168

A Study of the Adsorption of Dyes on Bovine Serum Albuminby the Method of Polarization of Fluorescence

[BY D. J. R. LAURENCE*Biochemical Laboratory, University of Cambridge

(Received 1 Augugt 1951)

The ability of serum to bind small molecules wasbrought into prominence by the work of de Haan(1922) on the excretion of vital dyestuffs by thekidney. Grollman (1925) later showed by ultra-filtration that serum albumin possesses the propertynot possessed by ovalbumin, serum globulin, orgelatin, of adsorbing phenol red at alkaline pH. Heshowed that all the proteins studied bound the dyewhen the dye and the protein were of oppositeelectrical charge, but that serum albumin could alsobind the negatively charged dye when it was itselfnegatively charged, adsorption decreasing rapidly,however, above pH 8. Bennhold (1932) showedthat a variety of dyes and substances of pharmaco-logical value was carried by the albumin fraction ofserum and proposed for the protein a transportingfunction in the animal body. Recent studies on thebinding by the serum albumin fraction (Smith &Smith, 1938; Rawson, 1943; Chow & McKee, 1945)and especially the detailed studies of Murray Luckand his co-workers on the stabilization of serumalbumin (Boyer, Lum, Ballou, Luck & Rice, 1946)and the intensive physicochemical studies of Klotz(1949) have gone far to elucidate both the conditionsunderwhich adsorptiontakes place and the practicalimportance ofthephenomenon (Edsall, 1947). Mean-while, Scatchard, Scheinberg & Armstrong (1950)and Edsall, Edelhoch, Lontie & Morrison (1950) haveshown that serum albumin also binds a number ofnegatively charged inorganic ions.The present paper is concerned with a study

similar in principle to those previously carried out,of the binding of cQrtain fluorescent dyes to bovineserum albumin. The method used was that de-veloped by Dr G. Weber in this department de-pending on a measurement of the polarization ofthe light emitted by excited fluorescent molecules(Weber, 1952).

THEORETICALThe polarization of the fluorescent light from a dyemolecule depends on the relaxation time of rotaryBrownian movement of the dye in a given environ-ment and on the time which the dye molecule takesto emit its light following an excitation. If the dye

* Present address: Postgraduate Medical School, DucaneRd., London, W. 12.

molecules are able to rotate extensively betweenexcitation and emission, the fluorescent light haslittle overall polarization while if the moleculesrotate very little in this time the emitting oscillatorswill remain appreciably orientated and will producean overall polarization of the fluorescent light(Perrin, 1936). For the majority of dyes in aqueoussolution the relaxation times of the molecules aresmall compared with the time taken to emit thefluorescence, and so the polarization of the fluor-escence is small (usually less than 1 %). The relaxa-tion time may be increased by placing the dye in aviscous medium such as glycerol (Weigert, 1920) or,as Weber has shown, by attaching the dye to amacromolecule such as a protein, which rotates moreslowly in aqueous solution than does the free dye.The result ofimportance in the present study is thatthe fluorescence ofthe dye adsorbed on the protein ispartially polarized, while the fluorescence of thefree dye is usually unpolarized. Therefore, in anymixture of a fluorescent dye with a protein, where apart of the dye is bound to the protein and theremainder is free, the polarization P observed willbe less than that for complete binding of the dye PBdepending on the relative contribution which thefree and the bound dye make to the total fluorescentintensity. One can say that the polarized fluor-escence of the bound dye is diluted by the unpolar-ized fluorescence ofthe free dye. A knowledge oftherelative intensities of fluorescence of the dye, freeand bound, and of the measured polarization of thedye-protein mixture enables the amount of dyebound to be easily calculated as follows. (Forderivations of these formulae see Appendix.)Assuming that the fluorescence of the free dye in

solution is unpolarized, the ratio ofP to PB will varybetween 0 and 1 as the degree of adsorption of thedye on to the protein is varied. Calling this ratio pand supposing the intensity of the fluorescence ofthe dye-protein system relative to that of the dyealone at the same concentration and pH is R it isshown that the fraction of dye bound (x) is

x=1pR-R+ 1. (1)If the dye does not change its fluorescent intensityon adsorption (i.e. R= 1) the simple formula

x=p (la)

I952

Vol. 5I ADSORPTION OF DYES

will hold. On the other hand, if the free dye inaqueous solution possesses a measurable polariza-tion a second modification of Eqn. 1 is needed,namely that, where II =PpPB

(I1- I) x=pR-R+ 1-HI' (I1 b)and P. is the polarization ofthe free dye in solution.

Only one case was found in this work for which thesimple formula (1 a) could be used (see below); mostof the dyes studied showed either a marked increaseor a decrease in fluorescence when bound to bovineserum albumin. The significance of these changes influorescence will be discussed later, but it is enoughto note here that if a marked increase in fluorescentyield occurs on adsorption (as in the case of thesulphonic acid derivatives of 1-naphthylamine),formula (1) ceased to be of practical use since theobserved polarization is little diluted by the re-latively poor fluorescence of the free dye. In thesecases a formula R -1

X R -1 (2)

has been used, in spite of the fact that this formuladoes not distinguish between a change in the fractionbound (x) and a redistribution of the already boundmolecules among adsorption sites which affect thefluorescence to different degrees. A limitation ofthis sort is implicit in any indirect method such asthis, especially where only oneexperimental variableis used, and may equally enter into the spectro-photometric technique (Klotz, 1946b). Sinceformula (1) includes two experimentally determinedvariables it is less likely to suffer from limitations ofthis sort and the formula has been used to detect onecase (eosin) where the sites are of different kinds.

In order to determine the number of adsorptionsites on the albumin and their affinity for the dye theequation derived by Klotz (1946a)was used, namely,

Si1 K- ~~~~(3)Dx n nD(l-x)

The known quantities in this equation are S and D,the total protein and dye concentrations respec-tively, and x, the fraction of dye bound, calculatedas described above. The unknown quantities K, theequilibrium constant of dissociation of the eye-protein complex, and n the number of adsorbingsites, may easily be derived from the experimentalcurves, since n is the reciprocal of the intercept atthe axis and K/n the slope of the graph. A greaterslope means a more dissociable complex. When thedye concentration is small compared with the pro-tein concentration the simple formula

(1- x)nS =K (3a)x

will apply. This is a useful formula since, even if theadsorbing sites differ in combining power, the formof the equation remains the same.

