GERTRUDE E . PERLMANN 77 THE GLOBULAR-FIBROUS PROTEIN TRANSFORMATION BY E. BARBU AND M. JOLY Service de Chimie-Physique et Service de Chimie Biologique, Institut Pasteur, Paris Received 7th May, 1952 Some globular proteins can be transformed into fibrous proteins by moderate heating, salt effect or high pressure. It has been studied by electron microscopy, viscosity, electrophoresis, ultracentrifuge, paper electro- phoresis, fractional precipitation and streaming birefringence. The problem is to know whether this transformation is due to any unfolding of the polypeptide chains or to the end-to-end aggregation of protein molecules that remain globular. A great number of experimental facts support the second assumption. The denaturation of protein molecules is an activation process.The molecules thus activated aggregate when they collide. The kinetics of this process has been studied both experimentally and theoretically. This transformation is partially reversible. The transformation of globular proteins into fibrous is generally performed under very drastic conditions,l-4 e.g. at pH 10-1 1, in the presence of urea or deter- gents, by violent extruding and precipitation. The analogy of corresponding X-ray diagrams with those of stretched keratin prompted the authors to interpret the globular-fibrous (g-f) protein transformation as an unfolding of polypeptide chains. Previous research has shown that less powerful treatment of protein solutions can cause the formation of thread particles by end-to-end aggregation of globular molecules. This occurs, for instance, in the first stages of the sol-gel transformation of gelatin,s* 6 in the thermal denaturation of horse 7-9 or human 10 serum albumin and of egg albumin,ll-13 in insulin gelation 14-16 or in fibrinogen association in- duced by thrombin.17 We have recently studied the effects of different physical and chemical factors on the process of linear aggregation in the thermal denaturation of horse serum alb~min.7~ 8 The purpose of the present paper is to give some new experimental and theoretical information on the g-f transformation and to extend this investi- gation to other proteins and denaturing agents.78 GLOBULAR-FIBROUS PROTEIN TRANSFORMATION EXPERIMENTAL MATERIALS.-The horse serum albumin and the hen egg albumin were prepared by salting-out with (N&)2so4 and by successive crystallization according to Sorensen.18 The conservation of our sample of Cohn's crystalline bovine serum albumin (no. 59, control 128-1 65) was tested by electrophoresis.The horse y-pseudo-globulin was pre- cipitated three times by 33 % saturated (NH&S04, dialyzed against H20 and centrifuged to eliminate the euglobulins. Rabbit actin was prepared according to Straub 19 and dissolved in twice distilled water cooled in absence of CO2. METHODS.-(a) Streaming birefringence.-The capacity of the apparatus (with moving inner drum) used for the streaming birefringence measurements was 5 cm3 ; the cylinders were 7 cm high with a gap of 0.025 cm ; the velocity gradient could be varied between 80 sec-1 and 6,000 sec-1.The extinction angle X was determined with an accuracy of f 0.1" by the differential method of Frey-Wyssling and Weber.20 For elongated particles the apparent lengths Z(g) for each value of the velocity gradient g were determined from the X (g) values using the tables of Scheraga, Edsall and Gadd.21 The size I f of the most frequent particles and the polydispersity P of the solutions were derived using the semi-empirical relationships given by one of us.% 23 The accuracy was 5 % for I f and 10 % for P. The precise significance of I f and P was given in a preceding paper.22 We shall use later the rate of polydispersity T which is, by definition, the quotient P/If. For globular particles the computation of precise relationships giving the most fre- quent diameter df and the polydispersity is now in progress ; here we shall give only the mean diameter d obtained empirically by extrapolating the d(Z/g) curve against Z/g to 0 (the validity of this method was checked by electron microscopy).From the experimental values of the streaming birefringence A = n, - no, we can deduce for very dissymmetrical monodisperse particles the order of magnitude of the intrinsic particle anisotropy A = n12 - n22, 11: and 112 being the principal refractive indexes of the particles. We have A N A/cf(X), f(X) being a function of the extinction angle, and c the volume concentration of the particles. This quotient is independent of c and g and so bas real significance. For polydisperse systems A/cf(X) is not independent of g and of the solution composition ; we can only determine apparent values A(g) of the anisotropy, of which we shall give the mean value A-for velocity gradients ranging from 80 sec-1 to 5,500 sec-1.In a mixture of small globular and large elongated particles, only the threadlike particles give streaming birefringence, and so, by comparing the values of 2 during a g-f transformation, we can estimate the rate of transformation. By this means we find, for instance, with horse serum albumin solutions (c = 2.4 %, pH = 4.15, acetate buffer M/50) heated for 10 min at 65" C and 80" C, a ratio of the transformation rates equal to 60 % (in good agreement with the value 61 % given by precipitation in 16 % Na2S04). If the particles are not very elongated, A/cf(X) + n12 - n22 and the determination of f ( X ) requires a knowledge of the axial ratio p of the particles.But for any system we can arbitrarily define, by means of the relationships valid for very elongated particles, an equivalent length of the most frequent particles [b], an equivalent rate of polydispersity [;z] and an equivalent mean anisotropy [a; these parameters have no direct physical significance, but they are very useful in the study of globular or quasi-globular aggrega- tion 24 and of photoelastic effect or dynamic turbidity.25 For instance, if we know If and A for a given protein under such conditions that the fibrous state of the particles is com- plete, and if for the same protein under different conditions the apparent values [ I f ] > r f and [A] < 2, we can conclude that the second type of particle is much less dissymmetrical, and from these values it is possible by successive approximations to obtain the order of magnitude of the axial ratio p of the quasi-globular aggregates. (b) Other teclzniques.