IsothermalTitrationCalorimetryof ProteinProtein Interactions MichaelM. Pierce, C. S.Raman, 1 and Barry T.Nall 2 Department of Biochemistry, Universityof Texas Health Science Center, 7703 Floyd CurlDrive,San Antonio,Texas782847760 The interaction of biological macromolecules, whether protein DNA, antibodyantigen, hormonereceptor, etc., illustrates the complexityanddiversityofmolecularrecognition.Theimportance of such interactions in the immune response, signal transduction cascades,andgeneexpressioncannotbeoverstated.Itisofgreat interest to determine the nature of the forces that stabilize the interaction.Thethermodynamicsofassociationarecharacterized by the stoichiometry of the interaction ( n ), the association con- stant ( K a ), the free energy ( D G b ), enthalpy ( D H b ), entropy ( D S b ), and heat capacity of binding ( D C p ). In combination with structural information, the energetics of binding can provide a complete dissection of the interaction and aid in identifying the most im- portant regions of the interface and the energetic contributions. Various indirect methods (ELISA, RIA, surface plasmon reso- nance, etc.) are routinely used to characterize biologically impor- tant interactions. Here we describe the use of isothermal titration calorimetry(ITC)inthestudyofproteinproteininteractions.ITCis the most quantitative means available for measuring the thermo- dynamicpropertiesofaproteinproteininteraction.ITCmeasures the binding equilibrium directly by determining the heat evolved on association of a ligand with its binding partner. In a single experiment, the values of the binding constant ( K a ), the stoichi- ometry( n ),andtheenthalpyofbinding( D H b )aredetermined.The free energy and entropy of binding are determined from the association constant. The temperature dependence of the D H b parameter, measured by performing the titration at varying tem- peratures, describes the D C p term. As a practical application of the method, we describe the use of ITC to study the interaction between cytochrome c and two monoclonal antibodies. 1999 Academic Press How are foreign molecules recognized by the im- mune system? How are signals mediated by small mol- ecules translated into the expression and regulation of specic genes? These are just two of the fundamental questions currently being investigated by many re- search groups. Although distinctly different concepts, thesetwoexamplesarefundamentallyrelatedthrough the necessity of forming associations between biologi- cal molecules. For the scope of this paper, we limit the discussion to proteinprotein interactions. One impor- tant aspect of the study of proteinprotein interactions is in the application to protein folding. The exceedingly complex process through which a polypeptide assumes its native, functional conformation shares some simi- larity with the process of proteinprotein interactions. Both events involve the burial of solvent-exposed sur- face,theformationofhydrogenbonds,saltbridges,and van der Waals interactions. It should be noted though that while protein folding appears to be driven by the burial of hydrophobic residues (the hydrophobic effect) itislessclearwhetherproteinproteininteractionsare dominated by the same driving force. The energetics of the two processes also share similarities, most notable of which is the large change in C p which is typically negative in sign. Classic transfer free energy experi- ments have shown that the transfer of a nonpolar solute from an aqueous solvent to a nonpolar solvent is characterized by a large negative change in C p (1, 2). These studies have been proposed to be analogous to the sequestration of nonpolar amino acids in the inte- rior of proteins. Thermodynamics of protein folding as studied by calorimetry as well as other indirect tech- niques support the proposal that a decrease in heat capacity is indicative of a decrease in exposure of hy- drophobicsurface(3). 1 Present address: Department of Molecular Biology and Biochem- istry,Universityof California, Irvine,CA 926973900. 2 To whom correspondence should be addressed. Fax: (210) 567 8778.E-mail:nall@uthscsa.edu. METHODS 19, 213221 (1999) Article ID meth.1999.0852,available online athttp://ww.idealibrary.com on 213 1046-2023/99$30.00 Copyright1999by Academic Press All rights of reproduction in any form reserved. " *® @3 The formation of proteinprotein complexes is often accompanied by a large negative change in C p yet it is not certain whether this decrease in heat capacity can also be correlated with a decrease in the exposure of nonpolar surface. In many cases, well-resolved solvent molecules can be identied at the interface of a proteinprotein complex (when a three-dimensional structure is available at atomic resolution). In cases where the observed heat capacity change cannot ac- count for the amount of nonpolar surface buried, it has been suggested that buried solvent may be responsible for this effect (4). Although some features of protein