3 ON SERUM ALBUMIN

EXPERIMENTAL

DyesThe dyes used were chosen for their availability and possibleexperimental value. They can be classified chemically intothree classes: (i) sulphonic acid derivatives of 1- and 2-naphthylamine containing one, two or three sulphonic acidgroupings; (ii) derivatives offluorescein and rhodamine witha xanthydrol ring system; (iii) derivatives of acridine.Where necessary these dyes were recrystallized or convertedto crystalline derivatives and recovered. The purity of thedyes was examined by chromatography on paper. The term'dye' has been used for these fluorescent compounds tosimplify the nomenclature, although in some cases theabsorption maximum of the compounds lies in the nearultraviolet region of the spectrum.

Serum albuminThe protein used was the Armour crystalline bovine

serum albumin (batch no. 10522). It was found by com-parison with the Armour fraction 5 albumin (batch no.12 198) that the crystallization had not appreciably alteredthe adsorbing properties of the protein, but had removedsome yellow material whose absorption of the exciting lightmight complicate the fluorimetric studies. The absorption atthe excitation wavelength of 3600A. was not great enough inthe crystalline product to require any special corrections.The fluorescence of the crystalline protein was also unim-portant with the yellow filters (Corning 338, 351, etc.) usedfor examining fluorescence. The protein was made up in1% solution, dialysed for 24 hr. at 00 and filtered throughWhatman no. 50 filter paper to give a clear solution. Theprotein concentration was measured in a Zeiss interfero-meter using the diffusate in the control compartment. Theinterferometer was calibrated using Kjeldahl nitrogenvalues for ovalbumin solutions.

Measurement offluorescenceThe protein and dye were mixed and a suitable buffer

added. Experiments were made both at constant dye con-centration and at constant protein concentration. Whena series was carried out at constant dye concentration,only one comparison tube for free dye was required. Whenthe dye concentration was varied, a comparison tube wasmade up for each dye concentration used. Observationswere made in each case on the fluorescent intensity of thedye-protein mixture, as compared with the fluorescence offree dye; and on the polarization of the fluorescence of thedye-protein mixture. In cases where interest was found inthe absorption spectrum of the mixture, this was measuredin the Beckman spectrophotometer from 6000A. down to3100A.The limiting polarization PB was measured by so increas-

ing the protein concentration that no further increase inpolariza,tion resulted from addition of more protein. Thiscondition was obtained with the dyes at certain pH valueswhere binding was especially strong. The work of Weber(1952) suggests that the same value of PB would be valid atother pH values.

Since polarization measurements are most easily madevisually, the intensity measurements were also madevisually, the same filters for excitation and observationbeing used in both cases. Any changes in colour in the

169

D. J. R. LAURENCEfluorescence during adsorption would affect intensitymeasurements made with a photocell in an arbitrary waywith respect to the eye and make the use of Eqn. 1 moredifficult. Fluorescence of wavelength short enough to bereadsorbed by the dye was excluded from observation bychoice of filters. The visual fluorimeter was made by placinga Pulfrich photometer headpiece vertically at the end of anoptical bench and splitting the beam of the usual Siemensmercury lamp, placed on the bench, by means of a modifiedPulfrich beam-splitter. The cells containing the fluorescentsolutions, size 4 x 2 x 1 cm. were filled completely with thesolutions and covered with pieces ofoptical glass. They were

Fluorescein Ec

tion ofthe exciting light (Weber, 1952). In the measurementof polarizations the exciting light was polarized by a Nicolprism in a plane perpendicular to the axis ofobservation. Inthe fluorimeter, on the other hand, the exciting light wasunpolarized. To make the two sets of measurements com-parable the following formula (derived in the Appendix) wasused where I' is the observed and I is the corrected intensity:

I'1--1 -ip-

For the theoretical maximum of the polarization of 0-50 thiscorrection is 33%.

CH2COO-CH2

Br

0 0 0 0 0 0Br

)sin Sucecinyifluorescein

NH2

5-Aminoaaridine o-5-Acridylbenzoic acid I-Naphthylamine-8-sulphonic acid

Cresyl fast violet Mepacrine

Table 1. Combination offluore8cein in borate buffer at pH 9-1(Fluorescein concentration =0-13 x 10-5 mol./l.)

Relativeintensity

R pR0-61 0-130-51 0-170-37 0-17

* Where n is the number of binding sites.

Fraction ofdye bound (x)

0-520-660-80

Equilibriumconstant ( x 105)K =nS(I - x)/x*

1-35n1-50n1-45n

illuminated on their largest faces and observed through thetop. The fluorescent intensities of a series under test were RESULTSalways compared with the brightest fluorescence of theseries whether the free dye or the dye combined with the Table 1 gives some typical results the combina-

protein. tion of the bivalent negative ion of fluorescein with

A simple correction was made to the intensity measure. serum albumin in borate buffer at pH 9. Thements, since the intensity observed from a solution pro- fluorescein concentration was small compared withducing polarized fluorescence is dependent on the polariza- the protein concentration and so Eqn. 3a applies.

Protein concn.( x 105 mol./l.)

S1-462-925-84

Polarization( x 100)100P7-6

12-017-0

170

Vol. 51 ADSORPTION OF DYES ON SERUM ALBUMIN

The result of plotting according to Eqn. 3 is astraight line passing through the origin with slopedetermined by the equilibrium constant.

Table 2 illustrates a second simple case. This isthat of the tervalent negative ion of the 2-naph-thylamine-3:6:8-trisulphonic acid in unbufferedsalt-free protein solution at pH 5-4. In this casethere is strong binding to the protein so that over arange of concentration of both protein and dye twomolecules of the dye are bound by one molecule ofalbumin (molecular weight 70 000). Extrapolationfrom Table 1 shows that fluorescein is only 0-26times as fluorescent on the protein as in the solution.The fluorescence of the 2-naphthylamine-3:6:8-trisulphonic acid, on the other hand, is but littlealtered by adsorption.