-For the viscometry measurements we used a Baume capillary viscometer, the constant K of which was 4945 x 10-5, The electrophoresis was carried out in a Moore apparatus following the method of Tiselius, and the paper electrophoresis was performed according to the technique of Durrum as used by Macheboeuf and Rebeyrotte.26 The ultracentrifuge was of the Phywe type, running at 36,000 or 42,000 rev/min, with an optical design by Wielgosz.27 The precipitation at the isoelectric point was studied by N titration of the super- natant obtained by centrifugation, 3 h after putting the protein in a 14 % or 16 % Na2S04 -E.BARBU AND M. JOLY 79 solution at pH4.75. The solutions were compressed at high pressure in the apparatus built by Basset using the technique developed by this author.28 EVIDENCE OF AGGREGAnoN.-From the data of streaming birefringence we have shown in preceding papers7, 8 that the heating of dilute serum albumin solutions gives rise to filaments, the length of which can attain several thousand A, and that the properties of these filamentous particles (e.g. rheological behaviour and tensile strength) cannot be explained by the unfolding of initially folded polypeptide chains, but can be easily under- stood using the concept of end-to-end aggregates. On the contrary, large globular aggregates grow by adding tri- or tetravalent ions to the protein solution.We shall show, too, that the high pressures gave rise in the solu- tions to quasi-spherical aggregates with a diameter of several hundred A. For characterizing these different types of aggregation we examined some solutions of 2.5 % horse serum albumin at pH 7.3, without treatment, after heating for 10 min at 65" C and 80" C, and after compression for 1 h at 37" C under 10,OOO Kg/cm2. (a) Streaming birefringence.-The solutions before treatment do not show streaming birefringence. For the treated solutions we found : heated at 65" C, L) = 700& T = 1.7, = 2-4 X 10-3 ; heatedat 8O"C, b= SOOA, n = 1, A= 3 x 10-3; The quotient of the values of 2 at 65" C and 80" C was 0.8 in good agreement with the ratio 0.85 given by the solubility test.(6) Zsoelectric precipitution.-3 h after mixing with 14 % Na2S04 (PH 4.75) we found as percentage of soluble serum albumin : 100 % for the untreated solution, 17 % for the solution heated at 65" C, 2 % for that heated at 80" C and 18 % for the compressed solu- tion. All these results agree with the values of the mean intrinsic anisotropy. (c) Viscosity.-The measurements were carried out at 22°C in a phosphate buffer pH 7-3 (p = 0.15) with a protein concentration of 0.94 % for the solution heated at 80" C and 0.63 "/, for the other samples. The approximate values of the intrinsic viscosity were 12.9 for the solution heated at 65" C, 16.7 for that heated at 80" C and 11.6 for the compressed solution. The corresponding values of the axial ratios (following Guth's relationship 29, 30) are at a first approximation 13 and 15 for the solutions heated at 65" C and 80" C, 12 for the compressed solution.The results concerning the heated samples fit well with those of streaming birefringence and lead to a width of about 50A for the rod-shaped particles in satisfactory agreement with the approximate diameter of the serum albumin molecules.31 But the value for the compressed solution is hardly compatible with a suspension of rigid and compact spheres. ( d ) Electrophoresis.-The detailed electrophoresis study of the heated or compressed serum albumin solutions will be published in a forthcoming paper.32 We summarize here the results concerning the solutions we investigated. The electrophoresis was carried out after dilution of the protein to 1 % in phosphate buffer (PH 7.5 ; p = 0-15 NaCI) and dialysis against the same buffer.Fig. 1 shows the diagrams (ascending boundary) after 3 h electrophoresis (V = 110 V, i = 18 mA) of the untreated and heated solutions. The electrophoretic patterns are the same for the samples compressed or heated at 65" C . We found for the mobilities the following approximate magnitudes : untreated solution 6.2 (one fraction only) ; solution heated at 80" C, 5.8 (one fraction only) ; solution heated at 65" C, 6.5 (16 %) and 5.5 (84 %) ; solution compressed at 10,000 atm, 6.5 (15 %) and 5.5 (85 %). There is a good agreement of the transformation rates derived from electrophoresis with those obtained from the solubility experiments. The small difference between the mobility of the native protein and that of the heated or compressed, proves that the ratio charge/ frictional coefficient differ very little for the single protein molecules and for the linear or globular aggregates.We particularly verified that this ratio was, for very elongated rod-shaped particles, practically independent of the length of the particles. (e) Puper-electrophoresis.--'Fhe experiments were carried out with 2.5 % solutions. Fig. 2 shows the diagrams obtained by 6 h paper-electrophoresis. We note the great mobility and homogeneity of the untreated sample. The mobility of the particles heated at 80" C was much lower and the corresponding spot spread considerably. The solutions heated at 65" C or compressed at 10,000 kg/cm2 show two spots: one strong with a compressed at 10,000 kg/cm2, 2 =; 350 A.80 GLOBULAR-FIBROUS PROTEIN TRANSFORMATION mobility lower than that of the untreated solution and one weak with a mobility of the same order as that of the native protein.The discrepancy between the mobilities of un- treated and heated protein was much larger in paper-electrophoresis than in ordinary electrophoresis. The mobility of the main fraction of the solution heated at 65" C was a little lower than that of the corresponding fraction of the compressed solution. These differences between ordinary and paper-electrophoresis are related to the fact that a very dissymmetrical particle moves less easily in paper than a globular one does even though the ratio charge/frictional coefficient in water is the same for both the particles.(f) Ultracentrifuge.