tyrosine. There may also be a stabilizing interaction between the A17 glutamate carboxylate group and the N-terminal at B1 of the adjacent dimer.

FIG. 18 A view of the side chains close to OQ viewed along the 3-fold axis.

from top to bottom. Although the interactions between the dimers in the hexamer do not appear to be as strong as those between insulin molecules in the dimer, the packing achieved by use of a 3-fold symmetry leads to a very compact 3-dimensional structure.

VIII Insulin Structure in Solution

There is now a considerable body of evidence for the view that in solutions approximating physiological conditions insulin retains an ordered and stable structure and resembles crystalline insulin.

The insulin hexamer has a fairly compact, nonpolar core from which solvent is excluded and which is stabilized by disulfide bridges. In general, the surface residues of the hexamer are hydrophilic and most are surrounded by solvent in the crystal lattice. These features are also characteristic of enzymes where the results of X-ray analysis have been used extensively to rationalize the solution properties, and they suggest that the structure of insulin will also be similar in solution. This was also the conclusion of Praissman and Rupley (1968) as a result of comparative tritium exchange experiments in crystals and in solution.

We may also consider the results of structure studies in solution. These include experiments designed to examine the availability of various functional groups—whether they are buried or exposed—and also to study the geometric relations between parts of the molecule.

B29 (Africa and Carpenter, 1968; Bromer et al., 1967; Borras and Offord, 1970). This is consistent with their disposition in the dimer where the B1 amino group should be freely accessible. However, reaction at the B1 site, which is in the region of dimer-dimer contacts in the hexamer, might be expected to inhibit hexamer formation as previously suggested (Marcker, 1960). The A1 α-amino group is more tightly bound to the structure, and removal or modification of this group rather than B1 appears to result in greater conformational changes in the insulin molecule (Arquilla et alfrom the A1 amino group and extends into the solvent; however, the high pK appears to limit its reactivity. The proximity of the A1 and B29 amino groups has been repeatedly observed in solution. Zahn and Meienhofer (1958) found that the bifunctional reagent, l,5–difluoro–2,4–dinitrobenzene forms a preferential cross-link with these groups; and the small separation has been supported by immunochemical (Arquilla et al., 1969) and spectral studies (Mercola, 1969).

In general, esterification of the carboxylate groups is also consistent with their position on the surface of the dimer. However, two of these groups appear to be involved in interactions with other residues. The carboxylate group of A4 glutamate lies close to the lysine at B29 and the carboxylterminal of the A chain is close to the guanidinium group of the B22 arginine. These surface interactions may be disturbed on esterification and lead to conformational changes in the molecule (Massaglia et al., 1968). The close interaction of A21 may also be the reason for the limited cleavage of this residue by carboxipeptidase (Slobin and Carpenter, 1963).

The four tyrosine residues have unique environments which depend on the state of aggregation. The A14 tyrosine residues lie on the surface of the hexamer in the region of contacts between dimers (close to the OQ axis in Fig. 15). The A19 tyrosine residues are not directly affected by the level of quaternary structure and lie in a pocket in the surface of the molecule in such a way that the 3-position on the phenyl ring is in contact with the nonpolar residues of the core while the 5-position and the hydroxyl function are exposed to solvent. The tyrosines at B16 and B26 are less accessible in the hexamer. The phenyl ring of B26 is in the nonpolar environment between molecules in the dimer, with its hydroxyl function close to the α-nitrogen of B8 of the same molecule. B16 tyrosine is close to the B26 and B8 of the adjacent monomer, but the hydroxyl function appears to be more accessible. All tyrosine hydroxyl functions have associated solvent molecules in the crystal structure.

Several studies on structures in solutions indicate that one or more of the tyrosine residues are involved in various forms of interactions, and most can be correlated with the crystal structure. Dissociation of the dimers on dilution or by trypsin catalyzed removal of the C-terminal octapeptide of the B chain leads to characteristic changes in the tyrosine absorption spectrum (Rupley et al., 1967; Laskowski et al., 1960) and decreased circular dichroism of the tyrosines (Morris et al., 1968). The results are consistent with the transfer of B chain tyrosine residues from a nonpolar environment to aqueous solution as suggested by Yanari and Bovey (1960). Similarly, with iodination or nitration there is a preferential reaction with the tyrosine residues of the A chain (De Zoeten and De Bruin, 1961; Arquilla et al., 1969; Morris et al., 1970). These tyrosyl residues are more exposed in the dimer and higher states of aggregation than the B chain tyrosines. In particular, monoiodination of A19 compared to diiodination of A14 is consistent with the respective environments of these tyrosine residues (De Zoeten and De Bruin, 1961).

