thermodynamics of protein ligand interactions: history,...

JOURNAL OF RECEPTORS AND SIGNAL TRANSDUCTION

Vol. 24, Nos. 1 & 2, pp. 1–52, 2004

REVIEW

Thermodynamics of Protein–Ligand Interactions:

History, Presence, and Future Aspects

Remo Perozzo,* Gerd Folkers, and Leonardo Scapozza

Department of Chemistry and Applied BioSciences, Swiss Federal

Institute of Technology (ETH), Zurich, Switzerland

ABSTRACT

The understanding of molecular recognition processes of small ligands and

biological macromolecules requires a complete characterization of the binding

energetics and correlation of thermodynamic data with interacting structures

involved. A quantitative description of the forces that govern molecular

associations requires determination of changes of all thermodynamic parameters,

including free energy of binding (�G), enthalpy (�H ), and entropy (�S ) of

binding and the heat capacity change (�Cp). A close insight into the binding

process is of significant and practical interest, since it provides the fundamental

know-how for development of structure-based molecular design strategies. The

only direct method to measure the heat change during complex formation at

constant temperature is provided by isothermal titration calorimetry (ITC). With

this method one binding partner is titrated into a solution containing the

interaction partner, thereby generating or absorbing heat. This heat is the direct

observable that can be quantified by the calorimeter. The use of ITC has been

limited due to the lack of sensitivity, but recent developments in instrument design

permit to measure heat effects generated by nanomol (typically 10–100) amounts

of reactants. ITC has emerged as the primary tool for characterizing interactions

*Correspondence: Remo Perozzo, Department of Chemistry and Applied BioSciences, Swiss

Federal Institute of Technology (ETH) Zurich, Winterthurerstr. 190, CH-8057 Zurich,

Switzerland; Fax: þ41-1-6356884; E-mail: [email protected].

1

DOI: 10.1081/RRS-120037896 1079-9893 (Print); 1532-4281 (Online)

Copyright & 2004 by Marcel Dekker, Inc. www.dekker.com

Downloaded By: [University of Missouri] At: 19:42 13 May 2009

ORDER REPRINTS

in terms of thermodynamic parameters. Because heat changes occur in almost all

chemical and biochemical processes, ITC can be used for numerous applications,

e.g., binding studies of antibody–antigen, protein–peptide, protein–protein,

enzyme–inhibitor or enzyme–substrate, carbohydrate–protein, DNA–protein

(and many more) interactions as well as enzyme kinetics. Under appropriate

conditions data analysis from a single experiment yields �H, KB, the

stoichiometry (n), �G and �S of binding. Moreover, ITC experiments performed

at different temperatures yield the heat capacity change (�Cp). The informational

content of thermodynamic data is large, and it has been shown that it plays an

important role in the elucidation of binding mechanisms and, through the link to

structural data, also in rational drug design. In this review we will present a

comprehensive overview to ITC by giving some historical background to

calorimetry, outline some critical experimental and data analysis aspects, discuss

the latest developments, and give three recent examples of studies published with

respect to macromolecule–ligand interactions that have utilized ITC technology.

Key Words: Isothermal titration calorimetry; Protein–ligand interaction;

Thermodynamics.

INTRODUCTION

A fundamental principle of all biological processes is molecular organizationand recognition. Biological macromolecules are able to interact with various smalland large molecules, with a high degree of specificity and with high affinity,fascinating chemists and biologists from the very beginning of modern biochemistry.A prerequisite for a deeper understanding of the molecular basis of protein–ligandinteractions is a thorough characterization and quantification of the energeticsgoverning complex formation. Calorimetry is the only technique enabling us to studydirectly the basic physical forces between and within a macromolecule in sufficientdetail by measuring heat quantities or heat effects.

Historical Background to Thermodynamics and Calorimetry

The background of the development of calorimetry and thermodynamics hasbeen the subject of a variety of historical studies, and here we try to do a shortsummary of the most interesting aspects thereof (1–9). Calorimetry is a very oldscience. In principle, the historical development of calorimetry and thermo-dynamics began with the description and definition of temperature and heat. Thefirst known documents from the early 17th century witness for very crude attemptsto describe temperature, most of them derived by perception: ‘‘heat of a breedinghen, heat of boiling water, heat of glowing charcoal.’’ These estimations were toorough, and therefore it was necessary to develop objective standards. The inventionof the first thermometer had its origin in the same time period. The concept ofexpansion of gases and liquids due to heat was already known from the antiquity

2 Perozzo, Folkers, and Scapozza


ORDER REPRINTS

and was used by Galileo and Drebbel. They independently used a bulb with anopen-ended stem inverted over water to observe the expansion of air. The resultsobtained with these so-called thermoscops were very inaccurate, confused with theeffects of barometric pressure and lacking scaling. The next crucial step to makesatisfactory thermometers was the use of (pure) liquids instead of air (water,ethanol, mercury) in a closed compartment. By the end of the 17th century, areliable temperature scale was established by Fahrenheit and Celsius. It wasaround this time when the nature of heat and its quantitative aspects became ofinterest.

People had speculated on the nature of heat since ancient times. It was awidespread belief that heat was a substance, some held the view that it was composedof atoms. During the 18th century the foundations of calorimetry were laid byJoseph Black. He preferred an alternative explanation of heat being a fluid that canbe absorbed or squeezed out of bodies and can flow from one place to another. Herecognized that heat applied to melting ice did not change the temperature of themixture but was consumed for the solid–liquid phase transition, for the first timeclearly discriminating between the ‘‘strength’’ and ‘‘amount’’ of heat. Blackintroduced the concept of latent heat and showed that quantities of heat could beestimated from the amount of melted ice. This view brought him to the firstcalorimetric experiments with a simple phase-transition calorimeter. A warm probewas placed in the cavity of an block of ice, covered with a plate of ice and brought tothermal equilibrium. Furthermore, he adopted the idea of mixing water of differenttemperatures (mixing calorimeter) from Brooke Tylor (1723) to determine a series oflatent heats of different substances.

At the same time, A. L. Lavoisier and P. S. Laplace became interested in thetheory of heat. They considered the widespread mixing calorimeters as unsuitablebecause of several disadvantages: the need of delicate corrections for the heatcapacity of vessel and thermometer, heat loss by cooling, chemically reactingsubstances, inmiscible liquids. Moreover, this method did not allow themeasurement of the heat produced during combustion and other chemicalreactions, and during respiration, topics in which they were mostly interested.They developed the first convenient phase transition calorimeter that led toreproducing results. It was a simple but ingenious ice calorimeter, a device formeasuring heat release due to respiration and combustion (Fig. 1). The instrumentconsisted of a chamber surrounded by an ice-packed jacket, and the whole devicewas further insulated with another ice-packed jacket to improve accuracy. Theamount of water collected from the melted ice of the inner jacket was used as ameasure of the heat evolved in the chamber. The handling was difficult andexperiments could only be performed on days when the outside temperature was afew degrees above freezing. With this device, Lavoisier and Laplace determined thespecific heat of various substances and found fairly good results compared tomodern standards. The most famous experiments were conducted around 1780,when Lavoisier and Laplace measured the heat generated by a guinea pig anddetermined the amount of carbon dioxide in its exhaled air during the experiment.They compared it to heat release and carbon dioxide formation when burningcharcoal. The results were accurate enough to conclude that respiration was a formof combustion.

3


ORDER REPRINTS

Despite this interesting experimental work, the resulting interpretations of thenature of heat remained unclear. Lavoisier still treated heat as a weightless substanceand called it caloric, matter of fire. Laplace favored a mechanical explanation ofheat as motion of particles of matter, a view that emerged toward the end of the18th century out of experimental evidence provided by Count Rumford, formerlyknown as Benjamin Thompson. Rumford noticed that drilling a hole into metal tobuild a cannon barrel generated a large quantity of heat, and he noticed that therewas no limit to the amount of heat that could be produced by simple drilling. Heconcluded that heat was motion and not matter (or caloric), otherwise it had to stopwhen the cannon was running out of caloric. But there was a big controversy aboutthis theory, and it was not until the middle of the 18th century when the calorictheory was finally overthrown. The kinetic gas theory was established and theconcept of energy arose.

With the Industrial Revolution beginning in the 19th century, the nature ofmatter became of more than academic interest. With the realization that heat fromcombustion could produce work, the science of thermodynamics was born. It isconcerned with the rules governing the interconversion of energy and is able topredict the feasibility of chemical processes.

Calorimetric measuring techniques remained more or less the same duringthis time period, although there were some modifications and improvements. Inthe last few decades, calorimetric techniques have started to become of interest tobiochemists and biologists outside a few specialized laboratories. Since practically

Figure 1. The ice calorimeter of Lavoisier and Laplace (from Oeuvres de Lavoisier, Tome

Premier, Paris, Imprimerie Imperiale, 1862).



ORDER REPRINTS

every process, be it physical, chemical, or biological, is accompanied by heatchanges, it is obvious that calorimetry could serve as powerful analytical tool for avariety of applications, particularly in biological sciences. With concurrent advancesin molecular biology, expression and purification techniques, that made availablesignificant amounts of homogeneous protein, there was an increasing need for moreand reliable thermodynamic data. This gave the inputs to develop new and verysensitive calorimeters, requiring only small sample quantities and being able to detectaccurately very small heat quantities.

Since the middle of the 20th century several calorimetric principles of differentpractical design have emerged. But it is only since the last few years, with thedevelopment and improvement of sufficiently sensitive, stable, user friendly, andaffordable commercial calorimeters, that made calorimetry to become an almostroutine analytical procedure in biochemical and biophysical research. Since moderninstruments are very sensitive, detecting heat changes in the range of microcalories,requiring only 10–100 nmol of sample in a volume of 0.2–1.4mL, they are usuallydenominated as microcalorimeters.

CALORIMETRIC PRINCIPLES AND PROPERTIES

To avoid confusion in describing the principles of calorimetry and calorimetersit is useful to distinguish three important areas: the measuring principle, theoperating mode, and the type of construction (5,10–12).