Table 2. Combination of 2-naphthylamine-3:6:8-trisulphonic acid at the isolectric point

Molecules of dyeProtein conen. Fraction of bound per moleculex 106 mol./l. dye bound (x) of protein

(a) Dye concentration= 5 x 10-5 mol./l.0-60 0-26 2-151-21 0-52 2-152*42 0-92 1-89

Molecules of dyeDye concn. Fraction of bound per molecule

(x 105 mol./l.) dye bound (x) of protein(b) Protein concentration = 1-21 x 10-5 mol./l.

5.0 0.50 2-0610.0 0.258 2-1320-0 0-132 2-18

Table 3 gives data for the monovalent negativeion of 1-naphthylamine-8-sulphonic acid in phos-phate buffer, pH 6-9. In this case the fluorescenceof the dye is much increased on the protein (about

Table 3. Combination of 1-naphthylamine-8-8uphonic acid at pH 6-9

(Sulphonic acid concentration 4-15 x 10-5 mol./l.)

Protein conen.( x 105 mol./l.)

0-000-7251-452-907-25

Relativeintensity( . 20)*0-050-370-620-841-00

Molecules ofFraction of dye bounddye bound per molecule

(x) of protein0-0000 34 1-950-60 1-720-84 1-191-00 -

* I.e. relative to the same concentration of free dye,allowing for the increase of fluorescence on adsorption.

20-fold), and the polarization rises to a maximum

before the limit is reached in the increase of fluor-escence. Using Eqn. 2 it appears that about 2molecules of this substance are bound to the proteinin the concentration range studied.

Effect of pH

As the pH in the surrounding solution is de-creased and the charge of protein becomes morepositive it would be expected that there would be anincrease in the amount of a negative dye bound to

- n=l

0 n=4

1-0 2-0x105 mol-11 /[Free dYe]=-1 /D (1-X)

Fig. 1. Adsorption isotherms of 2-naphthylamine-3:6:8-trisulphonic acid.

1 /Free dye]l-1 /D (1-x)

Fig. 2. Adsorption isotherms of fluorescein.

the protein. This is borne out by the following datafor 2-naphthylamnine-3:6:8-trisulphonic acid (Fig. 1),fluorescein (Fig. 2) and 1-naphthylamine-3:8-disulphonic acid (Fig. 3). It should be noted thatfluorescein contains two titratable groups betweenpH 8 and 4, while the 1-naphthylamine-3:8-disul-phonic acid contains one with pK 3-9. These reducethe negative charge on the dye at acidpH relative toalkaline pH.

VoI. 5I 171

D. J. R. LAURENCE

Figs. 4 and 5 summarize the results obtained byplotting the polarizations obtained at an arbitraryprotein concentration against the pH. Fig. 4 is forpositive ions (basic dyes) and Fig. 5 is for negativeions (acidic dyes). The results illustrate the familiarobservation that an opposite charge on the proteinfavours the combination of a dye, but there is also

20r

a1-5

-o

'- 1.0

0-oC.t 050a-

)v0.

n.n

_ ~~002$P.~~~~~~~~bae ?70

n=2

n4 0-02 m-acetate. pH 3-5

0-5 10 1-5x105 mol.-'1 /[Free dye]-=l /D (1-X)

Fig. 3. Adsorption isotherms of 1-naphthylamine-3:8-disulphonic acid.

lipid-soluble) is more strongly adsorbed than thepositive ion, even though the average charge on theprotein at these pH's would be expected to favourcombination with the positive ion.

This is confirmed by the result of adding serumalbumin to o-5-acridylbenzoic acid at pH 6 wherethe protein is slightly negative and the dye mostlyin the zwitterion form with a negatively charged

0-6

x 0.50

C NNIO

C

~04500 4000 3500

Wavelength (A.)

Fig. 6. Effect of pH on the absorption spectrum ofo-5-acrid.ylbenzoic acid.

0-30 - 0-30 (d)

o 020 (a) (c

(d) aciybnocail(d)

(a) a)oeci,()lnptyamn-:-iupoi cd

(c)

0

3 5 7 9 1c1 3 5 7 9 1s1pH pH

Fig. 4. Fig. 5.

Fig. 4. Effect of pH on the adsorption of positive dyes

(a) mepacrine, (b) cresyl fast violet, (o)5-aminoacridine,(d) acridylbenzoic acid methyl ester.

Fig. 5. Effect of pH on the adsorption of negative dyes

(a) fluorescein, (b) 1-naphthylamine-3:8-disulphonic acid,

(t) 2-naphthylamine-3:6:8-trisulphonic acid, (d) eosin.

noticeable the special characteristic of serum

albumin. This is apparent in both graphs. In Fig. 5

it is seen that negative ions, even tervalent ions,

continue to combine with the protein on the alkaline

side of the isoelectric point (pH 5.4). In Fig. 4,

though the basic dyes combine in all cases with the

protein, there are marked changes in extent of

combination at the pK's of certain groups in thedyes. These changes lead to the conclusion that theneutral form of the dye (that form which is more

VO

0-45

0

C

o0.30

4500 4000 3500Wavelength (A

Fig. 7. Effect of the addition of serum albumin on theabsorption spectrum of acridylbenzoic acid at pHE 6-0.

carboxyl group and a positively charged hetero-cyclic ring. The serum albumin has the same effecton the dye as does a change of pH toward thealkaline side (Figs. 6 and 7). Addition of serumalbumin to the dye at pH 9 when the heterocyclicnitrogen is uncharged has no effect on the spectrum.This 'reversed protein error' can be due to forceslike those which produce the other effects discussedhere. Fig. 8 shows adsorption isotherms for this dye.That the neutral and ionic forms ofthe dyes differ

markedly in the PB values is made unlikely byexperiments on the polarization of the fluorescenceof the two forms in glycerol-water solutions ofvarious viscosities. The conclusion follows that

172 I952

ADSORPTION OF DYES ON SERUM ALBUMIN

while the presence of a charged group on the mole-cule is not a necessary condition for adsorption tooccur, a positive group in contrast with a negativegroup has a strong inhibitory influence on thecombination. This influence extends asfar as pH 9 3,the pK of the ring nitrogen in 5-aminoacridine(Albert & Goldacre, 1943) when the protein has anet charge of about -30 (Tanford, 1950).

4+0x

3-0

-o

cO 2-0D

a-cL

- I I-x A

lX6s - n=2

0 1-0 20 3C1/[Free dye]=1I/D(1-x)

Ox105 mol.t

Fig. 8. Adsorption isotherms of acridylbenzoic acid.