-The ultracentrifuge studies were carried out with solutions of 1 % of horse serum albumin, in phosphate buffer (pH 7.3; p = 0.15). Fig. 3 shows the sedimentation patterns after 1 h run at 42,000 revlmin for the untreated and at 36,000 revlmin for the compressed or heated samples. The mean values of the sedimentation constants were : for the slow fraction of all the samples (only this one for the untreated solution) s = 4.6 x 10-13 (in good agreement with the literature values) ; for the rapid fractions, 15 x 10-13 with the compressed solution, 20 x 10-13 with the solution heated at 65" C and 29 x 10-13 with that 9f 80" C. From the general relationship of the ultracentrifuge M = -____ ( p solvent density, V specific volume and D diffusion constant of the particles) we can derive the size of the aggregates according to the assumptions made on their shape. For instance, for linear aggregates made of only one row of serum albumin molecules, the length L of the particles is related to s by L N 25 exp (1.6 x 1012 s) A, and for linear aggregates made of two adjacent rows of molecules L N 25 exp (1.2 x 1012 s)A; for a globular but rigid and compact aggregate of serum albumin molecules, the diameter d is given by d N 1-14 108 6 A .By applying these equations to the preceding values of s and by assuming the particle to be a single row of spherical molecules of diameter 50 A, one finds L = 600 A for the solution heated at 65" C ; this is in good agreement with the value derived from streaming birefringence data ( I f = 700 A).For the solution heated at 80" C , the values found by ultracentrifuge and by streaming birefringence are the same (L = I f = SOOA), if we as- sume the particles to be composed of two adjacent rows of protein molecules. This as- sumption is consistent with the mean width of the particles as deduced from electron micrographs and agrees with the variation of the mean intrinsic anisotropy A with the temperature of heating. But for the globular aggregates, the sedimentation measure- ments gives d = 140 A, whereas streaming birefringence and electron microscopy lead to d= 350 A. This means (like the results of viscometry) that the globular aggregates are not rigid and compact globules and that the solvent can partly flow through them.(8) Electron microscopy.-Fig. 4 shows photographs of untreated, heated and com- pressed solutions. The two types of aggregation, filamentous and globular, are evident. One can always object that this effect is produced by some artefact in electron micro- scopy and that such an interpretation of the photographs should not necessarily carry weight. One knows also how uncertain is the discussion of viscometric results in the yet imperfect state of the theory of colloid viscosity. The precise explanation of the electrophoresis or paper electrophoresis patterns and of the sedimentation diagrams is frequently difficult. We have emphasized in preceding papers 22,23, 33 the caution necessary in interpreting streaming birefringence data.But when there is rather good agreement between the results that we obtain by applying such varied techniques, in which the fundamental assumptions are different, it seems sound to draw the following conclusion. Protein denaturation is frequently accompanied (according to the experimental conditions) by linear or globular aggregation of the molecules remaining individually globular. This nevertheless does not exclude partial modification of their shape. Particularly we can legitimately say that the globular-fibrous protein transformation, in all the cases we studied, proceeds by end-to-end aggregation of the globular molecules, and in no circumstance by complete unfolding of the chains. In the following we examine the details of this transformation for different proteins and for various treatments of the solutions.HORSE SERUM ALBUMIN.-The linear aggregation by heating of the horse serum albumin molecules was systematically studied,', 8 and we recall here only the main features of the results. If we consider the variation of the length rf of the most frequent particles during the thermal treatment, for a given temperature, pH, ionic strength and protein concentration, RTs D(I - Vp)'A B C A B C FIG. I . FIG. 3. FIG. 1 .-Electrophoretic diagrams (ascending boundary). t = 3 h ; Y = l l O V ; i = 1 8 m A ; c = 1 % of serum albumin ; phosphate buffer pH 7.5 ( p 7 0.1 5 ) . A, untreated solution. B, sample heated for 10 rnin at 80' C C , sample heated for 10 min at 65" C . FIG. 3 .-UI tracentrifuge patterns.t = 1 h ; c = 1 04; pH 7.3 ( p = 0.15). A, untreated sample B, sample compressed at 10,000 atm C, sample heated for 10 min at 80" C. FIG. 2.-Diagrams of paper electrophoresis. t = 6 h ; c = 2.5 %; pH 7.5 ( p = 0.15). A, solution heated for 10 min at 80 C B, solution heated for 10 min at 65' C C , solution compressed at 10,000 kg,icm2 D, untreated solution. [To face page 80FIG. 4.--Electron micrographs. A, untreated solution pH 7.3 B, solution at 2.5 %, pH 7.3, heated for 10 niin at 80 C and then diluted to 0.01 7; C, solution at 4 %, pH 7.3, compressed for 1 h at 37" C under 8000 kg/cm2 and then diluted to 1 % D, solution at 2.5 %, pH 4.3, heated for 10 inin at 80' C in acetate buRer and then diluted to 0-01 7:. [See puge 8 1E . BARBU AND M .JOLY 81 we observe that it attains after a time 0 a value Lf which then remains practically constant, often for a very long time (see table (1)). Lf depends upon concentration and pH ; 0 decreases when the temperature and the ionic strength increase. After this period of stationary length, If increases again more or less regularly and, according to the values of pH, concentration and ionic strength, coagulation or gelation of the solution can occur. The variation with the time t of the rate of polydispersity 77 depends greatly on the temperature of heating. The increase of the mean intrinsic anisotropy 2 during the first period of the particle growth (and, for low temperatures, at the beginning of the flat part of the If against t curve) corresponds to the increasing number of the particles large enough to promote streaming birefringence. The decrease of A along the flat part of the I f against t curve (or only at its end, for low temperatures) and during the second period of aggregate growth, seems to correspond to the increase of the particle width due to side-by-side aggregation of several rows of molecules, which is in agreement with the information from electron microscopy (see fig.4, D) and ultracentrifugation. One conceives for very elongated linear aggregates, that the polarizability in the length direction is inde- pendent of the number of the end-to-end aggregated molecules, while the polarizability in the width direction changes greatly when there are several side-by-side aggregated rows of molecules instead of one. If the solutions are heated in absence of a neutral electrolyte, 0 is significantly larger, but the flat part of the curve is often very short and sometimes even completely hidden by the second step of growth or by the effect of very strong interactions between the particles (see table 2).TABLE GROWTH OF THE PARTICLES OF SERUM ALBUMIN c = 2.4 % ; pH = 4-15 ; M/50 acetate buffer T("C) t (min) 1.0 57 I f ( & I A x 103 - 65 I,- - 2 x 103 - 80 r f - 90 I f - 2 x 103 - n - rr I 77 - 77 - 100 I f 650 rr 1.5 ;i x 103 - 5 - - - - - - 1100 0.1 1850 1.5 2.4 2600 0.95 - 10 - - - 950 - 4-2 2100 0.25 2200 1.2 3.5 2750 1.1 3.1 20 1000 - 1.5 1050 2.0 4.3 2450 0.55 2200 2.1 3.3 2800 1.6 - 40 I150 0.4 1.4 1100 0.9 4-4 2500 0.75 2450 2-8 - 2850 2.4 - 80 1250 0.5 1.9 1600 0.13 7.0 2400 1.7 2650 3-1 - 2800 3.0 - TABLE 2.