The general agreement of these studies of structure in solutions with the crystal structure does not preclude small changes of conformation upon crystallization or upon dissociation of the hexamers in solution to dimers, the species commonly studied in solution. In fact, a change in the position of the A14 tyrosine side chain by a single bond rotation upon dissociation of the hexamers may give rise to the reported anomalous tyrosyl ionization (Fredericq, 1954; Inada, 1961) and the selective o-cyanolation of residues A19 and B16 (Aoyama et al., 1965). This rather eclectic and brief description of the chemistry of insulin described here has served to illustrate our present impression of the nature of the structure and its properties in solution.

IX Insulin Structure and Evolution

The primary sequences of many insulins are known (Smith, 1966, 1970), and it is interesting to consider the variations in sequence in the light of the three-dimensional structure of porcine insulin hexamers found by X-ray analysis. Tables II and III show the sequence variations for the A and B chains, respectively.

TABLE II

Insulin A Chain Sequences

a Invariant residues are circled. The sequences are divided into three groups. Unless an entry is included, the residue is the same as in the sequence of pig insulin given on the top line.

TABLE III

Insulin B Chain Sequencesa

aSequences are given in the same convention as in Table II. Deletions are indicated by dash (—).

The totally invariant residues will be considered first. The positions of these residues are shown in Fig. 19. This is a projection of the insulin molecule down the 3-fold axis, and the diagram is directly comparable with Fig. 11. In this case, only the α-carbons of the backbone and the side chains of invariant residues are shown.

FIG. 19 The disposition of invariant residues, shown as bold lines, seen along the 3-fold axis. The polypeptide backbone also is indicated.

The disulfide bridges and lengths of polypeptide chain between them are the same in all species. Apart from this, the most obvious feature is the clustering of invariant residues in the core of the molecule. In fact, all residues of the hydrophobic core of the porcine insulin dimer are retained. The other invariant residues include those with side chains involved in stabilizing interactions on the surface such as A19 and A1. The invariance of residues in the core and on the surface that are important to the structure suggests that all insulins may have the same general three-dimensional architecture in the monomer. Furthermore, they should all form dimers.

There are several sequence variations which are unique to guinea pig insulin, as shown in Tables II and III. For instance, B10, B14, B17, and B20 are invariant in all species except guinea pig, and also coypu, the sequence of which has yet to be published in full (Smith, 1970). These residues have a very interesting relation to the structure. B10 histidine coordinates to the zinc atom in porcine insulin hexamers; in guinea pig this residue is asparagine. The other residues which are invariant in other insulins all lie in the region between dimers in the hexamer, close to the axis, OQ, as shown in Fig. 20. B14 alanine becomes threonine, B17 leucine changes to serine, and B20 glycine becomes glutamate in guinea pig insulin. These residues become larger or more hydrophilic or both in guinea pig, and we feel that these changes, like the change of the zinc coordinating residue at B10, would not favor the formation of hexamers. Furthermore, A13 leucine and B4 glutamine are replaced by arginine in guinea pig. These residues also lie in the region between dimers, as shown in Fig. 10, and must destabilize hexamers. Indeed, there are no reports of rhombohedral zinc insulin crystals of guinea pig. If hexamers are uniquely not required in any role of insulin in guinea pigs, this would allow these unusual changes in sequence, and this may be the reason why guinea pig appears to have mutated at a rate ten times greater than other insulins (Smith, 1966).

FIG. 20 The position of side chains in pig insulin, which are very different in guinea pig insulin. The polypeptide backbone, as viewed along the 3-fold axis, is again indicated.

Finally, the arrangement of the residues, which vary in the nature of their functional groups, is shown in Fig. 21 for the mammals (other than guinea pig) and the fishes.

FIG. 21 The polypeptide backbone and the highly variable residues, shown as bold lines, as seen along the 3-fold axis.

The highly variable residues are all on the surface of the hexamer also. For instance, A8, A9, and A10 are highly variable even among the mammals, as shown in the early studies of Sanger and his co-workers (Harris et al., 1956). These residues are all exposed to solvent and are not involved in stabilizing interactions. This is also true for B30.

