ARNE TISELIUS 33 POLARIZATION OF THE FLUORESCENCE OF LABELLED PROTEIN MOLECULES BY G. WEBER Sir William Dunn Institute of Biochemistry, Cambridge Received 8th May, 1952 The recent development of a method of determination of the average rotational relaxation time of protein molecules in solution by measurements of the polarization of the fluorescence of protein conjugates is reviewed. A method for the purification of the conjugates by ion exchange is described, together with the use of paper chromatography to distinguish between reversible adsorption and chemical combination of small mole- cules to proteins. The reversible dissociation of bovine serum albumin in acid first detected by observations of the changes in the polarization of the fluorescence with pH34 POLARIZATION OF FLUORESCENCE has been confirmed by measurements of sedimentation and diffusion at pH -1-8.A study of the effect of neutral salts on the dissociated protein shows that reversal of the dissoci- ation depends on the anion and increases in the order C1 < Br < NO3 < SCN < p - toluensulphonate. On this basis it is concluded that the electrostatic repulsion between the subunits is indispensable for the acid dissociation. The independence of the alkaline dissociation of bovine serum albumin from the ionic composition and strength of the medium is discussed. THEORETICAL.-The studies of Perrin 1, 2 have shown that the polarization of the fluorescence emitted by molecules in solution depends only on the relaxation time of the rotation, the lifetime of the excited state of the fluorescent molecules and the relative orientation of the virtual oscillators of absorption and emission of light.For spherical molecules a particularly simple relation obtains between these quantities : where Here p is the partial linear polarization of the fluorescent light emitted at right- angles to the direction of propogztion of the exciting light. The minus signs in l / p correspond to excitation by polarized light vibrating at right-angles to the directions of excitation and observation, the plus signs to ex- citation by natural light. T is the lifetime of the excited state of the fluorescence and p the rotational relaxation time ; po is the polarization obtaining when 3r/p -+ 0, and as 7 is a quantity of the order of 10-8 sec, po does not differ appreciably from the polarization observed at room temperature in glycerol solution for most fluorescent molecules.For ellipsoidal molecules the relations between these quantities are consider- ably more complex since the general equations developed by Perrin2 require a knowledge of the angles that the directions of the oscillators make with the axis of revolution of the ellipsoid. However, as I have shown elsewhere 3 a consider- able simplification is possible if the fluorescent units carrying the oscillators are attached to a much larger molecule with random orientation, that is, if all the angles between the oscillators and the axis of revolution are equally probable. In such cases eqn. (1) becomes 8 and l/po 'f 7 1 1 5 , 9T2 1 -i- - - < - T - PO 2 I n 2 n l ! 8 p2 in which po = 377 Y,/RT is the relaxation time of the rotation of a sphere of volume Ve equal to that of the ellipsoid.If p1 is the rotational relaxation time about the axis of revolution and p 2 the rotational relaxation time about a direction normal to it, then When the experimentally accessible values of T / ~ O are sufficiently small the last equation becomes n1 = pdpo; n2 = P2/PO.FIG. 1 . acetate 1 2 3 -Fluorescence of ascending paper chromatogram. Watman no. 1 ethanol + pH 4.8 ; fluorescence excited by Hg arc with Wood's filter ; camera filter : 2 cm 1, bovine serum albumin conjugate before filtering through resin. 2, same after filtering through resin. 3, 1-dimethylamino-naphthalene-5 sulphonate. layer of saturated NaN02 solution.[To face page 35G . WEBER 35 Therefore in general if a straight line is obtained on plotting l/p against absolute temperature/viscosity, the relaxation time of the rotation calculated from the slope and intercept is ph the harmonic mean of the principal relaxation times : 2 ph = ~ 1 1 ’ - + - P1 P2 (4) When the depolarization is due to more than one relaxation lime the plot of l/p against T/q may yield a curve concave towards the latter axis. If the solution is monodisperse it may safely be concluded that the emitting units are elongated molecules since the relaxation times of the rotation of an oblate ellipsoid never differ by more than 10 % from each other 4 and would in their effects be indis- tinguishable from a sphere even under the most favourable conditions.If the plot shows curvature the elongation may be determined by comparison with the theo- retical curves given by eqn. (2). If the axial ratio of the ellipsoid is greater than 10, po/p1 can be considered effectively 0 and po/p2 = 3. For such particles therefore the dynamic volume may be calculated directly, and for proteins an approximate molecular weight may be obtained if due allowance for hydration is