foldingandproteinassociationaresimilar,itisevident that the various forces involved in stabilizing native protein structures and proteinprotein complexes, such as hydrogen bonds, van der Waals interactions, and hydrophobic interactions, may have distinct roles in the two processes. GENERAL ASPECTS OF ITC Calorimetric methods have been an invaluable tool for understanding the forces that stabilize the folded conformations of proteins. Recently, the advent of sev- eral highly sensitive titration calorimeters has gener- ated much interest in this technique (57). A thorough description of the theoretical concepts of ITC has been provided in the above and other reports (8, 9) to which the reader is referred. The following discussion repre- sents a general description of titration calorimetry, independent of the specic instrument used. An ITC instrument consists of two identical cells composed of a highly efcient thermal conducting ma- terial (Hasteloy or gold) surrounded by an adiabatic jacket (Fig. 1). The jacket is usually cooled by a circu- lating water bath. Sensitive thermopile/thermocouple circuits detect temperature differences between the two cells and between the cells and the jacket. Heaters located on both cells and the jacket are activated when necessary to maintain identical temperatures between all components. In an ITC experiment, the macromol- ecule solution is placed in the sample cell. The refer- ence cell contains buffer or water minus the macromol- ecule. Prior to the injection of the titrant, a constant power ( , 1 mW) is applied to the reference cell. This signal directs the feedback circuit to activate the heater located on the sample cell. This represents the baseline signal. The direct observable measured in an ITC experiment is the time-dependent input of power requiredtomaintainequaltemperaturesinthesample and reference cell. During the injection of the titrant into the sample cell, heat is taken up or evolved de- pending on whether the macromolecular association reaction is endothermic or exothermic. For an exother- mic reaction, the temperature in the sample cell will increase,andthefeedbackpowerwillbedeactivatedto maintain equal temperatures between the two cells. For endothermic reactions, the reverse will occur, meaningthefeedbackcircuitwillincreasepowertothe sample cell to maintain the temperature. ITC instru- ments should be routinely calibrated by applying spec- iedelectricalpulsesofapproximately510 m cal/s.The total heat as determined by the area under the pulse should be within 2% of the expected value (Omega ITC Manual, Microcal Inc.). The performance of the calo- rimeter can also be tested by measuring the heat of a standard chemical reaction, such as the protonation of tris(hydroxymethyl)aminomethane (THAM) by HCl (10). To ensure optimal performance of the calorimeter both calibration methods are recommended. The heat absorbed or evolved during a calorimetric titrationisproportionaltothefractionofboundligand. Thus, it is of extreme importance to determine accu- rately the initial concentrations of both the macromol- ecule and the ligand. For the initial injections, all or most of the added ligand is bound to the macromole- cule, resulting in large endothermic or exothermic sig- FIG. 1. Schematic diagram of an ITC instrument. Two lollipop- shaped cells are contained within an adiabatic jacket. A small con- tinuous power is applied by the heater on the reference cell. Thermopile/thermocouple detectors sense temperature differences between the reference and sample cells. On interaction of ligand and macromolecule, heat is either taken up or evolved. Depending on the nature of the association, the feedback circuit will either increase or decrease power to the sample cell to maintain equal temperature withthereferencecell.Theheatperunittimesuppliedtothesample cell is the observable signal in an ITC experiment and a direct measure of the heat evolved on binding of a ligand to a macromole- cule. 214 PIERCE, RAMAN, AND NALL cell feedback heater heater ;·, calibration Y?}EE{€{§·€2i&?}€z§»? .h»‘h AT T?ifiEE@5:2;;;_€f;F{I1iE* hcam Reference Cell Sample Cell Adiabatic Jacket nals depending on the nature of the association. As the ligand concentration increases, the macromolecule be- comes saturated and subsequently less heat is evolved or absorbed on further addition of titrant. The amount ofheatevolvedonadditionofligandcanberepresented by the equation Q 5 V 0 D H b [M] t K a [L]/ ~ 1 1 K a [L] ! [1] where V 0 is the volume of the cell, D H b is the enthalpy of binding per mole of ligand, [M] t is the total macro- molecule concentration including bound and free frac- tions, K a is the binding constant, and [L] is the free ligand concentration. For a more general model of binding, the multiple independent sites model, the macromolecule contains multiple ligand binding sites that are noninteracting. The cumulative heat of bind- ing can be described by Q 5 V 0 [M] t (~ n i D H i K a i @ L #! / ~ 1 1 K a i @ L #! . [2] A thorough discussion of the derivation of the above equations is presented by Indyk and Fisher (9). The method of analysis will be specic to the system under investigation and the reader is advised to attempt t- ting according to a general model before the use of more specic models. RUNNING A TYPICAL ITC EXPERIMENT ITC instruments can be of either the single- or dual- injection variety (5, 6). In a single-injection instrument titrant is added to the sample cell containing the mac- romolecule. In a separate (control) experiment titrant is added to the sample cell in the absence of the mac- romolecule. In the dual-injection instrument, the ti- trant is simultaneously added to a reference cell con- taining buffer and to the sample cell containing macromolecule. The advantage of this type of instru- mentisthatthecontrolreactiontomeasuretheheatof dilution of the ligand is eliminated (6). For this review, wedescribetheuseofthesingle-injectionOmegatitra- tion calorimeter manufactured by Microcal (North- hampton, MA) (5). Extreme care must be taken in all aspects of an ITC experiment, from sample preparation to data analysis. In the following section we divide a typical ITC exper- iment into the following steps: sample preparation, sample and reference cell loading, injection syringe loading, experimental parameters, control experi- ments, data analysis, and troubleshooting. 1. Initial Considerations and Sample Preparation The concentrations of macromolecule and ligand are critical, especially when one or both partners of a com- plex are difcult to obtain in large quantities or are of minimal solubility. Additionally, for measurement of the association constant, the initial concentrations of both the ligand and the macromolecule should be de- termined with a high degree of accuracy. The experi- mental binding isotherm can be characterized by the unitlessvalue c , whichistheproductoftheassociation constant, K a , the concentration of macromolecule [M], and the stoichiometry of the reaction, n : c 5 K a [M] n . [3] For an accurate determination of the binding constant, a c value between 1 and 1000 is recommended (5). Large c values prohibit the determination of K a since the transition is very sharp and too few points are collectednearequivalence(saturationmaybeachieved in a single injection of ligand). Binding isotherms with low c values lose the characteristic sigmoidal shape and are very broad transitions that approach linearity and the equivalence point cannot be identied. For tight binding complexes such as antibodyantigen in- teractions with K a values in the range 10 9 10 10 Mitis difcult to obtain accurate association constants since lowering the antibody concentration to obtain an ap- propriate c value extends beyond the sensitivity of the calorimeter (Fig. 2). The choice of buffer is also critical when planning an ITC experiment. If on complex formation protons are takenuporreleased,theequivalentnumberofprotons will be taken up or released by the buffer. If the se- lected buffer has a large enthalpy of ionization, the measured enthalpy will reect both buffer ionization andcomplexformation.Thiscanbequiteinformativeif one wishes to determine whether protons are taken up or released on complex formation and is accomplished simply by performing equivalent experiments in a buffer with negligible enthalpy of ionization (e.g., so- dium phosphate) and one with a large enthalpy of ionization (e.g., TrisHCl). For initial experiments we recommend the use of a buffer such as sodium phos- phate with a low ionization enthalpy. Both titrant and macromolecule should be exhaus- tively dialyzed in buffer (preferably in the same ask) to minimize artifacts arising from mismatched buffer components. The nal dialysis buffer should be saved and used for any necessary concentration adjustments of the macromolecule or titrant solutions. Macromole- cule and ligand are ltered or centrifuged to remove any precipitated material. Immediately prior to load- ing the sample cell and injection syringe, the ligand and macromolecule solution are degassed to remove 215 THERMODYNAMICS OF PROTEINPROTEIN ASSOCIATIONS residual air bubbles. Even the smallest air bubbles remaining in the cell or injection syringe can interfere with the feedback circuit. Air bubbles can additionally lead to poor baselines. 2. Loading the Sample and Reference Cells The reference cell (Fig. 1) usually contains water or buffer with 0.01% sodium azide and need not be changed after every experiment. We recommend changing the solution on a weekly basis if the instru- ment is routinely used. The macromolecule (usually but not necessarily the larger component of the inter- action) is added to the sample cell (Fig. 1) of the calo- rimeter using a long needle glass syringe. Typically, 1.5 to 2 ml of the solution is prepared to ll a cell with a volume of 1.31.5 ml. The utmost care is required to ll the sample cell without introducing air bubbles. Once the cell is completely lled, several rapid addi- tions of solution will dislodge any residual air bubbles that cling to the side of the cell. Any excess solution remaining in the reservoir is removed. 