THE PHYSICAL CHEMISTRY OF PROTEINS SIXTH SPIERS MEMORIAL LECTURE THE MOLECULAR SHAPES OF CERTAIN PROTEINS AND SOME OF THEIR INTERACTIONS WITH OTHER SUBSTANCES BY JOHN T. EDSALL" University Laboratory of Physical Chemistry Related to Medicine and Public Health Harvard University Boston Mass. Received 30th September 1952 The Discussions of the Faraday Society have developed in recent years into events of international significance for many scientists. To-day as in a number of the previous Discussions the focus of interest is on the great border area between the provinces of physics chemistry and biology. In those earlier Discussions many ideas of extraordinary fruitfulness were presented and vigorously debated before they had become familiar and respectable.I have often turned back for illumination to the printed record of their proceedings. Three years ago I had for the first time the opportunity to take part in one of these Discussions-that on Lipoproteins at Birmingham. Hence I feel deeply the honour of giving the Sixth Spiers Memorial Lecture. It was never my good fortune to know Mr. Spiers personally but it was he especially who guided the pattern of this Society's development from small beginnings until it grew into an active and significant organization and its discussion meetings became as they are to-day the focal point for investigators from every land where science is free and active. The untimely death of Mr. Spiers occurred in 1926. At that time although the Faraday Society had held many conferences of international importance on the border line of physical chemistry and other sciences it is scarcely likely that it could have contemplated a two-day discussion on the physical chemistry of the proteins.It is doubtful indeed whether in 1926 more than half a dozen laboratories throughout the world could have made effective contributions to such a discussion. L can well remember that year for in the midst of it I returned to Harvard after two years of study here in Cambridge where I had imbibed the broad vision of the scope and the future of biochemistry that was to be found in Cambridge in Hopkins's laboratory. Also I had had the good fortune to know that great pioneer in the physical chemistry of the proteins Sir William Hardy then no longer active in that field but still with a freshness of mind a vitality and zest in life that I have seldom seen matched in other men.In those days we scarcely knew the molecular weight of a single protein. The view was indeed held by some distinguished chemists that proteins were actually very small molecules with molecular weights below 1000. This theory was short- lived being quite demolished late in 1926 by a few well-chosen freezing-point * Fulbright Visiting Lecturer University of Cambridge 1952. 9 A SIXTH SPIERS MEMORIAL LECTURE 10 measurements by Cohn and Conant. Sorensen indeed had carried out a decade earlier his classical investigations on egg albumin and his careful studies of osmotic pressure had given the basic information for the determination of its molecular size.Adair had propounded in 1925 the startling view that the molecular weight of mammalian haemoglobin was near 67,000 and that the molecule therefore contained four iron atoms. I can well remember the profound scepticism with which this report was received by more than one professor of great eminence; and the doubts persisted in high quarters until the ultracentrifugal measurements of Svedberg a year or two later gave complete support to Adair’s conclusions. The ultracentrifuge had indeed been under development since about 1922 but there were few at that time who anticipated its immense and revolutionary influence on protein chemistry. If relatively little was known of the sizes of protein molecules almost nothing was known about their shapes; indeed the problem of molecular shape was scarcely envisaged at that time by protein chemists.The air was still filled with the strife of the conflict between Jacques Loeb and the colloid chemists of the school of Wolfgang Ostwald. Loeb’s dzmonstration that some of the colloidal properties of proteins could be explained on simple chemical principles a demon- stration expounded brilliantly and with a fighting spirit appeared to many besides myself as if it were a new revelation. To-day I still appreciate the value of Loeb’s contribution. but I am also aware that T. B. Osborne in New Haven many years before had been quietly carrying on his work on the plant proteins treating them as chemical substances to be understood on general chemical principles and paying no attention to the clamour of conflicting schools.Also there was one idea expressed by distinguished protein chemists as late as twenty years ago that proteins furnished essentially the inert structural frame- work of the animal organism while the active pattern of biochemical trans- formations was determined by the interplay of other molecules of quite a different sort. Certainly this view was never upheld in the laboratory headed by Edwin Cohn. Nevertheless the fact that such a belief could have prevailed among many biochemists even for a few years seems astonishing to-day; and indeed it represented the denial of a much older tradition going back over a century to Liebig and Mulder which held that proteins were indeed of absolutely prime importance to the living organism.To Mulder in 1840 this must have been largely a matter of faith and intuition ; but now in the last quarter century we have seen that faith abundantly justified by works. I have tried to recall here something of the temper of the times the prevailing atmosphere of thought among the relatively small band of protein chemists in 1926 when I came back from Cambridge to Harvard as a third year medical student and started working in my spare time under the direction of Edwin Cohn on a viscous and peculiar globulin from skeletal muscle continuing work which had been begun by W. T. Salter. Later we called this globulin myosin and still later it was rechristened by Szent-Gyorgyi as a form of actomyosin. It proved a most refractory and difficult substance to work with but in the midst of our struggles Alexander von Muralt arrived from Switzerland and introduced the measurement of double refraction into the laboratory.His brilliant contributions both to experiment and to thinking brought us face to face with the problem of molecular shape; from then on we knew that although many proteins might be more or less globular some-and these indeed of supreme importance-were long and filamentous even in solution. In these casual remarks I shall not try to review the intervening years with the colossal advances that they have seen. Protein chemistry which was a rela- tively unfashionable subject a quarter of a century ago has now become almost embarrassingly fashionable.It does not seem likely to go out of fashion unless mankind decides to turn away from science and repudiate it altogether-a possible but I still hope a relatively improbable event. JOHN T . EDSALL 11 However of all the advances in this quarter century perhaps none will stand out more than two that have cccurred in the last two years-the elucidation of the complete sequence of amino acid residues in the A and B chains of insulin by Sanger Tuppy and Thompson,l and the detailed formulation of spatial con- figurations of the polypeptide chain by Pauling and Corey.2 The aspects of protein chemistry in which these developments are included will not be dealt with by other authors in the present Discussion but these achievements are bound to influence the thinking of all of us so profoundly that I cannot leave them without mention here.We cannot yet write a structural formula for insulin until the arrangement of the disulphide bridges is specified but that the goal has now been so nearly approached is an achievement which would have seemed incredible even five years ago. Protein chemists will now really begin to think for the first time in terms of structural organic chemistry and all of us here who are physical chemists will be permeated more and more by that influence. However in protein chemistry more than anywhere else structural chemistry is of little use without stereochemistry. Proteins with their marvellously specific patterns are the supreme examples of molecillar geometry. The work of Pauling and Corey is neither the beginning nor the end of the deciphering of these patterns; their achievement is built on the work of many and brilliant predecessors and what they have done is only a beginning.Their work still has its sharp critics; but I have been led more and more to the conclusion that their a-helix does quite closely represent the spatial pattern of long chain synthetic polyamino acids and not improbably also that of the fibrous proteins which are found in the cc configuration. Also-though there is less concrete evidence for this view-the pleated sheets which they have proposed appear to be by far the best existing models for the extended pattern of protein chains in the ,B configuration. In several crystals of corpuscular proteins the presence of intramolecular bundles of parallel chains has been inferred from the analysis of X-ray data.In the case of haemoglobin Bragg Howells and Perutz 3 inferred that these chains were polypeptide chains in the cc configuration. Well-defined chains were in- ferred by Kendrew 4 in the myoglobin molecule. The study of crystalline ribo- nuclease 5 has also suggested the presence of chains parallel to the c-axis of the crystal but the configuration of these chains seems still to be a matter for dispute. Some of the most striking evidence for chain structures in a protein crystal comes from the recent work of my colleague Barbara W. Low at Harvard on the ortho- rhombic crystals of acid insulin sulphate,6 originally prepared in our laboratory by Dr. Eric Ellenbogen working with Prof.Oncley. The three-dimensional Patterson diagrams of these crystals in the air-dried form show a most striking pattern of parallel rods with axes parallel to the a-axis of the unit cell which has a length of approximately 45 A. Dr. Low has constructed possible packing models for the polypeptide chains in this crystal taking account of Sanger’s findings on the sequence of amino acid residues. I should like to turn now from these brief remarks on protein structure at the deeper level to some of the problems related to the general size shape and inter- actions of proteins. I shall confine myself to two proteins with which I have become particularly familiar in recent years fibrinogen and serum albumin FIBRINOGEN SIZE AND SHAPE Fibrinogen occupies a place apart among the plasma proteins because of its high molecular asymmetry and its low solubility the latter property rendering it the first major constituent of plasma to be precipitated in almost any system of fractionation.Its high viscosity its marked flow birefringence its high frictional ratio and its significant angular dissymmetry of light scattering all mark it out as a molecule of shape very far from spherical. There are now data available on both human and bovine fibrinogen in solution which are too numerous to 12 S I X T H SPIERS MEMORIAL LECTURE discuss here in detail. They are summarized in table 1 and appear to indicate that human and bovine fibrinogen are remarkably similar in size and shape. The molecular weight indeed appears at present more uncertain than the absolute dimensions-many workers have accepted a value of 450,000-500,000 but the TABLE 1 .-PHYSICAL CONSTANTS OF HUMAN AND BOVINE FIBRINOGEN sedimentation constant S ~ O ~ (Svedberg units) partial specific volume V20 intrinsic viscosity Eio rotary diffusion constant 0 2 0 (sec-1) (from double refraction of flow) molecular weight (in thousands) osmotic pressure (Mo) sedimentation and viscosity (Mg) light scattering molecular length (A) double refraction of flow osmotic pressure and viscosity sedimentation and viscosity light scattering molecular diameter (A) H = Holmberg (Arkiv.Kemi Min. Geol. A 1944 17 no. 28). NoTEs.-T~~s table is also published in BZood Cells and Plasma Proteins ( J .L. Tullis ed. Academic Press New York 1953 in press) G. S. Adair obtained a molecular weight from osmotic pressure near 500,000 (species not recorded) as reported by Bailey Advances in Protein Chem. 1 308 (1944). Holmberg reports a translational diffusion constant for human fibrinogen which gives a molecular weight near 700,000 if taken with sedimentation constants given above. A new investigation of the diffusion constant would be desirable. ABMH = Armstrong Budka Morrison and Hasson (J. Amer. Ch~m. SOC. 1947,69,1747). EFS = Edsall Foster and Scheinberg (J. Amer. Chem. Soc. 1947 69 2731). HLS = Hocking Laskowski and Scheraga (J. Amer. Chem. SOC. 1952 74 775). OSB OSB - K = Koenig (Arch. Biochem. 1950 25 241). KGSF = Katz Gutfreund Shulman Ferry J.Amer. Chem. SOC. 1952 74 5706. KP = Koenig and Pedersen (Arch. Biochem. 1950 25 97). ML = Morrison and Lontie (unpublished measurements from this laboratory). N = Nanninga (Arch. neerland. physiol. 1946 28 241). OSB = Oncley Scatchard and Brown (J. Physic. Chem. 1947 51 184). SF = Shulman and Ferry (unpublished). SL = Steiner and Laki (Arch. Biochem. Biophys. 