Professor Kunitz, ladies and gentlemen, it is a very great pleasure to be here today and I must thank Professor Kunitz

for all of his kind words and particularly for his very useful introduction to the subject of my particular talk. Insulin, as no doubt you all know, is a hormone that is responsible

for part of the control of the metabolism of sugars and in its absence, the disease diabetes intervenes. It was first isolated in 1921

by a group in Toronto, McLeod, Banting and Fes of whom Best at the time was a graduate student aged 22 and they were able to isolate it following first of all the much earlier evidence

that the active hormone occurred in the pancreas and that it was a protein and so they were able to take effective measures against the protein being digested by the enzymes

that occurred in the pancreas. The next stage in the story, as far as this story of mine is concerned, was the crystallization of insulin in 1926 by Abel. Abel actually was professor at Johns Hopkins,

aged 67 and near retirement when Professor Noyes produced a small sum of $10,000 and invited him with this to Pasadena for a year to try to crystallize insulin

and the $10,000 allowed him to collect a small group of graduate students to help him and in the course of the year, they obtained the first crystals of insulin and a picture of these is on the first slide.

This first picture is taken from Abel's notebook and you see he looked up mineralogical texts and was able to describe the crystals as trigonal from Dana, rhombohedral class

and later he was able to obtain them looking somewhat more like the form that we usually see, little rhombs of very characteristic flat shape. There was some difficulty for two or three years over the crystallization of insulin.

Squibby insulin crystallized and lily insulin didn't and this was traced to the occurrence in squib insulin of a small amount of zinc which is necessary for the appearance of this particular modification of insulin.

The existence of zinc was shown by Scott in Toronto. It does also occur in the pancreas and Scott found that the same kinds of crystals would appear with certain other related metals.

Now, once the zinc was found to occur and be necessary in 1934, of course, all the different firms making insulin could obtain insulin crystals and a small sample of insulin crystals

was given to Sir Robert Robinson in Oxford by boots when they first made it and because he had no real intention of working on this subject himself, he gave them to me. I had then just come over from Cambridge

where I had been working with Bernal and had been involved with him in taking X-ray photographs of the crystalline enzyme pepsin and Robinson thought I might try to do the same, take photographs of insulin so I studied this back history

and tried the different methods that they had used to grow insulin crystals large enough for X-ray work and I must say that one of the most exciting nights of my life was when I saw the first spots on a photograph produced by passing X-rays

through a crystal of insulin. Now, the next photograph, next picture shows you an X-ray photograph of an insulin crystal very, very much better than the one that I first saw

but you can see the reflections here have a definite hexagonal pattern of intensities and a definite distance apart and from the distance apart of this photograph combined with others that cover the whole diffraction picture of insulin

one can calculate at first just the size of the repeating unit. The size varies a little according to whether the crystals are wet or dry and I will put up here the later measured value

for the side of the ROM of 49 angstroms and the angle 114 and a half degrees.

The first calculation that one can make is how much, oh, I don't, that's all I, the space I need. Thank you very much. The first calculation that one can make

is what is the weight of protein in this particular volume defined by the unit ROM and allowing for some 30% of water of crystallization the weight is just round about 36,000

and this seemed dramatically close to the weight of the protein molecule in solution found by the first experiments of Svedberg in the ultra centrifuge. He gave a weight of 35,000 as the molecular weight of insulin.

I was a little bit concerned. I discussed the problem with Harrington who worked on insulin, that in fact these crystals are trigonal and that this weight should therefore be divided into three equal parts, 12,000,

but I published the molecular weight as 35,000 possibly divided by three and this brought an immediate letter from Freudenberg in Heidelberg saying Svedberg doesn't really measure a molecular weight.

He measures a particle mass and the amount of chemical change that inactivates insulin is so small as to suggest the molecular weight is much smaller than 36,000. In fact, he gave an estimate of perhaps

from nine to 18,000 which of course was nearer to the possible 12,000, but as the years went on, it became clear that the molecule of insulin was smaller still.