Principles of Measurement

Calorimeters are instruments for quantification of heat effects. Several principlesof measurement have come into use. Amongst solution calorimeters there are twomain groups: adiabatic calorimeters and heat conduction calorimeters. With an idealadiabatic calorimeter there is no heat exchange between the calorimeter and thesurroundings, and the heat quantity Q evolved during the experiment is directlyproportional to the observed temperature change �T, and to the heat capacity " ofthe reaction vessel and its contents:

Q ¼ " ��T ð1Þ

Thus, in an experiment the heat quantity is determined by measuring thetemperature change. In an ideal heat conduction calorimeter the heat evolved isquantitatively transferred from the reaction vessel to the heat sink, a bodysurrounding the calorimeter which is usually made of metal. With this type ofcalorimeter, some property proportional to the heat flow between vessel and heatsink is measured. Normally the heat flow is recorded by placing a thermopile wallbetween the vessel and the surrounding sink. The temperature difference over thethermopile gives rise to a potential or voltage signal S that is proportional to the heatflow. The time integral for the heat flow, multiplied by a calibration constant ",

5


ORDER REPRINTS

is proportional to the heat quantity released in the experiment:

Q ¼ " �

ZS dt ð2Þ

The heat quantity is thus proportional to the area under the signal time curve.

Operating Mode and Construction Design of the Instrument

The most common type of calorimeter in use is the isoperibol calorimeter, alsocalled ‘‘constant temperature environment’’ calorimeter. The vessel is separated bythermal insulation from the surrounding thermostated bath, which formsthe isothermal jacket. The insulation is usually filled with air or vacuum.Exothermic or endothermic processes will result in a temperature change that isrecorded by a thermometer. In practice, there will always be a small heat loss fromthe vessel into the surrounding. Therefore, this calorimeters are not truly adiabatic,but quasi-adiabatic. The heat exchange cannot be neglected and must be correctedfor. Isoperibol calorimeters are very simple and for fast processes also very preciseinstruments. They are used as reaction or solution calorimeters and as combustioncalorimeters, but have not found widespread use in biochemical or biological studies.In an adiabatic shield calorimeters the reaction vessel is enclosed by an additionalthin-walled metal envelope, the adiabatic shield. It is placed in the vacuum or airspace between the reaction vessel and the thermostated bath. The temperaturedifference between the shield and the vessel is kept at zero during the experiment byautomatically applying a suitable heat effect on the shield.

Calorimeters can be in a single or a twin arrangement. Although the singlearrangement is simpler, the twin calorimeter has some advantage which makes itvery attractive for microcalorimetry. One of the calorimetric vessels, the reactioncell, contains the system of interest, whereas the other vessel, the reference cell,contains water or buffer. With such an arrangement the recorded signal is adifferential signal, of which the effects of thermal disturbances from thesurroundings are expected to cancel out.

ISOTHERMAL TITRATION CALORIMETRY

The main calorimetric techniques applied to investigate biological macromole-cules are differential scanning calorimetry (DSC) and isothermal titrationcalorimetry (ITC). DSC measures the enthalpy and heat capacity of thermaldenaturation, and researches have learned about stability of biological macro-molecule (proteins and nucleic acids) and of macromolecular assemblies (13–16).In contrast, ITC measures the heat evolved during molecular association. Thedirect thermodynamic observable is the heat associated with a binding event, i.e., aligand is titrated into a solution containing the macromolecule of interest andthe heat evolved or absorbed is detected. It allows the simultaneous determination ofthe equilibrium binding constant (KB) and thus the standard Gibbs free energychange (�G), the enthalpy change (�H ), the entropy change (�S ), as well as the



ORDER REPRINTS

stoichiometry (n) of the association event. Moreover, experiments performed atdifferent temperatures yield the heat capacity change (�Cp) of the binding reaction(17–19). As almost any interacting system is characterized by changes in enthalpy,there is a vast range of potential ITC applications.

ITC Instrumentation

A number of suppliers offer microcalorimetric instruments with sufficientsensitivity for the determination of binding reactions, but most studies published sofar have used instruments from MicroCal (Northhampton, MA, USA). With theintroduction of the commercially available Omega titration calorimeter in 1989 (19),titration calorimetry has had a broad impact throughout biotechnology, which isreflected by a large body of publications. The Omega unit has been subsequentlymodified and automated during the last decade, offering now the most sensitiveinstrument that is able to measure interaction heat effects of reactant concentrationsas low as 1–10 nmol. This type of calorimeter is based on a cell feedback thatmeasures the differential heat effects between a reference and sample cell. A constantpower applied to the reference cell activates the power feedback circuit that in turnregulates the temperature in the sample cell, thus slowly increasing the temperatureduring a measurement (typically less than 0.1�C/h). The resting power applied to thesample cell is the baseline signal.

Exothermic reactions as a result of the addition of ligand will decrease thenecessary feedback power, and endothermic reactions lead to an increase in feedbackpower. The enthalpy of reaction for each injection is obtained by integration of thedeflections from the resting baseline. Both cells are accessible by long narrow accesstubes through which samples are introduced or removed using long-needled syringes.Typically, the reference cell is filled with water and the sample cell with the system ofinterest. The ligand is applied by injection syringes with long needles having astirring paddle attached to the extreme end. The syringe is continuously rotatedduring an experiment, leading to complete mixing in the cell within a few secondsafter an injection. The mechanical heat of stirring is constant and becomes part ofthe resting baseline. With optimal performance (short equilibration time) a completebinding isotherm may be determined within 30min, although in practice it takesusually 80–100min to obtain reliable data.

It is worth emphasizing that calorimetric binding experiments are very challengingsince noncovalent binding heats are intrinsically small, typically in the range of5–10kcalmol�1, and must be liberated stepwise during the binding experiment.Furthermore, ligand addition produces additional heat effects arising from dilution andmixing, for which corrections must be made, and which are frequently comparable tothe binding heat of interest. Considering the case of a typical reaction of interestthat exhibits the heat effect of �5 kcalmol�1 in a 1–2-ml solution containing 10�7molof protein (a few mg), the experiment would liberate about 0.5mcal of heat aftercomplete saturation of all binding sites. Assuming that this heat is releasedupon 10 injections, the mean individual contribution would be 50mcal. Accuratedetection, with an accuracy of 10% or better, of such small quantities would requireinstrumental sensitivity and noise levels as low as 5mcal or less. This corresponds to

7


ORDER REPRINTS

temperature changes in the sample solution of just a few millionths of a degree andis comparable to the inevitable heat effects of dilution, mixing, and stirring. Theabsolute detection limit of the latest generation VP-ITC calorimeter, expressed asthe minimum detectable heat quantity, is reported to be 0.1mcal (20).

EXPERIMENTAL DESIGN

General Experimental Setup

The setup of an ITC experiment is largely dependent on the thermodynamiccharacteristics of the system of interest, i.e., the expected binding affinity and theheat effect of the interaction. The appropriate concentration range for themacromolecule placed in the cell depends on the binding constant of the reaction.The shape of the binding curve is dependent on the product of the binding constantKB (in M�1) and the molar concentration of macromolecule [MT] being titrated (19):

C ¼ KB½MT � ð3Þ

The sensitivity of the shape of the binding isotherm to the dimensionlessparameter C is crucial for determination of the binding constant. At high values forthe so-called C-value (C> 500), the shape of the curve approaches a step functionand becomes increasingly insensitive to changes in KB (Fig. 2). Experience showsthat conditions should be chosen to have a C-value in the range of 10–100 for anaccurate determination of KB. Therefore, to measure at these C-values, very strong

0 1 2-11

-10

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

1

C = KB[MT]

C=1000C=500

C=50C=10

C=1

kcal

/mol

of i

njec

tant

Molar Ratio

Figure 2. Simulated calorimetric binding curves illustrating the dependence of the shape of

the curve on the product of the association constant KB and the total macromolecule

concentration MT (C¼KB [MT]). The curves are simulated for several C-values (as indicated

in the plot) according to Eqs. (16) and (18) for �H¼�10 kcalmol�1. For high C-values the

binding isotherms approach a step function, becoming increasingly insensitive to changes

in KB. At low C-values, the binding curve becomes a horizontal trace that yields very little

information about KB, making it necessary to use high macromolecule concentrations to

obtain suitable binding isotherms.



ORDER REPRINTS

binding (107–108M�1) requires low concentrations of the macromolecule. Withdecreasing concentrations of the reactants, the signal arising from the interactionswill also become smaller, leading to the detection limit of ITC. This gives thetechnical limit of the highest affinity constant that can be determined.

At low C-values (C< 10), the binding curve forms a horizontal trace that againyields very little information about KB. Consequently, it is necessary to use highmacromolecule concentrations to obtain informative binding isotherms if low bindingaffinity is expected. In principle, there is no lower limit to lower binding affinity eventsthat can be determined, but in practice there are problems with solubility, stability,and often with availability of the macromolecule in order to measure in the C-valuerange of 10–100. From Fig. 2 it is evident that even at low ligand concentration, only aminor fraction of ligand is bound to the macromolecule, making it difficult to detectsufficient heat and to determine �H accurately. This often sets the limit of the lowestaffinity constant measurable in the range of 104M�1.

The correct choice of reactant concentrations depends not only on themagnitude of KB, but also on the objective of the experiment. If it is of interest tosimultaneously determine KB, �H, and n for a binding event, a complete bindingisotherm must be recorded. Considering the limiting sensitivity of 0.1 mcal, eachinjection should produce an average heat change of 1–2 mcal in the 1.4-ml cell. Fora series of 10 injections, each of 10 mL, a total Q of 20 mcal in the sample volumeare required to define a total binding curve:

Q ¼ �H½MT �V0 ð4Þ

where�H is the enthalpy of binding and V0 is the reaction volume of the sample cell.Solving Eq. (4) for [MT] predicts a minimum concentration of about 1.4 mM fora protein with a �H of �10 kcalmol�1 needed to generate a complete bindingisotherm to yield n, KB, and �H. According to Eq. (3), an arbitrarily chosen KB of106M�1 would result in a very low C-value of 1.4. In practice it will be necessary toincrease the protein concentration to perform the experiment in the ideal range forC-values of 10–100. As a desirable side effect the heat signals will become larger. Ifthe same calculation were done using higher affinity (107M�1), the resulting C-valueof 14 would be sufficient for a complete deconvolution of the binding isotherm.Taken together, the almost 10-fold increase in sensitivity achieved from first to thirdgeneration microcalorimeters allows now to measure higher affinity (up to 109M�1)faster, more accurately, and with less material.

