Temperature and pH Dependence of the Binding of Oligosaccharides to Lysozyme*

SUMMARY Association of lysozyme with the following ligands was measured at 0.1 ionic strength, using absorbance and fluorescence procedures: (a) the p(1 + 4)-linked trimer of N-acetylglucosamine, at pH 0 to 9 and 8-60’; (b) the dimer, at pH 0 to 9 and 30” and at 9-60” and pH 5.3; (c) N-acetylglu-cosamine, at 9-60’ and pH 5.3. The complex-free protein difference spectrum varied with temperature and pH. Analysis of the association data indicated that trimer binding perturbs the pK of three ionizations, glutamic 35, 6.1 to 6.4; aspartic 101,4.3 to 3.4; aspartic 66, 1.9 to 1.5. aspartic interact the saccharide, its pK shift presumably reflects the change that has been


Association
of lysozyme with the following ligands was measured at 0.1 ionic strength, using absorbance and fluorescence procedures: (a) the p(1 + 4)-linked trimer of Nacetylglucosamine, at pH 0 to 9 and 8-60'; (b) the dimer, at pH 0 to 9 and 30" and at 9-60" and pH 5.3; (c) N-acetylglucosamine, at 9-60' and pH 5.3. The complex-free protein difference spectrum varied with temperature and pH. Analysis of the association data indicated that trimer binding perturbs the pK of three ionizations, glutamic 35, 6.1 to 6.4; aspartic 101,4.3 to 3.4; aspartic 66, 1.9 to 1.5.
Since aspartic 66 does not interact directly with the saccharide, its pK shift presumably reflects the change in conformation that has been seen in crystallographic analysis ( The enthalpy of association increases by 5.9 kcal between pH 5 and 2 and is compensated by a nearly equal increase in entropy. This behavior may be related to desolvation of aspartic 101 in complex formation.
The apparent average enthalpy of hydrogen-bond formation in the trimer complex is -1.5 kcal.
A considerable number of publications have reported data on the free energies of binding of oligosaccharides to lysozyme (References l-5; other data are reviewed in Imoto et nl. (6)). Several reports have given enthalpy values (1,(7)(8)(9)(10)(11) at' pH near 5. This paper describes the dependence on both pH and temperature of the association of lysozyme with the /3(1 + 4)linked dimer and trimer of N-acetylglucosamine. Particular attention is paid to the pH-dependence of the enthalpy of binding of trisaccharide, and to changes in free energy of binding at pH below 2. The enthalpy of association changes much more strongly between pH 2 and 5 than does the free energy. This * This work was supported by research grants from the Bmerican Cancer Society and the National Institutes of Health. $ To whom inquiries concerning this paper should be addressed. behavior appears to be associated with the withdrawal of aspartic 101 from solvent when saccharide binds. EXPERIMENTAL (16) ; formation of the trimer complex was followed at 340 nm for pH greater than 7 or 370 nm for pH below 7, and formation of the dimer complex at pH below 5 was measured at 360 nm.
For determination of equilibrium binding paramet'ers the fluorescence intensities of lysozyme solutions (approximately 0.05 mg per ml) were determined as a function of increasing saccharide concentration.
The highest concentration of ligand always gave at least 80 "/0 saturation.
The change in fluorescence was assumed to depend linearly on the amount of complex formed.
Association constants were calculated for simple bimolecular association between lysozyme and saccharide: ES is the molar concentration of complex, and E. and X0 are the total molar concentrations of lysozyme and saccharide. The association constant and change in fluorescence at saturation (AB',n,,) were obtained from a Scatchard plot, of AF versus AF/(S, -IS').
Protein concentration was 1 mg per ml in a l-cm cell or, for measurement with (Glcr\jAc),, 0.1 mg per ml in a lo-cm cell. Fig. 1 gives typical spectra for trisaccharide complexes. The 293 to 289 nm peak-trough difference is more precisely determined than the height of the 293 nm peak above a long wave length base, and the former value routinely was the measure of extent of association.
Values of K"ssoc and AE,,, were calculated as for fluorescence measurements. Proton Binding-The pH dependence of an association process can be determined by measurements of the effect of complex formation upon proton binding (17). A Radiometer Titrator This concentration of trimer is sllfficient to nearly saturate the enzyme at all pH.