There was evidence that the particle of 36,000 would break down in solution and gradually the evidence accumulated that the weight might even be not 12,000 but 6,000 though that this was so was not really clear until on the next slide,

Fred Sanger found in detail the structure of insulin and I'm sorry this is a little bit small to read from a distance as I realized when I looked at it a moment ago but essentially Sanger showed that the insulin molecule consisted of two chains

and the amino acids of which luckily you were shown larger pictures today are here given their abbreviated names, leucine, valine, tyrosine, alanine and a definite order along these two chains,

one of 21 residues, one of 30 residues joined by disulfide bridges at particular points. And of course as soon as this structure was produced, this chemical sequence, it became possible to synthesize insulin

and three different groups set out to synthesize insulin and succeeded, Zahn in Aachen, Katsoianis in Pittsburgh and a group of Chinese scientists in Shanghai and Peking. And the next slide shows a little bit

what is involved in this process. Could I have the next one please? This is taken from one of their papers. One has, they all synthesized separate peptides, different choices in the different laboratories and when they had synthesized

the A chain and the B chain and the sulfur, the cysteine residues at the right points, they reduced to the SH form and then left them in solution hopefully together with in the presence of air and some insulin formed.

Not a good organic chemists synthesis, just hoping that nature would work and insulin would form out of these separate chains. But it worked to a small degree. The first products got by this reaction

had only a few percent activity but out of them by suitable manipulation as on the next slide, crystals of insulin could be obtained. Again, not perhaps such very good looking crystals but you can see this very characteristic shape.

These are taken from the Chinese paper. Now, so you might say we know the structure of insulin but of course, it's a problem when one thinks of it, these two separated chains, how are they really disposed in three dimensions

and throughout this long period from the isolation of insulin and our first work, we really hoped to find the actual three dimensional arrangement of the insulin molecule in space.

Could I have the next slide please? We grew very much better crystals than the first ones. These are actually very beautiful crystals taken from a paper by Schlittkrull in Copenhagen who had developed crystallization techniques

to a fine art because he used actual crystalline insulin, very finely crystalline insulin as the basis of a method for giving slow acting insulin to diabetic patients, the novo insulins. He also found that in these particular crystals,

there were in general, two atoms of zinc per 36,000 molecular weight. How to find in detail the arrangement of the atoms

in space given these crystals was in principle, of course, known when we started out, when I started out in 1934. The next slide gives a letter, a little extract from a letter and I think Professor Ruzichka may recognize the writing,

although the signature isn't there. It's J.D. Bernal and he was very excited because of the x-ray photographs of insulin and he immediately suggested that one should change the zinc

for other heavier atoms and the implication of this was that we should be able to calculate the actual electron density in the crystal if we could observe changes in the intensity produced by these substitutions

and have an isomorphous series. Now actually, these are the substitutions that Scott had made in Toronto but the cadmium is rather little much heavier than zinc. And the actual differences that I could see at that time

were far too small to work on. Also, the method was untried for very much simpler compounds and it seemed desirable to have some simpler subject from which to try out the calculation

of the electron density by isomorphous replacement before we tackled insulin. The next slide just gives you, I think the background relations involved the actual form of the calculation

of the electron density in a crystal at a particular point, FHKL being derived from our measurements. Both FHKL and the phase angle appropriate alpha depend upon the positions of the atoms in the crystal

and if one can vary amongst these atoms that each has its appropriate scattering number and introduce heavier atoms, one can formally solve for the phase angle and this, of course, was what we hoped to set out to do.

And the first substance I tried this on in a very simplified form because the certain parts of the distribution were centrosymmetrical, which simplifies the phase relations, was penicillin.

And I think the next slide just shows you the look of the electron density in penicillin set up in a way that we used here for the first time but has become the way that is used in protein crystallography today. We calculate the electron density,

a series of figures at points over three-dimensional space from the measured intensities of the X-ray diffraction effects. We join up, we plot it out on sheets,

first of all in two dimensions, join contours of equal electron density to get representation of the atomic positions and so form a picture of actual individual atoms on sheets of transparent perspex

or glass or other material. And here, of course, you see separate atoms resolved and on the next slide, the formula that you could write as a result of this vision of the individual atomic positions in space.