For some applications it is preferred to obtain �H not as a fitting parameter,but to directly determine �H very accurately. In this case it is common practice tomeasure �H at concentrations when the binding partners are fully associated andthe saturation is still low, i.e., full association at partial saturation (21). At theseconditions (C-value>100), the amount of heat released or absorbed is directlydetermined by the amount of ligand injected:

Q ¼ �H½LT �Vinj ð5Þ

where [LT] is the concentration of the ligand solution in the syringe, and Vinj is theinjection volume. In principle, this allows the determination of �H with a singleinjection. In practice �H will be determined as the mean of a number of injectionsfrom the same experiment.

9


ORDER REPRINTS

To obtain high quality data, an appropriate protocol has to be established byoptimizing ligand and protein concentrations and the injection volume. Typically, theligand concentration is much higher since several equivalents must be added to thesample cell. The titration experiment should be planned to approach or reachcomplete saturation of the binding sites at the end of the experiment. To generate asufficient number of data points, which will improve data analysis, the ligand has to beadded in small aliquots. However, the heat signal should not become too small tomaintain high precision of each data point. If the interaction heat is small, it will benecessary to choose larger injection volumes. These prerequisites define the titrationprotocol, and it is up to the experimentator to find the ideal compromise. As a rule ofthumb, 25 injections, each of 5 mL, of a ligand solution with a concentration 25 timeshigher than that of the protein solution will result in an adequate binding isotherm. Ifthe ligand is poorly soluble, it is possible to place it in the sample cell and to inject themacromolecule. As long as the binding stoichiometry is 1:1, either interactingmolecule can act as the titrant without adjusting the binding model. For morecomplicated cases where this assumption does not hold, the model must be modifiedaccordingly (22).

The time between successive injections is another important parameter. Ifassociation is rapid, the instrument baseline will be equilibrated in a short time,depending on the response time of the calorimeter. Under such conditions 3–4minare sufficient to reach baseline again after injection. In contrast, heat signals of slowprocesses require much more time to reach thermal equilibrium. Several other issuesrelated to experimental design should be mentioned. It is crucial that solutions ofligand and macromolecule are pure and exactly match with respect to pH, buffercapacity, and salt concentration. This means that macromolecule and ligand arepreferably dissolved in the same buffer. To achieve this goal, it is good practice todialyze the protein prior to the experiment and dissolve the ligand in the dialysisbuffer. This procedure will prevent spurious heat effects resulting from mixing ofdifferent buffers. Both interacting components, often purified from biologicalsource, must be free of contaminating enzymatic activity that could affect theassociation event under investigation. Furthermore, the formation of air bubbles hasto be avoided. Thus it is very important to thoroughly degas all solutions prior to theexperiment. Any air in the syringe can cause variation in the injected volume or leadto additional heat signals, and bubbles in the sample cell interfere with the thermalcontact of solution and cell wall. Finally, in most experiments the heat effect of thefirst injection of a series of injections is obviously too small. This results fromdiffusion while equilibrating the system. Even if care is taken to avoid this leakage,the problem may persist. Therefore it is common practice to make a small firstinjection of 1 mL and then to remove the first data point before data analysis.

Control Experiments

Isothermal titration calorimetry not only measures the heat released or absorbedduring binding reactions, but it detects the total heat effect in the calorimetric cellupon addition of ligand. Thus, the experimental heat effect contains contributionsarising from nonspecific effects, such as dilution of ligand in the buffer, dilution of



ORDER REPRINTS

the protein sample, heat of mixing, temperature differences between the cell and thesyringe, and mixing of buffers of slightly different composition. These contributionsneed to be determined by performing control experiments in order to extract the heatof complex formation. This would need at least a further three titrations to measurethese effects (ligand into buffer solution, buffer into protein solution, buffer intobuffer solution). In practice however, the latter contributions are found to be smalland frequently negligible, whereas the heat of ligand dilution may be significant andneeds to be corrected for. Alternatively, if the titration experiment is designed toensure complete saturation of the enzyme before the final injection, and if the blankexperiments mentioned above show the heat of ligand dilution to be concentration-independent, then the nonspecific heat effects can be estimated very well byaveraging the small heats at the end of the titration.

Evaluation of Protonation Effects

Whenever binding is coupled to changes in the protonation state of the system,the measured heat signal will contain the heat effect due to ionization of buffer. If thebinding event changes the protonation state of free or bound ligand as well as of freeor complexed macromolecule, proton transfer with the buffered medium occurs. Asa consequence, the heat of protonation/deprotonation will contribute to the overallheat of binding and �Hobs will depend on the ionization enthalpy of the buffer(�Hion). Repeating the calorimetric experiment at the same pH in buffers of different�Hion allows to determine the number of protons nHþ that are released (nHþ> 0) ortaken up (nHþ< 0) by the buffer, and thus to calculate the intrinsic binding enthalpy,�Hbind, corrected for protonation heats:

�Hobs ¼ �Hbind þ nHþ�Hion ð6Þ

In practice, it is recommended to perform ITC experiments in a series of buffersof different ionization heats under otherwise the same conditions. Values of �Hion

have been described (23,24) or can be determined by ITC (25). If the same �Hobs isobserved, there is no protonation event coupled to binding. Deviations of �Hobs

with different buffers point to a protonation event, and the intrinsic enthalpy ofbinding (�Hbind) can be obtained from the intercept (�Hion¼ 0) of the regressionline described by Eq. (6) (26).

Although it would seem that such buffer effects would mean a nuisance, it is oneof the most powerful means to investigate binding mechanisms, and this propertycan be exploited to increase the signal strength of an otherwise undetectable bindingevent by simply changing the buffer system to a different pH and to buffers with highionization enthalpies (25).

DATA ANALYSIS

The signal monitored by ITC is the differential power applied to the sample cell.The total heat released or absorbed upon an injection of ligand into the cell

11


ORDER REPRINTS

corresponds to the area under the signal vs. time curve (Fig. 3, panel A). Theinstrumental baseline and other unspecific heat effects must be carefully subtractedfrom the raw data. Experimental data can be presented as a sigmoid plot (differentialmode) or as a hyperbolic saturation curve (integral mode). The differentialmode treats each injection as an independent point and is plotted as heat evolvedper injection vs. total ligand concentration or the ratio of the total ligandconcentration to the concentration of macromolecule (Fig. 3, panel B). In theintegral mode, the total cumulative heat is plotted against the total ligandconcentration. Fitting a binding model to the calorimetric data plotted in eithermode yields equivalent results. Generally, random errors tend to cancel out in theintegral mode, whereas systematic errors tend to be amplified. Comparativestatistical analysis of both modes can give information about the accumulation ofsystematic errors (27,28).

Ligand Binding in Titration Calorimetry

There are many techniques available to measure binding constants (KB).Equilibrium dialysis, radio-ligand binding assays or ultracentrifugation directly

-8

-6

-4

-2

0

-10 0 10 20 30 40 50 60 70 80 90 100

A

Time (min)

µcal

/sec

0.0 0.5 1.0 1.5 2.0 2.5-14-12-10-8-6-4-20

B

Ribonuclease ASingle site binding modelKB(105M−1) 2.51 ± 0.029

∆H (kcal/mol) −12.98 ± 0.02n 0.953 ± 0.001

Molar Ratio [2´CMP]/[RNASE A]

kcal

/mol

e of

inje

ctan

t

Figure 3. Calorimetric data for the exothermic binding of cytidine 20-monophosphate

(20CMP) to ribonuclease A (RNase A) at pH 5.5 (0.2M K-acetate, 0.2M KCl) and 28�C

(figure kindly provided by Dr. I. Jelesarov). A: raw data obtained for 25 automatic injections

of 5mL. Concentrations of RNase A and 20CMP are 0.145mM and 3.72mM respectively. The

area of each peak represents the total heat evolved upon addition of a single aliquot of 20CMP.

B: titration plot derived from the integrated heats of binding, corrected for heats of dilution.

The solid line represents the nonlinear best fit to the data assuming a single-site binding model.



ORDER REPRINTS

yield concentrations for the macromolecule [M], ligand [L], or complex [ML]to calculate KB. Spectroscopic methods are more indirect by detecting an observablep, the change of which is proportional to the degree of saturation (18,28–30).ITC is the most direct method to measure the heat change on complex formation.In general, a simple reversible association between a macromolecule M and aligand L,

M þ L $ ML ð7Þ

is characterized by its binding constant KB:

KB ¼½ML�

½M�½L�ð8Þ

The observable response of an ITC experiment is the heat change associatedwith each addition of ligand. For each injection, the heat released or absorbedis directly proportional to the total amount of formed complex. This can beexpressed by

q ¼ V0 �H�½ML� ð9Þ

where q is the heat associated with the change in complex concentration, �[ML],�His the molar enthalpy of binding, and V0 is the reaction volume of the sample cell.

In a calorimetric experiment, each addition of ligand gives rise to a heat changedepending on the reaction volume, concentrations, molar enthalpy, binding constant,heat of dilution, stoichiometry, and the amount of previously added ligand. As theconcentration of unoccupied binding sites begins to decrease, the heat changesdecrease correspondingly as ligand is added. The total cumulative heat after the ithaddition, Q, will be

Q ¼ V0�HX

�½ML�i ¼ V0�H½ML�i ð10Þ

where [ML]i is the total concentration of complex after the ith injection.Evaluation of microcalorimetric data requires the consideration of the

observable response in terms of total ligand added or the total ligand concentration.Therefore, the binding equations must be expressed as a function of total ligand andmacromolecule concentration:

½MT � ¼ ½ML� þ ½M� ð11Þ

½LT � ¼ ½ML� þ ½L� ð12Þ

where [MT] and [LT] are total macromolecule and ligand concentrations,respectively, and [M] and [L] are free concentrations of macromolecule and ligand,respectively. [ML] is the concentration of the formed complex.