Effect of modi,fications in 8tructure of the dyeSome results of interest to the general theory of

adsorption are obtained when the effect of minoralterations in the structure of the dye on its adsorp-tion properties is studied. Fig. 9 shows results withfluorescein as compared with eosin in neutralphosphate buffer. Eosin is a case where the in-tensity does not change linearly with the calculatedfraction bound (x) but first decreases and then in-creases as the protein concentration is increased.This unusual effect can be explained if only a few ofthe total binding sites yield an appreciably fluor-escent adsorption complex and if it is assumed thatthese are the most actively combining sites. Whilethe dye is in excess ofthe total sites, both the sites oflow and ofhigh fluorescent yield will be occupied. Ifan excess of sites is present, however, the sites ofhigher fluorescent yield will be favoured and thefluorescent intensity will increase again. There istherefore a minimum of fluorescence depending onthe ratio ofdye to protein. Using the formula 1 b theamountbound is found to increase continuously overthe range of protein concentration.The naphthylaminesulphonic acids tried, namely

the l-naphthylamine-5- and 8-monosulphonic acids,the 3:8-disulphonic acid and the 3:6:8- and 4:6:8-trisulphonic acids and the 2-naphthylamine-3:6:8-sulphonic acid, showed binding of about the sameextent at neutral pH. Methylation of the aminogroup increases the extent of combination of the1-naphthylamine-5-sulphonic acid with serumalbumin.

It can also be shown that fluorescein methylester and o-5-acridyl-benzoic acid methyl ester areadsorbed to about the same extent as the free acids.Sulphofluorescein is adsorbed similarly to fluor-escein, and succinylfluorescein, with a short sidechain instead of a ring, less than fluorescein.

30x

t 20 4C)

o ~~~0

1000

1.0 -d/ O n=1

c0 <* n=2

0 20 4-0 60x105mrol'1 /[Free dye]-1 /D (1-x)

Fig. 9. Adsorption isotherms of eosin and fluorescein in001 M-phosphate pH 6-9.

Competition experiment

As it has so far been described, the method of thepolarization of fluorescence is limited to studyingthe combination of fluorescent substances only.A knowledge of the ability of various substances tocompete with suitable test dyes leads, however, toan extension of the method in a more indirect form(Klotz, 1949). There are two easy ways of carryingout competition experiments:

(i) With a given quantity of serum albumin and of testdye the polarization P1 is recorded and, in addition, thepolarization P2 (always less than PF) which results when aknown further amount of the same dye is added. If, in-stead of this further addition, an equimolar amount of anon-fluorescent competitor is added and a polarization P3results we have either

P3 =PlP3 =P2P1 >P3>P2P3 <P2

The substance shows no competitionCompetes as strongly as the test substanceCompetes less than the test substanceCompetes more than the test substance

This is illustrated by Table 4, where various derivativesof naphthalene are used to compete with the 2-naphthyl-amine-3:6:8-sulphonic acid. The sulphonic groups, but notthe amino groups, seem necessary for competition. As the1- and 2-naphthylamines themselves have no effect on thepolarization, it seems likely that they do not combine at thesite ofcombination ofthe test substance at the concentrationused.

(ii) For substances whose combination with protein isfairly weak, competition can be studied by adding thecompetitor in great excess relative to the protein and notingthe decrease in the polarization.

Vol. 5I 173

D. J. R. LAURENCE

Table 4. Competition of related compounds with 2-naphthylamine-3:6:8-triulphonic acid (pH 6.7)

CompetitorNoneNone1-Naphthylamine2-NaphthylamineNaphthalene-l-sulphonic acidNaphthalene-2-sulphonic acidNaphthalene-3:6:8-trisulphonic acidN-Acetyl-l-naphthylamine-8-sulphonic acid

The present experiments have been confined tocomparing the relative affinities of related simplesubstances as revealed by this method, although anabsolute derivation should be possible.

Table 5 gives data for the ability ofvarious ions tocompete at the same molar concentrations, againusing the 2-naphthylamine-3:6:8-trisulphonic acidas test substance. The outstanding trend in thistable is that whereas alteration of the cation haslittle orno effect on the extent ofthe combination theability of different anions to compete is markedlydifferent. This agrees with the results found both inthis study and by other workers that preferentialcombination occurs with the negative ion ofan ioniccompound in solution.The competition with the series of univalent

anions F-< Cl-< Br- <NO3- < I- lies in their orderin the so-called Hofmeister series relating to gelatin(Loeb, 1924) and this shows that the simple assump-tions of the Debye-Huckel theory cannot in thiscase be sufficient to explain the effect of these salts.The effect of bromine atoms (see results with eosinabove) and of nitro groups (Teresi & Luck, 1948) inorganic moleculesseem to be paralleled by the strongbinding of the nitrate and bromide ions.

Table 6 gives a similar set of figures for someorganic anions and one cation. The differencesobserved between the different compounds in this

Amount ofAmount of competitor

test substance (ml. equimolar(ml.) solution)1.00-20-20-20-20-20-20-2

0-80-80-80-80-80-8

Polarization0.060 P20.116 P10'115 P30-117 P30-065 P30-052 P30*035 P30-071 P3

Table 5. Competition of some ion8 with 2-naphthyl-amine-3:6:8-tri8ulphonic acid (pH 6.7)

Competitor (m/15):NoneNaClKCINa2SO4K2SO4Sodium acetateSodium succinateSodium benzoateNH4ClCsClNaFKCIKBrKINaN,NaNO3

Competitor (M/30):K2SO4(NH4)2SO4Li2SO4MgSO0

Polarization

0-2020.1000.1000.0150-01501300 0150 0000.1000.1010.1810-0980-0500 0000-0450 015

00450-0470-0480 070

table are of importance for the theory of adsorptiondiscussed below.The 2-naphthylamine-3:6:8-trisulphonic acid was

used in these studies because its fluorescence in-tensity is little changed on adsorption and its

Table 6. Competition of organic ions with 2-naphthylamine-3:6:8-tri8ulphonic acid (pH 6.7)

Competitor conen. (M X 10-2)

Competitor

Sodium galacturonateSodium phenylacetateSodium cyclohexylacetateSodium caprylateOctylamine hydrochloride

0

0-2050-2050-2050-2050-205

1

0-2050-1620-07100-205

1-5 3-5Polarization

0-2050-1020

0

0-205

0-2050-0000

0

0-148

Competitor conen. (m x 10-3)

0

Sodium caprylate 0-208

1 1-5 2-5Polarization

0-140 0-067 0

* Attributable to the chloride ion alone.