And we were lucky that penicillin is at least a little bit connected, derived from a peptide kind of skeleton. In the case of insulin, as with other proteins, although the diffraction pattern that you see may look marvelous to your eyes, it doesn't really extend sufficiently far in space

to resolve individual atoms. The next picture shows you the consequences of this illustrated by a very simple molecule. Could I have the next one, please? If the space, in penicillin,

the actual pattern extended to about 0.8 of an angstrom resolution, this shows diketopiperazine seen at gradually reducing degrees of resolution.

This being the stage at which most proteins reach and at least a little bit better than the case of the insulin maps we were to calculate. You can see that some atoms almost appear individually, but in general, there is a lack of resolution

and one has to infer the positions of the atoms from the very characteristic shapes of the electron density peaks that one can calculate in a protein electron density map. Now our real trouble with insulin

turned out to be twofold. I must say, I was trying to introduce heavy atoms into insulin crystals parallel with John Kendrew and Max Proust at Cambridge in hemoglobin and myoglobin.

We had no single chemical point at which we could attach a heavy atom as in the case of hemoglobin, but we thought perhaps therefore we should follow

the line found by Kendrew and try to get just heavy atoms to pass into the crystals by other means. And actually it was Carlisle working with Bernal who first produced the idea of floating them into the crystal by just soaking them into solution.

So we tried a very great many experiments and we found it very difficult to find where the heavy atoms were. And here I have to start to take, go back a stage with my formulae.

The first thing one can calculate is not of course the electron density. One can only begin to calculate the Patterson taking the F squared values and obtaining a derived function of the electron density.

Well, the next slide I think should show you, could I have the next one please? A Patterson map for insulin. This is just a bit of a three-dimensional Patterson function and actually it's quite interesting. We started by thinking we could know nothing at all from it but you can see something from it.

This is just part of the function and in this pattern you do see in fact very roughly planes of symmetry which don't extend the whole way through the crystal but a very complicated pattern. The implication of those planes of symmetry is shown in the next slide.

It is that in insulin in the unit cell we have in this 36,000 mass six molecules and these six molecules, of course, three are arranged around the three-fold axis

standing in front of it, two, three but the other two, other triplet is arranged so that there's a two-fold axis relation approximately between them and so over here the Patterson function derived from this

shows planes of symmetry. The Patterson function is got by imagining you stand at each atom, put each atom in turn at the origin and plot the whole function around it so it's derived from it but much more complicated. At this point, two things became obvious.

First was that if there were only two zinc atoms in the crystal as Strickl had found they must lie on the three-fold axis and that was one great help and secondly that we would expect that other heavy atoms introduced into the crystal

should have this characteristic pattern and so if we did a difference map taking, measuring the intensities of the crystal and the heavy atom, we should see this sort of pattern in the difference map. The first point was quite all right,

two zinc atoms on the three-fold axis. The second turned out to be a small hair, they seldom if ever showed anything like the correct symmetry for reasons that are now obvious. The next slide I think just shows one of the difference, Patterson's of the heavy atoms,

a very complicated looking one but the little peaks, could you sharpen it a little, near the origin do imply that the heavy atom in this crystal has gone in near the three-fold axis. Now the heavy atom in this particular case was a very interesting one.

It was got by taking zinc insulin crystals, leaving them overnight in EDTA solution. This removed the zinc from the crystals and left us with crystals standing, looking rather sorry for themselves but still containing, but still giving x-ray

diffraction effects and if we then left them in lead overnight, the lead went into the positions occupied by the zinc and also into positions represented by this map and into some others

on the outside of the molecule and I will now cut a very long story short of how the different heavy atom positions were found and take you essentially to the answer. So on the next slide, I have first of all, the positions found for four different,

five different heavy atom containing crystals, plotted along the three-fold axis over a unit of volume, which covers a whole insulin hexamer.

And you can see, yes, sharpen a little. A lot of them come round the three-fold axis. The pattern doesn't have two-fold symmetry. There seem to be two general sorts of positions, one near the three-fold axis and one out on the edge of the molecule

as we later found it. And our difficulty was the sorting out of this very complicated pattern of heavy atoms before we could calculate the phase relations and the three-dimensional structure of insulin.