3. Filling and Attachment of the Injection Syringe Filling the injection syringe with the titrant or li- gand solution also requires great care. The concentra- tion of ligand solution should be such that the molar ratio of ligand to macromolecule, following the last injection, is approximately 2. Typically, a complete ti- tration will involve approximately fteen to twenty 5- to 10- m l injections of ligand. Handling of the injection syringe is extremely critical. Great care must be taken to avoid bending of the needle while the injection sy- ringe is loaded into place. Bending of the injection syringeneedlecanresultinsomeofthetitrantsolution being expelled into the macromolecule solution, caus- ing the rst injection to be unusable. Any minor bend- ing in the syringe can also result in poor baselines when the injection apparatus is stirring. 4. Experimental Parameters The parameters of the titration are input into the software program controlling data acquisition. The number, volume, and length of time of injections are critical and are discussed below. To determine accu- rately the enthalpy of binding, it is critical that the rst several shots dene a baseline region where all added ligand is bound to the macromolecule. The equivalence region should also be well dened by the concentration range spanned by the injections, to de- termine an accurate value of the association constant. It is necessary that concentrations be chosen so that measurable amounts of free and bound ligand are in equilibrium within the titration zone dened by the titrant injections. For characteristically tight binding afnities such as those exhibited by antibodyantigen complexes this can be an impossible task. In the tight binding limit, even when a binding constant cannot be determined,itisstillpossibletodetermineanaccurate value for the binding enthalpy. Several injections should be performed after complete saturation of the macromoleculebyligand.Theheatevolvedorabsorbed following saturation represents the heat of dilution of the titrant. The length of time of injection should be such that proper mixing is achieved. Typically, we rec- ommend 7- to 10-s injections of ligand. To ensure proper mixing the injection syringe is tted with a Teon paddle and attached to a stirring motor by a rubber belt. Stirring at approximately 400 rpm should ensure good mixing. 5. Control Experiments to Determine Heats of Dilution Theobservedbindingisothermisusuallynormalized as kilocalories per mole of ligand injected and plotted FIG. 2. Calorimetric titration of MAb 5F8 with cytochrome c in 0.1 M sodium phosphate, pH 7.0. The experiment consisted of 25 injec- tionsof5 m leachofa96 m Mstocksolutionofcytc.Cytochromecwas injected into a sample cell (volume 5 1.38 ml) containing 6 m M antibody combining sites at 25C. The injections were made over a period o f 9 s with a 2-min interval between subsequent injections. The sample cell was stirred at 400 rpm. (A) Differences between the sampleandreferencecellcontainingwaterwith0.01%sodiumazide. The heat of dilution of cyt c into buffer has been subtracted. (B) Enthalpypermoleofcytcinjectedversusinjectionnumber.Sinceno attempt was made to obtain a binding constant from these data, it was not necessary to plot the enthalpy as a function of the molar ratio of cyt c to MAb 5F8. Reprinted with permission from C. S. Raman, M. J. Allen, and B. T. Nall (1995) Biochemistry 34, 5831 5838. Copyright 1995 American Chemical Society. 216 PIERCE, RAMAN, AND NALL Timc (min) 0 l02030405060 -0.0 A 8 { -0.2 `E o :1. -0.4 4 E o B yay <¤ •-· -4 _3 • E ·8 "·5 -12 .2 0 -16 E -20 D ‘° ••}`h $2 -24 ’ 0 5 l0 15 20 25 Injection Numbcr versus the molar ratio of ligand to macromolecule. The observed heats of binding include contributions from the dilution of the titrant (ligand) and dilution of the macromolecule. A small contribution arising from stir- ring is also included in the observed binding enthalpy. Several control experiments must be performed to cor- rect for the heats of dilution. The heat of dilution of the ligand is usually the most signicant. This is generally true since the initial concentration and the dilution factor for the ligand are typically 1020 times greater than those of the macromolecule. The heat of dilution of the ligand can be determined by performing an iden- tical titration experiment in which ligand is injected into a sample cell containing buffer only (no macromol- ecule). The heat of dilution of the macromolecule (which is typically less signicant than the heat of dilution of the ligand) is determined by titrating buffer solution into the sample cell containing the macromol- ecule. The two heats of dilution are used to correct the concentration-normalized binding isotherm. Since in- jection of ligand following saturation of the molecule is essentiallyameasurementoftheheatofdilutionofthe ligand, the measured enthalpies of the last several injectionscanbeaveragedandsubtractedtocorrectfor heats of dilution. We recommend performing the con- troltitrationsforthemostprecisedeterminationofthe heats of dilution. 