STRUCTURE OF PROTEINS1

J. D. Bernal, London

The structure of proteins represents one of the most important unsolved problems in the borderland between chemistry and biology. We must admit that up to now we do not possess the key to the solution of this problem, although in recent years many facts have been obtained that make it possible to find new ways of approaching it.

The question of the structure of proteins falls into two parts: first, the question of the form and properties of the protein molecule, and second, the question of its internal structure. In view of the extreme instability of the protein molecule, its investigation is possible only by the most delicate physical methods. The greatest success has fallen to three methods: centrifugation, electrical, and roentgenographic methods. Fundamental results were obtained with the aid of ultracentrifuges, and almost all of them were achieved in Svedberg’s laboratory.[^2] In sufficiently strong centrifugal fields, protein molecules precipitate out of solution, and with the aid of the corresponding sedimentation constant sufficiently accurate values for molecular weights can be obtained. It is true that doubts still remain as to what exactly is measured in this case: whether it is the protein molecule, since one must always take into account the inevitable association with these molecules of the liquid in which the molecules are suspended.

Svedberg’s most striking discovery is the establishment of the fact that the resulting molecular weights of proteins are close to certain more or less constant numbers, which are related to one another as small whole numbers. This discovery makes very probable the idea that all proteins are composed of certain common units. Up to now, however, it is difficult to say what such a unit is. Until recently, the protein unit was assigned a molecular weight of 35,000; however, recently it has been possible to isolate proteins with molecular weights of 17,000 and even 10,000. In addition, critical consideration of the available data shows that the law of multiple ratios is not observed especially strictly, and that we have merely a concentration of molecular weights within certain intervals. There is no doubt, however, that in proteins close to one another in their chemical properties, multiple ratios are observed fairly strictly. Thus, for example, both hemoglobins and hemocyanins can be reversibly decomposed into 2, 4, or 8 parts. It still remains doubtful, however, whether these ratios can be extended to all proteins. In particular, the study of viscosity has shown that proteins characterized by one and the same molecular weight, such as, for example, insulin and gliadin, may differ from one another in form in an entirely substantial way: the first is characterized by a very flattened spheroidal form, whereas the second usually has the form of an elongated rod. It is therefore hardly possible to imagine a common physical structure for these two proteins.

It was suggested that the diversity of molecular weights of proteins within one and the same Svedberg class is due to differences in the number of amino-acid groups \(K\) contained in the elementary material and makes it possible to establish whether this explanation is correct. The number of 288 amino-acid residues in the class of proteins characterized by a molecular weight of 35,000 must in any case, at the present day, also be regarded as a sharply rounded-off assumption.

The second method of a more profound approach to questions of the structure of proteins is reduced to the study of the electrical properties of their molecules. The work of Cohn, Tiselius, and others has led to the conception of the protein molecule in solution as a particle whose surface is rather abundantly covered with positive and negative charges, conditioned by acid and basic groups of amino-acid side chains. The number of these groups is determined by the conditions of the medium. The protein molecule is thus, in its outer parts, of an essentially ionic character and probably carries around itself an ionic atmosphere, sufficiently widely extending in the solution (water) in which it is found.

The third method of approach to the structure of proteins is the roentgenographic method. The beautiful crystalline forms observed in proteins, which have been known for a century, have always attracted the attention of roentgenographers; but until the last five years all attempts at analysis failed because of the instability of protein crystals and their extraordinarily small dimensions. Only in 1934 was it finally possible to emphasize by roentgenographic study a considerable number of typical protein crystals. The corresponding experiments were carried out on crystals remaining in their mother liquor (in small tubes), since most, and perhaps all, proteins break down when one attempts to examine them in the dry state. Already the very first results of these investigations led to the establishment of very important and still unknown facts. First of all it turned out that roentgenograms of such crystals are distinguished by an extraordinary sharpness of pattern. All of them lead to very large elementary cells with a very large number of reflections on the oscillation roentgenogram; moreover, these reflections are obtained sharp even at very large angles corresponding to interplanar distances of \(2 \text{ Å}\) and less. This shows that protein molecules not only are approximately similar in form and size, but that in each separate case they are in general identical and possess a regular internal structure down to atomic dimensions.