The spectrum measured at pH 11.4 nas changing rapidly in the low xs-ave length region (260 nm).  Fig.  1 gives 19soz~rne-saccharide difference spectra over the pH range 0 to 11. The spectra resemble that described for p1-I near 5 by Hayashi el al. (18). Considering the three t,ryptophans in the lysozyme active site (19)) it is not surprising that the data typify tryptophan perturbation (20), with peaks at 293, 285, and 276 nm.

Spectra-
The difference spect,ra are more sharply structured than those obtained through solvent perturbation of lysozyme (e.g. using glucose or ebhylene glycol), iu particular with regard to the depth of the trough at 289 nm. The shallow trough at 304 nm found at low pH disappears as the pH goes to 5. At high pH a prominent peak develops at 302 nm.
These features corrclatc with carbosyl ionizations of pK approximately 3.5 and 6.5 (21). Fig. 2 shows the temperature and pH dependence of Aa,,,. The effect of increased temperature accords with the expected broadening of a spectrum line. An ionization of pK 3 to 4 like that seen in the 304 nm trough also affects Atmilx.
The slight pH dependence of AE,,, in very acid solution may be associated with a group of lower pK.
In this regard, the pH dependence of a saccharide binding (see below) and other experimental approaches determine an ionization of pK 1.5 to 2. Treatment oj Association Data-The pH dcpendencc of an association constant, can be described ( Ii""""" pII rci, nsa 2 is the association constant under reference coldit,ions of pH and temperature, hcrc arbitrarily set in the integration as (H+) >> 0 and T = 298.2; AH&a:~~ is the correspollding value of the cnthalpy of association; KL,, 298,2 and K'\, 298,2 arc apparent ionization constants at the reference temperature for groups of the enzyme-substrate complex and the enzyme; AII:: and AH: ' are the corresponding apparent enthalpies of ionization.
The sum is over all n ionizable groups i affected by saccharide binding (i.e. those for which KkS, T # Kb, T). The enthalpy of association is described by differentiation of Equation  2c with respect to l/T: The derivation of Equations 2c and 3 assumes that the mode of ligand binding is pH independent.' Dahlquist and Raftery (22) interpret NMR data for (G~cNA~)~ at pH 2 to 10 and 31-45" in terms of a single complex.
There is no evidence suggesting more than one rnode of (G~~NAc)~ binding (6). Equations 2c and 3 also require that each ionizing group affected by saccharidc binding have constant pK and enthalpy of ionization, at least over the range in which it contributes significantly to the pH dependence of log K""""~. These parameters might be altered owing to (a) a change in protein conformation; (b) a change in the state of ionization of a neighboring group; or (c) a noli-zero value of AC, (in the case of enthalpy parameters).' A variety of physical measurements have shown the Iysozyme conformation to be invariant over the pH and temperature range of this study (6) Table  I gives log Kassoc as a function of pH and tcmpcraturc for the (G~cNA~)~lysozyme system, and similar results for (G~cNAc)~ as a function of pH, at 30". The log K:*"""c -pH profile has three distinguishable regions (Fig. 3, pH 0 to 2.5, 2.5 to 5, and 5 to 7). At least three protein groups are perturbed by saccharide binding. There was no significant change in the parameters calculated if (a) a subset of the dat'a (15 and 30" values) was used; (b) reduced weight was given to data point,s that deviated more than the average; or (c) random error of range 0.1 in log Z<BSSor, 0.2 in ternperature and 0.1 in pH was introduced into each data point. It is important to note that data that are as accurate as can be obtained in measurements on protein systems (2 to 5c/ err01 in K~SSoc) cannot define pK values to better than ho.2 to 0.5 unit (or AH values to +2 to 5 kcal), if the titration ranges of the groups overlap and if three or more groups are involved. However, although complex pH profiles cannot accurately define values of pKi and AH', the changes in pK (ApK') and AH (AAH') brought about by saccharide binding depend little on the particular pKE est,imates. Literature data have suggested that ionizations with pKz different than given above also may be in the active site. NMR measurements have indicated a group of pKE 4.7 that is in or near the active site but is not perturbed by binding of P-methyl-GlcNAc (8). A group of pKE about 3.2 affects the lysozyme absorbance spectrum (21) and this effect is enhanced by saccharide binding wit,hout change in the observed pK (Fig. 2).4 Fluorescence rneasurements (16) show pKB8 3 to 3.5. Assuming three ionizing groups, satisfactory fits could not be obtained with pKi 3.2, nor with pKi 4.7 if pIi; was less than 2.5. The dependence of log Kassoc on pH does not require that saccharidc binding perturb more than three ionizable groups, but the cryst,allographic informat,ion indicates that four (glutamic 35 and aspartics 52, 66, and 101) might in fact be affected.