But now we pass to the actual electron density map. So the actual calculation came through the middle of July last year and we plotted it out and drew it up. And I won't show you the original map

because we have drawn it all much better since then. But the next slides are taken from a little model, very like the penicillin model, but representing the electron density

as we calculated it over the insulin hexamer. And because it's a large molecule and this hexamer is 30 angstroms deep and 46 angstroms across, the actual little blocks are broken up into five or six.

And I will just run the slides through so that you can imagine yourself looking at the molecule from top to bottom and thinking what you could see. The next, this is the top of it, a bit empty. Could we sharpen a little bit? Could we sharpen the, oh, that's better, yes.

This is the top of it looking a little bit empty, a rather complicated set of chains. These are half a dozen sheets projected together. Now on the next, could we pass on to the next one? Now you see we're thickening up tremendously

in one particular region, a lot of peaks coming together. And in this part, this is the beginning of the zinc atom and these three peaks are three peaks that are above it. And the next slide, it's becoming a bit more understandable,

sharpen up a bit more. Here's the zinc here and now these three peaks are three peaks that are underneath it. And they seem to run into a very definite chain structure here with peaks coming off it at certain intervals. And at this point, we began to be looking

at Sanger's sequence and seeing what we might, thinking what we might be seeing because we knew that there was chemical evidence that the histidine groups of insulin interacted with the zinc. And so we looked at what would be near the two histidine residues.

And in one case, histidine 10, if we take this as the histidine and come on on one side of it, it goes to a serine coming back towards the zinc. And on the other side, it goes out to a leucine and here is a peak with a kind of umbrella end

that could be a leucine. And from this point, we started tracing through the chain. Now I'll just show you the next two, if we could go on to the next one. This is passing below that particular point. And here we come to a highly empty region

in the middle of the crystal with chains coming in towards the central cavity. And this is the central cavity into which so many of the heavy atoms went in a somewhat random way in our substitution. And the next one passing down below this central point,

we begin to see another similar chain, the histidine on the other side coming in, the leucine on the other side, the serine on the other side. And they don't look absolutely the same, but they are sensibly the same sort of object.

And then down below that, the next slide, the sink with the other three little groups above it and the beginnings of the top or bottom of the molecule, the other side. So now you've passed through it all. And in about a week, we had traced very rapidly

where we thought everything was in the insulin molecule. But then we've taken a year going through everything, not quite a year actually, it isn't yet a while, it's only about eight months, isn't it?

Going through everything much more carefully, comparing the electron density map directly with a much more accurately built model, a model at two centimeters to the angstrom, which we could project directly into a map of the electron density at the same scale

by a mirror device produced in Oxford by Professor Fred Richards from Yale when he was visiting there for, which we call in Oxford, Fred's Folly, which doesn't mean a folly in a foolish sense,

but because a folly is a special kind of little home, house, a delight you see for all of us who work with it. So on the next slide, I think I have a little bit showing you how closely you can in fact fit

the appearance of the residues to the shape of the peaks as they have come out in this three-dimensional electron density map. And at the present moment, we have passed the whole way over the molecule

and recorded the positions in three dimensions to a first approximation for every atom in the molecule. So the next few slides, if I might have them, just show you the look of each chain in turn,

all the time projected in the same direction along the three-fold axis. And now this is one of the B chains of Sanger's molecule and it starts up at that end, B1, which is a phenylalanine group

and it runs in an extended form as far as this B10 histidine, which comes into the zinc and then it turns through three turns of a very good alpha helix and then it runs right down and right up in another extended chain.

It's better that afterwards you look at the model because then you can see what's happening. You come along the back, you go into this alpha helix straight down underneath and then you come up in a long chain right up here to the top again. And now the A chain on the next slide

just sort of nestles into the B chain. It just forms a little closed loop. You see starting here and going down there just lying on the face of the B chain and the disulfide links that link the two chains together link them,

one at this end to the alpha helix below and the other one at the other end to the other end of the alpha helix so that the alpha helix helical part of the B chain forms a kind of rigid skeleton on which the A chain lies.