1951 34 24). recent work of Hocking Laskowski and Scheraga 7 points to a value near 400,000 ; and very recently Katz Gutfreund Shulman and Ferry,8 in what appears to be a particularly careful study of light scattering have obtained a value of only 340,000 for bovine fibrinogen. From the angular dissymmetry of the scattering they also calculated a length of 520A for the molecule if it were assumed to be a rod or of 650A if it be taken as an ellipsoid of revolution.This marked difference shows how sensitive the calculated absolute dimensions are to the particular * Hocking Laskowski and Scheraga believe this value to be too high because of the form of the angular dissymmetry curves obtained. references numerical values bovine human 8.5-9 9 human OSB 0.725 bovine 8.4-8.6 7.9-8.2 0-706 0.25 0.25 H ABMH OSB EFS KP SF K N HLS 39,400 35,000 580 400 440 OSB - ML - - 440 - 540" 407 340 EFS - - N SL HLS KGSF HLS - N 700 - 700 670 725 - 840" - 38 SL HLS 38 JOHN T . EDSALL FIBRINOGEN AND FIBRIN INTERACTIONS 13 type of model assumed-a fact which if carefully borne in mind would help to obviate some of the disputes about molecular dimensions that now and then arise.Using the ellipsoidal model Foster Scheinberg and I 9 calculated a length near 700w for human fibrinogen and Hocking Laskowski and Scheraga7 more recently obtained a value of 670w for bovine fibrinogen. Both of these values obtained from flow birefringence measurements are in very satisfactory agree- ment with the light scattering value. Moreover the angular dissymmetry of scattering taken in conjunction with the order of magnitude of the molecular weight seems to be compatible only with a prolate or rod-shaped molecule; an oblate type of structure showing such high dissymmetry should have a much higher molecular weight.The dimensions of the fibrinogen molecule have aIso been studied by the electron microscopists. Hall 10 observed rod-shaped structures of varying length the most commonly found length being near 700A as might have been expected from the measurements in solution. The rods appeared to be built up of sub- units so that they looked a little like a string of beads tightly bound together. Except for the marked heterogeneity of the size of the molecules this was quite compatible with the studies on fibrinogen in solution. However to make the problem more complicated other electron microscopists have reported quite different findings. Porter and Hawn 11 reported particles which they believed to be disc shaped about 200w across from bovine fibrinogen; and Mitchell12 found only globular units about 50 in diameter and short filaments 100-300 A long in preparations from human fibrinogen.How to interpret these conflicting findings I do not know ; I must confess to a strong prejudice that Hall’s findings are closest to the truth and in any case the high asymmetry and rodlike shape of fibrinogen in solution seem well established whatever may happen to it in the drying process that is necessary before it can be examined in the electron microscope. The supremely characteristic property of fibrinogen is its interaction with thrombin to give an altered molecule which is capable of association with other like molecules to give the fibrin clot. Bailey and Bettelheim in Cambridge,l3 and Lorand and Middlebrook in Leeds,l4 have done much to eludicate the process.The transformation certainly involves the splitting-off of a peptide (fibrinopeptide) from the fibrinogen molecule and the unmasking of free terminal a-amino groups of glycine in the resulting reactive fibrinogen derivative from which the peptide has been dissociated. The subsequent association processes culminating in the fully developed fibrin clot have been studied in many ways. Here I shall only speak briefly of some influences that are exerted on the proccss by the very strong interactions of certain ions and neutral molecules with fibrinogen or fibrin or both. These effects were observed by Dr. Walter Lever and myseIf,ls studying the rate of clot formation from fibrinogen and thrombin by direct weighing of the fibrin formed and examining the general structure of the clot by measuring its turbidity.Ferry and Morrison,ls in the fundamental studies they had carried out in our laboratory during the war had already defined two extreme types of fibrin clots; the coarse type formed at low pH and low ionic strength which is opaque and synerizes readily under pressure and the fine type relatively transparent and very friable which forms at high pH or high ionic strength. Lever and I carried out our studies for the most part at constant pH and at a fixed ionic strength of 0.15. Even when these factors were fixed many ions and neutral molecules were found to exert profound and specific effects upon the process. Notable among these were urea and guanidine hydrochloride long of course well known as denatur- ing agents for proteins.In this system however these reagents were studied at 14 concentrations far below those ever reported to produce denaturation. Fig. 1 shows the effects of urea on the rate of fibrin formation. It will be seen that even at a concentration of 0.1 M the rate of fibrin formation in urea solution is distinctly lower than in the control solution without urea. At 0.5 M the decrease in the rate of fibrin formation is profound. Moreover the presence of urea affected the structure of the clots as well as the rate of their formation ; invariably the clots formed in the presence of urea were less turbid and more friable than the controls. It has become well known from the work of Laki and L0rand~17.18 that urea at much higher concentrations (above 3 M) actually dissolves fibrin clots formed from purified fibrinogen.It seems probable that this is due to its strong capacity for hydrogen-bond formation so that the hydrogen bonds between urea and the protein displace the protein-protein bonds that normally link the protein molecules together into the fibrin clot. Our measurements at very low urea concentrations which were carried out in 1946-47 before the work of Laki and Lorand was SIXTH SPIERS MEMORIAL LECTURE I 2 0 4 24 3 Time in Hours FIG. 1.-Effect of urea on the rate of fibrin formation from human fibrinogen (fraction I) and thrombin. Thrombin concentration 0.1 unitlml. All solutions contained 0.1 5 M sodium chloride at pH 6.3.The top curve is for a solution containing no urea; the others are for solutjons containing the amounts indicated on the diagram. From Edsall and Lever.15 published show the same effect in incipient form. The formation of highly turbid clots of the coarse type is the result of extensive lateral aggregation of the first thin strands developing in the early stages of clot formation.16 Increase of pH increases the negative net charge on the fibrin molecules and thus tends to decrease lateral association by electrostatic repulsion. Addition of urea blocks lateral association by hydrogen-bond formation between urea and protein leading to the same result. In either case the clot becomes more translucent and friable. The effect of guanidine hydrochloride is very different.Like urea indeed it greatly retards the rate of formation of fibrin even at 0.05 M there is a very marked retardation and at 0.075 M the effect is profound (fig. 2). However the effect on the structure of the clot is in the opposite direction from that of urea; as the structure becomes fully developed it becomes markedly more coarse and turbid than the control clot; indeed the pattern of development of the clot both as regards increase of turbidity and rate of fibrin formation is nearly the same for a control clot at pH 6-3 and for a clot at pH 7.3 which contains guanidine hydro- chloride at a concentration of 0-075 M at least in the earlier stages of the clotting process (fig. 2 and 3). Adding guanidine hydrochloride at this concentration thus has nearly the same effect as lowering the pH by one unit ; the initial rate of fibrin 15 0 0.135 M NaCl; pH 6.3.0.8 0-6 0.4 0.2 JOHN T . EDSALL formation is greatly retarded and the final turbidity of the fully developed clot is greatly increased. The results suggest that the positively charged guanidinium ion FIG. 2.-Effect of guanidine hydrochloride on rate of fibrin formation at pH 6.3 and 7-3. Total ionic strength in all solutions was 0.15. pH was controlled by imidazole buffers the concentration of imidazole hydrochloride being 0.01 5 in ail experiments except in the top curve where it was 0.03. 0 sodium chloride (0.12 M) plus imidazole buffer at pH 7.3. @ 0.075 M guanidine hydrochloride (C(NH&+Cl-) ; 0.06 M sodium chloride ; pH 7.3.0 0.05 M C(NH&+CI- ; 0.085 M NaCl ; pH 6.3. 0 0-075 M C(NH&+Cl- ; 0-06 M NaCl ; pH 6.3 (From measurements of Walter F. Lever.) Time in Hours FIG. 3.-Effect of guanidine hydrochloride on turbidity of fibrin clots at pH 6.3 and 7.3. The solutions had the same composition as those described in fig. 2 and corresponding curves are denoted by the same symbols in both figures. Note the high turbidity of the solution containing 0-075 M C(NH&+C1- at pH 7.3 as compared to the extremely low turbidity of the control solution containing only sodium chloride at the same pH. (From measurements of Walter F. Lever.) combines preferentially with the protein decreasing its negative net charge just as hydrogen ion would do.Quantitative studies of the binding of guanidinium ion to 16 fibrinogen and fibrin have not been carried out but these results suggest that the binding would be strong. K'I- pH 6*50,,+ .' :-- 0 / -0 "0 I / ,4 / */o/@--o- K+SCN- pH 6 - 2 5 ' 1 P / / I Time in Hours * FIG. 4.-Effect of iodide and thiocyanate on rate of fibrin formation from human 2 4 NaC1; middle curve 0.135 M KI + 0.015 M NaCI; bottom curve 0.135 M KSCN + fibrinogen and thrombin. Ionic strength in all solutions was 0.15. Top curve 0.15 M 0.015 M NaCl. Note that the pH in the iodide solution was significantly higher than in the other two. This increased pH accelerated the reaction; thus the rate of fibrin formation in the iodide solution would have been lower at lower pH.Control experi- ments showed that the effect of potassium salts was indistinguishable from that of sodium salts with the same anion with respect to rate of fibrin formation and clot turbidity. 5 (From measurements of Walter F. Lever.) -0.8 Y A7 0 SIXTH S P I E R S MEMORlAL LECTUKE I 9 J 3 2 Time in Hours FIG. 5.-Effect of iodide and thiocyanate on turbidity of fibrin clots as a function of time. Molar concentrations of salts in these experiments were identical with those denoted by corresponding symbols in fig. 4. (From measurements by Walter F. Lever.) * Acknowledgments are made to the Josiah Macy Jr. Foundation for permission to publish fig. 4 and 5 which are slightly modified versions of those appearing in Blood Clotting rmd Allied Psohltwis.Trmsnctions of' the Forirtk CorlJ;.renre ( I95 I) pp. 241-2. JOHN T . EDSALL SERUM ALBUMINS SIZE AND SHAPE 17 Certain anions such as iodide and thiocyanate-also well known as denaturing agents for many proteins at much higher concentrations-act in the opposite direction on the structure of the clot greatly decreasing its turbidity as compared with that of the control at the same pH and ionic strength (fig. 4 and 5). In these cases therefore there is suggestive evidence of selective anion binding. It should be noted that in all cases these reagents diminish the initial rate of fibrin formation as shown in fig. 2 and 4 whatever their effect on the structure of the clot. In no case however was there any destruction of thrombin by the added reagent as control experiments showed so that the decreased rate of reaction cannot be laid to this.Closely related to these studies are the extensive and important studies of J. D. Ferry and Sidney Shulman on the effects of many reagents in inhibiting the conversion of fibrinogen to fibrin. These have been presented by Dr. Shulman 19 in a paper to be discussed at this conference so that I shall not treat them further here. The serum albumins have been among the most intensively studied of all pro- teins ; yet even the size and general shape of the albumin molecule are still a matter of dispute. Let us consider some of the evidence. (i) By the use of dinitrofluorobenzene van Vunakis and Brand 20 could find only one free terminal a-amino group in human horse or bovine serum albumin per molecule of molecular weight approximately 70,000.In all three cases this amino group belonged to an aspartic acid (or asparagine) residue. Thus it would appear that the molecule consists of a single peptide chain. Yet Weber,21 in the paper he is presenting to this meeting has concluded from polarization fluorescence studies that serum albumin dissociates into smaller units in both acid and alkaline solution. There is no necessary incompatibility between these findings but inevitably they raise certain questions. Does the splitting of the albumin molecule involve the breaking of some exceptionally labile peptide linkages? If so what brings about the reassociation that Weber has found to occur on neutralizing the acid albumin solution? Or on the other hand are there actually two or more peptide chains within the albumin molecule only one of which-presumably for steric reasons-reacts with dinitrofluorobenzene? Or is it possible that Weber's findings can be explained in terms of strong electro- static interactions between molecules carrying a high net charge without actually having to assume molecular dissociation? The latter possibility seems to me relatively unlikely but it should not be altogether disregarded.(ii) The molecular weight of human and bovine serum albumin has generally been taken in recent years as 69,000 chiefly on the basis of the very careful osmotic pressure studies of Scatchard Batchelder and Brown.22 The recent X-ray measurements on albumin crystals in our laboratory by Barbara LOW,^^ combined with accurate determinations of density and composition definitely indicate a lower molecular weight near 65,500.Quite independently Creeth 24 has deduced an almost identical value for bovine serum albumin from very careful sedimentation and diffusion measurements. There is converging evidence indeed from several laboratories both in England and the United States that some of the generally accepted measurements of sedimentation constants are in fact too high and require some downward revision-a process which inevitably involves a downward re- vision of molecular weights as well. (iii) In 1947 Oncley Scatchard and Brown25 gave a detailed discussion of sedimentation viscosity and other data for serum albumin and-assuming 20 % hydration-described the molecule approximately as a prolate ellipsoid about I50 A long and 38 A in diameter at the middle.Their choicc of a prolate rather 18 t C SIXTH SPIERS MEMORIAL LECTURE than an oblate shape to describe the molecule was determined largely by Oncley’s dielectric dispersion data which indicated two distinctly different relaxation times -a finding incompatible with an oblate shape if the preparation consists of molecules which are all alike. Recently Oncley Dintzis and Hollies 25a have obtained dielectric measurements of much higher precision on carefully purified human mercaptalbumin and other serum albumin preparations freed from fatty acid and other impurities by treatment with ion exchange resins.These prepara- tions gave considerably higher dipole moments than earlier less pure samples. The form of the dielectric dispersion curve was quite sensitive to the protein con- centration ; but at sufficiently low concentrations the curve was entirely consistent with that to be expected for a prolate ellipsoid of molar volume 60 I. and an axial ratio of 4 to 1. No oblate ellipsoidal mode1 could be chosen to fit the data. The studies of Weber 26 on the polarization of fluorescence from albumin derivatives also give important evidence on this point and on the whole Weber’s findings appear to confirm those of Oncley. l 0 -1 0 2 0 L FIG. 6.