made. A curvature convex towards the T/q axis is only possible if T lengthens as the temperature increases or ifph is not alinear function of q/T. The former is extremely improbable on theoretical grounds. It has been suggested by Perrin5 that the latter obtains in the Brownian motion of solute molecules differing little in size from the molecules of a solvent of low viscosity.For macromolecules this possi- bility may be discarded and a curvature convex towards the T/T axis is only possible if the total number of rotational degrees of freedom increases with the temperature. Therefore either new modes of intramolecular rotation appear that were originally frozen at lower temperatures or dissociation into fragments takes place. In most cases it is possible to distinguish between these alternatives.4 EXPERIMENTAL To study the rotational diffusion of proteins by the depolarization of the fluorescence they must be converted into stable fluorescent conjugates. With this object the protein is reacted with 1 to 2 % of its weight of 1-dimethylamino-naphthalene-5-sulphonyl chloride.6 Part of the acid chloride combines with the protein, mainly or only with the primary amino groups yielding substituted sulphonamides, and the rest hydrolyzes giving the strongly fluorescent sulphonate.The separation of the conjugate from the latter may be accomplished by dialysis against salt solutions or by ethanol precipitation. Lately I have used a more rapid and very effective method : the reaction mixture is dialyzed against M/15 phosphate buffer at pH 7 for 24 h and then filtered slowly through a column of a basic exchange resin (Dowex 2, mesh 200). The strongly acidic sulphonate is retained while the protein conjugate passes through and is quantitatively recovered. Spectro- photometric measurements indicate that 50 to 60 % of the acid chloride reacts with the protein.6 The presence of free sulphonate as opposed to chemically bound sulphonamide may be detected by paper chromatography in a medium where the protein is insoluble while the sulphonate has some solubility.A mixture of 55 parts of ethanol and 45 parts of 0-2 M acetate buffer of pH 4-8 is very convenient. Here the Rf of the proteins exam- ined was effectively 0 and the Rf of the sulphonate nearly 0.9. Fig. 1 is the photograph of such a chromatogram. It shows the elimination of the sulphonate by the ion exchange resin and also the clear separation between the fluorescence of the conjugate and that of the sulphonate. There is little tailing of the fast spot in spite of the fact that serum albumin in solution adsorbs the sulphonate strongly.7 When 1-dimethylamino-naphthal- ene-5 sulphonate or any of several acidic dyes was dissolved together with unlabelled bovine serum albumin and then subjected to chromatography, complete separation of the dye occurred, all the fluorescence and/or the colour appearing in the moving spot.This method is clearly a general one by which reversible adsorption may be distinguished from the actual chemical binding of small molecules to proteins. It allows a study of the stability of the conjugates under a variety of conditions. Thus it is found that heating36 POLARIZATION OF FLUORESCENCE at neutral pH at 60" for 1 h results in hydrolysis of less than 1 % of the sulphonamide and that at pH 2 the same treatment may result in up to 5 % hydrolysis. DISCUSSION Two proteins, ovalbumin and bovine serum albumin have been studied in detail.6 On plotting l/p against T/v, straight lines are obtained for solutions in water or dilute electrolyte at neutral pH between 3" and 50".Moreover the extra- polated values of po agree within the precision of the measurements with the polarization observed in 60 % sucrose (w/v) where 3r/ph is effectively 0 (table 1). Therefore eqn. (3) may be used. If the values of ph obtained from the data of TABLE 1 .-FLUORESCENCE POLARIZATION OF PROTEIN CONJUGATES S is the calculated regression coefficient of l/p upon T/q. po (ext.) is the extrapolated value obtained from this regression coefficient. PO observed, is the polarization recorded in 60 % (w/v) sucrose at 4". ph at 20" has been calculated from eqn. (3) with 7 = 1.4 x lo-gsec. ph/pu is the ratio of the observed value of the rotational relaxation time to the value of a sphere of molecular weight (known or assumed) given in the second column. Notice the difference between pa (anhydrous sphere), and pa (hydrated sphere) appearing in eqn.(2). Therefore ph/po depends only on shape, ph/pu on both shape and hydration and has a significance similar to f/fo in the translational diffusion. The figures given for F- and G-actin are from observations of Dr. T.