6. Data Analysis The method of data analysis depends on the system of interest. For this article we briey describe the pro- cedure for tting data to the multiple independent binding site model using the analysis software ORI- GIN (Microcal, Northhampton, MA) provided with the Omega ITC. Prior to peak integration, the heats of binding are normalized as a function of ligand concen- tration. Additionally, a volume correction is also per- formed due to dilution of the macromolecule during each injection. The areas under the peaks are inte- grated in either a manual peak-by-peak fashion or automatically by routines provided in the software package. Baseline selection is an important factor in ITC data analysis and user input in the automated integration routine is limited. Therefore, we recom- mend manual peak-by-peak integration in which the operator denes the baseline regions used in the inte- gration step. To determine n , K a , and D H b Eq. [2] is represented in terms of the binding constant and total ligand concentration [L] T to obtain the quadratic: Q 5 ~ n @ M # t D HV 0 ! /2 $ 1 1 @ L # t / ~ n @ M # t ! 1 1/ ~ nK a @ M # t ! 2 @~ 1 1 @ L # t / ~ n @ M # t ! 1 1/ ~ nK a @ M # t !! 2 2 4 @ L # t / ~ n @ M # t !# 1/2 % . The n , K a and D H b parameters are then optimized using the standard Marquardt method with routines provided in the ORIGIN software. 7. Troubleshooting Routine problems are expected during ITC exper- iments, most of which are easily corrected through practice. One common problem frequently observed is that the enthalpy of binding measured for the initial injection is less than that of subsequent injec- tions. This is due to ligand solution slowly leaking from the injection syringe or due to the syringe plunger not being exactly ush with the driving pis- ton. To avoid slow leakage of ligand from the injec- tionsyringesimplyreducethelengthoftimethatthe injection syringe is in contact with the macromole- cule prior to the rst injection. Some equilibration following attachment of the injection apparatus is required. It is recommended that a baseline be initi- ated following the insertion of the injection syringe into the sample cell. The signal will level off on thermal equilibration of sample, reference cell, and jacket. The experiment should be started after the baseline has leveled and remained steady for several minutes. Low binding enthalpies measured after the initial injection can be observed if the injection sy- ringe and drive piston are not exactly aligned. Align- ing the drive piston exactly ush with the injection syringe can be a difcult task. The use of a magni- fyingglasswillgreatlyaidintheproperalignmentof the injection syringe and drive piston. Fortunately, morerecentversionsoftheOMEGAinstrumenthave solved this problem by the development of an auto- matic alignment device in which an infrared beam guides the alignment of the injection syringe and drive piston. These problems can also be corrected by injecting a small volume of the ligand for the rst addition and discarding the observed data. If a pre- injection is used, the concentration of added ligand must be taken into consideration during analysis. Baseline stability is also a common problem occur- ring during ITC experiments and can arise for several reasons. As mentioned previously, a bent injection sy- ringe can lead to poor baselines. Air bubbles can also result in reduced quality of baselines and can be cor- rected by degassing solutions longer (510 min should be sufcient) as well as by taking additional care in loading the sample cell. Condensation around the adi- abatic jacket may also lead to poor baselines. For ex- periments below room temperature, the jacket must be purged with dry nitrogen prior to low-temperature equilibration. 217 THERMODYNAMICS OF PROTEINPROTEIN ASSOCIATIONS AN ITC CASE STUDY: THE INTERACTION OF HORSE HEART CYTOCHROME c WITH MONOCLONAL ANTIBODIES 2B5 AND 5F8 Numerous examples of antibodyantigen binding have been characterized by isothermal titration calo- rimetry (1114). Studies such as these have been in- valuable for understanding the thermodynamic prop- erties of antibodyantigen association. Additional strategies including alanine scanning mutagenesis in combination with ITC have demonstrated that al- though the MAb antigen interface covers a substantial surface (6501000 2 ) only a small number of residues contribute to the energetics of the interaction. In most cases, an accurate estimation of the binding constant forantibodyantigenbindingisprecluded.Despitethis limitation, accurate binding enthalpies are readily de- termined. A complete description of the thermody- namic parameters can be obtained