-Model for possible modes of packing of human serum albumin in the crystal of mercaptalbumin mercury dimer (a wet crystal; b and c air dried crystals).All drawings represent the c plane projection. Two different modifications of Oncley’s model for the serum albumin molecule are indicated. From Low.23 (iv) As was discovered by Hughes,27 two molecules of human mercaptalbumin are readily linked by means of their sulphydryl groups through a single mercury atom to form a dimer. Oncley was led to reconsider his earlier model for the albumin molecule in order to account for the steric relations involved in the forma- tion of this dimer. This revised model has been employed by Low 23 to describe the packing of serum albumin molecules in the crystal of mercaptalbumin mercury dimer. The model consists of a prism approximately 145 A long 45A in dia- meter at the middle and 22A thick.Such a model can fit well into the unit cell of the crystal as determined by Low; and the indications of a layer structure which are given by the X-ray data are well accounted for by the piling into layers of such long thin molecules (fig. 6). It should be clearly understood that the X-ray data do not prove the structure but they are apparently compatible with it and the model itself-to be taken of course only as a very rough approximation -seems reasonably plausible on other grounds. (v) As was found by Foster and myself,2* human serum albumin shows flow birefringence although only at very high gradients in solvents of high viscosity. From our measurements we estimated a length at least as great as that of l50& inferred by Oncley on other grounds.However these measurements on so relatively short a molecule were exceedingly difficult and subject to considerable JOHN T . EDSALL 19 error ; they should not be taken as having precise quantitative significance. They do point however to a somewhat asymmetrical shape. (vi) In contrast to all these conclusions Riley and Oster 29 have inferred from low angle X-ray diffraction studies on very concentrated albumin solutions that the shape of the molecules cannot be very far from spherical. The theory under- lying this deduction is one which I do not feel competent to criticize ; but I feel very reluctant to accept the conclusion itself in view of the other evidence summarized above. (vii) Very recently Anderegg Beeman and Shulman30 have carried out low angle X-ray scattering studies on relatively dilute albumin solutions.Here the interpretation of the data is more straightforward than for the very concentrated systems studied by Riley and Oster. At vzry low angles the scattering curves indicate that the radius of gyration is definitely greater than for a sphere of the same molecular weight while the scattering at somewhat higher angles which is more characteristic of shape appears to be in better accord with that expected for an oblate than for a prolate ellipsoid of revolution. Assuming an oblate ellipsoid the ratio of the molecular axes is found to be near 3-5 to 1. These data must I think be taken very seriously; we cannot evade the attempt to find some explanation that will make them understandable in conjunction with all the other findings that seem to point to an elongated shape.Dr. Beeman has remarked in a recent letter to me on the subject “ We can say from our data that we are much more convinced that the molecule is not a prolate ellipsoid of revolution than that it is an oblate ellipsoid of revolution.” Indeed the major moral of all this may be that we should not take too seriously calculations based upon simple shapes. They are useful for representing the results of particular researches and for correlating the properties of proteins derived from hydrodynamic measure- ments but in detail the description of the shapes of protein molecules must involve something more complex than the very simple geometrical figures that are con- veniently used for preliminary descriptions.It would be a feasible though certainly a very tedious task to calculate the scattering function for the prismatic model of serum albumin previously discussed; and the curve so derived might be quite different from that of a prolate ellipsoid ; but the case seems a little too special at present to warrant the work involved. It does seem to me that the weight of the evidence does rather favour a general model of this sort and that an oblate ellipsoidal shape for albumin or anything very close to such a shape is unlikely. However I do not wish to be dogmatic on the subject ; what I hope to do in this lecture is to ask more questions than I can answer and I trust that others will come forward with the answers or will set to work to find them.ON THE POSSIBLE FLEXIBILITY OF PROTEIN MOLECULES In contrast to the arrangement of most synthetic polymers which coil about in random fashion in solution proteins have a highly ordered pattern. The evidence for this order is clearest from X-ray studies on protein crystals. Such order also implies a high degree of molecular rigidity which is shown clearly for proteins in solution by the dielectric studies of Oncley.31 Protein molecules orient as entire units in an electric field not as a series of more or less independent segments as do the polar polymers. All this is true and important ; yet we should certainly not assume that this rigidity is strict and absolute. Indeed Klotz and Ayers,32 in a paper presented at this meeting conclude that the molecule of serum albumin must be able to change its configuration with pH ; that it may swell or unfold somewhat as its net charge increases and that these efTects are reversible.Others I think have made similar suggestions before. Likewise the remarkable interactions involved in the combination of haemoglobin with oxygen have led Pauling,33 and also Wyman and Allen,34 to suggest that the addition of oxygen SIXTH SPIERS MEMORIAL LECTURE SERUM ALBUMINS INTERACTlONS WITH CERTAIN CATIONS 20 actually leads to a rather fundamental change in the configuration of the mole- cule as a whole. Pauling and Wyman hold rather different views as to the nature of this change and we certainly need more proof before we assume that the change really occurs but the suggestive evidence is strong.Even more striking I think are some recent unpublished studies here in Cambridge on the enzyme fumarase from pig heart recently crystallized by Vincent Massey 35 working with Dr. Malcolm Dixon. Combined observations by Massey on enzyme activity and by Weber using polarization of fluorescence suggest alterations of molecular size or shape or both with varying pH and ionic strength ; and these variations appear closely correlated with changes in enzyme activity and the formation of the enzyme-substrate complex. We have long been accustomed to change of shape on the part of the fibrous proteins as in the a-/3 transformation of keratin. Indeed the myosin system of muscle could not do its job unless it changed its shape radically and rapidly.Alterations of shape among the corpuscular proteins are presumably much less drastic-that is if they occur under conditions in which the protein still remains undenatured-but they may be of profound importance both chemically and biologically. The fluorescence polarization technique of Dr. Weber may be of great value in such studies for unlike the dielectric dispersion method it can be applied over a very wide range of pH and salt concentration to give information about the relaxation times of proteins. The serum albumins are famous for the strength and diversity of their inter- actions with anions; the story is a long and remarkable one and I have nothing to add to it to-day.Dr. Klotz and Ayers 32 are reporting some new developments in this area and I shall confine myself to a discussion of the interactions of serum albumin with two cations both from group 2 B of the periodic table namely zinc and mercury. The use of these two cations in the fractionation of blood plasma proteins is discussed at this meeting by Cohn Surgenor Schmid Batchelor and Isliker.36 The far-reaching developments which they are presenting speak for themselves; here I shall discuss only some recent work from our laboratory which bears on the chemical nature of the interactions involved. The tendency of metallic ions toward complex formation has been admirably surveyed by Jannik Bjerrum,37 who himself has done so much for the advance- ment of knowledge in this field.From his data it is apparent that the tendency of zinc ion to complex formation with the carboxylate ion is relatively weak (log k is approximately zero for zinc and acetate ion where k is the association constant). The association tendency with ammonia is much stronger (log k = 2-37) and this should give an approximate measure of the affinity of zinc for amino groups. However the affinity of amino groups for hydrogen ion is so strong that the zinc binding capacity at pH 7 or below is relatively small. One very important group found in proteins the irnidazole group of histidine had not been previously investigated; Gurd and Goodman undertook to study the combination of zinc ion and imidazole while Felsenfeld and I were studying its combination with cupric ion.33 Both these ions were found to bind four molecules of imidazole the successive values of log k being tentatively estimated (at 4.5") for copper as 4.58 3-76 3.14 and 2.18.For Zn2+ they were 2.76 2.38 2.40 and 2-22. These values of k for zinc ion are defined by the relation In these equations Im denotes uncharged imidazole. On correcting for the statistical factors involved it is interesting to note that although the first Cu2+ ion is bound much more strongly than the first Zn2' the binding of one imidazole JOHN T . EDSALL 21 by copper decreases thc probability of binding the next whereas the binding of the first imidazole by zinc increases the affinity for more imidazoles until saturation is achieved. While this work was proceeding Gurd and Goodman 39 also made a most careful study of the binding of zinc ion by serum albumin employing the dialysis FIG.7.-Effect of zinc ion concentration on zinc binding to human serum albumin. From Gurd and Goodman.39 Temp.= 0°C 0.15M NaNO -2.5% Albumin Q 7.5 o 5.8 9 14.0 10.0 OH- Added I / Albumin 4 I FIG. &-Dependence of n on v (see text) in the binding of zinc ions by human serum albumin. From Gurd and Goodman.33 equilibrium technique. This work required great care in setting up the experi- ments and the analysis of the data involved a quite elaborate examination of the competition between zinc and hydrogen ions for the various sites on the protein and of the influence of electrostatic forces due to the charge on the protein on the affinity constants.Some of their results are shown in fig. 7 and 8. The former SIXTH SPIERS MEMORIAL LEC'TIJRE 22 illustrates zinc binding as a function of free zinc ion concentration for various amounts of OH- ion added to the isoionic protein; while the latter shows the number n o€ imidazole groups available to combine with hydrogen ions as a function of C the number of Zn2+ ions bound per molecule of albumin. The conclusion from these studies was clear cut ; the binding of Zn2+ by albumin in the pH range 5.5 to 7.5 is due almost entirely to the 16 imidazole groups of the histidine residues. At more acid pH values very little zinc is bound and at more alkaline values the precipitation of zinc hydroxide renders experimental measurement impracticable.Moreover the association constant involved is defined by the relation log k" = 2-82 where k" denotes the association constant corrected for the influence of electro- static forces. This value agrees extraordinarily well with the first association constant in the relation between the zinc ion and the imidazole molecule (log kl = 2.76). Indeed such very close agreement could hardly be expected and is perhaps accidental. Nevertheless this value certainly adds strong confirmatory evidence for the significance of the imidazole groups in the binding of zinc by albumin. The carboxylate groups in spite of their much greater number-106 per albumin molecule-show no significant binding at least in aqueous solutions. However in the presence of concentrated urea as Cohn et al.36 are reporting at this meeting there is evidence from recent work of Dr.Gurd that carboxylate groups become available for binding. The amino groups would certainly combine at more alkaline pH values but their relatively strong basicity renders them unreactive with zinc at neutral pH. Quite independently of Gurd and Goodman Tanford 40 studied polaro- graphically the binding of zinc and several other ions by serum albumin and has come to identical conclusions concerning the significance of the imidazole groups for the binding. The mercuric ion in spite of its electronic relationship to that of zinc differs profoundly in its interactions with the groups found in proteins. Its affinity constant for amino groups is more than a million times as great as that of the Zn2+ ion; and its affinity for imidazole groups although the exact figure is not known to me is certainly far greater also.However its well-known affinity for sulphydryl groups is the dominating feature in its interactions with proteins as long as the protein has any free sulphydryl groups left to react. My colleague Dr. W. L. Hughes Jr. whose work in this area has been basic for all that the rest of us have done estimates that log k for the association of Hg2+ with a sulp- hydryl group is of the order of 17 to 19. On the other hand we must remember that the amount of free Hg2+ ion present in most aqueous solutions of a mercuric salt is extremely small because of its ex- tremely strong affinity for the hydroxyl ion and for all halogen ions except fluoride.The reversible combination of mercury with mercaptal bumin was discovered by Hughes,27 who formulated the essential features of the reaction and crystallized the mercury dimer of mercaptalbumin. Oncley by means of ultracentrifuge measurements showed the simultaneous presence of dimer and monomer in solutions derived from these crystals and obtained indications that they existed in a state of reversible equilibrium. Over a period of several years we have continued the study of the process by light-scattering measurements and have characterized in a more quantitative manner both the kinetics of the formation and dissociation of the dimer and the position of the equilibrium as a function of pH ionic strength the concentration of certain specific ions in the solution temperature and other variables.It was also discovered that certain organic mercurials containing two or more mercury atoms were also capable of linking two albumin molecules together to form a dimer. The one most thoroughly studied was the dioxane derivative introduced into the laboratory by Straessle 41 CH2-0 \CH . CH2. Hg+ +Hg . CH2. HC/ \O-CHz/ JOHN T. EDSALL ASH + HgC12 + ASHgCl + H+ + C1-. ASHgCl + ASH + ASHgSA + Hf + C1-. Kcl/RhO = M-1 + 2Bc1 111. ASHgSA + HgC12 + 2ASHgCl. The exact formulation of these reactions is open to some discussion. In the solution before reaction starts and at the concentrations used in our studies most of the mercury is in the form of undissociated HgC12.Some however must be present in combination with hydroxyl or acetate ion; some presumably as HgCl+ ion. The amount of free Hg2+ is certainly minute at equilibrium. Probably however the reactions I I1 and 111 represent the most convenient simple way of formulating the process. Dr. Hughes has demonstrated the release of hydrogen ions accompanying reactions I and 11. and for the dimer of molecular weight 2M Kc~/R$~ = (2M)-1 + 2B~2. 