X-Ray Analysis and the Structure of Insulin

In. a symposium on endocrinology it seems proper to begin by looking at insulin as it exists in the β granules of the pancreas. There have been a number of observations in the past made with the light microscope of single crystals, probably of insulin, visible in the granules, particularly in the pancreas of the dog. With the electron microscope it is possible to see dense, apparently crystalline, aggregates in the β granules of many animals and within these, compact particles, regularly arranged. A very good example, photographed by Greider, Howell and Lacy (1969), shows roughly spherical particles in lines about 50 Å apart in a β granule in rat pancreas. It may well be that the actual formation of crystals is a consequence of partial drying of the granules on isolation or preparation for electron microscopy; the individual particles of insulin that appear would however be almost certainly normally present in the living tissue.

The size of a single particle visible within the β granules is very similar to that of the hexamer of insulin molecules, 48 Å across, found in rhombohedral pig insulin crystals by X-ray analysis. Hexamers are shown in Fig. 1 in projection along the three fold axis of the rhombohedral crystals; the view is not unlike that of the rat islet ‘crystal’ though this may well have a different structure in detail from pig insulin. The appearance of hexamers both in the crystals and in the islets would be expected from experiments on insulin in solutions which show that, in the presence of zinc, six insulin molecules aggregate around two zinc ions; in most creatures, except perhaps the guinea pig and coypu, zinc is present in the β granules.

Fig. 2 shows a single hexamer of insulin, again observed in projection along the three fold axis of a rhombohedral insulin crystal. All of the atoms, except hydrogen atoms, are shown by circles varying in size with atomic number in the order zinc, sulphur, oxygen, nitrogen and carbon. Their positions are not at all precisely defined. The experimental evidence on which they are based – our next order of observation on insulin – is an electron density map derived by calculation from the intensities of spectra obtained by diffraction of X rays passing through the insulin crystals (Adams et al., 1969, Blundell et al., 1971). The spectra extend to spacings of 2·8 Å and the electron density map calculated accordingly provides a very blurred representation of the atomic positions. Its character can be demonstrated by plotting to scale contours of equal calculated electron density on sheets of perspex and stacking the sheets together to cover the unit volume in the insulin crystal. Groups of atoms, such as benzene rings, appear as single peaks, while peptide chains are represented by strands with higher density in the neighbourhood of, for example, carbonyl groups. A few overlapping sections of the electron density map are shown in Fig. 3 while Fig. 4 indicates how a part of the chemical structure of insulin is fitted within the electron density outlines in a single section of the map.

It will be clear from the character of the electron density map that while our knowledge of the atomic positions is, in detail, very imprecise, certain features of the arrangement of the peptide chains within the insulin molecules and of the six insulin molecules within the hexamer are very clear. Our confidence that many of the details of atomic positions now deduced are also reasonably correct depends on the close correlation of the electron density map with the chemical structure of pig insulin derived by Ryle, Sanger, Smith and Kitai (1955), shown in Table 1. It is also helped by other evidence, particularly on the titration of insulin in the presence and absence of zinc, which made it very probable that one of the histidine residues was attached to the zinc ions.

The electron density map shows that the two zinc ions are arranged along the three fold axis of the crystal and are about 17 Å apart. The insulin molecules are linked in threes around them through the B 10 histidine residues. Though the linking of the histidine residues to the zinc ions is similar, it does not appear quite identical in the two triplets in the projected view along the three fold axis. The molecules in one triplet are arranged relative to those in the second triplet nearly but again not quite exactly as required by two fold symmetry axes along the lines OP and OQ of Fig. 2 normal to the three fold axis. In certain regions the packing together of the six molecules in the hexamer is as close as is the packing of the amino acid residues and peptide chains within a single molecule. As a whole, accordingly the hexamer presents a compact spheroidal appearance to the external world, 48 Å in diameter, 35 Å in height. Viewed down the three-fold axis its circumference appears smoothly circular but the upper and lower surfaces of the spheroid are in fact pitted by deep grooves between projecting residues of the A chain loops.

Around one of the two fold axis, OP of Fig. 2, the contacts between the insulin molecules appear to be very close; it seems almost certain that these are the contacts responsible for the dimeric character frequently observed in molecular weight measurements of insulin in solutions in the absence of zinc. They are illustrated in Fig. 5 which shows a view of the atomic positions in the dimer along the ‘two’ fold axis. They include both non polar van der Waal’s interactions between, for example, the valine B 12 and phenyl alanine B 24 groups of the two molecules, and also hydrogen bonded contacts between the peptide carbonyl and NH groups, B 24 and B 26. The latter appear as part of a β-pleated sheet type of structure, formed by the antiparallel arrangement of the terminal residues of the B chain in the two molecules. In Fig. 5b, where this system is isolated, one can see that one reason for geometrical differences between the two molecules in the dimer may be the necessity for close packing of residues; thus the two phenyl alanine B 25 groups pack together, destroying the exactly symmetrical relation between them which might have appeared from the chain arrangement.

The two insulin molecules in the dimer are therefore very similar but not geometrically identical in every detail. They are illustrated in Fig. 6 in which they have been set side by side, for comparison. In each, the B chain starts out in an extended conformation from B 1 – B 8, turns sharply into an α helix from B 9 – B 20, and then through a U turn involving residues 21 – 23, ends in a further long extended region from 24 –...