6

0-2050000.095*

174 I952

ADSORPTION OF DYES ON SERUM ALBUMIN

polarization has, therefore, a linear relationship tothe fraction which is bound. The next section con-siders more generally the changes in fluorescentproperties and absorption spectra of the dyes.

Change influorescence andabsorptionspectra on adsorption

In the previous sections the change in fluorescentintensity on adsorption has been mentioned as animportant factor to be considered in applying themethod ofthe polarization offluorescence to a studyof adsorption isotherms. During adsorption thereare also changes in the absorption spectra of thedyes. The changes may be summarized convenientlyaccording to the chemical types of the dye mole-cules.

(i) Sulphonic acid derivatives of 1- and 2-naphthyl-amine. The fluorescence of the 2-naphthylamine-3:6:8-trisulphonic acid is bright sky blue and isslightly quenched when the dye is on the proteinwithout a change in colour. The fluorescences of thel-naphthylamine derivatives are greenish yellow inaqueous solution and this is also the case with thedimethylaminonaphthalene derivative. The fluor-escence of all these compounds is increased whenthey are adsorbed on the protein and in most casesit is changed to the bright blue of the 2-naphthyl-amine compound described above. This change ismost marked with the relatively weak fluorescenceof the 1-naphthylamine-8-sulphonic acid where a20-fold increase in brightness occurs. With thestronger fluorescence of the others in this series,increases of two or fourfold are observed. Thechange may be due to an increase in the dipolemoment of the amino group on adsorption, as it isknown that 2-naphthylamine possesses a higherdipole moment than 1-naphthylamine (Bergmann& Weizmann, 1936). Shifts to the red of 80A. wereobserved in the ultraviolet spectra of the 1-naph-thylamine compounds when adsorbed.

(ii) Xanthydrol derivatives. The fluoresceins andrhodamines tried were all quenched on the protein.For fluorescein itselfthe intensity was 0-26 ofthat ofthe free dye. In addition, characteristic shifts in theabsorption spectrum towards the red occurred. Thiseffect was most marked with the methyl ester offluorescein where the sharp absorption bands of thefree and of the adsorbed dyes are clearly separatedat the maxima, while the shape and height of thecurves are hardly altered (Fig. 10). The colours ofsolutions of the adsorbed methyl ester are brightpink while that of the free dye at alkaline pH isorange yellow. This is not an effect of pH, as theabsorption of the different ionic forms of fluoresceinmoves progressively towards the blue with increasedacidity. In the case of eosin the adsorbed dye has amoderate adsorption at the green line of mercury,namely, 5460A. while the free dye absorbs little at

this wavelength. The tubes which contain adsorbedmaterial can therefore be distinguished easily asthey are brightly fluorescent in this light, whereastubes with the free dye alone are by comparisonpractically non-fluorescent.

(iii) Acridine derivatives. Considerable variationwas found in the behaviour of the fluorescences ofthe acridines when adsorbed. The fluorescence ofo-5-acridylbenzoic acid was quenched to one-tenthof its value in solution without change in its ab-sorption spectrum. The fluorescence of mepacrineincreased slightly.

b8-0

C4)

0

C0V 4,0xa (a)_ *~~~~~~Free

~2-0 -(b)0 ~~~Adsorbed

05500 5000 4500 4000

Wavelength (A.)Fig. 10. Absorption spectrum of fluorescein methyl ester atpH 9 (a) free dye, (b) dye adsorbed on serum albumin orcetyltrimethylammonium bromide.

The relation of the changes in fluorescences andabsorption spectra to the state of the environment ofthe dyesSince the change in the fluorescences and the

absorption spectra of the dyes is likely to becharacteristic ofsome new electronic state due to anenvironment at the site of adsorption different fromthat in water, it was considered likely that a simu-lation of these changes in simple physico-chemicalsystems would lead to an understanding of thenature of the adsorbing region on the protein. A suit-able series of dyes was chosen to test this, namely,eosin, fluorescein methyl ester, 1 -dimethylamino-naphthalene-5-sulphonic acid, 1-naphthylamine-8-sulphonic acid and o-5-acridylbenzoic acid. By trialit was found that the kationic detergent, cetyl-trimethylammonium bromide (CTAB), gave re-

actions with these dyes very similar to those given bybovine serum albumin. That adsorption by thedetergent micelles had occurred was demonstratedby showing that the detergent in low concentrationincreased the polarization of eosin excited with the

VoI. 5 I 175

D. J. R. LAURENCE

green line of the mercury lamp and of fluoresceinexcited with the blue lines. Klotz (1947), using theazo-dye methyl orange, has shown a similar parallelbetween the effect of protein and of CTAB on theabsorption spectrum. The ring structure of the dyewhich acts as the oscillator (Benel, Kastler& Rousset,1940) islikelyto be inthemicelleratherthanadsorbedon its surface (Hartley, 1937), and hence ina substan-tially organic environment free from water dipoles.An important property of the interior of the

micelle is its low dielectric constant, and in a single-phase system the dieloctric constant of the solutionsof the dyes was lowered by adding dioxan (Akerlof& Oliver, 1936) which had been purified by the

Absorption8-0 maximum

A.5200-

x 5100

Fi. 11. Effec of dixncncnrtono hasrto

increasing dixacncntatf4900-

0 020 4060 80c Dielectric constant

Fig. lla

x

20-a0

5500 5000 4500 4000Wavelength (A.)

Fig. 11. Effect of dioxan concentration on the absorptionspectrum of fluorescein methyl ester (from right to left,increasing dioxan concentrations from 0 to 90%.)