23 which for brevity will be denoted as HgRHg. This compound as will be seen induces dimer formation far more rapidly than mercuric chloride itself. Other organic mercurials prepared by Dr. Howard Dintzis are now being studied. This work has been carried on by several investigators in our laboratory; its progress is due to the thinking and the experimental skill of Harold Edelhoch R.B. Simpson Rudolf Straessle Ephraim Katchalski and R. H. Maybury. The insight of Dr. Hughes has been invaluable to us throughout the progress of the work. The reaction may be described as proceeding in three stages. First is the combination of mercury with mercaptalbumin (denoted here as ASH) through the sulphydryl group of the latter (1) This reaction involves no significant change of the molecular weight of the protein ; it therefore cannot be measured by light scattering and it is in any case very rapid-probably too fast to measure by most techniques. The product of the first reaction then reacts with a second molecule of mercaptalbumin to form the dimer (11) This is the reaction which is directly observed by light scattering; it proceeds very slowly as compared to reaction I and is readily followed.Finally if an excess of mercuric chloride is added the dimer rapidly dissociates according to reaction (111) The interpretation of the light-scattering measurements in order to derive weight average molecular weights also raises certain problems. For the monomer of molecular weight M we can employ the usual equation for light scattering in dilute solutions (W (V) Here c1 and c2 denote the weight concentrations of monomer and dimer respectively expressed in g/ml RJO and R30 denotes the corresponding values for the reduced intensities of scattering at 90" from monomer and dimer and K denotes the factor K = 2~2n2(dn/dc)2/NoA-f where n denotes refractive index N is Avogadro's number and ho is the wave- length of the light in vacuo (Doty and Edsall42).The factor B is the interaction constant which is determined in large measure by the net charge on the protein and the ionic strength of the solution (Edsall Edelhoch Lontie and Morrison 43). (VI) If the protein is isoelectric B is zero or nearly zero for both monomer and dimer and the weight average molecular weight can be inferred directly from the reduced intensity. However it was of great interest to study the system under conditions of varying net charge on the protein and under these conditions B is SIXTH SPIERS MEMORIAL LECTURE (VW (VIII) Using this approach we have been able to obtain extensive information con- cerning the velocity of the reaction and the proportion of dimer which is present at equilibrium.Repeated experiments have shown that true equilibrium is actually attained. In a given medium one may start either from pure monomer with an appropriate amount of mercury added or from a solution which is nearly pure dimer ; and on adjusting to the same weight concentration of protein in each case one attains the same final equilibrium value. If the activity of hydrogen and of chloride ions is fixed this equilibrium may conveniently be defined in terms of an apparent equilibrium constant (ASHg SA) 24 in generai different from zero. Fortunately it proved in practice that in a given medium the value of B for the pure mononier and the pure dimer is the same within experimental error.We have made the assumption-and it can be justified by theoretical arguments developed by Dr. Katchalski which I need not discuss here-that the same value of B can then be used for any mixture of monomer and dimer in the same solution and we can then obtain the molecular weight by extra- polation. For the mixture of the two species then we may write in fig. 9. Kc/RgO = X-1 -t ~ B c where c = c1 + c2 and X = M(c1 -+ 2c~)/c. Since c and R90 are measured experi- mentally X may then be calculated from (VIl) provided that the value of B has been determined for the pure monomer or dimer in the same solvent in which the reaction is taking place. The weight fraction of dimer is then given by the equation (ASHgCI)(ASH) Kapp == and the variation of this apparent constant with change in pH and ionic strength may then be studied.To our surprise Kapp was found to vary little over the pH range 4-25-6 the mean value being 2.6 (& 0-4) x lo4 over the whole range. The measurements were carried out in phosphate buffers at pH 6 and in acetate buffers at the lower pH values. In simple terms this value of Kapp means that in a system containing 0.5 mole Hg per mole albumin if the total weight con- centration of protein is 10 g/l. then the weight fraction of dimer at equilibrium is near 0.5. Preliminary studies of the temperature coefficient of equilibrium indi- cate that rise of temperature favours dimer formation and therefore that the process is presumably endothermic.For a given protein concentration in a given medium the weight fraction of dimer at equilibrium varies in a characteristic fashion with the amount of mercury added. It is of course zero in the absence of added mercury when all mercapt- albumin is present as ASH and falls again to zero when one mole of mercury has been added per mole of albumin all of the mercaptalbumin being then in the form ASHgCI. The maximum amount of dimer formation always occurs when the molar ratio of mercury to albumin is 0.5. The value of the weight fraction of dimer at this point increases with the value Kapp. The form of the whole curve indeed can be calculated from the value of the equilibrium constant. The curve is symmetrical about its midpoint and is similar in charactcr to the curves theoret- ically deduced by Michaelis 44 for the fraction of semiquinone formed in the re- versible titration of an oxidation-reduction system as a function of the total of oxidant or reductant added and of the semiquinone formation constant.A family of some of the curves to be expected for different values of Kapp is shown We have also made extensive determinations of the velocity constant kf of the combination reaction between ASHgCl and ASH. This constant expressed in 1. mole-1 min-1 is equal about to 40 from pH 4.25-4.75. As the pH increases JOHN T. EDSALL 25 further however the value decreases progressively to 0.69 at pH 6 . I t seems probable that this variation at pH values above 4.7 can be at least partly explained by the electrostatic effect of the increasing negative net charge on the albumin molecules above pH 5.The treatment involved here would be essentially that developed by Bronsted N. Bjerrum Scatchard and others for simpler i0ns.~5 However since the equilibrium constant is essentially independent of pH the results indicate clearly that the dissociation of dimer into monomer is far slower at pH 6 or above than below pH 5. This has indeed been confirmed experi- mentally. A concentrated solution of the dimer may be prepared and adjusted to a pH somewhat above 6 in phosphate buffer. Even if the solution is then greatly diluted the dissociation of the dimer is found to be extremely slow as 100 80 L (Y E *6 60 .c. c 0 e 0 ) 2 40 20 0 Molar Ratio HgCI,/(ASH) at t= 0 FIG.9.-Percent of mercaptalbumin-dimer formed at equilibrium as a function of HgC12/ mercaptalbumin ratio. Curves are shown for various values of the parameter Kappa (apparent association constant multiplied by the molar concentration of albumin). compared to the rate below pH 5. In this way dilute solutions containing nearly pure dimer with very little monomer can be obtained for study over a considerable period of time as for instance in ultracentrifuge runs. Studies of interaction of mercaptalbumin with the organic dioxane derivative discussed above (HgRHg) indicate a striking difference from the interactions with mercuric chloride. The equilibrium at all the pH values studied seems to lie almost completely in the direction of dimer formation.At low pH values below 4.8 the velocity of the process is so great that no accurate velocity constants for dimer formation could be determined by our methods. Between pH 5 and 6 the value of kf decreases with rising pH in much the same manner as for the re- action with mercuric chloride. However the absolute values of kf at any pH in this range are very much greater-indeed several thousand times as high as those for the reaction involving mercuric chloride. A plot of the two sets of kfvalues is given in fig. 10. The great contrast between the rates and the equilibrium constants found in these two different types of reaction with mercury compounds is presumably to 26 be explained largely on steric grounds. When two albumin molecules are linked through their sulphydryl groups by a single mercury atom the juxtaposition of the adjoining surfaces of the two protein molecules must be extraordinarily close.It is indeed remarkable that the surfaces are so shaped that they can be fitted together at all. It will be of great interest to see whether chemical modification of other groups in the albumin will lead to alterations in the tendency to undergo this reaction. In any case the relative slowness and incompleteness of the reaction are not surprising. On the other hand for the dioxane derivative HgRHg in which there is a mercury atom at each end there is an effective distance of the order of lOA between the two mercury atoms. Thus the coupling of the two albumin molecules which are linked by these atoms is far looser than when a single mercury serves as the link.The conditions for the combination are much less critical and pre- sumably the reaction is possible in a much greater variety of relative orientations t PH SIXTH SPIERS MEMORIAL LECTURE 4; 6 A S Hq RHq SA \ A5 Hq SA - 5;8 5 4 log K f 10 I :2 5 0 FIG. IO.-Logarithm of the velocity constant kf for dimer formation of human Serum mercaptalbumin with mercurials. Lower curve reaction with mercuric chloride ; upper curve reaction with the dioxane derivative containing two mercury atoms. Phosphate buffers were employed near pH 6 acetate buffers at lower pH values; ionic strength 0.05 in acetate buffers; 0.05-0.15 in phosphate; temp. 25". (From studies of E.Katchalski H. Edelhoch and R. B. Simpson.) of the two albumin molecules than would be the case if the fitting were closer. Thus the reaction goes relatively rapidly and it goes largely to completion. One may suspect although this is not proved that the two halves of the dirner ASHgRHgSA may be much more loosely coupled than the halves of the dimer ASHgSA. Thus in the former case a much greater degree of free rotation of one-half of the dimer relative to the other should perhaps be expected than in the latter; a possibility which may perhaps be explored by such methods as dielectric dispersion. It should be added that in spite of the presumably very close fitting of the two halves of the dimer ASHgSA the mercury atom does not appear to be buried within the protein in such a way as to be inaccessible to the solvent.Addition of other ions such as bromide iodide or cyanide which have a strong affinity for mercury leads to quite rapid dissociation of the dimer. Evidently these ions can penetrate readily to the mercury atom which is then released from its com- bination from one of the sulphydryl groups to form the monomer ASHgBr ASHgI or ASHgCN. The reaction has been studied by us particularly with JOHN T. EDSALL 282 729. 3 Bragg Howells and Perutz Acta Cryst. 1952 5 136. 4Kendrew Proc. Roy. SOC. A 1950 201 62. 5 Carlisle and Scouloudi Proc. Roy. SOC. A 1951 207 496. 6 Low Nature 1952 169 955. 7 Hocking Laskowski and Scheraga J. .4mer. Chem. SOC. 1952 74 775. 8 Katz Gutfreund Shulman and Ferry J.Amer. Chem. SOC. 1952 74 5706. Bettelheim and Bailey ibid. 9 578. 14 Lorand Nature 1951 167 992. 15 Edsall and Lever J. Biol. Chem. 1951 191 735. 16 Ferry and Morrison J. Amer. Chem. SOC. 1947 69 388. 17 Laki and Lorand Science 1948 108 280. 18 Lorand Nature 1950 166 694. 19 Shulman this Discussion. 27 the iodide ion. The addition of even a single mole of iodide per mole of mercury in the dimer leads to a relatively rapid and very marked dissociation of the dimer. Bromide is inuch less effective and chloride far less than bromide. Indeed these studies appear to be significant as a method of evaluating the relative affinity for mercury of various inorganic ions and thus obtaining data which should be of interest for inorganic chemistry.This is of course not our primary concern; we are interested in these reactions chiefly because of what we can learn from them concerning the external form the surface properties and re- activity and perhaps something of the finer pattern of the configuration of the albumin molecule. The extensive studies carried out by Lewin 46 in our laboratory demonstrated that mercaptalbumin mercury dimer could be crystallized in association with many other metallic cations and also with anions containing atoms of high atomic number. Dimer formation of the type first observed by Hughes27 with serum albumin has not yet apparently been observed with other proteins. Several years earlier however the successful use of mercury in the crystallization of enolase was reported by Warburg and Christian.47 In this case the crystallized mercury derivative contained just one mole of mercury per mole of enolase and other derivatives have not been found.The interaction of mercury with proteins is of course not confined to a single group. However the affinity of mercury for the sulphydryl group is so much greater than for any other groups found in proteins that no other reaction will occur until all such groups are saturated with mercury. Beyond this point however important interactions begin to occur with imidazole amino and other groups. These interactions between mercury and proteins have recently been exploited with great success by Dr. Cohn and his associates in fractionation processes .36 1 Sanger and Tuppy Biochem.J. 1951,49,463,481. Sanger Thompson and Tuppy Symposiiiin sur les Hormones Proteines et Derivtes des Proteines (Paris 1952). 2 Pauling and Corey Proc. Nut. Acad. Sci. 1951 37 205 235 241 251 261 272 9 Edsall Foster and Scheinberg J. Amer. Chem. SOC. 1947 69 273. 10 Hall J. Biol. Chem. 1949 179 857. 11 Porter and Hawn J. Expt. Meci. 1949 90 225. 12 Mitchell Biochem. Bioplzys. Acta 1952 9 430. 13 Bailey Bettelheim Lorand and Middlebrook Nature 1951 167 233 ; Bettelheim and Bailey ResumL des Communications I P Congres Internationale de Biochemie 1952 411; Lorand and Middlebrook Biochem. Biophys. Acta 1952 9 581; 20 van Vunakis and Brand Abstr. 119th meeting Amer. Chem. SOC. April 1951 p. 28c. 21 Weber this Discussion. 22 Scatchard Batchelder and Brown J.Amer. Chem. SOC. 1946 68 2320. 23 Low J. Amer. Chem. SOC. 1952 74 4830. 24 Creeth Biochem. J. 1952 51 10. 25 Oncley Scatchard and Brown J. Physic. Chem. 1947 51 184. 28 SIXTH SPIERS MEMORIAL LECTURE 25a Oncley Dintzis and Hollies Abstr. 122nd meetirrg Amer. Chem. SOC. Sept. 1952 p. 12p; Dintzis Thesis Harvard University 1952. 26 Weber Biochenr. J . 1952 51 145 155. Biol. 1950 14 79. 27 Hughes J. Amer. Chem. Sac. 1947 69 1836 ; CoM Spring Harbor Symp. Quant. 28 Edsall and Foster J. Amer. Chem. SOC. 1948 70 1860. 29 Riley and Oster Furuday SOC. Discussions 1951 11 107. 30 Anderegg Beeman and Shulman paper presented to Amer. Physic. SOC. April 1952 ; and unpublished work. 31 Oncley Chem. Rev. 1942 30 433. 32 Klotz and Ayers this Discussion 33 St.George and Pauling Science 1951 114 629. 34 Wyman and Allen J. Polymer Sci. 1951 7 499. 35 Massey Biochem. J. 1952 51 490. 36 Cohn Surgenor Schmid Batchelor Isliker and Alameri this Discussion. 37 Bjerrum Chem. Rev. 1950 46 38 1. 38 Gurd Edsall Felsenfeld and Goodman Fed. Proc. 1952 11 224. 39 Gurd and Goodman J. Amer. Chem. SOC. 1952,74 670. 40 Tanford J. Amer. Chem. SOC. 1952 74 21 1. 41 Straessle J. Amer. Chem. SOC. 1951 73 504. 42 Doty and Edsall Advances in Protein Chem. 1951 6 35. 43 Edsall Edelhoch Lontie and Morrison J. Amer. Chem. SOC. 1950 72 4641. 44 Michaelis Chem. Rev. 1935 16 243 ; see especially fig. 4. 45 see for instance the discussion by Laidler Chemical Kinetics (McGraw Hill Pub- lishing Co. New York 1950) p.123. 46 Lewin J. Amer. Chem. SOC. 1951 73 3906. 