Knowing the dimensions of the elementary cell and the specific weights of the corresponding crystals, one may calculate the weight of the material inside each elementary cell and the corresponding numbers; however, they do not give direct molecular weights for the following reasons. First, we do not know the number of molecules in the elementary cell; secondly, it is difficult to determine what part of all the matter inside the elementary cell is formed by protein substance and what part is water bound to it. Measuring the number of the latter by means of some drying procedure, we nevertheless arrive at the figure of “dry weight”—a figure which is very unreliable and which can be compared with the results given by the centrifuge and chemical analyses. The corresponding results showed that the figures fit very well into the system, if it is accepted that in each cell there are either four molecules, or that they contain 2, 4, or 8. Thus, roentgenographic results make it possible to calculate molecular weights quite accurately, although they cannot determine the corresponding object and boundary unequivocally. One may think that most proteins are built of subunits which, although they have approximately identical chemical weights, are not chemically identical, and therefore the corresponding proteins are better called not molecular but macromolecular compounds.

The most remarkable result of the measurement of the elementary cells of crystalline proteins is the considerable change in these cells which occurs upon drying the proteins. In many cases the diminution of the cell reaches 50%. Still more remarkable is the fact that this compression of the cell very often is restricted to only one or two dimensions of the cell. Two explanations of this fact are possible. The first proceeds from the supposition that protein molecules are bound in aggregates in an exceedingly loose manner and that, when water is removed, these aggregates are destroyed, leaving a skeleton of molecules packed more closely. The other explanation proceeds from the idea of separate molecules which are divided from one another by ionic atmospheres determining the corresponding charges and actually being intermediate layers of free water molecules. Probably, in the general case, we have both factors, with a great predominance of one or the other in individual cases. Thus, for hemoglobin, characterized on drying by a very considerable compression of the elementary cell from 55 to 38 Å in one direction, it is very difficult to imagine the presence of an ionic atmosphere, since the corresponding crystals contain almost no layers of salts and are characterized by an isoelectric point pH = 6.8. Conversely, the remarkable properties of the tobacco mosaic virus are very difficult to explain by any other supposition. This virus is characterized by long thin particles which tend to arrange themselves parallel to one another, and under these conditions the water content reaches 1.3%; and, on the contrary, even up to 1.5%. It would be difficult to imagine any other cause of this regularity without the corresponding ionic atmospheres. Very recently we have shown that the equilibrium distance between particles is determined to a very considerable extent by pH and salt concentrations; thus, for example, we have 320 Å at pH = 7 and 206 at pH = 3—4. There is no doubt that a quantitative theory of these facts can already now be created, and we find its beginnings in the works of Langmuir⁴ and Levine⁵.

In any case, we can now more or less definitely imagine the general picture of the surface of protein molecules. We must think of these molecules as spheroidal bodies with dimensions from 30 to 100 Å, covered with hydrophilic groups which have charges of both signs. Moreover, each such molecule in solution carries with it an atmosphere of ions.

The question of the internal structure of protein molecules appears incomparably more complex. Any at all detailed picture of the structure of a protein must first of all explain those general properties which are characteristic of all proteins and which are manifested in the corresponding chemical and physical properties, and only further those specific properties by which the data or other proteins are characterized. It is very probable, although this is impossible to prove rigorously, that the first series of properties is determined by the sameness of the general character of the distribution of amino acids. We have more or less definite indications on this point as a result of the study of fibrous proteins. In this field, owing to the technical difficulties of the corresponding products and because of the greater convenience of working with them, we already have very considerable results, connected chiefly with the name of Astbury⁶, who conducts his investigations very persistently according to a well-developed plan.