Single Set of Independent Sites Model

In the simplest case of ligand binding, each macromolecule consists of only onetype of binding sites with a finite number of identical noninteracting binding sites,

13


ORDER REPRINTS

all of which exhibiting the same intrinsic affinity for the ligand. For such a system,the binding constant KB is given by

KB ¼�

ð1��Þ½L�ð13Þ

where � is the fractional saturation and [L] is the concentration of free ligand. It is

related to the total ligand [LT] and macromolecule concentration [MT], by massconservation:

½L� ¼ ½LT � � n�½MT � ð14Þ

Combining Eqs. (13) and (14) gives the quadratic equation

�2 �� 1þ1

nKB½MT �þ

½LT �

n½MT �

� �þ

½LT �

n½MT �¼ 0 ð15Þ

whose only meaningful root is

� ¼1

21þ

1

nKB½MT �þ

½LT �

n½MT ��

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1þ

1

nKB½MT �þ

½LT �

n½MT �

� �2

�4½LT �

n½MT �

s0@

1A ð16Þ

The integral heat of reaction Q after the ith injection is given by

Q ¼ n½MT �V0�H�i ð17Þ

where V0 is the cell volume and �H is the molar heat of ligand binding. The

differential heat of the ith injection is

qi ¼ n½MT �V0�H �i ��i�1ð Þ ð18Þ

A nonlinear fit based on Eq. (17) to the hyperbolic saturation curve in theintegral mode (Q vs. [LT]) yields the parameters KB, �H, and n from a singleexperiment. Based on Eq. (18), the titration data can be fitted to the sigmoid

saturation curve in the differential heat mode (qi vs. [LT], or vs. [LT]/[MT]). The sameparameters are obtained.

Complex Binding Models

Similar relationships as described above exist for other models, i.e., a model formultiple sets of independent binding sites, single set of interacting sites (cooperative

sites), multiple sets of interacting binding sites. By use of statistical thermodynamictreatment it is possible to deconvolute a binding isotherm of such complex systems(31–33). Instructive examples from the literature demonstrate the strength of this

approach (34–38). However, the success strongly depends on the quality andreliability of the experimental data.



ORDER REPRINTS

BASIC THERMODYNAMIC RELATIONSHIPS

The binding enthalpy of protein–ligand interactions can be determinedaccurately by means of ITC. The association constant KB is related to the Gibbsfree energy �G by the well-known relation

�G ¼ �RT lnKB ð19Þ

where R is the universal gas constant and equals to 1.987 calK�1mol�1 and T is thetemperature in degrees Kelvin. �G is again composed of an enthalpy term (�H) andan entropy term (�S), related by another fundamental equation:

�G ¼ �H � T�S ð20Þ

The Gibbs free energy is temperature dependent and is described by

�GðTÞ ¼ �HðT0Þ þ

Z T

T0

�Cp dT � T�SðT0Þ �

Z T

T0

�Cp dlnT ð21Þ

where �Cp is the heat capacity change and T0 is an appropriate referencetemperature. With �Cp being independent of temperature in the range of interest,Eq. (21) simplifies to

�GðTÞ ¼ �HðT0Þ � T�SðT0Þ þ�Cp T � T0 � T lnT

T0

� �ð22Þ

Equation (22) shows that enthalpy and entropy changes are dependent ontemperature through the heat capacity change �Cp:

�HðTÞ ¼ �HðT0Þ þ�CpðT � T0Þ ð23Þ

�SðTÞ ¼ �SðT0Þ þ�Cp lnðT=T0Þ ð24Þ

In a thermodynamic analysis the goal is to determine �G, �H, �S, and theirtemperature dependence by �Cp, since these four parameters provide a fulldescription of the energetics governing molecular interactions.

STRATEGIES FOR MEASURING

TIGHT BINDING AFFINITY

High-affinity binding constants for protein–ligand interactions are inherentlydifficult to measure. With increasing affinity, it becomes necessary to work at lowconcentration of macromolecule, leading to difficulties in detecting the signal specificto the analytical method. In the case of ITC, the largest binding constant that canbe measured reliably, approaches 109M�1 for a typical macromolecule–ligandinteraction (18,19,39).

The thermodynamic approach offers a powerful advantage for measuring tightbinding affinities and thus Gibbs free energy changes (�G). Free energy changes arestate functions, i.e., their values are defined by the initial and final thermodynamicstates, regardless of the pathway connecting the two states. This being the case, it is

15


ORDER REPRINTS

possible to determine the binding constant for a protein–ligand interaction under

different conditions that allow measuring the affinity. The result can be corrected to

other conditions of more physiological relevance, if it is known by what parameters

the binding free energy is linked with the conditions being varied. In principal, any

change of physical or chemical conditions that influence ligand association to a

macromolecule is useful, but in most cases linkage of binding with pH andtemperature are exploited.

Linked Protonation Effects in Ligand Binding

Often molecular interactions are very tight (>109M�1), and are not accurately

measurable even with the most sensitive calorimeters available. Generally, molecular

interactions occur to some degree in dependence of pH, reflecting the linkage

between the association of a ligand and the binding of protons (proton linkage). The

molecular basis of the linkage is the result of alterations of pKa values of ionizable

amino acid groups concomitant with binding.If ligand binding is coupled with uptake of a single proton (Fig. 4), the observed

ligand binding constant Kobs is given as

Kobs ¼ Kint1þ Kc

P10�pH

1þ KfP10

�pHð25Þ

where Kint is the intrinsic binding constant, KcP and K

fP are the proton binding

constants for the complex and free form of the protein and are equal to 10pKa,c and

10pKa, f, respectively, of the ionizing group (40,41). According to Eq. (25) proton

linkage can be viewed as change in proton affinity, thus protons will either be

released or absorbed due to ligand binding.If proton transfer occurs during binding, the �Hobs is determined by the

ionization enthalpy of the buffer and the enthalpy of binding corrected for

buffer effects see according to Eq. (6), and both the number of protons (nHþ) and

M++

M M:L

M:L+

Kpc

Kint

Kpf

Figure 4. Scheme for proton binding linked to binding of a ligand L to a macromolecule M.

Ligand binding reactions are shown in horizontal direction whereas proton binding occurs

in vertical direction (see text for details).



ORDER REPRINTS

the intrinsic binding enthalpy (�Hbind) will vary as a function of pH. Thus nHþ isgiven by

nHþ ¼ f c � f f ð26Þ

where f c and f fare the fractional saturation of protons at a given pH of the boundand free protein. In case of a single protonation event, f c and f f can be expressed as

f c ¼Kc

P10�pH

1þ KcP10

�pHð27Þ

and

f f ¼K

fP10

�pH

1þ KfP10

�pHð28Þ

The change in the number of protons bound by the protein upon binding of theligand is the difference between Eqs. (27) and (28):

nHþ ¼ f c � f f ¼Kc

P10�pH

1þ KcP10

�pH�

KfP10

�pH

1þ KfP10

�pHð29Þ

Equation (29) clearly shows that at a minimum of two pH values pKa,c of thecomplexed (Kc

P ¼ 10pKa,c) and pKa,f of free protein (KfP ¼ 10pKa,f ) can be calculated

by simultaneously solving Eq. (29) for nHþ determined at the corresponding pHvalues, even when the ligand affinity is too tight to be measured (40). In practice, nHþ

is determined in a series of buffers of different ionization enthalpies as a function ofpH. Equation (29) is used to determine the pKa of the protein in the free andcomplexed state. With these values KB at the tight binding conditions can becalculated by Eq. (25) (40,41). In general this treatment can be applied to morecomplicated systems that involve two protons linked with binding (31,41). A similarapproach based on free energy of binding linked to the number of protonstransferred as a function of pH has been developed (39).

The power of the calorimetric approach in evaluating proton linkage lies in thefact that �H can be determined with high precision under conditions where KB isnot measurable, and thus the contributions of the linkage.

Thermodynamic Linkage to Temperature

The fundamental Eqs. (19) and (20) demonstrate that the equilibrium constantfor a process is related to the standard entropy and enthalpy changes, and to theabsolute temperature. The temperature dependence of the changes in free energy(�G) of the Gibbs-Helmholtz equation for a thermodynamic system is described as

� �G=Tð Þ

�T

� �¼ �

�H

T2ð30Þ

where T is the absolute temperature and �H is the reaction enthalpy. Substitution ofEq. (30) with Eq. (19) yields the familiar van’t Hoff equation:

� lnKB

�ð1=TÞ¼

��H

Rð31Þ

17


ORDER REPRINTS

where KB is the binding constant. The temperature dependence of KB is commonlyanalyzed by means of the van’t Hoff plot, whose basis is Eq. (31). Measuring KB overa temperature range and plotting lnKB vs. 1/T yields the van’t Hoff enthalpy(�HvH). It is calculated from the slope of the plot, according to Eq. (31).

When �H is known, then integration of Eq. (31) gives the temperaturedependence of the equilibrium constant:

KBðTÞ ¼ KBðT0Þ exp��H

R

1

T�

1

T0

� �� ð32Þ

Equation (32) explicitly assumes that �H is constant over the temperature rangeT–T0. It has been shown for many biological protein–ligand interactions that thisassumption is not valid, with �H often being temperature dependent.

The temperature dependence of the binding enthalpy, �Cp, is described byEq. (23), and combination with Eq. (32) leads to the extended form of the van’t Hoffequation that accounts for the temperature dependence of �H:

KBðT Þ ¼ KBðT0Þ exp ��H T0ð Þ

R

1

T�

1

T0

� �þ�Cp

Rln

T

T0þT0

T� 1

� �� ð33Þ

where KB (T) is the binding constant to be calculated at temperature T, KB (T0) isthe binding constant experimentally determined at temperature T0, and �H(T0) isthe experimental enthalpy change at temperature T0.

In cases where the affinity is too tight to be measured directly by ITC underambient conditions, the binding constant becomes accessible by changing to atemperature whereby binding is lowered. After determining �H and the temperaturedependence thereof, the binding constant can easily be calculated for thetemperature of interest (42).

Displacement Experiments

As an alternative approach for measuring tight binding constants, a displace-ment experiment can be carried out. The protein of interest is presaturated with amore weakly binding ligand whose binding parameter can be determined directly,and this ligand is displaced by injecting a ligand that binds more strongly. The firstligand will compete with the second and thus reduce the apparent binding constant(43–46).

A first experiment yields the thermodynamic parameters for the first ligand A(�H1,K1), the second titration gives apparent values for the second ligand B(�Hobs

2 ,Kobs2 ). The observed association constant for binding of B ðKobs

2 Þ in thepresence of A is given by:

Kobs2 ¼

K2

1þ K1½A�ð34Þ

The observed binding enthalpy (�Hobs2 ) is given by:

�Hobs2 ¼ �H2 ��H1

K1½A�

1þ K1½A�ð35Þ



ORDER REPRINTS

Thus, from knowledge of �Hobs2 and Kobs

2 the affinity K2 and binding heat �H2

for the tight binding ligand B can be deduced using Eqs. (34) and (35). Bothequations assume that the concentration of B does not change during themeasurment. This can be achieved when B is also present in the syringe solution.However, it is also possible to include dilution effects into the fitting procedure. Anexact mathematical treatment for the analysis of competition ligand binding usingdisplacement ITC has recently been published (47).