Fig. Illa. Shift in the wavelength of the absorptionmaximum with increasing dlielectric constant of thedioxan-water mixtures.

method of Oxford (1934). The dyes behaved in away similar to that observed when they were ad-sorbed by serum albumin and CTAB, the moststriking similarities being shown by fluoresceinmethyl ester, 1-naphthylamine-8-sulphonic acidand acridylbenzoic acid (Figs. 10-13). The effect ofsolvents of low dielectric constant in imitating theeffect ofserum albumin is such that we could almostwithout exception predict how a given dye wouldchange its fluorescence and spectrum from itsbehaviour in dioxan solution. Evidence that thedielectric constant of the medium is the determiningfactor in the changes in spectrum of the xanthydrolcompounds may also be obtained from the results ofVaillant (1927) for erythrosin, though this authordoes not interpret his results in this way. He foundthat the largest shifts to the red in the spectrum oferythrosin were in benzene and 'essence', both sob-

vents of non-polar nature. Glycerol, concentratedsugar solution and ethanol gave smaller shifts to theredwith respect towater, although the refractive in-dices and hence the volume-polarizabilities of theseliquids was as high or higher than the benzene and'essence'. A similar view is admitted by Seshan(1936).

4Jc

4._5C4._c

El

0

El

(U

El

10Ia 0-6x

'c' 0-54._C 0*5El

a 040

c.2 0-3Idc

, 02aI-

0 0.1uE

O

20 40 60Dielectric constant

i

I1-

, _

, -

I

80

4000 3500 3000Wavelength (A.)

Fig. 12. Effect of dioxan on the absorption spectrum andfluorescent intensity of 1-naphthylamine-8-sulphonicacid. (a) spectrum ofdye in aqueous solution, (b) spectrumof dye adsorbed on serum albumin or in 90% dioxansolution.

It may be concluded that the behaviour of thefluorescent dyes when adsorbed on to serum albuminidentifies the site of adsorption with material of lowdielectric constant. The use of a series of these dyesshould enable other systems of amphipathic be-haviour to be easily distinguished ashas already beenstated (Sheppard & Geddes, 1945).

Relation to the work of Dr G. Weber with fluorescentgroups coupled by chemical reaction with proteins

Of the dyes used in this paper, two, namely the2-naphthylamine-3:6:8-trisulphonic acid and the1-dimethylaminonaphthalene-5-sulphonic acid have

176 I952

_

ADSORPTION OF DYES ON SERUM ALBUMIN

been used by Dr G. Weber as compounds suitablefor coupling with proteins by covalent bonding. It ispossible to carry out studies using serum albuminsimilar to those ofWeber (1952) by ensuring that allthe dye is combined with protein and plotting acurve of IIPB against T/a,, where T is the absolute

100r- p

._!

ci 80C

v 60vVI

0° 40

204 )

0

0

x

* 1*0v0C

0

x" 0 5tvo

eosin excited by Hg 5460A. where an extrapolationof I/PB to T/I = 0 gives the accepted value for eosinin very viscous solution. The colour of the fluor-escence of the serum albumin-sulphonamide con-jugate is greenish blue like the colour offluorescenceof the simple sulphonamides in 90% dioxan.Digestion of the albumin conjugate with A±mourcrystalline pepsin results in the fluorescence be-coming orange like that of the sulphonamides inwater. Even with native proteins the greenish-bluefluorescence of albumin conjugates is not universal(Weber, 1952) and the yellowing of the fluorescenceof the sulphonamides as the polarity of the mediumis increased may provide a convenient yardstickagainst which the polarity of the protein surface inthe neigbbourhood of the primary amino groupsmay be estimated.

5*0

0

-

r

2

m

CL0L-

v

00.

L-v4)a:

Wavelength (A.)

Fig. 13. Effect of dioxan on the absorption spectrum andfluorescent intensity of o-5-acridylbenzoic acid. (a)spectrum of dye in aqueous solution, (b) spectrum of dyein 90% dioxan.

temperature and X the viscosity. The intercept ofthis line with the ordinate gives the value of 1/PB inthe absence ofBrownian movement, and the slope isdetermined by the lifetime of the excited state andby the size of the protein. In either case the value of1/PB was found to be the same at T/lj = 0 for thecoupled as for the adsorbed dye. The slope of thegraph for the adsorbed 1-dimethylaminonaphtha-lene-5-sulphonic acid was found to be twice thatfor the coupled sulphonamide conjugate (Fig. 14).Since Weber (unpublished observations) has shownthat the lifetime of excited state of the acid in gly-cerol is about twice that of the derived sulphon-amide, these results show that the adsorbed dye hasa fixity in the albumin surface comparable with thatof the coupled dye. A similar result is obtained for

Biochem. 1952, 51

2-0 4-0

T/b7 x 10 4 I K./poiseFig. 14. Plot of polarization at different temperatures

according to the Perrin equation for the 1-dimethyl-aminonaphthalene-5-sulphonic acid attached to serumalbumin (a) covalently coupled (Weber's data), (b)adsorbed.

DISCUSSION

The evidence presented here and that obtained byother workers leads to the view that there areseveral kinds of physicochemical mechanismoperating in the process of adsorption on to serumalbumin.

It is clear that the charge on the molecule is ofimportance, negative ions being adsorbed inpreference to positive ions at the sites of adsorption.Superimposed on this is a variation in binding de-pending on the nature of the ion. It is often sup-posed that this effect is due to van der Waals forcesbetween the protein and the adsorbed molecule.Since, however, for compounds of about the samemolecular weight, namely, caprylic acid, cyclohexyl

12

VoI. 5 I 177

D. J. R. LAURENCE

acetic acid and phenyl acetic acid, one obtainsdifferent degrees ofbinding with the benzene moietyless active than the paraffin moieties, this does notseem to be an adequate hypothesis.Any organic molecule may be considered as an

assembly of polar and of non-polar parts, and inaqueous solution the relation of these parts with thewater will be very different. Non-polar materialintroduced into water produces a separation of thestrongly bonded water molecules and replaces theinteraction of water molecules with each other by aweaker interaction of water with the non-polarsubstance. Reversal ofthis process with recombina-tion of the separated molecules of water results in aliberation of energy which corresponds to theenergy of the surface of the non-polar material insolution. The polar parts of the molecule, in con-trast, can associate with the water as well as canwater with itself (Langmuir, 1925) and the adsorp-tion of the non-polar part of the organic moleculeon to or into a non-polar surface, leaving the polarparts in the aqueous phase, is therefore accompaniedby a liberation of energy, since the surface of thenon-polar material in contact with the water is de-creased.A semi-quantitative way of stating this is to

assign to water a dipole W and to the so-called non-polar material a smaller dipole P (Gent, 1948).Before adsorption the dipole interaction energy isproportional to 2PW and after adsorption top2+ W2. The change in energy is p2+ W2 - 2PW or(W- P)2 which increases as the difference betweenthe respective dipoles increases and is alwayspositive.From the behaviour of the fluorescent dyes when

adsorbed there is good reason to suppose thatadsorption does remove the dye from the water to asecond medium of lower dielectric constant, and soit seems plausible to attribute the adsorption ingreat part to a rejection from the water structurerather than an attraction by the protein. The con-cept of surface energy further accounts for therelatively low activity of a benzene moiety as com-pared with a paraffin, since the surface energv ofbenzene in water is less (34 ergs/cm.2) than for aparaffin (54 ergs/cm.2); benzene is more hydro-philic than paraffin. As would be expected, galac-turonic acid, although its molecular weight ishigher than most ofthe simple competitors includedin Table 6, is ineffective in competition, as it isstrongly hydrophilic and the force of rejection isweak.