-GROWTH OF THE PARTICLES OF SERUM ALBUMIN c = 2.9 % ; pH = 4.4 ; without buffer T("C) t (min) 2 2.5 5 6 10 20 35 - 800 - 1050 1250 - 2400 77 - 0.5 - 1.8 1.4 - 0.7 100 If 1750 - - - 2450 2400 - n 0.7 - - - 0.8 2.7 - 80 I f ( @ 160 1200 0.5 3.0 21 50 tom1 5.4 2450 3.0 3250 3-8 - 3050 4.7 - 40 - - 3350 0.55 We have described in detail the effect of salts in preceding papers.7.8 The variation of I f , n and 2 with the temperature of heating (all other conditions being the same) seems irregular if this is determined after a given time of treatment (see table 1) ; but the vari- ation of Lfis regular as is seen in table 3,82 GLOBULAR-FIBROUS PROTEIN TRANSFORMATION The dilution after heating does not affect the size of the particles. On the contrary, Ef varies considerably with the protein concentration for a given time of heating (all other conditions being the same), the law of variation not being simple, chiefly at the low concentrations, as we see from table 4 ; but Lf increases with q c a s follows from the theory of the end-to-end aggregation (see below).TABLE 3.-vARIATION OF Lf WITH T FOR SERUM ALBUMIN c = 2-4 % ; pH = 4.15 ; M/50 acetate buffer T ("C) 50 57 65 80 90 100 Lf (A) < 300 1200 1600 2400 2600 2800 TABLE 4.---VARIATION OF If WITH C OF SERUM ALBUMIN T = 100°C ; t = 20 min ; pH = 4.15 ; M/50 acetate buffer c (%I 0.3 0.6 1.2 2.4 r, (A) 3 50 1600 2000 2750 The variation of the aggregation as a function of the pH was systematically studied.7, 8 Generally speaking, the size and mainly the rate of polydispersity of the aggregates in- crease with proximity to the isoelectric point.If the pH of the solution is very near the isoelectric point, the aggregation is no longer linear but random, the solution becoming turbid and precipitation frequently occurs. The aggregates built up by heating are not stable indefinitely and the size of the particles spontaneously decreases ; when the aggregate dissociation starts, n decreases rapidly (see table 5). If the spontaneous change of an initially slightly aggregated solution occurs at moderate temperature, some particles dissociate while others aggregate and the system becomes more homogeneous, the particles being larger but less numerous (see table 6). TABLE 5.-sPONTANEOUS VARIATION OF AND 7~ FOR SERUM ALBUMIN c = 2.5 %; pH = 7 ; t = 10 min; T = 80" C ; without buffer after : Oh 15 h 40 h I f (A) 2600 2450 1000 71 1-35 1.0 - The results for compressed solutions will be reported in a forthcoming paper.24 Gener- ally high pressures lead to the formation of globular aggregates, the mean diameter of which varies between a few hundred and a few thousand A.Frequently the compressed solutions show photoelastic effect characterized by low values of the extinction angle which are nearly independent of the velocity gradient, and this is similar to the phenomenon observed with partly hydrolyzed alkaline solutions.34 TABLE 6.-sPONTANEOUS EVOLUTION AT MODERATE TEMPERATURE c=2*4%; pH=7; t = 10min; T = 6 0 " C ; withoutbuffer at 37" C during : Oh 1 h 5 h 71 8.7 1.1 0-1 2 x 104 5.7 12.0 8.5 I f (A) 250 700 1200 It was interesting to find the effect of successive heating and compression.Gener- ally, when a solution is compressed after heating the linear aggregation diminishes, as is seen in table 7. In some cases the decrease of the linear aggregation is accompanied by a globular aggregation as is shown in table 8. Conversely, by heating a previously compressed solution the globular aggregation diminishes and is replaced by a linear aggregation, but to a degree less important than by direct heating, as one can see in table 9. It is recalled that the horse serum albumin solutions frequently show photoelastic effect or dynamic t~rbidity.7.8~25E. BARBU AND M . JOLY 83 TABLE 7.-EFFECT OF PRESSURE A m R HEATING 2.4 % serum albumin heated for 10 rnin at To C and then for 1 h at 37" C under pressure p P H T ("C) P (atm> I/(& 7l 2 x 104 7-0 60 1 - - 1000 7.1 80 1 - - 1000 7.3 65 1 7.7 60 1 - I 1000 - 65 1 - - 1000 - - 1000 < 700 750 400 250 850 350 600 150 750 600 1.1 2-3 5.8 10.0 - - 2.3 16.0 1.8 3.0 12.0 4.8 8.3 1.0 24-0 4.7 6.3 3.3 4-5 - TABLE EFFECT OF PRESSURE AFTER HEATING 2.4 % serum albumin heated for 10 rnin at To C and then for 1 h at 37" C under pressure p (in M/50 acetate buffer for the sample at pH 4.4) PH T ("C) JZ ;? x 103 2 (A) - 2.5 - 600 - 750 - 750 1.7 - - 350 0.9 - - 450 BOVINE SERUM ALBUMIN.-The aggregation ability of the bovine serum albumin is much weaker than that of horse albumin as shown in table 10.Dynamic turbidity appears frequently and greatly increases the polydispersity.Near the isoelectric point the side- by-side aggregatior: grows, inducing a net decrease of 3 (e.g. for a 2.4 % solution at pH 4.4, heated I0 min at 80" C, r f = 900& n = 5.3, 2 = 4.2 X 10-4). The addition of electrolytes during the thermal treatment greatly increases the aggregation (e.g. at pH 4-2 in acetate buffer M/50, with a 2.4 % solution after 10 min heating at 80" C, 6.9, the heated solutions of bovine serum albumin show photoelastic effect. Thus, a solution at 2-4 % and pH 6.9 heated 10 rnin at 80" C does not show streaming birefringence if M/IO NaCI is added, which proves that (in the limits of the sensitivity of the method) it does not con- tain aggregates larger than 300A. On the contrary, when salt is not added for the measurement, the same solution shows streaming birefringence (see table 1 l), from which we deduce the equivalent values 25 [ I f ] = 4800 A, [n] = 6.7 and [A] = 2.8 x 10-5.= 2700 A, 'ii : 4.2, A = 3.9 x 10-3). For the pH values, < 3.9 or TABLE EFFECT OF HEATING AFTER COMPRESSION 2.4 % serum albumin at pH 7.3 compressed for 1 h at 37" C underp atm or heated for 10 rnin at 65" C P T V C ) +(A) A x 103 d(A) 8 50 l h 8000 10 min - 65 8 50 2.4 - 65 900 0.35 - - - - - - - aooo - r l h I+ 10min - Globular aggregation is induced by high pressure with greater difficulty than is found with horse serum albumin. Also aggregation by compression is more difficult than aggregation by heating. For instance, a solution of 1 % bovine serum albumin at pH 7-15 in phosphate buffer 0.= 0.15) does not produce aggregates after 1 h at 37" C under a pressure of 8000 kg/cm2 while the same solution heated 10 rnin at 80" C gives linear aggregates ( r f = 400 A, 3 = 2.6 s 10-2).84 GLOBULAR-FIBROUS PROTEIN TRANSFORMATlON TABLE 1 O.-HEAT AGGREGATION OF BOVINE SERUM ALBUMIN c = 2.4 ; t = 10 rnin ; T = 80" C ; without buffer PH lf& 2 x 103 4.2 850 0.8 4.3 6.3 1600 1.6 3.5 6.6 650 7.9 1.8 TABLE 1 1 .-PHOTOELASTIC EFFECT OF BOVINE SERUM ALBUMIN SOLUTIONS g (sec)-1 328 638 1195 2585 5500 X O 13.9 13.6 13.3 12.7 12.2 (ne-llO) x 107 1.0 1.2 1.2 1.6 1.8 iz ACTIN FROM RABBIT MUSCLE.