A restriction to the displacement method occurs when both ligands bind withexothermic �H, as the tight binder will be reduced by the endothermic contributionof the dissociation of the first ligand. Therefore, it is necessary that the interactionenthalpies for both ligands differ significantly.

INFORMATIONAL CONTENT OF ITC DATA

The gain of knowledge through thermodynamic data of binding reactions islarge and can have a dramatic impact on characterizing the molecular mechanism ofbinding. The thermodynamic profile of an interaction process reflects various typesof forces that drive binding, including enthalpic contributions of bond formation,entropic effects such as restrictions of degrees of freedom, the release and uptake ofwater and ion molecules, the burial of water-accessible surface area and changes invibrational content.

As outlined before, calorimetry measures a global property of a system, and thusreflects the sum of all concomitant phenomena which must be carefully analyzed andquantified in order to yield parameters conforming to the binding event properly.Moreover, the quality of the deconvoluted parameters depends on the appropriatemodel used. Nevertheless, once reliable data are available, the informational contentof thermodynamic data can have an intriguing impact on the characterization ofmolecular mechanisms of binding.

Binding Free Energy

The Gibbs free energy of binding is the most important thermodynamicdescription of binding, since it determines the stability of any given biologicalcomplex, and it has been (and still is) a useful analytical tool for the phenom-enological characterization of structure–function relationships. The typical analysisof calorimetric data involves fitting an appropriate model to the data, i.e., simplesingle-site or two-site binding model, which yields the binding constant. But often,the system under investigation exhibits a more complex behavior and moresophisticated models must be applied (multiple interacting-site models). Themacromolecule may undergo ligand-induced changes, be they conformationaladaptations in the binding site or self-association of the receptor, which willcontribute to the total free energy of binding. These effects should be evaluatedindependently with companion methods, i.e., analytical ultracentrifugation,analytical gel filtration chromatography, and spectroscopic methods.

19


ORDER REPRINTS

In several recent publications it has been proposed to dissect binding free energyinto several contributing terms. The total binding free energy contains a contributiontypically associated with the formation of secondary and tertiary structure (van derWaals interactions, hydrogen bonding, hydration, conformational entropy),electrostatic and ionization effects, contributions due to conformational transitions,loss of translational and rotational degrees of freedom, and others that mustbe accounted for on an individual basis (48–55). For example, an observed �G canbe the same for, an interaction with positive �S and �H (binding dominated byhydrophobic effect) and an interaction with negative �S and �H (when specificinteractions dominate). Moreover, interacting systems tend to compensate enthalpicand entropic contributions to �G, making binding free energy relative insensitive tochanges in the molecular details of the interactions process (55–57). Thus,consideration of �H and �S are crucial for a detailed understanding of the freeenergy of binding.

Binding Enthalpy

The observed heat effect of a binding reaction is a global property of the wholesystem under investigation, reflecting the total heat change in the calorimetric cellupon addition of ligand. On the one hand, the measured heat contains contributionfrom unspecific heat effects, on the other hand, there might be protonation effectscoupled to binding. Therefore it is of outmost importance to determine possiblecontributions to the intrinsic binding enthalpy and to correct for them. But evencorrected heat changes are composed of different contributions, which is the reasonwhy the enthalpy is an apparent or observed (�Hobs) quantity (29).

At first appearance, the physical meaning of �H seems to be simple: itrepresents the changes in noncovalent bond energy occurring during the interaction.This interpretation is too simple to describe observed �H values. The measuredenthalpy must be the result of the formation and breaking of many individual bonds,since it is barely conceivable to form bonds without breaking any others, especiallyin aqueous medium. The enthalpy change of binding reflects the loss of protein–solvent hydrogen bonds and van der Waals interactions, formation of protein–ligandbonds, salt bridges and van der Waals contacts, and solvent reorganization nearprotein surfaces. These individual components may produce either favorable orunfavorable contributions, and the resultant is likely to be smaller than the specificinteractions (30).

From calorimetric studies carried out in water and D2O it was concluded that alarge part of the observed enthalpy change is due to bulk hydration effect (58,59).Often, water molecules are placed in the complex interface, improving thecomplementarity of the complex surfaces and extending H-bond networks. Thiscan make enthalpies more favorable, but is often counterbalanced by an entropicpenalty (60–62). The role of interfacial water was directly examined by loweringwater activity by means of glycerol or other osmolytes. Complexes with a low degreeof surface complementarity and no change in hydration are tolerant to osmoticpressure (25,63–66).



ORDER REPRINTS

Besides the unspecific hydration effects, all direct noncovalent bonds at thebinding interface contribute to �H, actually reflecting the binding enthalpy in astrict sense. The dissection of each noncovalent interaction is very difficult since thenet heat effect of a particular bond is the balance between the reaction enthalpy ofthe ligand to the macromolecule and to the solvent. Moreover, structural alterationsat the binding site due to the binding event may contribute to the binding enthalpy.Several mutational approaches have been applied to investigate the energetics ofindividual bonds: alanine scanning mutagenesis (67), removal of particular H-bondsat the active site (68), construction of double mutant cycles (69). However, all theseapproaches suffer from the problem that a direct relation between the change in �Hand the removal of the corresponding specific contact in the active site cannot bemade a priori. On a theoretical basis it has been argued that decomposition of �H isnot possible (70), but others favor a dissection into specific contributions (71,72).

Binding Entropy

The entropy of binding is directly calculated from �G and �H according toEq. (20). In general, it represents all other positive and negative driving forces thatcontribute to the free energy. Recently, it has been proposed that the total entropychange associated with binding can be expressed as the sum of several contributingeffects. The main factor contributing to �S of complex formation is due tohydration effects. Since the entropy of hydration of polar and apolar groups is large,the burial of water-accessible surface area on binding results in solvent release whichcontributes often large and positive to the total entropy of interaction. Anotherimportant, though unfavorable contribution reflects the reduction of rotationaldegrees of freedom around torsion angles of protein and ligand side-chains. Anadditional entropy term accounts for the reduction of the number of particles insolution and their degrees of freedom (73–75).

A negative entropy change can entail different contributions, and it does notnecessarily indicate increased or unchanged hydration interfaces, but a positiveentropy change is a strong indication that water molecules have been released fromthe complex surface (76).

Heat Capacity Changes

If �H is determined at a range of temperatures (modern ITC instruments allowmeasurements between 2 and 80�C), the change in the constant pressure heatcapacity (�Cp) for an interaction is given by the slope of the linear regressionanalysis of �Hobs plotted vs. temperature. Often �Cp does not depend ontemperature within the small physiological temperature range, although severalpublications reported weak or strong dependencies (37,38).

For the binding reaction, �Cp is almost always negative when the complexedstate of the macromolecule is taken as the reference state. Its origin lies in the factthat there is strong correlation between �Cp and the surface area buried on forminga complex (77–81). It has been shown that the removal of protein surface area from

21


ORDER REPRINTS

the contact with solvent results in a large negative �Cp. The basis of this observationis the different behavior of solvent on the surface of a macromolecule and to that inthe bulk, particularly with respect to water molecules interacting with hydrophobicsurfaces. This means that for any process, in which water is released from thesurface, �Cp will be substantial and would be proportional to the amount of surfaceinvolved. Through this correlation of �Cp and burial of surface area, the heatcapacity provides a link between thermodynamic data and structural information ofmacromolecules. Therefore in the last years, there has been considerable progressin the parameterization of all thermodynamic parameters and the predictionsthereof (82,83).

Enthalpy–Entropy Compensation

In most thermodynamic binding studies of biological systems, the phenomenonof enthalpy–entropy compensation has been described (56,57). It is characterized bythe linear relationship between the change in enthalpy and the change in entropy,i.e., favorable changes in binding enthalpy are compensated by opposite changes inentropy and vice versa, resulting in small changes in binding affinity over a range oftemperature. It is assumed that enthalpy–entropy compensation is connected to theproperties of the solvent (water), and it appears also to be a general consequence ofperturbing weak intermolecular interactions (84,85). This is in agreement with thefact that increased bonding in a binding process, resulting in more negative �H, willbe at the expense of increased order, leading to more negative �S. As both �H and�S are connected to �Cp, it is not surprising that both parameters are correlated.

In terms of medicinal chemistry and rational drug design, this phenomenon is adifficult problem to overcome in order to increase binding affinity of a compound tothe protein of interest. The ideal optimization strategy requires the implementationof enthalpic or entropic contributions that result in a minimal entropic or enthalpicpenalty. This will induce the largest change in �G, thereby defeating the deleteriouseffects of enthalpy–entropy compensation at the thermodynamic level.

Binding Stoichiometry

The determination of binding stoichiometries (n) is of central importance for thecharacterization of binding mechanisms of biological macromolecules. ITC hasemerged as an important tool, since it enables high-precision analysis with highreproducibility because of the computer-controlled injection of definite volumes.Assuming that the concentrations of both interacting species are known, the bindingstoichiometry can be determined from the molar ratio of the interacting species atthe equivalence point. During fitting procedures the parameter n can either be fixedas equal to the number binding sites per macromolecule, or it can be treated as anadditional floating parameter that is determined by iterative fitting. There are anumber of possible sources which lead to deviation from the expected values of n,i.e., high experimental uncertainty of the data set, error in concentration of either theligand or the macromolecule, unspecific binding, degradation of ligand, and low



ORDER REPRINTS

protein quality (unfolded, missfolded). If these systematic errors can be ruled outand the fitted values for n still deviate significantly from expected values known fromadditional independent information, they should be reexamined. It has beenproposed to fit the data to the hyperbolic equation and then to reanalyze them bymeans of a double reciprocal plot (30). However, it is recommended to corroboratestoichiometric data obtained by microcalorimetry with additional independentinformation.

Once the model is verified, ITC can be used as an excellent quality control toolfor the analysis of the fractional binding activity of different lots of protein (e.g.,antibodies), for stability testing (freeze-thaw) and many more (86).