In terms of the serum albumin molecule theamino-acid composition ofthe protein is adequate toprovide the sort of surface against which theseforces can act. It is visualized that the region ofadsorption is one where a local excess of positivegroupings is backed by a local excess of the hydro-

phobic side chains of leucine, valine and other non-polar amino-acid residues. Because of the low di-electric constant of this microphase the local chargeis shielded from the general field ofthe protein and isnot as greatly affected as it would otherwise be byvariations in the net charge of the protein as a whole(Hartley, 1938). To account for the absence ofcorresponding sites of adsorption for positive ions,it must be supposed that the negative groups on theprotein are more diffusely arranged and lack thehydrophobic backing which the positive groupshave.For each molecule the extent of adsorption will

depend both on non-polar forces and attraction ofnegative ions. In the series of naphthylamine-sulphonic acids, for example, the addition of asulphonic group will decrease the former and in-crease the latter type of interaction, the effect nearthe isoelectric point being to keep the total bindingof these acids appreciably constant. An additionaltype of force may be important with bromine,iodine and the nitro group, both in inorganic ionsand in organic binding. This could be a polarizationin the field between the protein and the orientateddipoles ofwater. The low activity ofthe fluoride ion,as compared with the chloride, may be due to itsincreased hydration, keeping it in the aqueous phaseby means of hydrogen bonding.

In a recent review Klotz (1949) has described hisviews on the special structure of the serum albuminmolecule and has criticized the ideas ofDavis (1943).The present work supports and extends the ideas ofDavis and is in disagreement with the views ofKlotz, as it underlines the importance of the surfaceactive properties of those amino-acids residuesusuallyregarded as inert, namely, the leucines,valine,etc. Klotz, in contrast, believes the stoicheiometricrelations of hydroxy amino-acids with the di-carboxylic acids to be of paramount importance.The correlations given by Klotz are neverthelesssuggestive, as it is for those proteins where thehydroxyl groups are relatively less numerous thatthe hydrophobic backing of charged groups wouldbe most easily achieved and the forces described inthis paper most likely to be active. In view of thenon-polar nature ofthe active surfaces onthe proteinand ofthe adsorbed molecule the idea of 'combined'water molecules being displaced on adsorption(Klotz, 1949) would seem unlikely.The results with the dyes used in this work show

that about 2-4 molecules of dye are adsorbed byeach molecule of the protein at neutral pH, with theexception of eosin where more molecules combine.A decrease in the pH increases the number of sitesavailable to negative dyes. Rosenfeld & Surgenor(1950) have found 2 molecules of ferriprotopor-phyrin strongly bound to the human serum albuminmolecule, while Martin (1949) has found 3 molecules

178 I952

Vol. 5I ADSORPTION OF DYES ON SERUM ALBUMIN 179

of bilirubin bound per molecule of albumin. Theseresults contrast with the much higher values fromthe dialysis experiments of Klotz (1946a) and otherworkers, e.g. Teresi & Luck (1948) (10-25 moleculesbound per albumin molecule). Klotz (1946b) hasnever reported these high values as a result of hisspectrophotometric technique, and it may be thatthe dialysis technique when applied to higher dyeconcentration includes some type of interactionbetween the dye and the protein unlike that occur-ring at lower concentrations and measured by thepresent method. Certainly the low values are morein accord with the view that the effect ofthe positivegroups is reinforced as described in this paper in afew special patches on the protein, while the highervalues accord better with the view of Klotz that theamino groups need only to be freed from hydrogenbonding with other amino-acid residues in order tobe active. The results of a comparison with Weber'swork on covalently coupled dyes, discussed above,strongly support the idea of a special patch ofpositive groupings as opposed to a general activa-tion.Whether a similar active region is important in

cases of interactions of a more specific nature re-quires further evidence to decide. It is likely that theforces related to surface tension described here playa part in all interactions of non-polar material inaqueous solution, and the forces of van der Waalsshould therefore not be used indiscriminately toaccount for these interactions. The analogy betweenthe electronic state of the dye molecules at theprotein surface and in organic media should beborne in mind when attempting to formulate atheory of enzyme activation.

SUMMARY

1. A method is described for studying quanti-tatively the adsorption of fluorescent dyes bymacromolecules using measurements ofthe polariza-tion and intensity of the fluorescent light. Themethod is applied to the adsorption of dyes bybovine serum albumin.

2. The effect ofpH on the adsorption is discussed.Negatively charged dye ions are adsorbed evenwhenthe net charge on the protein is negative. A positivecharge on the dye molecule inhibits combination oneither side of the isoelectric point. Basic dyes areadsorbed as the free base but not as the positive ion.

3. Free carboxyl, sulphonic, hydroxyl or aminogroups in the dye are not necessary for combinationbetween dye and protein to occur. Bromine atomsincrease the extent of adsorption.

4. Addition of related but non-fluorescent com-pounds can decrease the adsorption of a fluorescentdye by competition. Competition is also foundbetween the dye and a variety of simple inorganicand organic ions. The activity of univalent negativeinorganic ions is in the order of the Hofmeisterseries and that of univalent negative organic ions inthe order of surface activity. Positive ions do notcompete with the dyes.

5. Characteristic changes in the absorptionspectra and in the fluorescence of the dyes occur onadsorption. These changes can be reproduced if thedyes are adsorbed by cetyltrimethylammoniumbromide micelles or dissolved in 90% dioxan.