Xo nontrivial four group trial gave as good fit as three groups.

Entkalpies of Association
and Ionization- Fig.  4 shows AH"~~LsSOO as a function of pH. The curve was computed using Equation 3 and the parameter estimates of Table II. Values of AHop nSSoC calculated from van't Hoff plots of data for pH 2, 5, and 7 agreed within 0.5 kcal T\-itll the results of Fig. 4. The compensation between AH0 and TAX" and the much greater changes in these parameters than in AF" are striking.     Values of AHo, 85soc are given in Table III.   Table II. See text for discussion of deviation from experiment. B, as above, but measurements at 30". pH 5.3 and 2 is 5.9 kcal for the trimer and 5.3 kcal for the dimer.
This similar effcc-t of p1-I on the enthalpy stands against the substantially smaller effect of pI-1 on the free energy for dimer compared with trimer over the same pH range.
It is noteworthy that measurements for GlcNAc are fit within experimental error by a van't Hoff line. The ol-GlcNAc and P-Glcn'Ac complexes with lysozyme have considerably different geomet'rics (19).
Thus, either both modes have about the same enthalpy of association, or the ternperature dependence of the anomeric equilibrium in solution compensates for a putative difference in enthalpy of association. Proton Binding- Fig.  6 shows changes in proton binding measured for the lysozyme-trisaccharide system at 15 and 30". The curves are calculated according to Equation 3 using the parameters of Table II. Except at pII above 6, the agreement between experiment and calculation is good, and the proton binding data confirm the constants obtained from fitting of the association data.
The following two fact.ors, which are important for proton release but not for saccharide binding measurements, can csplain the discrepancies at pH above 6. In this pH range, lysozyme undergoes pH-dependent association (24). Binding of saccharide dissociates the lysozyme dimer (25). Increase in the lysozyme concentration increased ApH+, which accords with t'he contribution expecated from self-association. At high pH and at tllc high trisaccharide concentration used in thcsr rspcrimerits, a second molecule of l&and has been shown to bind to lysozyrnc with proton relcasc (26).

Thertnodynamics of Oligosaccharide
Binding at pfI 5.S- Table  III gives thermodynamic ljarameters for association of lysozymc at pH 5.3 with the monomer, dimer, and trimcr of AT-acrtylgltlcosarnincl.
The free energy values of the present, work agrc'c well with literature data (6). The van't Hoff enthalpy valuc~s are for all three saccllaridrs about 10yO higher than the valorimetric values (9). The agreement between calorimetric and can't Iloff values is satisfying and thr simple equilibrimll model used to interpret the data apparently is correct. N--~lcctvlglucosamine bitids at site C with an enthalpy of association of -6.2 kcal. 'I'hfare arc four hydrogen bonds b(,twccn cnzymc and sac&aride. If one accepts the premisrx that nonpolar interact'ions contribute much less to the enthalp> change t,han do hydrogen bonds, then this result suggests that formation of an enzyme-saccharidc hydrogen bond (ac%ually cschangc of partuers between enzyme-water and saccharidcwater) is on the average associated with an enthalpy change of -1.5 kcal.
l'hese numbers agree with the more negative of the estimates for the enthalpy of hydrogen bond formation rnatlc from model aqueous systems of several types (27)(28)(29). This agreement is comforting collsiderillg the difficulty in drawing analogies between reactions of small molecules and macromolrcules.
For reasons discussed below, the enthalpy of interaction at sites A and B cannot be simply related to the rlunrbct~ of hydrogen bonds.
The differences betwvccn thermodynamic parameters for association of lysozyme with different oligosaccharides have bc~n used to estimate the thermodynamic values for interaction of monosaccharide units at the A, 13, and C regions of the active site (for a discussion, see Imoto et a/. (6)).
Johnson et al. (30) noted the good correlation between number of atorn-to-atom contacts and AF,, for the several subsites.
In contrast,, the cnthalpy changes do not correlate well. It can be seen in Table  III that t,here is a much larger difference in enthalpy than in f'rcc eilergy between binding of the disaccharide at regions 1< and (," and of the monosaccsharide at region C. The estimated enthalpy of binding of a monosaccharide unit at 13 is -5.2 kcal (diffrretlcc between (G~~NAc)~ and GlcNAc), compared with -6.2 for C. The substrate forms one hydrogen bond at the 13 region and four at the C region, and the llurnbcr of atomto-atom contacts are in the ratio 1:3.