PREDICTION OF BINDING ENERGETICS

A long-standing goal of biophysical chemistry is the prediction of bindingenergetics from the 3D-structure of protein–ligand complexes, and it is a key elementin the field of structure based drug design. The rapidly increasing availability of high-resolution protein structures from X-ray crystallography and nuclear magneticresonance (NMR) opened the field to combine structural information withthermodynamic data of the binding process.

In the past few years, considerable progress has been made in characterizingmolecular aspects of protein–ligand interactions by means of thermodynamic data.It has been shown that the major polar and apolar contributions to the enthalpy,entropy and heat capacity changes for protein folding and unfolding can bedescribed in terms of changes in solvent-accessible polar and apolar surface area(77,80,81,87–91).

The empirical parameterization based on calculation of changes in solventaccessible surface areas was first applied to the prediction of protein foldingenergetics (77,80,81,87,89,91). Since the atomic interactions involved in associationsreactions are similar (92), it was suggested to apply this approach to protein–proteininteractions, peptide binding to proteins, and to small ligand binding. Theparameterization has now reached the state in which accurate prediction of proteinfolding energetics and binding energetics is possible (73,79,93,94).

Solvent-Accessible Surface Areas

According to Lee and Richards (95), the solvent-accessible surface area (ASA) isdefined as the surface traced out by the center of a solvent probe (frequently taken asa sphere with a radius of 1.4 A) as it moves over the surface of the protein. There area number of implementations described and used to determine ASA (95,96), and it isvery important to recognize that each implementation yields slightly different results.The original description of the predicting parameters is based on the Lee andRichards algorithm as implemented in the program ACCESS, using a probe radiusof 1.4 A and a slice width of 0.25 A, but there exist reparameterized values for otherimplementations (82). When performing calculations, it must be assured that theappropriate parameters are used, because they are dependent on the algorithm used.

23


ORDER REPRINTS

Calculation of Thermodynamic Parameters

In general, changes in solvent-accessible surface area (�ASA) are determined asthe difference of ASA of the final state and of the initial state. For a molecularinteraction process, this is the difference between the ASA of the complex and thesum of the ASA of the macromolecule and the ligand, resulting in negative values of�ASA. �ASA is further subdivided into nonpolar and polar contributions bysimply defining which atoms take part in the surface. Oxygen, nitrogen and sulfurare treated as polar and all carbon atoms as apolar. If structured water moleculestake part in the binding interface, they must be accounted for, since their presencewill contribute to the amount and type of surface area buried (82,97).

In the ideal case, structural information will be available for the complex and forboth interacting species, i.e., the free macromolecule and the ligand. This willaccount for any structurally defined conformational differences that occur onbinding. If there are no conformational changes linked with the association (rigidbody binding), it is good practice to extract the free state of the protein by removingthe coordinates of the ligand from the complex. However, the assumption of rigidbody binding must be verified by other independent methods because any structuralrearrangement will contribute to the energetics. For the case where binding is linkedto order/disorder of particular regions which are not defined in the coordinate file ofthe macromolecule, it may be necessary to add a model of the region. This isespecially found in protein peptide interactions, for which it is necessary to definea solution structure of the free peptide (73,79).

Calculation of DCp

The largest contribution to the �Cp for a binding process arises fromdehydration of protein and ligand surface with negative contributions due toburial of apolar surfaces (�ASAnp) and positive contributions due to burial of polarsurfaces (�ASAp). From studies of the dissolution of solid model compounds,the following relationship has been proposed (77,80):

�Cp ¼ �cnp�ASAnp þ�cp�ASAp ð36Þ

The parameters �cnp and �cp that are suitable for calculations based on ASAsdetermined by different programs are given in Table 1. Equation (36) is used topredict the �Cp for ligand binding.

Calculation of DH

The change in enthalpy of binding reflects the loss of protein–solvent hydrogenbonds, van der Waals interactions, and formation of protein–ligand bonds, saltbridges and van der Waals contacts, and solvent reorganization near proteinsurfaces. �H is calculated with reference to the temperature at which the apolarcontribution is assumed to be zero ðT�

HÞ. The value of T�H is the enthalpy convergence



ORDER REPRINTS

temperature, which is obtained from protein unfolding studies and is determined tobe 100�C (373K) (98). Thus the enthalpy change is calculated from:

�H ¼ �H� þ�Cp T � T�H

� �ð37Þ

where �H�is the polar contribution to �H at T�

H The value for �H�is directly

proportional to the burial of polar surface area (77,80) and is described by

�H� ¼ 35ð�6Þ�ASAp ð38Þ

where �ASAp is the change in polar accessible surface area and is negative for abinding reaction.

Above dissection contains a linear extrapolation of protein unfolding enthalpiesas a function of polar and apolar surface area (77,80,91). A regression analysis at themedium unfolding temperature of proteins (60�C) minimizes the extrapolationerror and yields the elementary contributions per A2 of apolar (�hap) and polar(�hp) surface to the enthalpy function at the reference temperature of 60�C(�Hbind(60

�C)):

�Hbindð60�CÞ ¼ �hap�ASAþ�hp�ASAp ð39Þ

Values for �hap and �hp are given in Table 1. At any other temperature T,�H(T) is given by the standard equation

�HbindðTÞ ¼ �Hbindð60�CÞ þ�CpðT � 333:15Þ ð40Þ

Above calculations were shown to hold reasonably well at predicting the bindingenthalpy for protein–protein and protein–peptide interactions, but correlation ofstructure and enthalpy has yielded inconsistent results for interactions of smallmolecules (MW< 800) with biological macromolecules. A first attempt for empiricalparameterization of �H for small ligands has been published recently (97). As a first

Table 1. Empirical parameters for calculation of �Cp and �H for three different

implementations (units for �c in calK�1 (molA2)�1; for �h in cal (molA2)�1; probe radius

1.4 A; slice width 0.25 A).a

Parameter

ACCESS

(Presnell)bACCESS

(Richards)cNACCESS

(Hubbard)d

�cap 0.45 0.36 0.43

�cp �0.26 �0.25 �0.26

�hap �8.43 �3.63 �7.26

�hp 31.29 23.75 29.14

�hap(25) �7.35e

�hp(25) 31.06e

aValues from Ref. (82)bRef. (141)cRef. (95)dRef. (96)eRef. (97).

25


ORDER REPRINTS

approximation the experimental binding heat (�Hexp) can be considered as thecombination of atleast three terms:

�Hexp ¼ �Hint þ�Hconf þ�Hprot ð41Þ

The intrinsic enthalpy (�Hint) reflects the interaction of the ligand with theprotein and corresponds to the situation when ligand and protein adopt the sameconformation in the free and the bound state. These contributions to the enthalpyare assumed to scale with changes in accessible surface area that can be param-eterized. Contributions from conformational changes (�Hconf) cannot be easilydescribed in terms of �ASA. �Hprot contains binding enthalpy that is due toprotonation effects that can and need to be dissected experimentally by performingexperiments in buffers of different ionization enthalpies. Once corrected for �Hprot,the calculated �Hbind at 25�C, the temperature at which most binding studiesare performed, will be the sum of the constant term �Hint and a term that is specificfor each ligand:

�Hbindð25Þ ¼ �Hconf ð25Þ þ�hapð25Þ�ASAþ�hpð25Þ�ASAp ð42Þ

The intrinsic enthalpy is now described in terms of �ASA with �hap(25) and�hp(25) as the scaling coefficients that yield the elementary contributions per A2

of apolar and polar surface, respectively. The values for �hap(25) and �hp(25) aregiven in Table 1.

In practice, a system under investigation will be analyzed using the scalingparameters, �ASA and leaving �Hconf as the only adjustable parameter that needsto be determined for each protein separately from a training set of known protein–ligand structures. This new approach is based on a HIV-1 protease database andhas been validated by six different datasets for which structural data of complexedand free protein was available (97).

Calculation of DS

The total entropic contributions associated with binding reactions can beexpressed as the sum of three terms (75):

�Stot ¼ �Ssolv þ�Sconf þ�Sr=t ð43Þ

where �Ssolv describes the change in entropy resulting from solvent release uponbinding, �Sconf is a configurational term reflecting the reduction of rotationaldegrees of freedom around torsion angles of protein and ligand. �Sr/t entails the lossof translational and rotational degrees of freedom when a complex is formed fromtwo molecules free in solution.

The most important contribution to the entropy change arises from thesolvation term (�Ssolv), primarily due to burial of apolar surface area and isapproximated for any temperature (T ) by following equation:

�Ssolv ¼ �Cp ln T=T�S

� �ð44Þ



ORDER REPRINTS

where T�S is the temperature at which there is no solvent contribution to the

hydrophobic entropy change and is equal to 112�C (385K) (74,99,100).The translational/rotational entropy term (�Sr/t) accounts for the reduction of

the number of particles in solution and their degrees of freedom. Very differentestimates of the magnitude of �Sr/t have been discussed (73,101). Empirical andtheoretical considerations have suggested that this term contributes �4 to�10cal K�1mol�1 to the overall entropy of a bimolecular binding event (74,102,103).This value is numerically close to the cratic entropy of�8 calK�1mol�1(104) even ifthere is no sound physical reason to equate translational/rotational entropy to thestatistical part of the mixing entropy (105).

Finally, the configurational entropy �Sconf reflects contributions from changesin side-chain conformational entropy as well as all other structural rearrangement ofprotein and ligand induced by complex formation. Since �Stot is experimentallyaccessible, and �Ssolv and �Sr/t can be estimated, �Sconf is calculated from:

�Sconf ¼ �Stot ��Ssolv ��Sr=t ð45Þ

The contribution to�S from side chains involved in binding can be estimated byconsidering an average contribution of �4.3 calK�1mol�1 per residue to �Sconf

(73,100,107). On the assumption that only minor configurational entropiccontributions of ligand occur, a rough estimate of the number of amino acids(Xres) participating in the interaction is available by

Xres ¼�Sconf

�4:3 calK�1mol�1ð46Þ

If the amino acid side chains directly participating in the interaction processare known from theoretical or experimental studies, it is possible to calculateeach contribution to �Sconf as the sum over the amino acids involved in binding(108–111). A binding process involves two contributions to �Sconf: (i) restrictionsaround side-chain torsion angles and (ii) immobilization of the peptide backbone. Ifbinding is not involved in order/disorder transitions, only the side-chain componentapplies. The side-chain contribution is then calculated by assuming that the entropyis zero when fully buried and scales linearly with ASA, resulting in a maximum valuewhen fully exposed. By applying this model, a term �Sbu!ex, i must be introduced toaccount for the buried-to-exposed entropy gain which differs for each side chain.The contribution is then calculated as

�Sconf ¼Xi

�ASAsc;i

�ASAAXA;i�Sbu!ex;i ð47Þ

where �ASAsc,i is the change in ASA of side chain i on binding, and �ASAAXA,i isthe ASA of the side chain in an extended Ala-X-Ala tripeptide (82). The summationis carried out over all side chains participating in the interface. Values for �ASAsc,i

are determined from structural data, estimates for �ASAAXA,i and �Sbu!ex,i (111)are available, and they have been adapted for different implementations (82).These data are presented in Table 2.