47 Warburg and Christian Biochem. Z. 1941 310 384. THE PHYSICAL CHEMISTRY OF PROTEINS SIXTH SPIERS MEMORIAL LECTURE THE MOLECULAR SHAPES OF CERTAIN PROTEINS AND SOME OF THEIR INTERACTIONS WITH OTHER SUBSTANCES BY JOHN T. EDSALL" University Laboratory of Physical Chemistry Related to Medicine and Public Health Harvard University Boston Mass. Received 30th September 1952 The Discussions of the Faraday Society have developed in recent years into events of international significance for many scientists. To-day as in a number of the previous Discussions the focus of interest is on the great border area between the provinces of physics chemistry and biology. In those earlier Discussions many ideas of extraordinary fruitfulness were presented and vigorously debated before they had become familiar and respectable.I have often turned back for illumination to the printed record of their proceedings. Three years ago I had for the first time the opportunity to take part in one of these Discussions-that on Lipoproteins at Birmingham. Hence I feel deeply the honour of giving the Sixth Spiers Memorial Lecture. It was never my good fortune to know Mr. Spiers personally but it was he especially who guided the pattern of this Society's development from small beginnings until it grew into an active and significant organization and its discussion meetings became as they are to-day the focal point for investigators from every land where science is free and active. At that time although the Faraday Society had held many conferences of international importance on the border line of physical chemistry and other sciences it is scarcely likely that it could have contemplated a two-day discussion on the physical chemistry of the proteins.It is doubtful indeed whether in 1926 more than half a dozen laboratories throughout the world could have made effective contributions to such a discussion. L can well remember that year for in the midst of it I returned to Harvard after two years of study here in Cambridge where I had imbibed the broad vision of the scope and the future of biochemistry that was to be found in Cambridge in Hopkins's laboratory. Also I had had the good fortune to know that great pioneer in the physical chemistry of the proteins Sir William Hardy then no longer active in that field but still with a freshness of mind a vitality and zest in life that I have seldom seen matched in other men.The view was indeed held by some distinguished chemists that proteins were actually very small molecules with molecular weights below 1000. This theory was short-lived being quite demolished late in 1926 by a few well-chosen freezing-point * Fulbright Visiting Lecturer University of Cambridge 1952. The untimely death of Mr. Spiers occurred in 1926. In those days we scarcely knew the molecular weight of a single protein. A 10 SIXTH SPIERS MEMORIAL LECTURE measurements by Cohn and Conant. Sorensen indeed had carried out a decade earlier his classical investigations on egg albumin and his careful studies of osmotic pressure had given the basic information for the determination of its molecular size.Adair had propounded in 1925 the startling view that the molecular weight of mammalian haemoglobin was near 67,000 and that the molecule therefore contained four iron atoms. I can well remember the profound scepticism with which this report was received by more than one professor of great eminence; and the doubts persisted in high quarters until the ultracentrifugal measurements of Svedberg a year or two later gave complete support to Adair’s conclusions. The ultracentrifuge had indeed been under development since about 1922 but there were few at that time who anticipated its immense and revolutionary influence on protein chemistry. If relatively little was known of the sizes of protein molecules almost nothing was known about their shapes; indeed the problem of molecular shape was scarcely envisaged at that time by protein chemists.The air was still filled with the strife of the conflict between Jacques Loeb and the colloid chemists of the school of Wolfgang Ostwald. Loeb’s dzmonstration that some of the colloidal properties of proteins could be explained on simple chemical principles a demon-stration expounded brilliantly and with a fighting spirit appeared to many besides myself as if it were a new revelation. To-day I still appreciate the value of Loeb’s contribution. but I am also aware that T. B. Osborne in New Haven, many years before had been quietly carrying on his work on the plant proteins, treating them as chemical substances to be understood on general chemical principles and paying no attention to the clamour of conflicting schools.Also there was one idea expressed by distinguished protein chemists as late as twenty years ago that proteins furnished essentially the inert structural frame-work of the animal organism while the active pattern of biochemical trans-formations was determined by the interplay of other molecules of quite a different sort. Certainly this view was never upheld in the laboratory headed by Edwin Cohn. Nevertheless the fact that such a belief could have prevailed among many biochemists even for a few years seems astonishing to-day; and indeed it represented the denial of a much older tradition going back over a century to Liebig and Mulder which held that proteins were indeed of absolutely prime importance to the living organism.To Mulder in 1840 this must have been largely a matter of faith and intuition ; but now in the last quarter century we have seen that faith abundantly justified by works. I have tried to recall here something of the temper of the times the prevailing atmosphere of thought among the relatively small band of protein chemists in 1926 when I came back from Cambridge to Harvard as a third year medical student and started working in my spare time under the direction of Edwin Cohn on a viscous and peculiar globulin from skeletal muscle continuing work which had been begun by W. T. Salter. Later we called this globulin myosin and still later it was rechristened by Szent-Gyorgyi as a form of actomyosin.It proved a most refractory and difficult substance to work with but in the midst of our struggles Alexander von Muralt arrived from Switzerland and introduced the measurement of double refraction into the laboratory. His brilliant contributions, both to experiment and to thinking brought us face to face with the problem of molecular shape; from then on we knew that although many proteins might be more or less globular some-and these indeed of supreme importance-were long and filamentous even in solution. In these casual remarks I shall not try to review the intervening years with the colossal advances that they have seen. Protein chemistry which was a rela-tively unfashionable subject a quarter of a century ago has now become almost embarrassingly fashionable.It does not seem likely to go out of fashion unless mankind decides to turn away from science and repudiate it altogether-a possible but I still hope a relatively improbable event JOHN T . EDSALL 11 However of all the advances in this quarter century perhaps none will stand out more than two that have cccurred in the last two years-the elucidation of the complete sequence of amino acid residues in the A and B chains of insulin by Sanger Tuppy and Thompson,l and the detailed formulation of spatial con-figurations of the polypeptide chain by Pauling and Corey.2 The aspects of protein chemistry in which these developments are included will not be dealt with by other authors in the present Discussion but these achievements are bound to influence the thinking of all of us so profoundly that I cannot leave them without mention here.We cannot yet write a structural formula for insulin until the arrangement of the disulphide bridges is specified but that the goal has now been so nearly approached is an achievement which would have seemed incredible even five years ago. Protein chemists will now really begin to think for the first time in terms of structural organic chemistry and all of us here who are physical chemists will be permeated more and more by that influence. However in protein chemistry more than anywhere else structural chemistry is of little use without stereochemistry. Proteins with their marvellously specific patterns are the supreme examples of molecillar geometry. The work of Pauling and Corey is neither the beginning nor the end of the deciphering of these patterns; their achievement is built on the work of many and brilliant predecessors and what they have done is only a beginning.Their work still has its sharp critics; but I have been led more and more to the conclusion that their a-helix does quite closely represent the spatial pattern of long chain synthetic polyamino acids and not improbably also that of the fibrous proteins which are found in the cc configuration. Also-though there is less concrete evidence for this view-the pleated sheets which they have proposed appear to be by far the best existing models for the extended pattern of protein chains in the ,B configuration. In several crystals of corpuscular proteins the presence of intramolecular bundles of parallel chains has been inferred from the analysis of X-ray data.In the case of haemoglobin Bragg Howells and Perutz 3 inferred that these chains were polypeptide chains in the cc configuration. Well-defined chains were in-ferred by Kendrew 4 in the myoglobin molecule. The study of crystalline ribo-nuclease 5 has also suggested the presence of chains parallel to the c-axis of the crystal but the configuration of these chains seems still to be a matter for dispute. Some of the most striking evidence for chain structures in a protein crystal comes from the recent work of my colleague Barbara W. Low at Harvard on the ortho-rhombic crystals of acid insulin sulphate,6 originally prepared in our laboratory by Dr. Eric Ellenbogen working with Prof.Oncley. The three-dimensional Patterson diagrams of these crystals in the air-dried form show a most striking pattern of parallel rods with axes parallel to the a-axis of the unit cell which has a length of approximately 45 A. Dr. Low has constructed possible packing models for the polypeptide chains in this crystal taking account of Sanger’s findings on the sequence of amino acid residues. I should like to turn now from these brief remarks on protein structure at the deeper level to some of the problems related to the general size shape and inter-actions of proteins. I shall confine myself to two proteins with which I have become particularly familiar in recent years fibrinogen and serum albumin, FIBRINOGEN SIZE AND SHAPE Fibrinogen occupies a place apart among the plasma proteins because of its high molecular asymmetry and its low solubility the latter property rendering it the first major constituent of plasma to be precipitated in almost any system of fractionation.Its high viscosity its marked flow birefringence its high frictional ratio and its significant angular dissymmetry of light scattering all mark it out as a molecule of shape very far from spherical. There are now data available on both human and bovine fibrinogen in solution which are too numerous t 12 S I X T H SPIERS MEMORIAL LECTURE discuss here in detail. They are summarized in table 1 and appear to indicate that human and bovine fibrinogen are remarkably similar in size and shape. The molecular weight indeed appears at present more uncertain than the absolute dimensions-many workers have accepted a value of 450,000-500,000 but the TABLE 1 .-PHYSICAL CONSTANTS OF HUMAN AND BOVINE FIBRINOGEN numerical values references sedimentation constant S ~ O ~ (Svedberg partial specific volume V20 intrinsic viscosity Eio rotary diffusion constant 0 2 0 (sec-1) (from molecular weight (in thousands) units) double refraction of flow) osmotic pressure (Mo) sedimentation and viscosity (Mg) light scattering molecular length (A) double refraction of flow osmotic pressure and viscosity sedimentation and viscosity light scattering molecular diameter (A) human 8.5-9 9 0.725 0.25 35,000 580 400 440 700 700 38 --bovine 8.4-8.6 7.9-8.2 0-706 0.25 39,400 440 540" 407 340 670 725 840" 38 --human H OSB ABMH OSB EFS OSB ML ---EFS OSB OSB --bovine KP SF K N HLS N SL HLS KGSF HLS N SL HLS --NoTEs.-T~~s table is also published in BZood Cells and Plasma Proteins ( J .L. Tullis ed. Academic Press New York 1953 in press), G. S. Adair obtained a molecular weight from osmotic pressure near 500,000 (species not recorded) as reported by Bailey Advances in Protein Chem. 1 308 (1944). Holmberg reports a translational diffusion constant for human fibrinogen which gives a molecular weight near 700,000 if taken with sedimentation constants given above. A new investigation of the diffusion constant would be desirable. ABMH = Armstrong Budka Morrison and Hasson (J.Amer. Ch~m. SOC. 1947,69,1747). EFS = Edsall Foster and Scheinberg (J. Amer. Chem. Soc. 1947 69 2731). HLS = Hocking Laskowski and Scheraga (J. Amer. Chem. SOC. 1952 74 775). KGSF = Katz Gutfreund Shulman Ferry J. Amer. Chem. SOC. 1952 74 5706. KP = Koenig and Pedersen (Arch. Biochem. 1950 25 97). ML = Morrison and Lontie (unpublished measurements from this laboratory). OSB = Oncley Scatchard and Brown (J. Physic. Chem. 1947 51 184). H = Holmberg (Arkiv. Kemi Min. Geol. A 1944 17 no. 28). K = Koenig (Arch. Biochem. 1950 25 241). N = Nanninga (Arch. neerland. physiol. 1946 28 241). SF = Shulman and Ferry (unpublished). SL = Steiner and Laki (Arch. Biochem. Biophys. 