The fully extended fiber of α-keratin is characterized by a very sharply expressed identity period of 3.5 Å, determining the length of one amino-acid residue, and, besides this, by two periods, 10 and 4.5 Å, corresponding to directions at right angles to the axes of the chains. One of Astbury’s most important discoveries was the establishment of the fact that these latter directions of periodicity also make right angles with one another. In other words, the repeating unit in the structure of a fully extended protein is characterized by definite dimensions in all three dimensions. The values of these periods

also apparently raises no doubts; namely, the distance of 10 Å corresponds to the length of the side chains of amino-acid residues, whereas 4.5 Å is the distance between the main chains (“backbones”) connected with one another by CO and NH groups. In the shorter α-form the distance of 10 Å is retained, while the distance of 4.5 Å disappears, which must be regarded as indicating that the chains “fold up,” but that this folding takes place in one plane, and not in two, as occurs in rubber. Up to now it remains unclear what the actual mechanism of this folding is, but it is beyond doubt that it is extremely important, since it is probably the very same mechanism that operates in muscle fibers.

The significance of these results for the general problem of protein is determined, first, by the fact that all soluble proteins so far studied, as a result of denaturation, are transformed into a fibrous material which can be oriented and which, apparently, in the basic features of its structure is identical with β-keratin. And, secondly, it is beyond doubt that the real changes which occur in a crystalline protein during its denaturation must be very considerable. This follows from the close similarity between the Debyegrams of crystalline proteins before and after denaturation, and also from the observations of Astbury, Dickinson, and Bailey^7, that monocrystals of excelsin, upon partial denaturation, yield fibers oriented along the axes of the crystal. Perutz^8 has shown by an entirely different method, namely by studying the extinction spectrum in different directions of the crystal, that the position of the prosthetic groups in hemoglobin does not change upon denaturation.

We have a very large number of works with photographs of crystalline proteins, showing characteristic differences of individual proteins, but a more detailed approach to the corresponding photographs is still beyond our power: we are not yet in a position to interpret the corresponding diffraction patterns. The roentgenograms of crystalline proteins are characterized by hundreds of individual spots with very considerable differences in intensity. In spite of this, direct analysis of these roentgenograms is impossible because we cannot establish the phases for reflections corresponding to individual spots. Apparently, however, the corresponding uncertainty can be bypassed with the aid of purely physical devices, for example by introducing heavy atoms, or by observing those changes in intensity which occur upon dehydration. Up to now, however, in practice this has not been done. Moreover, for all the proteins studied, with the exception of one case, the question is complicated by the presence in the cell of more than one Svedberg unit.

The corresponding roentgenogram in these cases is determined both by the arrangement of the molecules in the cell and by the internal structure of each molecule. Fortunately, we have one exception, namely insulin, which is characterized by only one molecule in the cell. Crowfoot^9 carried out an X-ray study of dry crystals of insulin with determination of the intensities of the reflections. Unfortunately, it was not possible to study wet crystals; dry crystals, however, do not give reflections corresponding to intermolecular distances below 8 Å. As a result, we do not have the initial data for judging the fine structure of the molecule. The results obtained with respect to the coarse structure, however, are very convincing and are also confirmed by the corresponding projections of Patterson’s double series^1).

^1) Patterson’s double series is the name given to a two-dimensional Fourier series whose coefficients are the experimentally measured intensities of X-ray diffraction spots. Patterson showed that the sum of such a series possesses the property that its maxima indicate the distances between the places of greatest concentration of scattering mass. Patterson’s series gives sufficiently precise indications of ...

Generally speaking, it is quite natural to assume—although in fact it will be only an arbitrary simplification—that the peaks in a Patterson projection correspond to the distances between a small number of points with high concentrations of scattering matter. Attempts have been made to reduce the analysis to finding that spatial pattern of points which gives maxima in the required places. All such attempts must be regarded as unsatisfactory, since it is easy to find a very large number of such point patterns which will give the correct distances on the basic projection and which, however, will not correspond to other sections of the cell. All attempts made up to now2 ultimately come down to an arbitrary choice of a certain number of vectors, and the regularities discovered disappear when the latter are chosen differently.[^11] The failure with which attempts to find a point solution of the problem have so far ended leads to the conclusion that we are dealing with groups characterized by dimensions of the same order as the mutual distances, and that, in particular, the observed strengthening of reflections at distances of 10 and 4.5 Å corresponds to the arrangement of side chains or to backbone distances analogous to those with which we are concerned in fibrous proteins.