using association constants determined from other more sensitive meth- ods. Two monoclonal antibodies 2B5 and 5F8 and their association with horse heart cytochrome (cyt) c have been studied by isothermal titration calorimetry (11). Each antibody recognizes a distinct antigenic epitope on cytochrome c. Monoclonal antibody 2B5 binds to a crevice where the covalently bound heme cofactor is partially exposed. It is known that Pro 44 in cyt c is a critical residue in the epitope, since 2B5 binding does notoccurintheabsenceofthisresidue.MAb5F8binds totheoppositesideofcytcandrequiresthepresenceof a lysine residue at position 60. Calorimetric titrations involved the addition of cyt c to the MAb contained in the sample cell (Fig. 2). Both antibodies exhibit tight associations with cyt c which precluded the determina- tion of the association constant. Fortunately, binding constants could be determined by independent experi- ments. The binding constant for the MAb 5F8cyt c interaction was determined from the association and dissociation rates ( K a 5 k on / k off ) (15). The binding con- stantforMAb2B5wasalsodeterminedexperimentally from an equilibrium titration monitoring uorescence (11). From these association constants, the free energy of binding was determined from the well-known rela- tion D G 52 RT ln K a . The entropy of binding at 25C was determined us- ing the free energy and enthalpy of binding. The free energies and enthalpies of binding for MAb 2B5 and 5F8weresimilar(Table1).Thenegativesignsindicate that the binding enthalpy contributes favorably to the free energy of binding. The decrease in entropy is as- sociatedwithconformationalrestrictionsofsidechains in the complex and contributes unfavorably to D G b . The free energies of binding of MAb 2B5 and 5F8 to cyt c are both favorable and of similar magnitude ( D G b 5 2 12.6 kcal mol 2 1 for 2B5 and D G b 52 13.9 kcal mol 2 1 for 5F8). Although the thermodynamic parameters were similar at 25C, the temperature dependences of the binding enthalpy and entropy were quite different (Fig. 3). For binding of both MAb 2B5 and 5F8 to cyt c, the Gibbs energy of binding exhibits a negligible de- pendence on temperature. Interestingly, the reasons for this effect are unique for each MAb. For MAb 5F8 cyt c binding, the observed temperature independence of D G b is due to the temperature independence of D H b and 2 T D S b . A very different situation is observed for the MAb 2B5cyt c interaction. In this case the tem- peratureindependenceof D G b isduetothecompensat- ing effects of the D H b and 2 T D S b parameters. The temperature dependence of the binding enthalpy and entropy terms leads to the observed differences in the heat capacity change on binding measured for the two MAbs (Table 1, Fig. 3). The most interesting features of 2B5 and 5F8 asso- ciation with cyt c are the differences observed in both D C p and protonation effects that occur on complex for- mation. The binding enthalpy of MAb 2B5 exhibited a strong dependence on temperature, becoming increas- ingly exothermic at higher temperatures. D C p , de- scribed by the slope of the linear dependence of D H b with temperature, was determined to be 2 580 cal mol 2 1 K 2 1 (Table 1). It should be noted that D H b may not necessarily exhibit a linear dependence on temper- TABLE 1 Thermodynamic Parameters for Association of Monoclonal Antibodies and Cytochrome c a Reaction K a (M 2 1 ) D C p (cal mol 2 1 K 2 1 ) D G 0 (kcal mol 2 1 ) D S b 0 (cal mol 2 1 K 2 1 ) D H b 0 (kcal mol 2 1 ) n H 1 MAb 2B5cyt c 2 3 10 9 2 580 2 12.6 2 28.2 2 21.0 1 0.73 MAb 5F8cyt c 1.4 3 10 10 2 172 2 13.9 2 26.3 2 21.7 1 0.02 Source. Raman et al. (11). a Values are for 0.1 M sodium phosphate, pH 7, 25C. 218 PIERCE, RAMAN, AND NALL ature in which case D C p is also temperature depen- dent. In contrast to 2B5, the enthalpy of MAb 5F8cyt c binding exhibited only a modest dependence on tem- perature ( D C p 52 172 kcal mol 2 1 K 2 1 ). These differ- ences observed in D C p values indicate probable differ- ences in the binding processes for the two MAbs to cyt c. The change in heat capacity on binding has been used to estimate the amount of polar and nonpolar surface buried on formation of the complex. Using the methods of Murphy and Freire (16) and Spolar and Record (17), the calculated apolar and polar surface was calculated for both 2B5cyt c and 5F8cyt c com- plexes (11). The two methods used gave essentially identical values of buried polar and nonpolar surface for the respective interaction. For MAb 2B5cyt c in- teraction, the amount of buried apolar surface calcu- latedwas88%oftheinterfacialsurface.Theamountof polar surface buried according to the calculation is negligible. For the interaction of cyt c with MAb 5F8, the calculated fractions of buried polar and apolar sur- facewereverysimilar.Unfortunately,intheabsenceof the crystal structure of either MAbcyt c complex, a comparison of buried surface and D C p is only specula- tive. Hibbits