-In its globular form (G-actin), the freshly prepared actin does not show streaming birefringence. It in fact comprises small globular par- ticles 35,36 with a molecular weight 19 of about 70,000.By storage at low temperature (near 0" C) linear aggregation (with high polydispersity) appears spontaneously in the actin solutions. For instance, a 0.017 % solution at pH 6.6 after a 2-days' storage contains aggregates in which If = 3950A, T = 6-6 and 2 = 9 x 10-3 ; a 0.034 % solu- tion at pH 6.6 gives after 10 days - 6150 A, T = 8.2 and A = 3.6 x 10-3. This spon- taneous aggregation is diminished by heating (e.g. a 0.068 % solution at pH 6.95, heated 10 rnin at 100" C after 3 days' storage, gives If= 2950A, T = 6.1 and ;? = 4.5 x 10-3). Likewise the aggregates are partly destroyed by high pressure (e.g. the same solution as above compressed 1 h at 37" C under 8000 atm gives lf.=2200 A, T-6 and ;?=56x 10-3).The modes of transformation G-actin --f F-actin have been studied by many authors 33-42 and the formation, probably in two steps,40,43 of very long filaments seems now well established by electron microscopy, and is in agreement with our streaming birefringence data. Thus a freshly prepared solution at 0.017 % of G-actin at pH 6.85 gives on addi- tion of M/10 KCl (after 30 min at room temperature) very polydisperse linear aggregates of F-actin (b= 113OOA, T = 1 1 , 2 = 4.7 x 10-2). With the same solution one finds by viscometry an axial ratio equal to 64, leading to a mean diameter of 170A (electron micrographs give 100-200 A). The decrease of the rate of the transformation G-actin --f F-actin after storage in an ice-box was reported by Straub.44 We found that the number, as well as the length, of the aggregates diminishes ; e.g.for a 0.017 % solution at pH 6.3 after 4 days' storage the aggregates obtained by adding M/10 KCl (after 24 h at room temperature) correspond to If= 9250b1, T = 8.9 and 2 = 8.1 x 10-3). The effect of KC1 concentration is summarized in table 12. It was interesting to discover the effect of heat or pressure on the G-actin -+ F-actin transformation induced by KC1. Generally both factors reduce the length of the aggregates; e.g. the most TABLE 12.-THE G-ACTIN -+ F-ACTIN TRANSFORMATION c = 0.034 % actin stored for 8 days ; pH 6.5 ; 24 h of contact with KCl - [KCU 'f (A) 7c A X 103 M/100 8750 8.7 3.3 M/10 9250 8-9 8.6 M/2 3950 6.5 5-0 frequent length I f = 9250 A of the aggregates obtained by adding M/10 KCl to a 0.034 % solution at pH 6.5 after 4 days' storage is reduced to 5800 8, by 10 min heating at 65" C, to 2050A by 10 min heating at 100" C and to 4050 8, by compressing for 1 h at 37" C under 8000 atm.The effect is similar if KCl is added after heating or compression ; e.g. the addition of M/lO KC1 (24 h of contact) to 0.034 % solutions of G-actin at pH 6.8 (after 10 days' storage) untreated, heated for 10 rnin at 60" C and compressed for 1 h at 37" C under 8000 atm, gives rise (at practically the same rate of transformation) to elongated aggregates of Factin, the most frequent lengths of which are respectively 8900 A, 7000 8, and 5800A. HEN EGG ALBUMIN.-The heat denaturation of egg albumin has been exhaustively studied by Foster and his co-workers,'L 12 who have shown the importance of the heat aggregation at different pH values and the existence of a minimum particle length (about 300 8, for c -- 0.39 %, t -- 15 min, T = 100" C) near pH 2-2.5.These authors suggest an unfolding of the egg albumin molecules to explain the behaviour of the solutions.E . BARBU AND M. JOLY 8 5 Our experimental conditions are a little different from their conditions : our egg albumin was prepared by their method but was neither lyophilized nor diluted with glycerol. We have only studied solutions which have been heated for 10 min at 80' C and at pH values between 2 and 4. The behaviour of the egg albumin differs largely from that of serum albumin. A 2.4 % solution (heated for 10 min at 80" C) of serum albumin was clear even at pH 4.5 (0.3 below the isoelectric point), while the egg albumin solution was very turbid from pH 4 (0.5 below the isoelectric point) and shows dynamic turbidity.The serum albumin did not give rise to aggregates detectable by streaming birefringence by heating below pH 3.5, while the egg albumin was greatly aggregated at very low pH value. 2 was always much weaker for egg albumin than for serum albumin; its maximum was at pH 3.95 (2 = 1.1 x 10-3) for particles of I f = 2100 8, and n 1 1.8. At pH 4 the par- ticles were a little longer ( I f 2 2500 A) but a little less dissymmetrical (2 = 9 x 10-4). Between pH 3.8 and 2, the relatively low values of the birefringence n,-no associated with the low values of the extinction angle X suggest the behaviour of large rather sym- metrical particles, the size of which reaches a minimum at pH 2-7.Table 13 gives a few values obtained by the extrapolation method.24 TABLE 13 .-HEAT AGGREGATION OF EGG ALBUMIN c = 1.6 %; f = 10min; T = 80°C PH 3.8 3-7 3.5 3.2 2.8 2.5 2.0 d (4 1050 750 700 650 600 600 800 - [A] x 104 0.9 2.8 0.8 1.7 2.1 2.0 1.4 By adding NaCl to solutions at pH 2.5 after heating, the particle size did not change as long as c[NaCl] < M/20. For M/10 NaCl the size decreased slightly (li = 550A) while the dissymmetry increased ([A] = 4.9 x 10-4 instead of 2 x 10-4). In M/6 NaCl large globular aggregates grow (d N 850A, [A] = 1.3 x 10-4); in M/5 NaCl the solu- tions became extremely turbid, and later the precipitation occurred. The dilution of these solutions gave rise to a large increase in the size of the aggregates : e.g.d was 850 8, instead of 600 8, by dilution to 0-8 % in H20 and nwas 900 8, instead of 550 8, by dilution to 0.8 % in M/10 NaCl, but without any striking change in [A]. The particles obtained by heating a more dilute solution (c = 0.4 %) at pH 2.5, with or without the addition of M/20 NaCl after the heat treatment, were practically of the same length as those produced in a 1.6 % solution, but were much more dissymmetrical. Therefore the increase of concentration greatly promotes the side-by-side aggregation. At pH 10 a solution at 1.6 % egg albumin showed without heating a very marked photoelastic effect (&] = 5350& [n] = 5.5, [A] = 9 x 10-5) of the same type as shown by alkaline serum albumin solutions.34 HORSE y-PSEUDO-GLOBULIN.