And now over the other side, I'll show you the other two molecules. This is the second molecule in the asymmetric unit. Again, the B chain starting out a long stretch of extended chain, the alpha helix coming down here

and then we're seeing this long end of the B chain coming up on end. And on the next slide, the second A chain again forming this small, very compact loop within the,

over the surface of the B chain. And now you put these two together and in the next slide you have the same projection of the insulin dimer with the two molecules together. And I think you can begin to see

that whereas this line is an approximate two-fold axis, it isn't by any means an exact two-fold axis, that there are slight differences in the relative positions of the groups on either side of it. Now, the actual shape of this dimer,

and I'm sorry that I haven't got it here in bodily presence before you, is a rather elongated oval. The next slide shows you a view along the side of the dimer

as seen in this more accurate model that we have in Oxford, which the boys won't let move away from Oxford even to be shown at the Royal Society at the present. We're gonna say it regardless of such a delicate object that has to be preserved in the exact form

in which they have built it. But anyway, here is one molecule and the second molecule packed together with the two zinc atoms along the three-fold axis, each of which has directed at it a histidine residue.

And if this is viewed from the direction of the zinc atoms, and the next slide shows you the view from the direction of, from the outside of the molecule. And now if you could, perhaps we might, yes,

have a little dimming and the rest of it at this stage. This shows you that these two molecules are fitted together in a very intricate way. That's better. The main groups that pack together are nonpolar residues, but there is a very long region here

where hydrogen bonds form between the extended ends of the B chain. This is the extended end from one molecule going up in this direction, and this is the extended end of the other one coming down in this direction. And the hydrogen bonds form across here,

you can see them in red, so that this is part of an anti-parallel pleated sheet, and it's between the two insulin molecules. Now the next slide shows this just drawn out so that you can see the way that it goes

with the hydrogen bonds forming on either side of the two-fold axis in this generally anti-parallel pleated sheet form. But now if we look down the two-fold, onto the two-fold axis, that's to say the way we've been looking most of the time,

we can see that while these groups are quite nicely related, the individual residues are not. Could we have the next slide? This is looking onto it, and now the residues that project are ones like here are the phenylalanines,

and really these phenylalanines to be related by the two-fold axis, one should turn away in one direction and the other one the other direction. But in fact, the second one turns across it in order to pack side by side nicely as phenyl groups like to do

disregarding this two-fold symmetry. And up here again, the valine groups don't seem to have the same orientation, and again, it looks as if they just turn to fit nicely into one another. There's a non-polar interaction at that point.

And at the end, we have the B13 glutamate ions going into the center, and they are the ones which attract these heavy atoms into the central pool of the insulin molecule. Now, the next slide I think shows you,

just to give you an impression, to go on from the dimer to the hexamer. This is a projection of all of the atoms in the unit cell of insulin as so far placed by us. That's to say the atoms belonging to the molecule.

And you can see that these six insulin molecules have formed a very good spheroidal object. And these spheroidal objects pack together in the next slide in a way that just is clearly quite good close packing of,

it's actually cubic body-centered close packing, but you can see the way these spherical objects, the insulin hexamers, fit together in the crystal structure. So now we should think in more detail

about the complications of this molecule and what it is doing in nature. And this is still a very great mystery. We have this very complicated arrangement of chains within the molecule, which yet ends up, you see, producing

a very remarkable, smoothly spheroidal object. And this smoothly spheroidal object, it seems, does exist in the living body. Of course, the processes by which the body makes insulin

are certainly quite different from those used in the test tube. And to start with, evidence has been acquired very recently that insulin is not synthesized in the body as two separate chains, but by a single chain, pro-insulin,

as discovered by Steiner in Chicago. Could I have the next slide? This is the way that the body synthesizes insulin as a single chain governed by the usual processes

of no doubt genetic processes that lead to a single chain. And then at a point in the pancreas, when it is transferred from the site at which it is synthesized to the storage cells,

the eyelid cells of Langerhans, a proteolytic enzyme chips off this very long peptide fragment that connects the two chains. There is a sort of curious point about this

if one looks at the molecule, because one might think that this connecting peptide had to bring certain parts of the molecule near together. Actually these parts are very close together, the actual beginning of the A chain at the end of the B chain in space.