The X-ray material available at the present time still does not provide any more or less detailed picture of the structure of proteins. It may be said that, up to now, the principal significance of the X-ray methods has consisted in refuting those hypothetical structures which had been proposed earlier. Nevertheless, even now it is possible to formulate, in more general terms, the various possible modes of grouping and to propose certain working hypotheses which might serve as guides in further research work.

The formal questions which arise in the study of the structure of the protein molecule are as follows.

a) What is the nature of the bond between amino-acid residues?

b) Do these bonds participate throughout the entire mass of the molecule, or do they correspond only in certain parts? In other words, is the entire protein molecule connected only by primary valence bonds, or is this molecule composed of subunits connected in some other way?

c) If these subunits exist, what is the nature of the bond between them?

As regards the first question, the difficulty in explaining the formation of spherical molecules from linear peptide chains has led to the idea of another mode of bonding, permitting each amino-acid residue to be connected not with only two neighbors, but with four. Although at first glance this hypothesis seems to have theoretical advantages, there are nevertheless no chemical grounds for it, and, on the contrary, from the chemical-experimental side it meets with serious objections. In protein chemistry we already stand before a large number of unknown factors, which makes the introduction of assumptions doubtful from the chemical point of view undesirable.

The question of the character of the units composing the protein molecule remains, independently of whether the bond between amino-acid residues is peptide or not. Of course, with multiple bonding it is easier to construct models of the cellular or spheroidal type, but, as has already been pointed out, such continuous structures are difficult to reconcile with the 10 Å gaps which the X-ray diagrams systematically show. If a peptide bond is assumed, however, it becomes difficult—

To be continued.

by isolating individual units. It is difficult to imagine any arrangement or loop that would allow a separate chain to occupy the corresponding section of space and at the same time would not be so complex as to make its formation by a natural process exceptionally improbable. However, even now we possess abundant evidence in favor of the fact that, at least, large protein molecules are not monomeric formations. This follows first of all from the fact that many such molecules in solution can be split into particles with a molecular weight of about 10,000, and it is probable that this figure is not limiting and that smaller particles are simply more difficult to isolate and measure.

Two characteristic features of the X-ray method also point to these sub-units, namely: the high symmetry of protein crystals and periods of 10 Å. The symmetry of protein crystals is considerably higher than might be expected statistically for such complex compounds. This symmetry undoubtedly indicates that, although each molecule may be assembled from asymmetric sub-units, the latter are oriented in a symmetrical fashion. The dimensions of these sub-units must lie between the dimensions of the smallest observed protein molecules, i.e., those with a molecular weight of about 9,000, and the dimensions of individual amino acids with a molecular weight, on average, of about 120. Thus each sub-unit is some divisor of the number 72 of amino-acid residues. The uncertainty is caused by the fact that such sub-units need not all be identical, although the condition of symmetry limits this inequality to a few sorts. The characteristic trigonal symmetry of the crystal permits the conclusion that the asymmetric unit must contain one third or an even smaller part of this number, i.e., must contain 24, 12, or 8 amino-acid residues.

The question of the structure of the sub-unit at first glance presents the same difficulties as the structure of the molecules themselves. In reality, however, with a decrease in the number of residues the difficulties and improbabilities in the folding process become considerably smaller, especially if we postulate (with considerable justification) that the sub-units are closed peptide rings. Such rings must necessarily coil up owing to the mutual attraction of the positively and negatively charged amino-acid categories, and models of such closed chains can easily be constructed while preserving all those distances that have been found for similar groups in other compounds. This time the chief difficulty is reduced to the fact that we have just a very large number of such models and have no criterion for choosing among them. The folding method in chains within the sub-units is, very probably, similar to that which characterizes the compressed forms of fibrous proteins.