For interaction processes involving transitions, two additional terms associatedwith the backbone entropy (�Sbb) and with the entropy of the transition from

27


ORDER REPRINTS

the exposed/folded state to the exposed/unfolded state (�Sex!u) must be included(82). Estimates of these contributions are also available (110), and are summarizedin Table 2.

For nonpeptide ligands a special empirical parameterization has been proposedto account for changes in conformational degrees of freedom between free andcomplexed forms of ligand (112). As a first approximation, it is assumed that theconformational entropy will be proportional to the number of rotatable bonds. Sinceeffects from excluded volume increase with the number of atoms for a given numberof rotatable bonds, the conformational entropy change for a nonpeptide ligand(�Sconf, np) is considered to be a linear function of the number of rotatable bonds(Nrb) and the total number of atoms (Nat):

�Sconf;np ¼ k1Nrb þ k2Nat ð48Þ

Table 2. Total side-chain ASA (�ASAsc) of Ala-X-Ala tripeptides used in parameterization

for three ASA implementations (units in A2), and side-chain and backbone conformational

entropy values (�Sbu!ex, �Sbb, �Sex!u; units in calK�1mol�1).a

ACCESS

(Presnell)bACCESS

(Richards)cNACCESS

(Hubbard)d

�ASAsc �ASAsc �ASAsc �Sbu!ex �Sbb �Sex!u

Ala 52.1 60.4 54.6 0.00 4.11 0.00

Arg 187.9 210.2 199.6 7.10 3.39 �0.84

Asn 113.8 123.0 112.9 3.30 3.39 2.25

Asp 102.4 106.5 99.4 2.01 3.39 2.15

Cyse 70.8 69.1 70.3 3.56 3.39 0.62

Cysf 91.9 94.5 92.0 3.56 3.39 0.62

Gln 128.7 142.8 122.3 5.02 3.39 2.13

Glu 117.5 135.8 124.5 3.54 3.39 2.27

Gly 0.0 0.0 0.0 0.00 6.50 0.00

His 144.3 152.2 144.9 3.44 3.39 0.79

Ile 123.8 138.5 135.9 1.74 2.17 0.67

Leu 134.5 150.6 143.8 1.62 3.39 0.24

Lys 156.9 177.7 155.6 5.85 3.39 1.03

Met 158.5 160.8 158.0 4.54 3.39 0.57

Phe 176.7 179.4 172.0 1.41 3.39 2.89

Pro 90.7 104.8 90.7 — — —

Ser 68.6 76.6 71.7 3.68 3.39 0.55

Thr 105.3 112.1 105.4 3.30 3.39 0.48

Trp 222.7 218.7 222.4 2.75 3.39 1.15

Tyr 188.5 196.4 190.2 2.77 3.39 3.13

Val 105.6 118.0 105.6 0.12 2.17 1.29

aValues from Ref. (82)bRef. (141)cRef. (95)dRef. (96)eValues for disulfide-bonded cystine;fValues for free cysteine.



ORDER REPRINTS

To date, Eq. (48) has not yet been widely applied, and in fact it has only beentested for the analysis of HIV-1 protease inhibitors. However, for these interactionsk1 was found to be�1.76 calK�1mol�1, whereas k2 equals 0.414 calK�1mol�1

(83,112).

Linkage Effects

Above calculations apply for systems that do not involve linked equilibria. Ifbinding of a second ligand is coupled to binding of a first one, the contributionsof the second equilibrium must be considered. Protonation linkage is a commonphenomenon in interaction processes. Its contributions to the binding energetics canbe determined experimentally, using a global analysis of experimental data as afunction of pH, temperature, and buffer ionization enthalpy (40,94).

THERMODYNAMICS AND RATIONAL DRUG DESIGN

The rapidly increasing availability of high-resolution protein structures hasopened the possibility to use structural information in the design of new drugs. Thecentral problem of structure-based design studies is the understanding of the featuresdictating the energetics of the interaction of a ligand with a macromolecule, i.e., theaccurate prediction of the Gibbs free energy that determines the binding affinity.

The Thermodynamic Approach

The prediction of binding energetics is greatly complicated by the effects ofenthalpy–entropy compensation (84,85) which means that an increase in �H doesnot contribute to the binding affinity, as the improvement is only achieved by acompensation cost in the T�S term. The sheer number of entropic and enthalpiceffects contributing to the observed �G makes it difficult to rationalize and predictbinding affinities. Additional energetic effects will arise from any differences inligand conformations in the free state in solution and bound to the macromolecule.Moreover, water molecules play an important role in adapting the binding pocketto different ligands.

The strength of the thermodynamic approach is that it has become possible todissect the contributing forces (�H, �S, �Cp) which make up the free bindingenergy of the interaction by direct calorimetric measurements, yielding effectiveenergetic contributions, including all interactions that are found in the system(protein, ligand, and solvent interactions).

Current Status of the Thermodynamic Approach

Until recently, thermodynamic data have not played an important rolein molecular design, since no theoretical framework relating structural and

29


ORDER REPRINTS

thermodynamic data has been established. The situation has changed with the

realization that changes in solvent-accessible surface area (�ASA) are related to the

thermodynamic parameters. The semi-empirically derived set of structural

energetic parameters (summarized in Table 1), based on the parameterization of

�G, �H, �S, and �Cp in terms of �ASA, is a promising thermodynamic approach

to drug design.It is possible now to predict the energetics of folding of a globular protein

(77,80,81,87,89,91) with an accuracy within 9–12% (82). The parameterization has

reached the state in which accurate predictions of binding energetics is possible as

well, as shown for peptide–protein and protein–protein association (73,79,93,94),

and also for nonpeptide ligand–protein interactions (112). However, the thermo-

dynamic approach is still in its infancy, as there are no comprehensive reports

documenting the design of a high-affinity drug using a ‘‘thermodynamic-directed

rational drug-design’’ approach.Furthermore, current parameterization is still based on protein unfolding data,

and there is need for more thermodynamic and structural characterization of

protein–ligand interactions, including not only �H, �S, and �Cp, but also

thorough investigation of phenomena linked to binding, i.e., protonation, ion

binding, and conformational changes. As the research in this field makes progress,

the structural thermodynamic database is expanded and the set of energetic

parameters can be refined to the point that will allow accurate prediction of effects of

small molecular changes. Thus, this will allow a much easier assessment of better

binding drugs.

CASE STUDIES FOR PROTEIN–LIGAND

INTERACTIONS STUDIED BY ITC

The term ‘‘protein–ligand interactions’’ stands for the association of a

biological macromolecule (a protein) with any molecule (the ligand). This is an

arbitrary restriction to biological system and does not mean that other binding

partners do not exist or cannot be measured. As mentioned before, any binding

event that will produce sufficient binding heat will be amenable to the calorimetric

approach. This has been shown for a variety of interacting systems (for a list see

http://www.calorimetry.com) that are not subject of the present discussion.

Whereas the biological systems under investigation deal with proteins as one

binding partner, e.g., antibodies, enzymes, or receptors (soluble and membrane

bound), the ligand will include all possible interaction partners for the protein.

These can be proteins, peptides, DNA, carbohydrates, lipids, metal ions, inhibitors,

substrates, or cofactors. At this point we describe three representative examples for

biological systems, i.e., enzymes, soluble receptor moieties, and membrane-bound

receptors that have been investigated using the direct thermodynamic approach.

This will give a deeper understanding to the reader with respect to the

informational content, possibilities, and potential drawbacks for the calorimetric

experiments.



ORDER REPRINTS

Substrate Binding to HSV1 Thymidine Kinase

Introduction

An example of high medicinal interest where thermodynamic information isessential is thymidine kinase from Herpes simplex virus type-1 (HSV1 TK). Thestructure of this enzyme is known at high resolution in complex with a series ofligands, including various substrates (natural and non-natural) and inhibitors(113–117).

Thymidine kinases (EC 2.7.1.21) catalyze the phosphorylation of dT tothymidine monophosphate (dTMP) in presence of magnesium ions by transferringthe �-phosphate group of adenosine triphosphate (ATP) to the 50-OH group of dT.Herpes viruses encode their own thymidine kinases, which differ considerably fromthe enzyme of the human cellular host (hTK1). While the human enzyme is highlyspecific, HSV1TK is a multifunctional enzyme of broad substrate specificity andrequires low stereo-chemical specificity. Therapeutic applications involvingHSV1TK make use of the broad substrate diversity of the viral enzyme in thebackground of strict substrate selectivity of the host cell enzyme. Therefore, adetailed thermodynamic analysis of substrate binding to the viral kinase is aprerequisite for the successful design of new therapeutically useful compounds. ITChas been used to investigate the thermodynamic parameters of the bindingof thymidine (dT) and adenosine triphosphate (ATP) to wild type and mutatedHSV1TK (118,119).

Thermodynamic Parameters for dT and ATP Binding to HSV1TK

In the ternary complex TK:dT:ATP, the substrate dT and the cofactor ATP arelocated in separate and well defined binding pockets of the enzyme. Formation of theternary enzyme–substrate complex may either proceed through an obligatorysequential pathway or by a random mechanism. Two sequential pathways arepossible: TK!TK:dT!TK:dT:ATP (reactions (i) and (ii) of Fig. 5), or TK!TK:ATP!TK:dT:ATP (reactions (iii) and (iv) of Fig. 5). In a random mechanism,

TK:ATP

TK TK:dT

TK:dT:ATP

(ii)

(i)

(iii)

(iv)

Figure 5. Formation of the ternary enzyme–substrate complex TK:dT:ATP. The two ordered

sequential pathways are (i), plus (ii) and (iii), plus (iv), respectively. In a random binding

mechanism, all four reactions will take place.