-C. Tsao. protein molecular weight solvent PH S X 105 po(ext) bovine serum 0.06 M phosphate 7.0 4-26 0-258 0.257 1.42 2.22 bovine serum - distilled water 1.8 11.6 0.245 0.244 0.56 - bovine serum - N/10 NaOH 13.0 11.9 0.233 0.230 0.58 - albumin 69,000 buffer albumin + HC1 albumin ovalbumin 45,000 0.06 M phos- 7.0 7.5 0.236 0.236 0.88 2.12 phate buffer ovalbumin 45,000 N/lONaOH 13-0 7.5 0-236 - 0.88 2.12 ovalbumin heat - 0.06 M phos- 7.0 - - 0.253 2-2 to - F-actin 147,000 0.1 M KCl 7.0 2.21 0.193 0.195 3.6 2.66 denatured phate buffer 4.8 0.06 M phosphate 10-4 M ATP G-actin 71,000 lO-3M Na 7.5 4.84 0.202 0.203 1.6 2.44 G-actin (dimer) 143,000 water, salt-free 7.0 2.30 0.193 0.193 3.5 2.66 (monomer) versinate Oncley8 using the dielectric dispersion method are introduced in eqn.(3) it is found that r has the same value in both conjugates, namely, 1-4 x 10-8 sec. Reproducible results are obtained without difficulty in the range of pH 1-5-14. Between these limits no conspicuous changes in the lifetime of the excited state of the fluorescence seem to occur, so that changes in the observed slope in the plot of l / p against T/y may be attributed to real changes in the mean harmonic relaxation time of the protein particles.It is thus possible to show that bovine serum albumin dissociates into subunits outside a stability region which at room temperature extends between pH 3.9 and 9. Both dissociations in acid and in alkali are rever- sible, a t least as regards the relaxation time of the rotation. They differ consider- ably in certain respects. The dissociation in acid is strongly dependent upon the ionic composition and strength of the solvent, the alkaline dissociation is practically independent of it. The lowest relaxation times in acid are observed in proteinG . WEBER 37 solutions dialyzed against distilled water for a period of 100 to 200 h and subse- quently brought to pH 1-8-2.2 by addition of a small amount of HCl.On addition of neutral salts to these solutions the polarization of the fluorescence increases with ionic strength and with most salts tends asymptotically to the polarization observed in neutral solution. Different salts vary in their ability to induce re-aggregation of the acid protein. Table 2 gives the concentrations of electrolyte in which a rotational relaxation time midway between those of the neutral and salt-free acid TABLE 2 Electrolyte concentration at which ph at 20" = 1.02 x 10-7 sec; 0.2 % bovine serum albumin dialyzed for 100 h against distilled water and subsequently adjusted to pH 2. with HCI. concentration electrolyte (equiv. /I.) electrolyte (equiv./ 1 .) KCI 0.18 KBr 0.075 N-CI 0.18 &So4 0.032 LiCl 0.12 KNO3 0-025 MgC12 0.16 KSCN 0.0082 CaC12 0.17 p-KSO3. C6H4. CH3 0.0059 concentration BaCl2 0.21 protein is observed. This characteristic concentration is independent of the cation and dependent on the anion, increasing for the latter in the series, Cl < Br < NO3 < SCN < p . CH3. C&4. sO3. The series is the same as is given by Scatchard and Black 9 for the relative adsorption of anions by iso-ionic serum albumin. The conclusion may be drawn that re- combination of the subunits takes place when the electrostatic repulsion between them is sufficiently decreased by the adsorption of the ion of opposite sign. The acid dissociation of the protein has been confirmed by sedimentation and diffusion measurements.Preliminary observations made by Dr. R. A. Kekwick on bovine serum albumin dissolved in 0-2 M KCl adjusted with HCl to give pH 1-82 showed in the ultracentrifuge only one component with 320 = 2-7 X 10-13. (For the neutral protein, s20 = 4.3 x 10-13.) Measurements of the free diffusion of the protein in the same solvent yielded a mean value for D20 of 6.85 x 10-7. (For the neutral molecule, D20 = 6.1 x 10-7.) The calculated molecular weight is 36,000 with f l f o = 1-43. Observations of the polarization of the fluorescence in the same medium gave (fig. 2) at 20" ph = 7.9 x 10-8. For the quoted molec- ular weight phipa = 2.4, while for the neutral molecule ph/pa = 2.2 and f / f o = 1.34. Therefore neglecting improbable changes in hydration, both translational and rotational diffusion indicate an increase in assymetry on dissociation.This is best explained by the splitting lengthwise of the neutral molecule into two equal or nearly equal parts. Fig. 2 shows that shorter relaxation times are observed in the absence of salt as compared with the medium in which the sedimentation study was carried out, suggesting a further splitting of the molecule under these conditions. After dialysis against distilled water for 48 h the relaxation time in HC1 solution was (pH 2), ph = 6-5 x 10-8 sec.5 After dialysis for a further 100 h, ph = 5.6 x 10-8sec in the same medium. Longer diaylsis periods did not result in any further change in the relaxation time. In N/10 NaOH the calculated relaxation time at 20" was 5.8 x 10-8 sec which does not differ significantly from 5-6 x 10-8 observed in acid.Apparently both dissociations result in the same products. The polarization of the fluorescence in alkali is not modified by the addition of neutral salts up to molar concentration. This difference with the acid dissociation may give an indication of the nature of the groups that bind the subunits together. If some of these have a pK of 9-11, the binding form being the one present at the lower pH, the disappearance of this38 POLARIZATION OF FLUORESCENCE form would result in dissociation even if the electrostatic repulsion is nzinimized by the addition of salt. On the other hand, in the acid dissociation the bonds would be broken by the electrostatic repulsion when the dissociation of the carboxyls is suppressed but could be restored if the approach of the subunits is made possible by addition of salt since the binding groups are present in the appropriate ionic form.The results just quoted and others obtained by Dr. T.