The beginning of the A chain is just down about here and the end of the B chain here in the molecule. See, only about two residues apart so that one feels that the whole of this long peptide isn't necessary only for bringing the chain structure

for encouraging the particular folding of the chain that we have in the insulin molecule. There are very interesting points about proinsulin. When it does seem that if you break the disulfide bridges in proinsulin,

the chains form again much better than in insulin and perhaps the long chain provides some kind of a template on which the molecule lies which encourages the correct form of the insulin molecule

or it provides some kind of protection for the molecule in its transport to the storage site. Now the next slide shows a picture of the actual storage of insulin in the islet cells

and now this is taken actually from rat islet cells by Dr. Howells in Sussex and I think you can see that the little insulin granules look sensibly as if they were crystalline. They look like little crystals and indeed in some animals you can see quite large crystals

but here they have the form very much the same form as the crystals that we have been working on. If slices of these little crystals viewed in higher power in the electron microscope in the next slide, you can see a repeat distance across them.

In some views you get a view of lines as of molecules and if you look at the size of the molecule and the distance across of these lines, the distance is about 50 angstroms which is sensibly the side of the rhombohedral unit,

the distance apart of repeating hexamers. So it seems almost certain that in the actual insulin, the islet cells, the insulin is stored in these zinc containing hexamers.

But now the problem is what is it doing? And here I have to throw this open really to the audience. The next slide just shows some of the different kinds of biological processes in which insulin has been implicated

by a large number of experiments and of these, of course, the original ones that seemed most important were ones involving the metabolism of glucose and particularly, possibly the transport of glucose through membranes.

We invite observations on this molecule at this point but I might just add one more piece of evidence to it. We thought it would be interesting to look into the actual position at which the insulin molecule

appeared to be able to change in different species without losing its effect. That's to say to map the species differences in insulin following the same kinds of ideas

that Dr. Singh was talking about earlier. So in the next slide, I have not all but quite a number of the, if it could be sharpened a little bit more, yes, that's better. You can see that many of the residues are allowed to change.

And when you look at the ones that don't change, they are generally, well, here the cysteine hasn't changed, the leucine hasn't changed, A19 tyrosine, A21 aspartine, asparagine hasn't changed. But a tremendous lot of changes have been observed.

In the next slide, we have just got a little map showing where the invariant residues are in the molecule as we see it, this is a single molecule. And they're very largely ones that seem to be involved

in pulling the molecule into its actual conformation. The cysteine residues, of course, but also two or three leucines that are involved in, as you might say, the heart of the molecule.

And then this leaves us with just one or two on the outside, and A19 tyrosine is one particularly, and the asparagine, which are also invariant so far. The last slide just gives you a picture

of the molecule, it's just this molecule as you look at it here, but drawn from the actual electron density map, the actual positions taken out of the electron density map.

And it's really to give you an impression of what a approaching reagent would be seeing. You see this face full of detail, and one asks oneself what parts of it are important for the actual biological reaction. Is it a particular region on this face,

or is it the whole organization, and is it the monomer, a single insulin molecule, as Sanger believes, or the dimer, as they believe in York, or the hexamers, we sometimes believe ourselves when we see

the very beautifully organized object that it is with pockets in it that might carry things about, you know. But this is still to be found out. And before I leave this subject, I should say two things. There are other crystalline modifications of insulin,

very interesting ones, which are being worked on at the moment. The orthorhombic form, which contains no zinc in Columbia by Dr. Barbara Lowe, and the monoclinic form, which contains hexamers as its asymmetric unit by Michael Rossman in Purdue.

And at the end, I should also say that in this particular research, the sum total of the coworkers was much, the part played by the coworkers,

the sum total was much greater than the part played by the professor. Do take your words from you, Professor Rizhichka. And that I should mention, particularly those who carried through the last stages of the analysis. It had rather a long history.

Margaret Adam and Tom Blundell from Oxford, Guy Dodson from New Zealand, Eleanor Dodson from Australia. They met and married in Oxford. And Vijayan from Bangalore. And I must say we partly chose him because his name meant victory.

Thank you.