et al. have shown that the heat capacity change observed on hen eggwhite lysozyme (HEL) binding to the monoclonal antibody HyHel 5 is well correlated to the amount of apolar surface buried. In thisparticularcasethecrystalstructureoftheHyHel HEL complex has been determined and the interfacial surface can be well quantitated. Several groups have reported values of D C p that do not correlate with the amount of buried apolar surface (1820). The methods of calculating the amounts of apolar and polar surface from the observed D C p values have been developed from the investigation of protein unfolding transitions. In contrast to protein folding, it appears that other factors must be taken into account to correlate heat capacity changes and surface exposure in protein asso- ciation. The binding properties of the two antibodies also differedwithrespecttotheinvolvementofprotonation- linked equilibria. Simply by performing identical titra- tionexperimentsinasecondbuffersystemwithalarge enthalpy of ionization, it is possible to determine whether protonation/deprotonation events occur dur- ing complex formation (21). The enthalpy of ionization of sodium phosphate buffer is signicantly less than that of TrisHCl buffer (1 kcal mol 2 1 vs 11 kcal mol 2 1 ). In 0.1 M sodium phosphate at 25C, the binding en- thalpy measured for MAb 2B5cyt c association is 2 19.3kcalmol 2 1 .In0.1MTrisHClbufferthebinding enthalpy of 2B5cyt c association is 2 11.7 kcal mol 2 1 . Since D H b in 0.1 M TrisHCl is less exothermic it can be concluded that protons are taken up on association. The apparent binding enthalpy comprises the binding enthalpies due to association and ionization of the buffer according to D H b app 5 D H b 1 n H 1 D H i . [4] It cannot be determined whether protons are being taken up by cyt c or by MAb 2B5 but the total number of protons can be determined by the relation n H 1 5 D H b app ~ P i ! 2 D H b app (Tris) D H i P i 2 D H i Tris [5] where n H 1 is the number of protons taken up, D H b app ( P i ) is the enthalpy of binding in sodium phosphate, D H b app (Tris) is the enthalpy of binding in 0.1 M Tris FIG. 3. Temperature dependence of the thermodynamic parame- ters for binding of cytochrome c to (A) MAb 2B5 and (B) MAb 5F8. The data points measured directly by isothermal titration calorime- try are included for D H b . The heat capacity change associated with antibody binding to cytochrome c was determined by linear regres- sion analysis as the slope of the plot of D H b versus temperature. Valuesof D G b ( T )arecalculatedfromthethermodynamicparameters in Table 1 using the equation D G b ( T ) 5 ( T /298) D G b 0 1 [1 2 ( T / 298)] D H b 0 2D C p [298 2 T 1 T ln( T /298)], where D G b 0 and D H b 0 are the thermodynamic parameters under standard conditions: 0.1 M sodium phosphate, pH 7.0, 25C. D C p is assumed to be independent of temperature. The values of 2 T D S b ( T ) are calculated from 2 T D S b ( T ) 5D G b ( T ) 2D H b ( T ). Reprinted with permission from C. S. Raman, M. J. Allen, and B. T. Nall (1995) Biochemistry 34, 58315838. Copyright 1995 American Chemical Society. 219 THERMODYNAMICS OF PROTEINPROTEIN ASSOCIATIONS zo .2 10 EG Fig 0 gg ’ 5:*1** nth J? -zo ~ @1 -so zvo zso zoo soo 310 T u<> zo .2 10 B -TAsb ECI GQ 0 >~n: "éz 10 E»Z 4-; -20 " -ao zvo zso zoo soo sxo T u<> HCl, D H i P i is the enthalpy of ionization of sodium phos- phate and D H i Tris is the enthalpy of ionization of Tris HCl(21).ForMAb2B5cytcbinding n H 1 51 0.73.For MAb 5F8, D H b is identical in 0.1 M sodium phosphate and 0.1 M TrisHCl, indicating no net changes in pro- tonation during the association reaction. From the dif- ferences in protonation and in the observed D C p it is evident that MAb 2B5 and MAb 5F8 bind cyt c in distinctly different fashions. A potential candidate re- sponsible for protonation during MAb 2B5cyt c bind- ing is His-33 on cyt c. The reasons for considering His-33 are as follows. The pH chosen to study the interaction is pH 7, the p K a of histidine is near neutral pH, and His-33 lies adjacent to Pro-44 in the crystal structure of cyt c, and as previously noted, this residue is required for MAb 2B5 binding. Also, it was recently shownfortheassociationofporcinepancreaticelastase with the serine protease inhibitor turkey ovomucoid third domain that the p K a of a histidine residue in the protease is shifted from 6.7 to 5.2 (22). Given this evidence it is assumed that His-33 on cyt c is the only residue undergoing a protonation change on complex formation. The contribution of His-33 protonation to the binding enthalpy is estimated by multiplying the number of protons taken up by the enthalpy of ioniza- tion of a His side chain. Using a value of 1 6.0 kcal mol 2 1 for D H i,His (23)thecontributionoftheprotonation toward MAb 2B5cyt c binding is 2 4.4 kcal mol 2 1 . Unfortunately, protonation cannot fully explain the differences in the observed D C p values for the two antibodies. If it is assumed that protonation