-The behaviour of the globulins is very different from that of the albumins.We have only reported in the present paper results from rela- tively concentrated solutions on the alkaline side of the isoelectric point. One obtains exceptional and unreproducible solutions which do not contain aggregates and do not show streaming birefringence. Generally the concentrated globulin solutions contain, without any treatment, large globular aggregates of small equivalent anisotropy and probably of very weak compact- ness. With the addition of electrolytes (NaC1) the size of these aggregates decreased slightly at low salt concentrations, but increased at high concentrations as it can be seen on table 14. The dilution of the solutions gives rise to a striking increase of the relative number of aggregates, as shown by the variation of [ I ] ; e.g.for a 1.1 % solution at pH 6.1 in M/10 NaC1, [A] = 3.2 x 10-4. The main effect of heat is to augment the dissymmetry of the particles and, in fact, to tiansform the globular into elongated aggregates. Thus, for 3.3 % solutions at pH 8.3 heated for 10 min, the shape and the size of the particles remain constant until 50" C with 2 = 850 A and [A] = 6.6 x 10-5 ; then the aggregates lengthen progressively and at SO" C (always for 10 min heating) one observes elongated particles with I ) = 1900 A, T = 5.9 and 3 = 2.5 x 10-4.86 GLOBULAR-FIBROUS PROTEIN TRANSFORMATION TABLE 14.-THE EFFECT OF ELECTROLYTES ON THE SPONTANEOUS AGGREGATION OF THE y-PSEUDO-GLOBULIN c = 3.3 % PH [NaCl] 2 (A) [s] X 105 6.1 0 900 6.8 - M/5 700 8.2 - M12.5 900 5-7 7.9 0 800 6.4 - M/5 900 4.3 The solutions heated above 80" C show dynamic turbidity; the flow induces the formation of large aggregates of low dissymmetry, of which the diameter is about 1000 A.Likewise a 3.3 % solution at pH 7-9 containing initially globular aggregates with 2 = 800 A and [z] = 6-4 x 10-5, shows after 10 min heating at 80" C elongated particles with & = 1450A, T = 5.5 and ;? = 2-5 x 10-4. By adding a small amount of electrolyte after heating, the length of the most frequent particles decreases slightly but their number increases (e.g. for the preceding solution, on adding M/10 NaC1, &varies from 1450A to 1350A and 3 from 2.5 x 10-4 to 3.9 x 10-4). On the contrary, in the presence of a large amount of electrolyte the solutions become turbid and sometimes precipitate (e.g.in M/5 NaC! the preceding solution is completely opaque). THEORETICAL TREATMENT OF THE AGGREGATION PROCESS The first theory of the aggregation of colloidal particles was proposed by Smoluchowslu 45 for the globular aggregation of spherical particles, the process being considered as a diffusion process. By assuming the probability E for two particles to remain aggregated after collision to be independent of the particle size, and by setting nk the number per cm3 of particles comprising R molecules, Rk and Dk being the action radius and the diffusion constant of the particles, the variation of nk as a function of time is given by i = l with h defined by 2 ini = no, no being the initial number of molecules per cm3.Smoluchowski assumes in a first approximation (Q + Dk)(Ri + Rk) = 2 DR, D and R being the diffusion constant and the diameter of the initial molecules. By setting i- 1 p = 47rDRE and i - A ni = N* i - 1 the total number of particles per cm3 at the time t , we find and ylk rises to a maximum at the timeE . BARBU AND M . JOLY 87 With numerous assumptions Kleczkowski 10 showed that Smohchowski’s theory fits for the aggregation (studied by salt precipitation) of diluted (0.25 - 1 %) solutions of human serum albumin (PH 6-8, phosphate buffer M/15) heated for 10 min at 83” C. La Mer46 extended the collision theory to charged particles and ions, and Collins,47, 48 by a more rigorous treatment, has shown that the calculation of Smoluchowski is valid only for low values of E and for small enough particles.Donnet and Sadron 49 have checked experimentally the validity of the Stokes relationship and of the equations of Brownian motion and diffusion for particles as small as 100 A, which justifies the application of the diffusion theory to the ag- gregation process. But in spite of these improvements, the theory of Smoluchowski does not suffice to explain all the experimental features of the protein aggregation. Particularly this theory cannot explain the spontaneous dissociation of the aggre- gates, the limited values of the particle sizes, the relative homogeneity of the solu- tions and particularly the linear aggregation in the globular-fibrous protein trans- formation.By plotting the equipotential curves of the interaction between one charged particle and one aggregate made of two such particles, Rees 50 has shown that the potential barrier to be surmounted by the single particle in order to collide with the aggregate falls to a minimum at the ends of the double particle. Therefore, in explaining the linear aggregation, we shall follow a treatment analogous to the polymerization kinetics developed by Tobolsky.51 With the same notations as above and by setting Aim equal to the addition fre- quency of the particles containing I and m molecules, and Bq to the dissociation frequency of the particles made of q protein molecules, we can write He found k < 7 and E N 10-8. The dissociation of a linear aggregate corresponds to the breaking of the temporary bonds between two elementary joined particles; one can therefore put as a first approximation Bq = B, independent of the particle size.By generalization of Frenkel’s relationship 52 we can write N being Avogadro’s number, A4 the molecular weight of the protein, k Boltzmann’s constant, and W the mean value of the interaction energy between one molecule and all its neighbours in the same aggregate. We have now to calculate Aim. Consider two rod-shaped particles AB and CD (fig. 5), whose centres are 0 and P. For linear aggregation of these particles it is necessary that : (i) B and C collide, i.e. they diffuse towards each other and they surmount the intervening potential barrier; (ii) the angle AB-CD is smaller than a given angle a ; (iii) the protein molecules in B and C are activated at the time of collision so that the two particles form one single aggregate.If AFis the free energy of this activation, corresponding probably to the freeing of one or several functional groups initially coupled inside the molecule, the prob- ability that the molecules in B and C will be activated at collision time is wa = exp (- AF/kT). We shall call this activation “ predenaturation ”. If U is the height of the potential barrier between the particles, only kinetic energies greater than U allow the particles to collide. Following Rees,sO U is a minimum at the ends of the particles and, according to the bead structure of the rodlike particles, we can suppose that this minimum value is independent of the88 GLOBULAR-FIBROUS PROTEIN TRANSFORMATION particle length.By assuming the validity of the Maxwell-Boltzmann law for the whole of the aggregates, the probability for a particle to surmount the potential barrier of height U is Wb = 1 - - (kT)-”’[ exp (- z/kT) 42 dz = F(U). 