Residues B1, B2, and B3 are very different in the fish and mammal insulins. There is also a residue attached to the N-terminal at B1 in many fish insulins. This part of the molecule is on the surface of the dimer and is not closely involved in the general three-dimensional arrangement of the rest of the molecule. The changes observed, which involve substitution by proline, may be accommodated without a gross structural change in the dimer. However, while these residues are on the outside of the dimer, they are in the region of contacts between the dimers in the hexamer. It is interesting that residues A13, A14, and A17 also vary in the fish insulins. These are residues which pack closely with the N-terminal residues of the B chain in the adjacent dimer. There is evidence that cod insulin crystallizes as hexamers in the presence of zinc ions (Baker, 1970), though in a different space group and this suggests that the change in the amino acid sequence is such as to allow aggregation into a hexameric state.

X Speculations on the Role of the Tertiary Structure of Insulin in Biology

We have suggested that the structure of zinc insulin hexamers as determined by X-ray analysis is very similar in solution, and also that the general three-dimensional structure of at least the dimers may be conserved in all the sequenced insulins. What can be said about the role of this structure in biology?

It is convenient to consider the question separately for the different stages—synthesis, storage, circulation, and action—of the life cycle of insulin.

A PROINSULIN

Insulin is synthesized as a single chain precursor, proinsulin (Steiner et al., 1969), and the sequences of bovine (Nolan and Margoliash, 1969) and porcine (Chance et al., 1968) proinsulins are now known. Their connecting peptides contain thirty-three and thirty residues, respectively, and differ in composition by 30%. The variability in length is to some extent consistent with the structure. The B30 alanine and A1 glycine are close to each other as shown in Fig. 12. This might imply that the complete sequence of the connecting peptide is not important to the arrangement of these groups relative to each other. In fact Steiner suggested at a previous Laurentian Conference that this connecting peptide has few specific structural requirements (Steiner et al., 1969); the lack of structural constraints would allow a high rate of mutation.

The obvious function of proinsulin connecting peptide to convert the bimolecular reaction of chain recombination to a unimolecular reaction (Steiner et al., 1969) does not explain its great length. However, the connecting peptide may be important in protecting the B22–B23 peptide bond from enzymatic cleavage during activation (cf. Wang and Carpenter, 1969). Our model shows that a peptide of about 30 residues could easily cover this part of the molecule, which is on the outside of the hexamer. Protection in this manner may occur until the insulin has formed hexamers which may be more stable to tryptic digestion. It is interesting that guinea pig insulin which may not form hexamers has B22 arginine replaced by aspartate which would make the adjacent peptide more stable toward enzymatic cleavage.

B STRUCTURE AND STORAGE IN β-CELLS

Howell et almay correspond to the diameter of the hexamer. However, the great variety of periodicities seen in the storage granules of other species makes it necessary to emphasize the speculative nature of conclusions based on these features.

We have suggested that guinea pig insulin cannot easily form hexamers, and it therefore could not be stored in this way. This is consistent with a report that guinea pig has no zinc in its β-granules.

C STRUCTURE AND IMMUNOLOGY OF INSULIN

Yagi et al. (1965) demonstrated that most antigenic determinants on insulin are disrupted when the separated A and B chains are tested against insulin antiserum, and Arquilla et al. (1969) have studied a series of insulin derivatives in reaction with various insulin antisera. They found that antibodies from strain 2 guinea pigs bind strongly to the N-terminal of the A chain and the ε-amino group of the lysine at B29 whereas strain 13 antibodies bind to the C–terminal A chain and the C-terminal octapeptide sequence of the B chain. They suggested that these associations represent tertiary relationships of A and B chains and that these relationships must be maintained for high immunoreactivity. The folding of the polypeptide chains in crystalline insulin is consistent with the proximity of the groups found necessary for these various antigenic determinants.

Also the variations in sequence between the mammals at A8, A9, and A10 change the surface of the molecule and could be involved in species recognition as some authors have suggested (Berson and Yalow, 1959). This area is also close on the surface of the molecule to the residues B1 to B3, which may be important to the immunochemistry of fish insulins.

D STRUCTURE AND ACTIVITY OF INSULIN

In order to identify the amino acids which are essential to the activity of insulin, we can consider not only the natural variation of sequences, but also the activities of synthetic and chemically or enzymatically modified insulins. The results of these studies can now be considered in relation to the structure of insulin found by X-ray analysis.