The postulation of sub-units leads to a number of further questions concerning the structure of the molecules. In fact, the latter must be bound together sufficiently strongly that such molecules should not disintegrate either in water or in ionic solution. Chemistry gives us a very limited number of such bonds. Ionic bonds may be disregarded, since the latter will be hydrated and destroyed in aqueous solution. There remains the possibility of aminocarboxyl bonds between the ends of side chains. But this possibility has much against it. It seems more probable that the bonds will be of one of the following two types: an S—S bond and association of hydrophobic groups. In some cases one may expect both types of bond. With a small number of sub-units, the number of S—S bonds necessary for fastening the entire molecule is provided by the number of sulfur atoms found in all proteins studied so far. And the extraordinary change in activity that proteins undergo when the S—S bond is broken shows that this bond indeed plays a fundamental role in the structure of the protein molecule. The characteristic properties of hydrophobic groups in proteins undoubtedly

are also an important factor in mutual association. As Danielli and Langmuir^14 have shown on the basis of their work on surface films, globular molecules in solution must arrange themselves in such a way that their hydrophobic groups are out of contact with water, i.e., the latter must come into contact with one another. At the surface of the solution the molecule spreads out into a film 10 Å thick, in which all the hydrophobic groups are pushed outward so as not to be in contact with particles of water. Thus, a considerable associating force is indeed present here, which in the final analysis is due not so much to the mutual attraction of the hydrophobic groups as chiefly to their repulsion by the aqueous medium.

Langmuir uses only the picture just described to support the hypothesis of cyclol cells; however, it is clear that these are two different and independent things, and that the model, as developed by us, can in particular explain the phenomena of denaturation very satisfactorily, especially in surface layers. At the moment when subunits appear at the surface, their rings fall into one plane, and different rings can interact in accordance with their well-known process of polymerization of cyclic chains, which leads to the formation of those fibers whose presence was shown by Astbury in films. Such a process of polymerization takes place rather slowly, as Danielli’s work has shown.

The picture we have drawn is still far from final or even simply satisfactory. The principal point requiring explanation is the more or less exact mechanism of folding and unfolding of peptide chains; to obtain a corresponding solution, we shall probably have to wait a considerable time, until the technique of radiography and other analogous methods has advanced far enough. The question of the structure of proteins we may, however, already now regard as having a concrete solution; but substantial progress can be obtained only when greater agreement is achieved among individual investigators, which we do not have at present. The work remains not only uncoordinated, but different authors are investigating different proteins by different methods, whereas it is beyond doubt that a concentrated and strictly planned attack on the problem of protein structure would save much effort and would lead to a more rapid solution of the task.

LITERATURE

1. T. Svedberg, Proc. Roy. Soc., A 170, 40, 1939.
2. J. D. Bernal and D. Crowfoot, Nature, 133, 794, 1934.
3. J. D. Bernal et al., Nature, 141, 521, 1938.
4. I. Langmuir, J. Chem. Phys., 6, 873, 1938.
5. S. Levine, Proc. Roy. Soc., A 170, 145, 1939.
6. W. T. Astbury, Fundamentals of Fibre Structure, Phil. Trans., 232, 333, 1933.
7. Biochem. J., 29, 2351, 1935.
8. Private communication.
9. D. Crowfoot, Proc. Roy. Soc., A 164, 580, 1938.
10. J. Am. Chem. Soc., 60, 2247, 1938.
11. J. D. Bernal, Nature, 143, 74, 1939.
12. A. Neuberger, Proc. Roy. Soc., A 170, 64, 1939.
13. J. F. Danielli, Proc. Roy. Soc., A 170, 73, 1939.
14. I. Langmuir and D. Wrinch, Nature, 143, 49, 1939.

1. Nature, 143, 663, 1939. Translated by N. V. Belov. ↩

2. —in the structure only when there is a very large number of members of the series and for crystals that are not very complex. Bernal’s criticism of attempts to modify Patterson’s method for the analysis of protein structure is entirely well founded. Undoubtedly, the problem of protein structure cannot be solved by this route. ↩