-C. Tsao in his study of muscle proteins indicate that observations of the polarization of the fluorescence of labelled protein molecules is a particularIy useful tool in the investigation of those r A I I / t z 1 3 14 1 5 1 6 I FIG. 2.-Polarization of the fluorescence of 0-2 % bovine serum albumin conjugates. 1, pH 7.0, 0.1 M phosphate buffer. 2, pH 1-5, 0.2 M acetate + HCl buffer. 3. pH 1.82, 0.15 M HCl + KCl. 4, conjugate dialyzed for 150 h against distilled water and brought to pH 1-8 by 2 and 3 are curves convex towards the T/v axis showing thermal dissociation ; 4, where addition of HCl.presumably dissociation is complete, shows no detectable curvature. protein reactions that result in conspicuous changes of the molecuiar size and/or shape. Perhaps the most important feature of this method is that it can be used in the absence or presence of salt, and also over a wide range of both pH (1.5-14) and temperature (0-60"). Also the study of protein interactions is considerably facilitated because it is possible to label one molecular species and study its behaviour in a complex protein system provided this is non-fluorescent. Within wide limits the polarization of the fluorescence is independent of the concentration so that the protein concentration may be ignored in all cases where interactions between the labelled protein molecules is negligible. As indicated by Perrin 4 the molecular rotations are independent of the simultaneous translations although the converse is not true.It may therefore be expected that interactions which appear as changes in the translational diffusion with concentration may not be observed in a studyG . WEBER 39 of the rotational diffusion alone. This is exemplified by the successful use of solutions of high protein concentration in the study of dielectric dispersions.8 Conjugates of native bovine serum albunlin and ovalbumin showed polarizations which were independent of the protein concentration at all the concentrations investigated (2 % to 0.05 %). When p1/7 is greater than 50 an increase in temperature over the experimental range cannot be expected to increase materially the depolarization.The range of molecular sizes that can be studied depends upon the lifetime of the excited state of the fluorescence of the label. With l-dimethylamino-naphthalene-5-sulphonyl chloride the upper limit corresponds to globular molecules of molecular weight 300,000. It is possible that by the use of other labels with longer lifetime of the excited state molecules of 2 to 3 times this size may be studied. Fluorescent conjugates have been used by Coons 10 to trace the fate of anti- bodies in animal tissues. Besides the general use of labelled proteins as tracers other applications may ultimately be of importance. A study of the absorption spectra of the conjugates may give indications regard- ing the nature of the protein environment that surrounds the attached groups, and its changes under varying conditions.Such changes have been observed following the denaturation of ovalbumin conjugates by heat.6 Parallel observations on the fluorescent spectrum and yield may give much important information on the same matter. The accessibility of the protein surface to certain ions and molecules that are quenchers of the fluorescence may also be explored. For example, H ions which quench the fluorescence of the naphthalene sulphonates and its sulphonamido derivatives at pH 4 fail to affect the fluorescence of the conjugates even at a much lower pH, but do so on addition of neutral salts. A quantitative study of this effect will no doubt give valuable evidence relating to the ionic atmosphere surround- ing the protein. The use of the polarization measurements in the study of the reversible adsorp- tion of fluorescent dyes to macromolecules has been discussed recently by Laurence. By combining the polarization measurements with observations of the absorption spectrum and the yield of the fluorescence Laurence was able to characterize the binding sites in bovine serum albumin as basic groups imbedded in a lipophilic environment. 1 Perrin, J . Physique Rad., 1926, 7, 390. 2 Perrin, J. Physique Rad., 1936, 7, 1. 3 Weber, Biochem. J . , 1952, 51, 145. 4 Perrin, J. Physique Rad., 1934, 5, 497. 5 Perrin, Acta Physic. Polon., 1936, 5, 335. 6 Weber, Biochem. J., i952, 51, 155. 7 Laurence, Bioclzem. J., 1952, 51, 168. 8 Oncley, Chem. Rew., i942, 30, 433. 9 Scatchard and Black, J. Physic. Chein., 1949, 53, 88. 10 Coons and Kaplan, J . Expt. Met/., 1650, 91, 1.