6. The significance of these results is discussedand a comparison made between the adsorbed dyeand the dye coupled covalently to serum albumin bythe methods of Dr G. Weber. It is concluded thatthe results of this work can best be explained by thepresence in serum albumin ofa few regions ofsurfacewhere a local excess of positively charged amino-acid residues is backed by residues with non-polarside chains.

I am indebted to Dr G. Weber of this laboratory for initi-ating this work, for critical advice during its progress, andfor making available the apparatus for measurement ofpolarization of fluorescence and many of the organicchemicals used. I would also like to thank Dr M. R. J.Sultan of the Department of Colloid Science, Cambridge, forproviding the cetyltrimethylammonium bromide, Dr F. A.Isherwood of the Low Temperature Research Station,Cambridge, for a gift of purified galacturonic acid andMessrs Imperial Chemical Industries Ltd. for the naphthyl-aminesulphonic acids. The work was carried out during thetenure of a grant from the Medical Research Council.

REFERENCESAkerlof, G. & Oliver, A. (1936). J. Amer. chem. Soc. 58,1241.Albert, A. & Goldacre, R. (1943). J. chem. Soc. p. 454.Benel, H., Kastler, A. & Rousset, A. (1940). C.B. Acad. Sci.,

Paris, 211, 595.Bennhold, H. (1932). Ergebn. inn. Med. Kinderheilk. 42,

473.Bergmann, E. & Weizmann, A. (1936). Trans. Faraday Soc.

32, 1318.Boyer, B. D., Lum, F. G., Ballou, G. A., Luck, J. M. &

Rice, R. G. (1946). J. biol. Chem. 162, 181.Chow, B. F. & McKee, C. M. (1945). Science, 101, 67.Davis, B. D. (1943). J. clin. Invest. 22, 753.

Edsall, J. T. (1947). Advanc. Prot. Chem. 3,463.Edsall, J. T., Edelhoch, H., Lontie, R. & Morrison, P. R.

(1950). J. Amer. chem. Soc. 72, 4641.Gent, W. L. G. (1948). Quart. Rev. 2, 383.Grollman, A. (1925). J. biol. Chem. 64, 141.de Haan, J. (1922). J. Physiol. 58, 444.Hartley, G. S. (1937). Symposium on Wetting and Deter-

gency. British Section of the International Society ofLeather Trade's Chemists.

Hartley, G. S. (1938). Tran8. Faraday Soc. 31,31. 'Added inDiscussion.'

Klotz, I. M. (1946a). J. Amer. chem. Soc. 68, 1486.12-2

180 D. J. R. LAURENCE I952Klotz, I. M. (1946b). J. Amer. chem. Soc. 68, 2299.Klotz, I. M. (1947). Chem. Rev. 41, 373.Klotz, I. M. (1949). Cold Spr. Harb. Sym. quant. Biol. 14,97.Langmuir, I. (1925). CoUoid Symp. Monogr. 3, 48.Loeb, J. (1924). Proteins and the Theory of Colloidal Be-

haviour. 2nd ed. p. 141. New York: McGraw Hill.Martin, N. H. (1949). J. Amer. chem. Soc. 71, 1230.Oxford, A. E. (1934). Biochem. J. 28, 1325.Perrin, F. (1936). Acta phy8. polon. 5, 335.Rawson, R. A. (1943). Amer. J. Phy8sol. 138, 708.Rosenfeld, N. & Surgenor, D. M. (1950). J. biol. Chem. 183,

663.

Scatchard, G., Scheinberg, I. H. & Armstrong, S. H. (1950).J. Amer. chem. Soc. 72, 535.

Seshan, P. K. (1936). Trans. Faraday Soc. 32, 689.Sheppard, S. E. & Geddes, A. L. (1945). J. chem. Phys. 13,

63.Smith, W. W. & Smith, H. W. (1938). J. biol. Chem. 124,

107.Tanford, C. (1950). J. Amer. chem. Soc. 72, 445.Teresi, J. D. & Luck, J. M. (1948). J. biol. Chem. 174, 653.Vaillant, M. P. (1927). J. Phys. Radium. (6), 8, 393.Weber, G. (1952). Biochem. J. 51, 155.Weigert, F. (1920). Verh. dtsch. phys. Gem. 23, 100.

APPENDIX

Derivation of equation8

Forrp.ulae (1) are derived as follows: consider asolution of dye with fluorescent intensity Io andpolarization PF, to which protein is added. As theconcentration of protein increases the fraction ofdye bound to the protein increases. Asmall additionof protein will remove an amount of dye Sx, offluorescent intensity 1o Sx, from solution. Attachedto the protein this amount ofdye will have a differentintensity A9,I &x, say, where Ax may vary during thecourse ofa large addition ofprotein. It will alsohavea polarization PB larger than PF. When a fraction xofthe dye isbound the intensity ofthe mixture I willbe rx rxb1=10- Iodx +j Ajodx

o o

(1$ ) ~~~~~~~~~(i)The polarization p is the average for free and

bound dye weighted according to their respectiveintensities, i.e.

(1 -x) PF+f PBAxdxP -x+ x (ii)

I-X+l Adx0

With PB not dependent on x, the integral fZA,dx

is eliminated from (ii) using (i). Calling

PIPB=P,PF/PB= "I I/IO= ;_(1-x) I+R+ I (iii)

or (1-II) x=pR-R+1-H. (Ib)

Equations (1) and (1 a) are derived from this byputting P.= II = 0 and R= 1 progressively.Formula (2) is based on a simple proportionality

argument. Formula (4) is derived as follows:Suppose the fluorescence is excited by light planepolarized at an angle 0 to the axis of observationand has a polarization p. The observed intensity I'will be the sum of three components, namely,

I' =11 sin00+12 cos2 0+ I2, (iv)where I1 and Ia are the intensities ofthe componentsof fluorescence parallel and perpendicular to the

exciting vector and so P = 1 + 2 by definition, or

12=-+P) I . Unpolarized excitation is equivalent

to averaging for all angles 0 or since sin' 0= cos'0= i;I = I + ft2

I1 31-p\221+p/

=(1+P) (v)For excitation perpendicular to the direction ofobservation on the other hand 0= 900, and

II + 12

(1 +p)(i

Combining (v) and (vi)

(4)1 -IP