2 U d; 0 Table 15 gives a few values of this function. TABLE 15.-vALUES OF Wb = F(U) UIkT wb UIkT *b UIkT wb 0.0 1.0 1 0.57275 5 0.0 195 0.01 0.99925 2 0.2625 7 0.003 15 0.1 0097765 3 0.1 125 10 0~00022 0.5 0.80135 4 0.04655 20 0*000004 Without the potential barrier, the collision frequency of particles having diffusion constants Di and Dm and an action radius RIm is (Dl + Dm)Rlm. For a rod-shaped particle made of 1 spherical molecules of diameter R aggregated end-to-end, the mean value of DI is - 21 (7 being the solvent viscosity).As the efficient 3 q l R ’ A I FIG. 5.-End-to-end aggregation of two rod-shaped particles. collisions are only by the ends, and as the angle a is assumed to be small, and therefore kT(1 + m) (- In 21 + !!!?), 6n7 1 m (01 + Dm)RIm which we can replace in practice by kT(1 + m)2 27~7Im ’ Among all the particles which have diffused towards 0, only those having their centres at the portion of the spherical surface PP’ (see fig. 5) can collide by their end with B, and the corresponding probability is f = a2/(1 + m)2. Thus we can write For greater simplification, according to the dissymmetry of the potential energy around each particle, we can introduce an effect of mutual orientation of theE . BARBU AND M. JOLY 89 particles in contact, this effect being stronger the longer the particles.It is therefore not necessary to keep cc constant and it seems sufficient for a linear aggregation that cc is small as m and 1 are small. So we can, for instance, set ct2 = (ml, 5 being a numerical factor about 10-2. At least, independent of I and m, and the general equation for end-to-end aggregation is For globular aggregation, we can assume with Smoluchowski that in a first approxi- mation and thus i.e. the same relationship as above, but with 4 = 4/3. It is only by a rather rough approximation we can assume B to be independent of the size of a globular aggre- gate and use the same equation for the globular and the linear aggregation. By setting the integration of the general equation leads 51 to nk = noPk-l (1 - p)2 1 with p = - noAt 2 A + (A2 - l)* coth - (A2 - 1)i If we consider the average number x of protein molecules per particle, 1 knk 1 1 - -- 1 x=- 5 nk 1 - p p ) 1 we can in a first approximation substitute it for the molecule number of each particle of the most frequent size and write x == l f / R for the linear and x = __ for the globular aggregates. We find then 4 2 ~ 3 B 2B , B2 ~ Q A 2B B2 + as (a T n$A2 --)* coth - 2 (--- + n$A2 -)*t B 2B B2 )* coth + -)'t ' x = ____ noA 2B B2 n$A2 and as the aggregation is measureable only for B/noA < 1, we have to consider in mactice90 GLOBULAR-FIBROUS PROTEIN TRANSFORMATION in good agreement with the experimental If against t curves, and the value reached at infinite time x, N dg, related to Lf by xoo -= LfIR.than in5--is given by t, = By substituting A and B for x, in their expression as a function of W, U, AF, molecular weight M and concentration c of the protein, one finds (R being the gas constant and k Boltzmann's constant) : - The time ts at which the ratio x/x, rises to a given value s-sufficiently larger J' noAB arctan s. and x, = w d / c exp (W - AF/2kT)F* ( U ) (x) RT f , with w N 0.46 and T N 5.8 for the linear aggregation and w N 5.2 and T N 0.49 for the globular aggregation. Therefore RTctsF(U) arctan s exp (AF/kT) = #--- Y M x, and with C$ N 0.38 and $ N 0.078 for the linear aggregation, and 4 N 0.4 and 4 N 0.9 for the globular aggregation. By applying these results to the experimental data obtained with a solution of 2.4 % horse serum albumin at pH 4.15 heated in M/50 acetate buffer at 65" C or 80" C (see table l), we find W N 16 kT and AF < 10kT.DISCUSSION AND CONCLUSION.-we think that a more systematic study by this method of the cohesion energy W of the aggregates will give a lot of information on the nature of the bonds in the large protein particles. We see that the free energy of the activation, which we called predenaturation, is considerably smaller (3 or 4 times) than that found in the literature 53-56 for the free energy of the protein denaturation. We may inquire if in the ordinary denaturation experiments some secondary process (such as the aggregation) does not increase the apparent free energy. Indeed, if instead of writing that the probability for a particle to collide and to be aggregated as A = a F( U) exp (- AF/kT) (a being a constant), we consider a free energy of activation AF' for the whole process of aggregation by setting A = exp (- AF'lkT), we find, in the same case as before, AF' of the order of 40 kT, in good enough agreement with the values generally given for the free energy of de- naturation.Thus the real protein denaturation (an intramolecular phenomenon) is perhaps only the activation process which we have described above as pre- denaturation, and all the other features generally included in the word denaturation are only secondary effects, involved in all the experiments on the denaturation, but related to the presence of other molecules. The chemical aspect of the protein denaturation was extensively studied,5g9 59 but the physical mechanism of this process was less investigated.In agreement with many authors 53-55,57 we have suggested 7 9 8 that the protein denaturation proceeds by the freeing of groups initially coupled inside the molecule and this allows the formation (in particular places of each molecule) of weak intermolecular bonds like Van der WaaIs forces, hydrogen or salt bridges. This freeing of functional groups certainly corresponds to a profound perturbation of the molecularE . BARBU AND M. JOLY 91 structure but does not necessarily involve considerable unfolding of polypeptide chains. Several authors suggest even for moderate denaturation, extensive unfolding but, it would seem, without very convincing evidence. The existence of particles practically all of the same size does not involve, as suggested by Foster,lz that each comprises only one kind of molecule ; we have seen from the above calculations that the aggregation of molecules does not lead to a random size distribution and that the heated solutions can show relatively low polydispersity.In addition, one has to be cautious in the interpretation of the streaming birefringence data for very acid or very alkaline protein solutions, because they show photoelastic effect depending upon the structure of the solution as a whole instead of upon the particle size. We have noted elsewhere 7,s that many features are in contradiction with a large unfolding of the chains, e.g. variation of Z j and with the pH, tensile strength of the elongated particles, partly reversible