1951 34 24). recent work of Hocking Laskowski and Scheraga 7 points to a value near 400,000 ; and very recently Katz Gutfreund Shulman and Ferry,8 in what appears to be a particularly careful study of light scattering have obtained a value of only 340,000 for bovine fibrinogen.From the angular dissymmetry of the scattering they also calculated a length of 520A for the molecule if it were assumed to be a rod, or of 650A if it be taken as an ellipsoid of revolution. This marked difference shows how sensitive the calculated absolute dimensions are to the particular * Hocking Laskowski and Scheraga believe this value to be too high because of the form of the angular dissymmetry curves obtained JOHN T . EDSALL 13 type of model assumed-a fact which if carefully borne in mind would help to obviate some of the disputes about molecular dimensions that now and then arise.Using the ellipsoidal model Foster Scheinberg and I 9 calculated a length near 700w for human fibrinogen and Hocking Laskowski and Scheraga7 more recently obtained a value of 670w for bovine fibrinogen. Both of these values, obtained from flow birefringence measurements are in very satisfactory agree-ment with the light scattering value. Moreover the angular dissymmetry of scattering taken in conjunction with the order of magnitude of the molecular weight seems to be compatible only with a prolate or rod-shaped molecule; an oblate type of structure showing such high dissymmetry should have a much higher molecular weight. The dimensions of the fibrinogen molecule have aIso been studied by the electron microscopists. Hall 10 observed rod-shaped structures of varying length, the most commonly found length being near 700A as might have been expected from the measurements in solution.The rods appeared to be built up of sub-units so that they looked a little like a string of beads tightly bound together. Except for the marked heterogeneity of the size of the molecules this was quite compatible with the studies on fibrinogen in solution. However to make the problem more complicated other electron microscopists have reported quite different findings. Porter and Hawn 11 reported particles which they believed to be disc shaped about 200w across from bovine fibrinogen; and Mitchell12 found only globular units about 50 in diameter and short filaments 100-300 A long in preparations from human fibrinogen. How to interpret these conflicting findings I do not know ; I must confess to a strong prejudice that Hall’s findings are closest to the truth and in any case the high asymmetry and rodlike shape of fibrinogen in solution seem well established whatever may happen to it in the drying process that is necessary before it can be examined in the electron microscope.FIBRINOGEN AND FIBRIN INTERACTIONS The supremely characteristic property of fibrinogen is its interaction with thrombin to give an altered molecule which is capable of association with other like molecules to give the fibrin clot. Bailey and Bettelheim in Cambridge,l3 and Lorand and Middlebrook in Leeds,l4 have done much to eludicate the process. The transformation certainly involves the splitting-off of a peptide (fibrinopeptide) from the fibrinogen molecule and the unmasking of free terminal a-amino groups of glycine in the resulting reactive fibrinogen derivative from which the peptide has been dissociated.The subsequent association processes culminating in the fully developed fibrin clot have been studied in many ways. Here I shall only speak briefly of some influences that are exerted on the proccss by the very strong interactions of certain ions and neutral molecules with fibrinogen or fibrin or both. These effects were observed by Dr. Walter Lever and myseIf,ls studying the rate of clot formation from fibrinogen and thrombin by direct weighing of the fibrin formed, and examining the general structure of the clot by measuring its turbidity. Ferry and Morrison,ls in the fundamental studies they had carried out in our laboratory during the war had already defined two extreme types of fibrin clots; the coarse type formed at low pH and low ionic strength which is opaque and synerizes readily under pressure and the fine type relatively transparent and very friable, which forms at high pH or high ionic strength.Lever and I carried out our studies, for the most part at constant pH and at a fixed ionic strength of 0.15. Even when these factors were fixed many ions and neutral molecules were found to exert profound and specific effects upon the process. Notable among these were urea and guanidine hydrochloride long of course well known as denatur-ing agents for proteins. In this system however these reagents were studied a 14 SIXTH SPIERS MEMORIAL LECTURE concentrations far below those ever reported to produce denaturation.Fig. 1 shows the effects of urea on the rate of fibrin formation. It will be seen that even at a concentration of 0.1 M the rate of fibrin formation in urea solution is distinctly lower than in the control solution without urea. At 0.5 M the decrease in the rate of fibrin formation is profound. Moreover the presence of urea affected the structure of the clots as well as the rate of their formation ; invariably the clots formed in the presence of urea were less turbid and more friable than the controls. It has become well known from the work of Laki and L0rand~17.18 that urea at much higher concentrations (above 3 M) actually dissolves fibrin clots formed from purified fibrinogen.It seems probable that this is due to its strong capacity for hydrogen-bond formation so that the hydrogen bonds between urea and the protein displace the protein-protein bonds that normally link the protein molecules together into the fibrin clot. Our measurements at very low urea concentrations, which were carried out in 1946-47 before the work of Laki and Lorand was 0 I 2 3 4 24 Time in Hours FIG. 1.-Effect of urea on the rate of fibrin formation from human fibrinogen (fraction I) and thrombin. Thrombin concentration 0.1 unitlml. All solutions contained 0.1 5 M sodium chloride at pH 6.3. The top curve is for a solution containing no urea; the others are for solutjons containing the amounts indicated on the diagram. From Edsall and Lever.15 published show the same effect in incipient form.The formation of highly turbid clots of the coarse type is the result of extensive lateral aggregation of the first thin strands developing in the early stages of clot formation.16 Increase of pH increases the negative net charge on the fibrin molecules and thus tends to decrease lateral association by electrostatic repulsion. Addition of urea blocks lateral association by hydrogen-bond formation between urea and protein leading to the same result. In either case the clot becomes more translucent and friable. Like urea indeed it greatly retards the rate of formation of fibrin even at 0.05 M there is a very marked retardation and at 0.075 M the effect is profound (fig. 2). However the effect on the structure of the clot is in the opposite direction from that of urea; as the structure becomes fully developed it becomes markedly more coarse and turbid than the control clot; indeed the pattern of development of the clot both as regards increase of turbidity and rate of fibrin formation is nearly the same for a control clot at pH 6-3 and for a clot at pH 7.3 which contains guanidine hydro-chloride at a concentration of 0-075 M at least in the earlier stages of the clotting process (fig.2 and 3). Adding guanidine hydrochloride at this concentration thus has nearly the same effect as lowering the pH by one unit ; the initial rate of fibrin The effect of guanidine hydrochloride is very different JOHN T . EDSALL 15 formation is greatly retarded and the final turbidity of the fully developed clot is greatly increased.The results suggest that the positively charged guanidinium ion FIG. 2.-Effect of guanidine hydrochloride on rate of fibrin formation at pH 6.3 and 7-3. Total ionic strength in all solutions was 0.15. pH was controlled by imidazole buffers the concentration of imidazole hydrochloride being 0.01 5 in ail experiments, except in the top curve where it was 0.03. 0 sodium chloride (0.12 M) plus imidazole buffer at pH 7.3. @ 0.075 M guanidine hydrochloride (C(NH&+Cl-) ; 0.06 M sodium chloride ; pH 7.3. 0 0.135 M NaCl; pH 6.3. 0 0-075 M C(NH&+Cl- ; 0-06 M NaCl ; pH 6.3, 0 0.05 M C(NH&+CI- ; 0.085 M NaCl ; pH 6.3. (From measurements of Walter F. Lever.) 0.8 0-6 0.4 0.2 Time in Hours FIG. 3.-Effect of guanidine hydrochloride on turbidity of fibrin clots at pH 6.3 and 7.3.The solutions had the same composition as those described in fig. 2 and corresponding curves are denoted by the same symbols in both figures. Note the high turbidity of the solution containing 0-075 M C(NH&+C1- at pH 7.3 as compared to the extremely low turbidity of the control solution containing only sodium chloride at the same pH. (From measurements of Walter F. Lever.) combines preferentially with the protein decreasing its negative net charge just as hydrogen ion would do. Quantitative studies of the binding of guanidinium ion t 16 SIXTH S P I E R S MEMORlAL LECTUKE fibrinogen and fibrin have not been carried out but these results suggest that the binding would be strong. .' K'I- pH 6*50,,+ :--0 P / / / 1 K+SCN- pH 6 - 2 5 ' */o/@--o- / -0 Time in Hours I / "0 I ,4 5 2 4, * FIG.4.-Effect of iodide and thiocyanate on rate of fibrin formation from human fibrinogen and thrombin. Ionic strength in all solutions was 0.15. Top curve 0.15 M NaC1; middle curve 0.135 M KI + 0.015 M NaCI; bottom curve 0.135 M KSCN + 0.015 M NaCl. Note that the pH in the iodide solution was significantly higher than in the other two. This increased pH accelerated the reaction; thus the rate of fibrin formation in the iodide solution would have been lower at lower pH. Control experi-ments showed that the effect of potassium salts was indistinguishable from that of sodium salts with the same anion with respect to rate of fibrin formation and clot turbidity.(From measurements of Walter F. Lever.) -0.8 Y A7 0 J I 2 3 9 Time in Hours FIG. 5.-Effect of iodide and thiocyanate on turbidity of fibrin clots as a function of time. Molar concentrations of salts in these experiments were identical with those denoted by corresponding symbols in fig. 4. (From measurements by Walter F. Lever.) * Acknowledgments are made to the Josiah Macy Jr. Foundation for permission to publish fig. 4 and 5 which are slightly modified versions of those appearing in Blood Clotting rmd Allied Psohltwis. Trmsnctions of' the Forirtk CorlJ;.renre ( I95 I) pp. 241-2 JOHN T . EDSALL 17 Certain anions such as iodide and thiocyanate-also well known as denaturing agents for many proteins at much higher concentrations-act in the opposite direction on the structure of the clot greatly decreasing its turbidity as compared with that of the control at the same pH and ionic strength (fig.4 and 5). In these cases therefore there is suggestive evidence of selective anion binding. It should be noted that in all cases these reagents diminish the initial rate of fibrin formation, as shown in fig. 2 and 4 whatever their effect on the structure of the clot. In no case however was there any destruction of thrombin by the added reagent as control experiments showed so that the decreased rate of reaction cannot be laid to this. Closely related to these studies are the extensive and important studies of J. D. Ferry and Sidney Shulman on the effects of many reagents in inhibiting the conversion of fibrinogen to fibrin.These have been presented by Dr. Shulman 19 in a paper to be discussed at this conference so that I shall not treat them further here. SERUM ALBUMINS SIZE AND SHAPE The serum albumins have been among the most intensively studied of all pro-teins ; yet even the size and general shape of the albumin molecule are still a matter of dispute. Let us consider some of the evidence. (i) By the use of dinitrofluorobenzene van Vunakis and Brand 20 could find only one free terminal a-amino group in human horse or bovine serum albumin per molecule of molecular weight approximately 70,000. In all three cases this amino group belonged to an aspartic acid (or asparagine) residue. Thus it would appear that the molecule consists of a single peptide chain. Yet Weber,21 in the paper he is presenting to this meeting has concluded from polarization fluorescence studies that serum albumin dissociates into smaller units in both acid and alkaline solution.There is no necessary incompatibility between these findings but inevitably they raise certain questions. Does the splitting of the albumin molecule involve the breaking of some exceptionally labile peptide linkages? If so what brings about the reassociation that Weber has found to occur on neutralizing the acid albumin solution? Or on the other hand are there actually two or more peptide chains within the albumin molecule only one of which-presumably for steric reasons-reacts with dinitrofluorobenzene? Or is it possible that Weber's findings can be explained in terms of strong electro-static interactions between molecules carrying a high net charge without actually having to assume molecular dissociation? The latter possibility seems to me relatively unlikely but it should not be altogether disregarded.(ii) The molecular weight of human and bovine serum albumin has generally been taken in recent years as 69,000 chiefly on the basis of the very careful osmotic pressure studies of Scatchard Batchelder and Brown.22 The recent X-ray measurements on albumin crystals in our laboratory by Barbara LOW,^^ combined with accurate determinations of density and composition definitely indicate a lower molecular weight near 65,500. Quite independently Creeth 24 has deduced an almost identical value for bovine serum albumin from very careful sedimentation and diffusion measurements.There is converging evidence indeed from several laboratories both in England and the United States that some of the generally accepted measurements of sedimentation constants are in fact too high and require some downward revision-a process which inevitably involves a downward re-vision of molecular weights as well. (iii) In 1947 Oncley Scatchard and Brown25 gave a detailed discussion of sedimentation viscosity and other data for serum albumin and-assuming 20 % hydration-described the molecule approximately as a prolate ellipsoid about I50 A long and 38 A in diameter at the middle. Their choicc of a prolate rathe 18 SIXTH SPIERS MEMORIAL LECTURE than an oblate shape to describe the molecule was determined largely by Oncley’s dielectric dispersion data which indicated two distinctly different relaxation times -a finding incompatible with an oblate shape if the preparation consists of molecules which are all alike.Recently Oncley Dintzis and Hollies 25a have obtained dielectric measurements of much higher precision on carefully purified human mercaptalbumin and other serum albumin preparations freed from fatty acid and other impurities by treatment with ion exchange resins. These prepara-tions gave considerably higher dipole moments than earlier less pure samples. The form of the dielectric dispersion curve was quite sensitive to the protein con-centration ; but at sufficiently low concentrations the curve was entirely consistent with that to be expected for a prolate ellipsoid of molar volume 60 I.and an axial ratio of 4 to 1. The studies of Weber 26 on the polarization of fluorescence from albumin derivatives, also give important evidence on this point and on the whole Weber’s findings appear to confirm those of Oncley. No oblate ellipsoidal mode1 could be chosen to fit the data. t l C 0 0 2 0 -1 L FIG. 6.-Model for possible modes of packing of human serum albumin in the crystal of mercaptalbumin mercury dimer (a wet crystal; b and c air dried crystals). All drawings represent the c plane projection. Two different modifications of Oncley’s model for the serum albumin molecule are indicated. From Low.23 (iv) As was discovered by Hughes,27 two molecules of human mercaptalbumin are readily linked by means of their sulphydryl groups through a single mercury atom to form a dimer.Oncley was led to reconsider his earlier model for the albumin molecule in order to account for the steric relations involved in the forma-tion of this dimer. This revised model has been employed by Low 23 to describe the packing of serum albumin molecules in the crystal of mercaptalbumin mercury dimer. The model consists of a prism approximately 145 A long 45A in dia-meter at the middle and 22A thick. Such a model can fit well into the unit cell of the crystal as determined by Low; and the indications of a layer structure which are given by the X-ray data are well accounted for by the piling into layers of such long thin molecules (fig. 6). It should be clearly understood that the X-ray data do not prove the structure but they are apparently compatible with it and the model itself-to be taken of course only as a very rough approximation -seems reasonably plausible on other grounds.