of His-33 is the only ionization occurring on complex formation, thenthecontributionofprotonationtoward D C p isonly 2 40 cal mol 2 1 . Clearly, this is a small contribution given the large difference in D C p observed for the two MAbs. It is evident that an enormous amount of informa- tion about the association of biological macromolecules can be determined from isothermal titration calorime- try. Recently the crystal structure of the MAb E8-cyt c complex was determined at 1.8- resolution which paves the way for correlating structure with the ener- getics of binding (24). When combined with structural information obtained from X-ray crystallography or NMR spectroscopy, an understanding of the details of the association is even further enhanced. CONCLUSIONS The primary advantage of ITC is that the observable signal is the heat evolved or absorbed on complex for- mation. The only limiting requirement for study by ITC is a measurable enthalpy change on binding. This is in contrast to a number of techniques that require modication of components with uorescent tags or require immobilization on plates. In a single ITC ex- periment, the binding constant and the stoichiometry andenthalpyofbindingcanbereadilydetermined.For complexesthatexhibitverytightafnities,thebinding constantcannotbeaccuratelydetermined.Thisrestric- tion does not limit the determination of very precise measuresofthebindingenthalpy.Althougheasilyper- formed, an ITC experiment requires great care in con- centration determination and sample preparation. As an application of the technique, the interaction of cyt c with two MAbs was observed by ITC. On the surface, the binding parameters appeared to be quite similar. Values of D H b , D G b , and D S b were nearly identical for the two antibodies at 25C. Determination of the bind- ing enthalpies as a function of temperature and buffer ionization indicated signicant differences in the modes of binding. MAb 2B5cyt c interaction was ac- companied by a large D C p and the net uptake of one proton. For the interaction of MAb 5F8 with cyt c, the D C p was small and no changes in protonation were observed. In combination with structural information, isothermal titration calorimetry provides a thorough description of the interactions of biological macromol- ecules. The information obtained from such studies should aid in the development of pharmacological com- pounds as well in the elucidation of factors that deter- mine the specicity of an interaction. ACKNOWLEDGMENTS This work was supported by grants from the National Institute of General Medical Sciences (GM32980), the National Center for Re- search Resources (RR05043), and the Robert A. Welch Foundation (AQ838). REFERENCES 1. Tanford, C. (1980) The Hydrophobic Effect: Formation of Mi- cellesandBiologicalMembranes,WileyInterscience,NewYork. 2. Nozaki, Y., and Tanford, C. (1971) J. Biol. Chem. 246, 2211 2217. 3. Makhatadze, G. I., and Privalov, P. L. (1990) J. Mol. Biol. 213, 375384. 4. Bhat, T. N., Bentley, G. A., Boulot, G., Greene, M. I., Tello, D., Dall'Acqua, W., Souchon, H., Schwarz, F. P., Mariuzza, R. A., and Poljak, R. J. (1994) Proc. Natl. Acad. Sci. USA 91, 10891093. 5. Wiseman, T., Williston, S., Brandts, J. F., and Lin, L. N. (1989) Anal. Biochem. 179, 131137. 6. Freire, E., Mayorga, O. L., and Straume, M. (1990) Anal. Chem. 62, 950959. 7. Gill, D. S., Roush, D. J., Shick, K. A., and Willson, R. C. (1995) J. Chromatogr. A 715, 8193. 8. Fisher, H. F., and Singh, N. (1995) Methods Enzymol. 259, 194 221. 220 PIERCE, RAMAN, AND NALL 9. Indyk, L., and Fisher, H. F. (1998) Methods Enzymol. 295, 350 364. 10. Hopkins, H. P., Jr., Fumero, J., and Wilson, W. D. (1990) Biopolymers 29, 449459. 11. Raman, C. S., Allen, M. J., and Nall, B. T. (1995) Biochemistry 34, 58315838. 12. Hibbits, K. A., Gill, D. S., and Willson, R. C. (1994) Biochemistry 33, 35843590. 13. Weber-Bornhauser, S., Eggenberger, J., Jelesarov, I., Bernard, A., Berger, C., and Bosshard, H. R. (1998) Biochemistry 37, 1301113020. 14. Murphy, K. P., Freire, E., and Paterson, Y. (1995) Proteins 21, 8390. 15. Raman,C.S.,Jemmerson,R.,Nall,B.T.,andAllen,M.J.(1992) Biochemistry 31, 1037010379. 16. Murphy, K. P., and Freire, E. (1992) Adv. Protein Chem. 43, 313361. 17. Spolar, R. S., and Record, M. T., Jr. (1994) Science 263, 777784. 18. Frisch,C.,Schreiber,G.,Johnson,C.M.,andFersht,A.R.(1997) J. Mol. Biol. 267, 696706. 19. Pearce, K. H., Jr., Ultsch, M. H., Kelley, R. F., de Vos, A. M., and Wells, J. A. (1996) Biochemistry 35, 1030010307. 20. Schwarz, F. P., Tello, D., Goldbaum, F. A., Mariuzza, R. A., and Poljak, R. J. (1995) Eur. J. Biochem. 228, 388394. 21. Beres, L., and Sturtevant, J. M. (1971) Biochemistry 10, 2120 2126. 22. Baker, B. M., and Murphy, K. P. (1997) J. Mol. Biol. 268, 557 569. 23. Shiao,D.D.,andSturtevant,J.M.(1976) Biopolymers 15, 1201 1211. 24. Mylvaganam, S. E., Paterson, Y., and Getzoff, E. D. (1998) J. Mol. Biol. 281, 301322. 221 THERMODYNAMICS OF PROTEINPROTEIN ASSOCIATIONS