An anomaly that involves a charged group that interacts with substrate can explain the large enthalpy change for binding at region U (see below). dssignwlent of pK Values lo Side Chuin Croups--l'ablc IV correlates pK values of Table II   kcal per mole, which is close to the value (3 kcal) estimated by 1)onovanef al. (21). They also concluded that the pK of aspartic 52 is llot changed through substrate or inhibitor binding (31). A group of pK below 2 has been sc'en in one previous report on (GlcXhc);I binding (1). '1'1 lrse measurements are confirmed and are estcllded to the disaccharide. The assignment of the low l)K group as aspartic 66 follows most directly from the crystallographic observation that this is the only carboxylic acid residue, other than the above three, that has environment perturbed by saccharide binding.
Sspartic 66 is completely buried within the wing of the cleft that moves slightly when saccharidc binds. Scvrral points arc of interest in connection with this abnormal ionization assigned to aspartic 66. This carbosyl lies near argilline 68; it participates in screral hydrogen bonds; one of the hydrogen bonds involves tyrosinc 53, which is of abnormally high pK ( > 12; Tojo et al. (32)) ; d cnaturation studies show one or more groups of pK <2 (for review, see Imoto et a/. (6)).
Since aspartic 66 does not contact saccharidc, the pK shift (pKE,~ < pKB z 1.9), which as expected is the same for dimer and trimer, 1)resumably reflects small changes in conformation associated with movement of the structure about tryptophan 62. The complete withdrawal of a charged group from solvent is noteworthy.
hpparently a hydrogerl-bond network can adequately disperse charge without ion pair formation (arginine 68 is 7 A from aspartic 66). It is also noteworthy that the pK is shifted to a lolver value.
Regardless of the free energy cost of burying a charged group, the hydrog-cii-bond network within the protein favors carboxylate over carbosyl more than does water. Ikeda and Hamaguchi (5) report pK 2.5 for GlcNhc binding, which was not observed with the fl-methyl-glycoside (8). Vaildenhoff (33) could not confirm the pK 2.5 for GlcNAc but observed a group of pK 1.5 to 2 for GlcSAc and its fl-methyl-glycoside, which is in accord with the dimer and trimer result's, ilnomalous Enthalpy of Protonation 0.f Aspartic 101 in Complex -Formation of the trisaccharidc-lyaozyme complex shifts the l)K of aspartic 101 down by 0.9 ullits, a loss through protonation of 1.2 kcal in AFO, assoc (Table IV).
Although this number is substantial, the corresponding change in AHo, -7.3 kcal, is notably larger.
It is balanced by an almost equivalent change of -21 e.u. in ASO. Fig. 4  llimer perturbs the pK of aspartic 101 about half as much as does trimer, which accords with only one hydrogen bond between dimer and aspartic 101. However, change in pH from 5.3 to 2 increases AHo* assoc by 5.3 kcal for dimer compared to 5.9 kcal for trimer.
The enthalpy anomaly found for the trimer complex exists also with dimer.
The strong compensation between changes in ellthalpy and entropy focuses attention on the solvent. hspartic 101 is the most exposed rarboxylate of the lysozyme molecule; binding of trimer reduces its exposure to solvent to below that of all carboxylatcs except aspartic 66 (6). Aspartic 101 in the complex is nearly enclosed in a rigid matrix of protein and saccharide atoms, which unlike a water shell, cannot rearrange during proton dissociation.
This does not conform to naive expectation, which because of the presumably greater electrostatic work required to charge in the medium of lower dielectric constant, is that AAHO should be positive and the principal determinant of AAF'O. The fact that this is not true has been explained, as have the negative values for AASO, in terms of changes in solvent structure (37). The point made here is that acetic and other acids give unexpected behavior in a transfer like the protein process of interest. Specifically AAHO is more negative than AAFO for increased concentration of nonaqueous solvent. Also, transfer of an ion to a mixed aqueous solvent should show residual solvation effects that would be absent for transfer into the more rigid protein ligand complex, and the absence of these should magnify for the protein the negative contribution of ion transfer to AAHO and AAP.
Since charged side chains frequently 1)articipate in protein interactions, large negative enthalpies of ionization generally may be a factor in llrotein reactions.
Cnfolding of the structure would remove anomalies, aud ionizable groups perhaps contribute more substantially than has been thought to enthalpics of proteill denaturation. 11EF15IZENCXS