(v) As was found by Foster and myself,2* human serum albumin shows flow birefringence although only at very high gradients in solvents of high viscosity. From our measurements we estimated a length at least as great as that of l50& inferred by Oncley on other grounds. However these measurements on so relatively short a molecule were exceedingly difficult and subject to considerabl JOHN T . EDSALL 19 error ; they should not be taken as having precise quantitative significance. They do point however to a somewhat asymmetrical shape. (vi) In contrast to all these conclusions Riley and Oster 29 have inferred from low angle X-ray diffraction studies on very concentrated albumin solutions that the shape of the molecules cannot be very far from spherical.The theory under-lying this deduction is one which I do not feel competent to criticize ; but I feel very reluctant to accept the conclusion itself in view of the other evidence summarized above. (vii) Very recently Anderegg Beeman and Shulman30 have carried out low angle X-ray scattering studies on relatively dilute albumin solutions. Here the interpretation of the data is more straightforward than for the very concentrated systems studied by Riley and Oster. At vzry low angles the scattering curves indicate that the radius of gyration is definitely greater than for a sphere of the same molecular weight while the scattering at somewhat higher angles which is more characteristic of shape appears to be in better accord with that expected for an oblate than for a prolate ellipsoid of revolution.Assuming an oblate ellipsoid the ratio of the molecular axes is found to be near 3-5 to 1. These data must I think be taken very seriously; we cannot evade the attempt to find some explanation that will make them understandable in conjunction with all the other findings that seem to point to an elongated shape. Dr. Beeman has remarked, in a recent letter to me on the subject “ We can say from our data that we are much more convinced that the molecule is not a prolate ellipsoid of revolution than that it is an oblate ellipsoid of revolution.” Indeed the major moral of all this may be that we should not take too seriously calculations based upon simple shapes.They are useful for representing the results of particular researches, and for correlating the properties of proteins derived from hydrodynamic measure-ments but in detail the description of the shapes of protein molecules must involve something more complex than the very simple geometrical figures that are con-veniently used for preliminary descriptions. It would be a feasible though certainly a very tedious task to calculate the scattering function for the prismatic model of serum albumin previously discussed; and the curve so derived might be quite different from that of a prolate ellipsoid ; but the case seems a little too special at present to warrant the work involved. It does seem to me that the weight of the evidence does rather favour a general model of this sort and that an oblate ellipsoidal shape for albumin or anything very close to such a shape is unlikely.However I do not wish to be dogmatic on the subject ; what I hope to do in this lecture is to ask more questions than I can answer and I trust that others will come forward with the answers or will set to work to find them. ON THE POSSIBLE FLEXIBILITY OF PROTEIN MOLECULES In contrast to the arrangement of most synthetic polymers which coil about in random fashion in solution proteins have a highly ordered pattern. The evidence for this order is clearest from X-ray studies on protein crystals. Such order also implies a high degree of molecular rigidity which is shown clearly for proteins in solution by the dielectric studies of Oncley.31 Protein molecules orient as entire units in an electric field not as a series of more or less independent segments as do the polar polymers.All this is true and important ; yet we should certainly not assume that this rigidity is strict and absolute. Indeed Klotz and Ayers,32 in a paper presented at this meeting conclude that the molecule of serum albumin must be able to change its configuration with pH ; that it may swell or unfold somewhat as its net charge increases and that these efTects are reversible. Others I think have made similar suggestions before. Likewise the remarkable interactions involved in the combination of haemoglobin with oxygen have led Pauling,33 and also Wyman and Allen,34 to suggest that the addition of oxyge 20 SIXTH SPIERS MEMORIAL LECTURE actually leads to a rather fundamental change in the configuration of the mole-cule as a whole.Pauling and Wyman hold rather different views as to the nature of this change and we certainly need more proof before we assume that the change really occurs but the suggestive evidence is strong. Even more striking I think are some recent unpublished studies here in Cambridge on the enzyme fumarase from pig heart recently crystallized by Vincent Massey 35 working with Dr. Malcolm Dixon. Combined observations by Massey on enzyme activity and by Weber using polarization of fluorescence, suggest alterations of molecular size or shape or both with varying pH and ionic strength ; and these variations appear closely correlated with changes in enzyme activity and the formation of the enzyme-substrate complex.We have long been accustomed to change of shape on the part of the fibrous proteins as in the a-/3 transformation of keratin. Indeed the myosin system of muscle could not do its job unless it changed its shape radically and rapidly. Alterations of shape among the corpuscular proteins are presumably much less drastic-that is if they occur under conditions in which the protein still remains undenatured-but they may be of profound importance both chemically and biologically. The fluorescence polarization technique of Dr. Weber may be of great value in such studies for unlike the dielectric dispersion method it can be applied over a very wide range of pH and salt concentration to give information about the relaxation times of proteins.SERUM ALBUMINS INTERACTlONS WITH CERTAIN CATIONS The serum albumins are famous for the strength and diversity of their inter-actions with anions; the story is a long and remarkable one and I have nothing to add to it to-day. Dr. Klotz and Ayers 32 are reporting some new developments in this area and I shall confine myself to a discussion of the interactions of serum albumin with two cations both from group 2 B of the periodic table namely, zinc and mercury. The use of these two cations in the fractionation of blood plasma proteins is discussed at this meeting by Cohn Surgenor Schmid Batchelor, and Isliker.36 The far-reaching developments which they are presenting speak for themselves; here I shall discuss only some recent work from our laboratory which bears on the chemical nature of the interactions involved.The tendency of metallic ions toward complex formation has been admirably surveyed by Jannik Bjerrum,37 who himself has done so much for the advance-ment of knowledge in this field. From his data it is apparent that the tendency of zinc ion to complex formation with the carboxylate ion is relatively weak (log k is approximately zero for zinc and acetate ion where k is the association constant). The association tendency with ammonia is much stronger (log k = 2-37) and this should give an approximate measure of the affinity of zinc for amino groups. However the affinity of amino groups for hydrogen ion is so strong that the zinc binding capacity at pH 7 or below is relatively small.One very important group found in proteins the irnidazole group of histidine, had not been previously investigated; Gurd and Goodman undertook to study the combination of zinc ion and imidazole while Felsenfeld and I were studying its combination with cupric ion.33 Both these ions were found to bind four molecules of imidazole the successive values of log k being tentatively estimated (at 4.5") for copper as 4.58 3-76 3.14 and 2.18. For Zn2+ they were 2.76 2.38, 2.40 and 2-22. These values of k for zinc ion are defined by the relation : In these equations Im denotes uncharged imidazole. On correcting for the statistical factors involved it is interesting to note that although the first Cu2+ ion is bound much more strongly than the first Zn2' the binding of one imidazol JOHN T .EDSALL 21 by copper decreases thc probability of binding the next whereas the binding of the first imidazole by zinc increases the affinity for more imidazoles until saturation is achieved. While this work was proceeding Gurd and Goodman 39 also made a most careful study of the binding of zinc ion by serum albumin employing the dialysis FIG. 7.-Effect of zinc ion concentration on zinc binding to human serum albumin. From Gurd and Goodman.39 OH- Added Albumin Temp.= 0°C 0.15M NaNO, -2.5% Albumin 9 14.0 I / 10.0 o 5.8 Q 7.5 I 4 FIG. &-Dependence of n on v (see text) in the binding of zinc ions by human serum albumin. From Gurd and Goodman.33 equilibrium technique. This work required great care in setting up the experi-ments and the analysis of the data involved a quite elaborate examination of the competition between zinc and hydrogen ions for the various sites on the protein, and of the influence of electrostatic forces due to the charge on the protein on the affinity constants.The former Some of their results are shown in fig. 7 and 8 22 SIXTH SPIERS MEMORIAL LEC'TIJRE illustrates zinc binding as a function of free zinc ion concentration for various amounts of OH- ion added to the isoionic protein; while the latter shows the number n o€ imidazole groups available to combine with hydrogen ions as a function of C the number of Zn2+ ions bound per molecule of albumin. The conclusion from these studies was clear cut ; the binding of Zn2+ by albumin in the pH range 5.5 to 7.5 is due almost entirely to the 16 imidazole groups of the histidine residues.At more acid pH values very little zinc is bound and at more alkaline values the precipitation of zinc hydroxide renders experimental measurement impracticable. Moreover the association constant involved is defined by the relation log k" = 2-82, where k" denotes the association constant corrected for the influence of electro-static forces. This value agrees extraordinarily well with the first association constant in the relation between the zinc ion and the imidazole molecule (log kl = 2.76). Indeed such very close agreement could hardly be expected and is perhaps accidental. Nevertheless this value certainly adds strong confirmatory evidence for the significance of the imidazole groups in the binding of zinc by albumin.The carboxylate groups in spite of their much greater number-106 per albumin molecule-show no significant binding at least in aqueous solutions. However, in the presence of concentrated urea as Cohn et al.36 are reporting at this meeting there is evidence from recent work of Dr. Gurd that carboxylate groups become available for binding. The amino groups would certainly combine at more alkaline pH values but their relatively strong basicity renders them unreactive with zinc at neutral pH. Quite independently of Gurd and Goodman Tanford 40 studied polaro-graphically the binding of zinc and several other ions by serum albumin and has come to identical conclusions concerning the significance of the imidazole groups for the binding.The mercuric ion in spite of its electronic relationship to that of zinc differs profoundly in its interactions with the groups found in proteins. Its affinity constant for amino groups is more than a million times as great as that of the Zn2+ ion; and its affinity for imidazole groups although the exact figure is not known to me is certainly far greater also. However its well-known affinity for sulphydryl groups is the dominating feature in its interactions with proteins as long as the protein has any free sulphydryl groups left to react. My colleague, Dr. W. L. Hughes Jr. whose work in this area has been basic for all that the rest of us have done estimates that log k for the association of Hg2+ with a sulp-hydryl group is of the order of 17 to 19.On the other hand we must remember that the amount of free Hg2+ ion present in most aqueous solutions of a mercuric salt is extremely small because of its ex-tremely strong affinity for the hydroxyl ion and for all halogen ions except fluoride. The reversible combination of mercury with mercaptal bumin was discovered by Hughes,27 who formulated the essential features of the reaction and crystallized the mercury dimer of mercaptalbumin. Oncley by means of ultracentrifuge measurements showed the simultaneous presence of dimer and monomer in solutions derived from these crystals and obtained indications that they existed in a state of reversible equilibrium. Over a period of several years we have continued the study of the process by light-scattering measurements and have characterized in a more quantitative manner both the kinetics of the formation and dissociation of the dimer and the position of the equilibrium as a function of pH ionic strength the concentration of certain specific ions in the solution, temperature and other variables.It was also discovered that certain organic mercurials containing two or more mercury atoms were also capable of linking two albumin molecules together to form a dimer. The one most thoroughly studied was the dioxane derivative introduced into the laboratory by Straessle 41 CH2-0 \O-CHz/ +Hg . CH2. HC/ \CH . CH2. Hg JOHN T. EDSALL 23 which for brevity will be denoted as HgRHg. This compound as will be seen, induces dimer formation far more rapidly than mercuric chloride itself.Other organic mercurials prepared by Dr. Howard Dintzis are now being studied. This work has been carried on by several investigators in our laboratory; its progress is due to the thinking and the experimental skill of Harold Edelhoch, R. B. Simpson Rudolf Straessle Ephraim Katchalski and R. H. Maybury. The insight of Dr. Hughes has been invaluable to us throughout the progress of the work. First is the combination of mercury with mercaptalbumin (denoted here as ASH) through the sulphydryl group of the latter : The reaction may be described as proceeding in three stages. ASH + HgC12 + ASHgCl + H+ + C1-. (1) This reaction involves no significant change of the molecular weight of the protein ; it therefore cannot be measured by light scattering and it is in any case very rapid-probably too fast to measure by most techniques.The product of the first reaction then reacts with a second molecule of mercaptalbumin to form the dimer : (11) This is the reaction which is directly observed by light scattering; it proceeds very slowly as compared to reaction I and is readily followed. Finally if an excess of mercuric chloride is added the dimer rapidly dissociates according to reaction 111. ASHgCl + ASH + ASHgSA + Hf + C1-. ASHgSA + HgC12 + 2ASHgCl. (111) The exact formulation of these reactions is open to some discussion. In the solution before reaction starts and at the concentrations used in our studies, most of the mercury is in the form of undissociated HgC12. Some however must be present in combination with hydroxyl or acetate