behaviour of the denaturation (and even in certain cases, conservation of the immunological properties).Moreoever if the rod-shaped particles were made of one single unfolded chain we could expect an intrinsic anisotropy considerably greater than that observed. If, instead of having a relatively rigid structure, the molecules of denatured protein were randomly folded chains, even for pH values rather far from the iso- electric point (but without hydrolysis), its mean configuration would be quasi- globular .6*-63 To explain the high values of the viscosity and Maxwell effect, it would be neces- sary to consider them as due to the flow ; but to reach the observed particle length, the corresponding velocity gradient should rise to values considerably higher than used in the measurements.64 On the contrary, one frequently finds by examination of the X against g and A against g curves that the particles obtained by moderate heating of dilute solutions or by compression after heating, show a dissymmetry increasing with the velocity gradient and probably comprise more or less folded end-to-end aggregates, which the flow progressively unrolls.It is therefore possible when fibres are ob- tained by extruding and precipitation of concentrated protein solutions at high temperature in alkaline medium or with detergents,l-4 that stretching, unfolding and side-by-side union of such aggregates occur and these comprise end-to-end aggregated globular molecules, instead of the unfolding of the chains of each protein molecule.X-ray diagrams of keratin would then be due to the general orientation of the molecules and not to the unfolding of their polypeptide chains. One has probably the same frame in the natural protein fibres as is observed in electron micrographs .65-68 The linear aggregation of the globular proteins seems to play an important part in the formation of the protein natural or artificial fibres. It is still difficult to say whether the aggregation process (concomitant or not with the denaturation) occurs in one or two steps ; it seems to be a two-step mechanism for certain concentrated proteins.16-403 43 To emphasize the biological importance of the globular-fibrous protein trans- formation is redundant. As a general conclusion perhaps we may quote a sentence of Szent-Gyorgyi : 44 ‘‘ It may be asked whether the globular form is not the basic form of protein and whether all fibrous proteins are not built up of globular par- ticles.” Our sincere thanks are due to Prof.Schapira, who provided us wth the actin sample, to Miss Croissant who carried out the electron microscopy work, to Mr. Wielgosz for his aid with the ultracentrifuge, to Mr. Basset, who helped us greatly with the high pressure experiments, and to Mr. Bjormholm and Mr. Rebeyrotte for their experimental assistance in the viscometric and electrophoretic measurements.92 GLOBULAR-FIBROUS PROTEIN TRANSFORMATION Palmer and Galvin, J . Amer. Chem. SOC., 1943, 65, 21 87. Senti, Eddy and Nutting, J . Amer. Chem. Soc., 1943, 65, 2473. 2 Putnani and Neurath, J. Biol. Chern., 1943, 150, 263.4 Senti, Copley and Nutting, J. Physic. Chem., 1945, 49, 192. 5 Joly, Bull. SOC. Chim. biol., 1948, 30, 398 ; 1949, 31, 105. 6 Bourgoin and Joly, J . Chim. Phys. (in press). 7 Joly and Barbu, Bull. Soc. Chim. biof., 1949, 31, 1642 ; 1950, 32, 116, 908. 8 Joly and Barbu, J. Chim. Phys., 1951, 48, 636. 9 Joly, Kolloid-Z., 1949, 115, 83. 10 Kleczkowski, Biochem. J., 1949, 44, 573. 11 Foster and Samsa, Science, 1950, 112,473. 12 Foster and Samsa, J . Amer. Chem. Soc., 1951, 73, 3187, 3190, 5388. 13 Jaggi and Waugh, Fed. Proc., 1950, 9, 66. 14 Waugh, Amer. J. Physiof., 1941, 133, 484. 15 Waugh, J. Amer. Chem. Soc., 1946, 68, 247. 16 Farrant and Mercer, Biochim. Biophys. Acta, 1952, 8, 355. 17 Waugh and Livingstone, J. Physic. Chem., 1951, 55, 1206.18 Sorensen and Hoyrup, C. Lab. Carlsberg, 1915-17, 12, 12. 19 Straub, Inst. Med. Chem. Univ. Szeged, 1942, 2, 3 ; 1943, 3, 23. 20 Frey-Wyssling and Weber, Helv. chim. Acta, 1941, 24, 278. 21 Scheraga, Edsall and Gadd, J. Chem. Physics, 1951, 19, 1101. Z2 Joly, Trans. Faraday Sac., 1952, 48, 279. 23 Joly, J. Chim. Phys., 1951, 48, 536. 24 Barbu, Basset and Joly, to be published. 25 Joly and Barbu, to be published. 26 Macheboeuf, Rebeyrotte and Brunerie, Bull. Soc. Chim. biol., 1951,33, 1953. 27 Wielgosz, to be published. 28Basset and Macheboeuf, Compt. rend., 1932, 195, 1931 ; 1933, 196, 1540; 1933, 197, 796 ; 1935, 200, 1072. 29 Guth, Kolloid-Z., 1936, 74, 147 ; 1936, 75, 5. 30 Guth, Gold and Simha, Kolloid-Z., 1936, 74, 266. 31 Dervichian, Fournet and Guinier, Bull. Soc. Chim. biof., 1949, 31, 101. 32 Barbu, Macheboeuf and Rebeyrotte, to be published. 33 Joly, J. Physique, 1951, 12, 900. 34 Barbu and Joly, Bull. Soc. Chim. biol., 1950, 32, 123. 35 Jakus and Hall, J. Biol. Chem., 1947, 167, 705. 36 Astbury, Perry, Reed and Spark, Biochem. Biophys. Acta, 1947, 1, 379. 37 Feuer, Molnar, Pettko and Straub, Hung. Acta Physiol., 1948, 1, 150. 38 Hall, Jakus and Schmitt, Biol. Bull., 1946, 90, 32. 39 Snellman and Erdos, Biochim. Biophys. Acta, 1948, 2, 660. 40 Rozsa, Szent-Gyorgyi and Wyckoff, Biochim. Biophys. Acta, 1949, 3, 561. 41 Dubuisson, 1st Int. Congr. Biochem. (Cambridge, 1949), p. 132. 42 Dubuisson, Biochim. Biophys. Acta, 1950, 5, 426. 43 Szent-Gyorgyi, Faraday Soc. Discussions, 1951 , 11, 199. 44 Szent-Gyorgyi, Acta Physiol. Scand., 1945, 9, suppl. 25, 33. 45 Smoluchowski, Z. physik. Chenz., 1918, 92, 129. 46 Umberger and La Mer, J. Amer. Chem. Soc., 1945, 67, 1099. 47 Collins and Kimball, J. Colloid Sci., 1949, 4, 425. 48 Collins, J. Colloid Sci., 1950, 5, 499. 49 Donnet and Sadron, Compt. rend., 1952, 234, 69. 50 Rees, J. Physic. Chem., 1951, 55, 1340. 51 Blatz and Tobolsky, J. Physic. Chem., 1945, 49, 77. 52 Frenkel, Kinetic Theory of Liquids, Oxford, 1946, p. 2. 53 Eyring and Stearn, Chem. Rev., 1939, 24,253. 54 Bull, Adv. Enzym., 1941, 1, 1. 55 Stearn, Adv. Enzym., 1949, 9, 25. 56 Sizer, Adv. Enzym., 1943, 3, 35. 57 Mirsky and Pauling, Proc. Nut. Acad. Sci., 1936, 22, 439. 58 Neurath, Greenstein, Putnam and Erikson, Chem. Rev., 1944, 34, 157. 59Anson, Adv. Protein Chem., 1945, 2, 361. 60 Sadron, J. Chim. Physique, 1946, 43, 142 and 1947,44, 22. 61 Guinand, Boyer, Kawenoki, Dobry and Tonnelat, Compt. rend., 1949,229, 143.E. BARBU AND M. JOLY 62 Kuhn, Bull. SOC. Chim. Belg., 1948, 57, 421. 63 Hermans, Bull. SOC. Chim. Belg., 1948, 57, 164. 64 Frenkel, Acta Physicochim., 1944, 19, 51. 65 Astbury, Brit. J. Derm. Syph., 1950, 62, 1. 66 Noda and Wyckoff, Biochem. Biophys. Acta, 1951, 7, 494. 67 Pratt and Wyckoff, Biochem. Biophys. Acta, 1950, 5, 166. 68 Weibull, Faraday Soc. Discussions, 1951, 11, 195. 93