A remarkable feature of insulin’s biology is, with a few interesting exceptions, the great similarity in potency that insulins from different animals exhibit in each other. We have indicated that in general the variations of sequence would not affect the stabilizing interactions which appear important in the crystal structure, and so retention of the general three-dimensional architecture may be correlated with high activity. The same conclusion can be reached by consideration of variations of sequence in active, synthetic insulins (Weitzel et al., 1970; Katsoyannis, 1969). Wherever sequence variations involving substantial modification of the nonpolar core or stabilizing residues are made, the insulin is found to be inactive. The general shape of the molecule is probably retained in all active species.

Activity studies on modified insulins also emphasize the importance of the integrity of the entire structure. Consideration of the tertiary structure found in the X-ray analysis suggests that many derivatives, such as iodinated, di and trisubstituted thiocarbamyl, des-asn A21, deoctapeptide and fully methylated insulins might introduce conformational changes. This has been indicated by several observations including spectral (Carpenter, 1966; Mercola et al., 1967; Africa and Carpenter, 1970), chemical reactivity (Massaglia et al., 1968), and immunochemical studies (Arquilla et al., 1969). All these modified insulins have reduced activity. In fact, studies of the modified insulins have so far failed to identify any specific functional groups which are important to the molecule’s activity but are not implicated in maintaining its structure. Indeed it has been emphasized that any change in conformation results in a decrease in biological activity of insulin (Arquilla et al., 1969).

A natural explanation for this apparent delicate dependence on conformation is that a local area, whose conformation is defined by the overall structure, is important to the activity. Such an area might be involved in an interaction with the insulin-responsive cells.

Proceeding tentatively, we may speculate about several regions on the surface of the molecule that are particularly interesting to us. One of these is shown in Fig. 22. Here the molecular structure has the A chain N-terminal residues involved with the tyrosine at A19. The C-terminal asparagine of the A chain is on the surface and is closely associated with the folding of the ten C-terminal residues of the B chain. The residues B24, B26, B12, and B16 form a nonpolar surface adjacent to this region in the monomer. The disulfide bonds at the top and bottom of this region probably play an important part in this arrangement.

FIG. 22 An interesting region on the surface of the molecule viewed along a direction perpendicular to the 3-fold axis.

This area on the surface is fascinating not only because there are complex interrelationships between residues, but also because the residues are nearly all invariant. Furthermore, the activity is extremely sensitive to modification or deletion of residues such as A1, A19, A21, and the nonpolar groups in the terminal octapeptide of the B chain. It seems a very attractive hypothesis that this area of the molecule should be involved quite specifically in the action of insulin. Moreover, if the monomer is the active species, the active site could involve many nonpolar groups that are otherwise important in the formation of dimers.

XI Future Work

spacing. This will be particularly valuable in trying to understand further the intricate packing of the molecules in the dimer and the hexamer. Studies on the other closely related rhombohedral form, 4 Zn insulin, have been begun again and it will be interesting to see how the zinc and chloride ions are involved in this structure and how the dimer and hexamer adjust to the changes in local symmetry.

It will be far more difficult to identify the region or regions of the molecule important to its activity but perhaps more light can be thrown on this problem by carrying out crystal structure analyses on insulins from other species and on insulins that have been modified in a way that affects their biological activity.

ACKNOWLEDGMENTS

The authors would like to acknowledge the help they have had from Dr. Dan Mercola in writing this article and are also grateful to Mrs. Lesley Blundell and Mrs. Heather Baker for the many diagrams they have drawn.

REFERENCES

Abel, J. J. Proc. Nat. Acad. Sci. U.S. 1926; 12:132.

Africa, B., Carpenter, F. H. Fed. Proc., Fed. Amer. Soc. Exp. Biol. 1968; 27:766.

Africa, B., Carpenter, F. H. Biochemistry. 1970; 9:1962.

Aoyama, M., Kurehara, K., Shibata, K. Biochim. Biophys. Acta. 1965; 10:257.

Arquilla, E. R., Bromer, W. W., Mercola, D. Diabetes. 1969; 18:193.

Baker, E. N. (1970). Private communication.

Berson, S. A., Yalow, R. S. J. Clin. Invest. 1959; 38:2017.

Borras, F., and Offord, R. E. (1970). Private communication.

Bromer, W., Sheehan, S., Berns, A., Arquilla, E. R. Biochemistry. 1967; 6:2378.

Carpenter, F. H. Amer. J. Med. 1966; 40:750.

Cecil, R., Wake, R. G. Biochem. J. 1962; 82:401.