ion; some presumably as HgCl+ ion.The amount of free Hg2+ is certainly minute at equilibrium. Probably, however the reactions I I1 and 111 represent the most convenient simple way of formulating the process. Dr. Hughes has demonstrated the release of hydrogen ions accompanying reactions I and 11. The interpretation of the light-scattering measurements in order to derive weight average molecular weights also raises certain problems. For the monomer, of molecular weight M we can employ the usual equation for light scattering in dilute solutions : Kcl/RhO = M-1 + 2Bc1, Kc~/R$~ = (2M)-1 + 2B~2. (W (V) and for the dimer of molecular weight 2M: Here c1 and c2 denote the weight concentrations of monomer and dimer respectively expressed in g/ml RJO and R30 denotes the corresponding values for the reduced intensities of scattering at 90" from monomer and dimer and K denotes the factor : K = 2~2n2(dn/dc)2/NoA-f (VI) where n denotes refractive index N is Avogadro's number and ho is the wave-length of the light in vacuo (Doty and Edsall42).The factor B is the interaction constant which is determined in large measure by the net charge on the protein and the ionic strength of the solution (Edsall Edelhoch Lontie and Morrison 43). If the protein is isoelectric B is zero or nearly zero for both monomer and dimer and the weight average molecular weight can be inferred directly from the reduced intensity. However it was of great interest to study the system under conditions of varying net charge on the protein and under these conditions B i 24 SIXTH SPIERS MEMORIAL LECTURE in generai different from zero.Fortunately it proved in practice that in a given medium the value of B for the pure mononier and the pure dimer is the same within experimental error. We have made the assumption-and it can be justified by theoretical arguments developed by Dr. Katchalski which I need not discuss here-that the same value of B can then be used for any mixture of monomer and dimer in the same solution and we can then obtain the molecular weight by extra-polation. For the mixture of the two species then we may write (VW where c = c1 + c2 and X = M(c1 -+ 2c~)/c. Since c and R90 are measured experi-mentally X may then be calculated from (VIl) provided that the value of B has been determined for the pure monomer or dimer in the same solvent in which the reaction is taking place.The weight fraction of dimer is then given by the equation Kc/RgO = X-1 -t ~ B c , (VIII) Using this approach we have been able to obtain extensive information con-cerning the velocity of the reaction and the proportion of dimer which is present at equilibrium. Repeated experiments have shown that true equilibrium is actually attained. In a given medium one may start either from pure monomer with an appropriate amount of mercury added or from a solution which is nearly pure dimer ; and on adjusting to the same weight concentration of protein in each case one attains the same final equilibrium value. If the activity of hydrogen and of chloride ions is fixed this equilibrium may conveniently be defined in terms of an apparent equilibrium constant (ASHg SA) (ASHgCI)(ASH) Kapp == and the variation of this apparent constant with change in pH and ionic strength may then be studied.To our surprise Kapp was found to vary little over the pH range 4-25-6 the mean value being 2.6 (& 0-4) x lo4 over the whole range. The measurements were carried out in phosphate buffers at pH 6 and in acetate buffers at the lower pH values. In simple terms this value of Kapp means that in a system containing 0.5 mole Hg per mole albumin if the total weight con-centration of protein is 10 g/l. then the weight fraction of dimer at equilibrium is near 0.5. Preliminary studies of the temperature coefficient of equilibrium indi-cate that rise of temperature favours dimer formation and therefore that the process is presumably endothermic.For a given protein concentration in a given medium the weight fraction of dimer at equilibrium varies in a characteristic fashion with the amount of mercury added. It is of course zero in the absence of added mercury when all mercapt-albumin is present as ASH and falls again to zero when one mole of mercury has been added per mole of albumin all of the mercaptalbumin being then in the form ASHgCI. The maximum amount of dimer formation always occurs when the molar ratio of mercury to albumin is 0.5. The value of the weight fraction of dimer at this point increases with the value Kapp. The form of the whole curve indeed can be calculated from the value of the equilibrium constant.The curve is symmetrical about its midpoint and is similar in charactcr to the curves theoret-ically deduced by Michaelis 44 for the fraction of semiquinone formed in the re-versible titration of an oxidation-reduction system as a function of the total of oxidant or reductant added and of the semiquinone formation constant. A family of some of the curves to be expected for different values of Kapp is shown in fig. 9. We have also made extensive determinations of the velocity constant kf of the combination reaction between ASHgCl and ASH. This constant expressed in 1. mole-1 min-1 is equal about to 40 from pH 4.25-4.75. As the pH increase JOHN T. EDSALL 25 further however the value decreases progressively to 0.69 at pH 6 .I t seems probable that this variation at pH values above 4.7 can be at least partly explained by the electrostatic effect of the increasing negative net charge on the albumin molecules above pH 5. The treatment involved here would be essentially that developed by Bronsted N. Bjerrum Scatchard and others for simpler i0ns.~5 However since the equilibrium constant is essentially independent of pH the results indicate clearly that the dissociation of dimer into monomer is far slower at pH 6 or above than below pH 5. This has indeed been confirmed experi-mentally. A concentrated solution of the dimer may be prepared and adjusted to a pH somewhat above 6 in phosphate buffer. Even if the solution is then greatly diluted the dissociation of the dimer is found to be extremely slow as 100 80 L (Y E *6 60 .c.0 0 ) c e 40 2 20 0 Molar Ratio HgCI,/(ASH) at t= 0 FIG. 9.-Percent of mercaptalbumin-dimer formed at equilibrium as a function of HgC12/ mercaptalbumin ratio. Curves are shown for various values of the parameter Kappa (apparent association constant multiplied by the molar concentration of albumin). compared to the rate below pH 5. In this way dilute solutions containing nearly pure dimer with very little monomer can be obtained for study over a considerable period of time as for instance in ultracentrifuge runs. Studies of interaction of mercaptalbumin with the organic dioxane derivative discussed above (HgRHg) indicate a striking difference from the interactions with mercuric chloride. The equilibrium at all the pH values studied seems to lie almost completely in the direction of dimer formation.At low pH values, below 4.8 the velocity of the process is so great that no accurate velocity constants for dimer formation could be determined by our methods. Between pH 5 and 6, the value of kf decreases with rising pH in much the same manner as for the re-action with mercuric chloride. However the absolute values of kf at any pH in this range are very much greater-indeed several thousand times as high as those for the reaction involving mercuric chloride. A plot of the two sets of kfvalues is given in fig. 10. The great contrast between the rates and the equilibrium constants found in these two different types of reaction with mercury compounds is presumably t 26 SIXTH SPIERS MEMORIAL LECTURE be explained largely on steric grounds.When two albumin molecules are linked through their sulphydryl groups by a single mercury atom the juxtaposition of the adjoining surfaces of the two protein molecules must be extraordinarily close. It is indeed remarkable that the surfaces are so shaped that they can be fitted together at all. It will be of great interest to see whether chemical modification of other groups in the albumin will lead to alterations in the tendency to undergo this reaction. In any case the relative slowness and incompleteness of the reaction are not surprising. On the other hand for the dioxane derivative HgRHg in which there is a mercury atom at each end there is an effective distance of the order of lOA between the two mercury atoms.Thus the coupling of the two albumin molecules which are linked by these atoms is far looser than when a single mercury serves as the link. The conditions for the combination are much less critical and pre-sumably the reaction is possible in a much greater variety of relative orientations A S Hq RHq SA \ log K f t: 10 A5 Hq SA I -PH :2 4; 6 5 0 5 4 5;8 , FIG. IO.-Logarithm of the velocity constant kf for dimer formation of human Serum mercaptalbumin with mercurials. Lower curve reaction with mercuric chloride ; upper curve reaction with the dioxane derivative containing two mercury atoms. Phosphate buffers were employed near pH 6 acetate buffers at lower pH values; ionic strength 0.05 in acetate buffers; 0.05-0.15 in phosphate; temp.25". (From studies of E. Katchalski H. Edelhoch and R. B. Simpson.) of the two albumin molecules than would be the case if the fitting were closer. Thus the reaction goes relatively rapidly and it goes largely to completion. One may suspect although this is not proved that the two halves of the dirner ASHgRHgSA may be much more loosely coupled than the halves of the dimer ASHgSA. Thus in the former case a much greater degree of free rotation of one-half of the dimer relative to the other should perhaps be expected than in the latter; a possibility which may perhaps be explored by such methods as dielectric dispersion. It should be added that in spite of the presumably very close fitting of the two halves of the dimer ASHgSA the mercury atom does not appear to be buried within the protein in such a way as to be inaccessible to the solvent.Addition of other ions such as bromide iodide or cyanide which have a strong affinity for mercury leads to quite rapid dissociation of the dimer. Evidently these ions can penetrate readily to the mercury atom which is then released from its com-bination from one of the sulphydryl groups to form the monomer ASHgBr, ASHgI or ASHgCN. The reaction has been studied by us particularly wit JOHN T. EDSALL 27 the iodide ion. The addition of even a single mole of iodide per mole of mercury in the dimer leads to a relatively rapid and very marked dissociation of the dimer. Bromide is inuch less effective and chloride far less than bromide. Indeed, these studies appear to be significant as a method of evaluating the relative affinity for mercury of various inorganic ions and thus obtaining data which should be of interest for inorganic chemistry.This is of course not our primary concern; we are interested in these reactions chiefly because of what we can learn from them concerning the external form the surface properties and re-activity and perhaps something of the finer pattern of the configuration of the albumin molecule. The extensive studies carried out by Lewin 46 in our laboratory demonstrated that mercaptalbumin mercury dimer could be crystallized in association with many other metallic cations and also with anions containing atoms of high atomic number. Dimer formation of the type first observed by Hughes27 with serum albumin has not yet apparently been observed with other proteins.Several years earlier however the successful use of mercury in the crystallization of enolase was reported by Warburg and Christian.47 In this case the crystallized mercury derivative contained just one mole of mercury per mole of enolase and other derivatives have not been found. The interaction of mercury with proteins is of course not confined to a single group. However the affinity of mercury for the sulphydryl group is so much greater than for any other groups found in proteins that no other reaction will occur until all such groups are saturated with mercury. Beyond this point, however important interactions begin to occur with imidazole amino and other groups. These interactions between mercury and proteins have recently been exploited with great success by Dr.Cohn and his associates in fractionation processes .36 1 Sanger and Tuppy Biochem. J. 1951,49,463,481. Sanger Thompson and Tuppy, 2 Pauling and Corey Proc. Nut. Acad. Sci. 1951 37 205 235 241 251 261 272, 3 Bragg Howells and Perutz Acta Cryst. 1952 5 136. 4Kendrew Proc. Roy. SOC. A 1950 201 62. 5 Carlisle and Scouloudi Proc. Roy. SOC. A 1951 207 496. 6 Low Nature 1952 169 955. 7 Hocking Laskowski and Scheraga J. .4mer. Chem. SOC. 1952 74 775. 8 Katz Gutfreund Shulman and Ferry J. Amer. Chem. SOC. 1952 74 5706. 9 Edsall Foster and Scheinberg J. Amer. Chem. SOC. 1947 69 273. 10 Hall J. Biol. Chem. 1949 179 857. 11 Porter and Hawn J. Expt. Meci. 1949 90 225. 12 Mitchell Biochem. Bioplzys. Acta 1952 9 430. 13 Bailey Bettelheim Lorand and Middlebrook Nature 1951 167 233 ; Bettelheim and Bailey ResumL des Communications I P Congres Internationale de Biochemie, 1952 411; Lorand and Middlebrook Biochem. Biophys. Acta 1952 9 581; Bettelheim and Bailey ibid. 9 578. Symposiiiin sur les Hormones Proteines et Derivtes des Proteines (Paris 1952). 282 729. 14 Lorand Nature 1951 167 992. 15 Edsall and Lever J. Biol. Chem. 1951 191 735. 16 Ferry and Morrison J. Amer. Chem. SOC. 1947 69 388. 17 Laki and Lorand Science 1948 108 280. 18 Lorand Nature 1950 166 694. 19 Shulman this Discussion. 20 van Vunakis and Brand Abstr. 119th meeting Amer. Chem. SOC. April 1951 p. 28c. 21 Weber this Discussion. 22 Scatchard Batchelder and Brown J. Amer. Chem. SOC. 1946 68 2320. 23 Low J. Amer. Chem. SOC. 1952 74 4830. 24 Creeth Biochem. J. 1952 51 10. 25 Oncley Scatchard and Brown J. Physic. Chem. 1947 51 184 28 SIXTH SPIERS MEMORIAL LECTURE 25a Oncley Dintzis and Hollies Abstr. 122nd meetirrg Amer. Chem. SOC. Sept. 1952: 26 Weber Biochenr. J . 1952 51 145 155. 27 Hughes J. Amer. Chem. Sac. 1947 69 1836 ; CoM Spring Harbor Symp. Quant. 28 Edsall and Foster J. Amer. Chem. SOC. 1948 70 1860. 29 Riley and Oster Furuday SOC. Discussions 1951 11 107. 30 Anderegg Beeman and Shulman paper presented to Amer. Physic. SOC. April, 1952 ; and unpublished work. 31 Oncley Chem. Rev. 1942 30 433. 32 Klotz and Ayers this Discussion, 33 St. George and Pauling Science 1951 114 629. 34 Wyman and Allen J. Polymer Sci. 1951 7 499. 35 Massey Biochem. J. 1952 51 490. 36 Cohn Surgenor Schmid Batchelor Isliker and Alameri this Discussion. 37 Bjerrum Chem. Rev. 1950 46 38 1. 38 Gurd Edsall Felsenfeld and Goodman Fed. Proc. 1952 11 224. 39 Gurd and Goodman J. Amer. Chem. SOC. 1952,74 670. 40 Tanford J. Amer. Chem. SOC. 1952 74 21 1. 41 Straessle J. Amer. Chem. SOC. 1951 73 504. 42 Doty and Edsall Advances in Protein Chem. 1951 6 35. 43 Edsall Edelhoch Lontie and Morrison J. Amer. Chem. SOC. 1950 72 4641. 44 Michaelis Chem. Rev. 1935 16 243 ; see especially fig. 4. 45 see for instance the discussion by Laidler Chemical Kinetics (McGraw Hill Pub-46 Lewin J. Amer. Chem. SOC. 1951 73 3906. 47 Warburg and Christian Biochem. Z. 1941 310 384. p. 12p; Dintzis Thesis Harvard University 1952. Biol. 1950 14 79. lishing Co. New York 1950) p. 123