Chance, R., Ellis, R. M., Bromer, W. Science. 1968; 161:3837.

Crowfoot, D. M. Nature (London). 1935; 135:591.

De Zoeten, L. W., De Bruin, O. A. Rec. Trav. Chim. Pays. Bas. 1961; 80:907.

Dodson, E. J., Harding, M. M., Hodgkin, D. C., Rossmann, M. G. J. Mol. Biol. 1966; 16:227.

Fredericq, E. J. Polym. Sci. 1954; 12:287.

Fredericq, E. J. Amer. Chem. Soc. 1957; 79:599.

Harding, M. M., Hodgkin, D. C., Kennedy, A. F., O’Connor, A., Weitzman, P. D. J. Mol. Biol. 1966; 16:212.

Harfenist, E. J., Craig, L. C. J. Amer. Chem. Soc. 1952; 74:3087.

Harris, J. I., Sanger, F., Naughton, M. A. Arch. Biochem. Biophys. 1956; 65:427.

Howell, S. L., Kostianovsky, M., Lacey, P. E. J. Cell Biol. 1969; 42:695.

Inada, Y. J. Biochem. (Tokyo). 1961; 49:217.

Katsoyannis, P. G. Excerpta Med., Intern. Cong. Ser. 1969; 172:379.

Laskowski, M., Jr., Leach, S. J., Scheraga, H. A. J. Amer. Chem. Soc. 1960; 82:571.

Lindsay, D. G., Shall, S. Biochem. J. 1971; 221 [(in press)].

Low, B. W., Einstein, J. R. Nature (London). 1960; 186:470.

Marcker, K. Acta Chem. Scand. 1960; 14:2071.

Markus, G. J. Biol. Chem. 1964; 239:4163.

Massaglia, A., Pennisi, F., Rosa, U., Ronca-Testoni, S., Rossi, C. A. Biochem. J. 1968; 108:247.

Mercola, D. (1969). Ph.D. Dissertation, University of California, Los Angeles, California.

Mercola, D., Morris, J. W.S., Arquilla, E. R. Biochim Biophys. Acta. 1967; 133:224.

Morris, J. W.S., Mercola, D., Arquilla, E. R. Biochim. Biophys. Acta. 1968; 160:145.

Morris, J. W.S., Mercola, D., Arquilla, E. R. Biochemistry. 1970; 9:3930.

Nolan, C., Margoliash, E. Recent Progr. Horm. Res. 1969; 25:250.

Praissman, M., Rupley, J. A. Biochemistry. 1968; 7:2431.

Richards, F. M. J. Mol. Biol. 1968; 37:225.

Ryle, A. P., Sanger, F., Smith, L. F., Kitai, R. Biochem. J. 1955; 60:541.

Rupley, J. A., Renthal, R. D., Praissman, M. Biochim. Biophys. Acta. 1967; 140:185.

Schiffer, M., Edmundson, A. B. Biophys. J. 1967; 7:121.

Schlichtkrull, J. Acta Chem. Scand. 1956; 10:1455.

Slobin, L., Carpenter, F. H. Biochemistry. 1963; 2:16.

Smith, L. F. Amer. J. Med. 1966; 40:662.

Smith, L. F. (1970). Private communication.

Steiner, D. F., Clark, J. L., Nolan, C., Rubenstein, A. H., Margoliash, E., Aten, B., Oyer, P. E. Recent Progr. Horm. Res. 1969; 25:207.

Tanford, C., Epstein, J. J. Amer. Chem. Soc. 1954; 76:2163.

Wang, S., Carpenter, F. H. J. Biol. Chem. 1969; 244:5537.

Weitzel, G., Erselle, K., Zollner, H., Weber, V. Hoppe-Seyler’s Z. Physiol. Chem. 1970; 351:1480.

Yagi, Y., Maier, P., Pressman, D. Science. 1965; 147:3658.

Yanari, S., Bovey, F. A. J. Biochem. (Tokyo). 1960; 235:2818.

Zahn, H., Meienhofer, J. Makromol. Chem. 1958; 26:153.

Zahn, H., Oregor Gattner, H. Hoppe-Seyler’s Z. Physiol. Chem. 1968; 349:359.

DISCUSSION

J. G. Pierce: Did you state that there are, in addition to the hydrophilic residues on the surface, areas of hydrophobic side chains which could also interact with a potential receptor site?

T. L. Blundell: The question whether surface hydrophobic residues might be able to interact with a receptor site begs a further, more difficult,