31


ORDER REPRINTS

binding of one substrate is not a prerequisite for binding of the other, and all four

reactions of Fig. 5 can take place. To distinguish between ordered and random

binding, HSV1 TK was titrated with dT and with ATP, respectively. The titration

with dT was characterized by a significant exothermic heat effect. The nonlinear least

squares fit for a single-site binding model gave a KB of 1.9� 105M�1 and �Hbind

of�19.1kcalmol�1 (Table 3). Under identical conditions ATP did not bind to empty

HSV1 TK. An ordered binding mechanism was confirmed by titrating the preformed

TK:dT binary complex with ATP (reaction ii of Fig. 5). The reaction was again

exothermic and yielded a KB of 3.9� 106M�1 and �Hbind¼�13.8 kcalmol�1.

Moreover, titration of TK with a 1:1 mixture of dT and ATP yielded within error a

heat change corresponding to the sum of the heat changes for reactions (i) and (ii) of

Fig. 5. The ITC experiment provides the binding constant KB for a single-site

reaction, and �G of reactions (i) and (ii) were calculated from Eq. (19). �S was

obtained from Eq. (20). Reactions (i) and (ii) were driven by favorable negative

changes in binding enthalpy and strongly opposed by unfavorable entropic

contributions. Although the reaction with the 1:1 mixture of dT and ATP was

more complex, it could still be treated as a single-site reaction if one considered

dTþATP as one ligand. ITC measurements performed in the temperature range of

10–25�C showed strong dependence of �H and T�S on temperature while �G was

almost insensitive due to enthalpy–entropy compensation. Values of �Cp were

calculated from the slopes of the regression lines of �Hbind vs. temperature. Binding

of dT to the free enzyme was characterized by �Cp of �360 calK�1mol�1. For ATP

binding to the TK:dT complex, �Cp of �140 calK�1mol�1 was determined,

and for the titration of the enzyme with a 1:1 mixture of dT and ATP, a �Cp

Table 3. Thermodynamic parameters for the binding of thymidine and ATP to HSV1TK

at pH 7.5 and 25�C.ab Values of �H are corrected for protonation effects by Eq. (6).

Temp.

(�C)

TKþ dT (reaction (i)) TK:dTþATP (reaction (ii)) TKþ dT/ATPc

�H �G T�S �H �G T�S �H �G d T�S

10 �13.6 �7.1 �6.5 �11.7 �8.9 �2.8 �25.4 �18.9 �6.5

15 �15.8 �7.0 �8.8 �12.1 �8.9 �3.2 �27.9 �18.9 �9.0

20 �17.4 �7.1 �10.2 �13.0 �9.1 �3.9 �30.5 �17.2 �13.3

25 �19.1 �7.2 �11.9 �13.8 �9.0 �4.8 �33.1 �16.8 �16.3

�Cp �0.36 �0.14 �0.51

aValues of �G, �H, and T�S in kcalmol�1, �Cp in kcalK�1mol�1.bValues are the mean of triplicates. �G was calculated from �G¼�RT lnKB, where KB is the

binding constant determined by ITC. Uncertainty of �G is within �0.35 kcalmol�1 of the

mean. Errors of �H are about �5% and mainly reflect the error in ligand concentration.

Maximal possible errors of T�S are 1.5 kcalmol�1. Errors of �Cp were estimated by

reduction of the data set by one data point at a time and were, on average, �0.02

kcalK�1mol�1, i.e., within 5–15% of the reported mean.c1:1 mixture of dT and ATP.dCalculated from �G¼�RT ln(KB)

2.



ORDER REPRINTS

of �510 kcalK�1mol�1 was measured. The latter value was very close to the sumof �Cp’s of reactions (i) and (ii).

The stoichiometry of binding obtained from all experiments was in the range0.7–0.8mol ligand per mol of HSV1TK monomer. Deviation from a value of 1 wasdue to the presence of inactive protein (118).

Protonation Effects

Titration experiments were repeated in various buffers of different �Hion. Theintrinsic enthalpy of binding, �Hbind, was obtained from the intercept (�Hion¼ 0) ofa plot according to Eq. (6). Protonation/deprotonation was negligible in the case ofdT binding to the free enzyme (�Hobs¼�Hbind). An uptake of 0.31 protons wasobserved with ATP binding to the TK:dT complex. Titration with the 1:1 mixture ofdT and ATP lead to the uptake of 0.35 protons. It follows that proton uptakeoccurred with ATP binding but not with dT binding. Heat changes from ITCexperiments were corrected accordingly (118).

Decomposition of Entropy Changes

The results for the decomposition of�Stot for substrate binding to HSV1TK aresummarized in Table 4. It is evident that the favorable solvent contribution �Ssolv isovercompensated by the large unfavorable conformational entropy change �Sconf.The contribution by �Sr/t is unfavorable but small. In accordance with athermodynamic cycle, the decomposed entropy changes of reactions (i) and (ii)add up to the changes calculated for the titration of the enzyme with the 1:1 mixtureof dT and ATP.

Solvent reorganization provided a substantial gain in binding entropy due towater release, about 90 calK�1mol�1 for dT binding and about 35 calK�1mol�1

for ATP binding. This considerable �Ssolv is in agreement with fact thatphosphorylation of an acceptor OH-group must occur in the absence ofcompeting water molecules. After subtracting from �Stot the favorable �Ssolv

Table 4. Decomposition of entropy changes for substrate binding to HSV1TK at 25�C.a

Reaction �Stotb �Ssolv

c �Sconfd �Sr/t

e

TKþ dT, reaction (i) �39.9 92 �124 �8

TK:dTþATP, reaction (ii) �16.1 36 �44 �8

TKþ dT/ATP �54.7 133 �172 �16

aIn calK�1mol�1.bCalculated from values of T�S in Table 3.cCalculated from Eq. (44) using �Cp from Table 3.dCalculated from Eq. (43).eValue for bimolecular and trimolecular reaction, respectively (104).

33


ORDER REPRINTS

term and the small unfavorable cratic �Sr/t, a large unfavorable entropiccontribution �Sconf remained. �Sconf may have two main origins, the first being‘‘freezing’’ of bond rotations, i.e., amino acid side chains which are in directcontact with the bound substrate or involved in building a closed or morecompact conformation, become less mobile. The other contribution to �Sconf ispartial folding or tightening of domains, particularly in the area of substratebinding. Interdomain movements within the enzyme dimer also can contribute to�Sconf if the HSV1TK dimer becomes more closed in the substrate-bound form.An estimation of these contributions according to Eq. (47) showed thatapproximately 95% of �Sconf may be due to conformational domain movementsand partial folding, or refolding, of domains, in line with the general concept thatnucleoside–nucleotide kinases undergo conformational changes during theircatalytic cycle (120,121).

Correlation Between �Cp and Surface Area Buried on Substrate Binding

The experimentally determined parameters �Cp and �Hbind (Table 3) wereused to simultaneously solve Eqs. (36) and (39), yielding the total areaapparently buried (�ASAtot) during the binding reactions (Table 5). �ASAtot

for the reaction with a 1:1 mixture of dT and ATP was�5100 A2. This changein area corresponded very well to the sum of the surface changes for theindividual reactions (i) and (ii) (Fig. 5). These figures seem to be very large forthe binding of the two small substrate molecules and are comparable to thesurface buried on folding of a small globular protein of 50–60 residues. As nostructural information is available for HSV1TK in the ligand-free state, �ASAap

Table 5. Changes of solvent accessible surface area (�ASA) caused by substrate binding to

HSV1TK.a

Calculation Reaction �ASAap �ASAp �ASAtot �Cpdcalc �Cpeexp

Rigid bodyb TKþ dT �400 �200 �600 �130 �360

TD parametersc Reaction (i) �1750 �1500 �3250

Rigid bodyb TK:dTþATP �300 �450 �750 �20 �140

TD parametersc Reaction (ii) �850 �850 �1700

Rigid bodyb TKþ dT/ATPf�700 �650 �1350 �150 �510

TD parametersc �2650 �2450 �5100

a�ASA in A2, �Cp in calK�1mol�1.b�ASA calculated by removing dT, ATP, or both, from the crystal structure of the

TK:dT:ATP complex (113).c�ASA calculated by simultaneously solving Eqs. (36) and (39) using experimental values of

�Cp and �Hbind from Table 3.dCalculated from Eq. (36) with �ASA calculated from the crystal structure of TK:dT:ATP

(rigid body assumption).eFrom Table 3.fReaction with 1:1 mixture of dT and ATP.



ORDER REPRINTS

and �ASAp could not be tested against actual structural data. As anapproximation, �ASA values were obtained for the ternary crystal complexafter removal of dT and ATP. The buried surface area calculated in this waycorresponds to a rigid body binding model in which no conformational changestake place when dT and ATP bind. As seen in Table 5, �ASAtot was very muchsmaller, resulting in calculated �Cp values, using Eq. (36), that werecorrespondingly smaller than experimental �Cp. Even so the calculated surfacearea changes may be rough estimations, the significant discrepancy to a rigidbinding model is an indisputable sign for significant conformational rearrange-ments accompanying substrate binding to HSV1TK. Moreover it is in very goodagreement with the finding that the major unfavorable entropic contribution isdue to loss of conformational entropy.

Mutational Studies

The mechanism of the broad substrate diversity observed with HSVTK1 wasinvestigated by a thorough mutational study, kinetic measurements and ITC. Theresidue triad H58/M128/Y172 was identified to confer distinctive binding of anexceptionally large variety of substrates to HSVTK1 and to guide catalyticproperties. Mutations in this triad (H58L, M128F, Y172F) have been preparedand were analyzed by ITC (Table 6).

The mutation M128F with a phenylalanine in this position resulted in acompletely inactive enzyme when examined by kinetic measurements and an HPLCassay. It was shown that the lack of activity for the M128F mutant was not due tohydrophobic collapse initiated by the mutation, but was the result of disturbed dTbinding (200-fold reduced affinity compared to the wild-type enzyme) that yielded anunproductive orientation of the substrate. This view is corroborated by the stronglyreduced binding enthalpy, in combination with the more than seven-fold reductionof unfavorable entropy. It is evident that the major structural changes needed forcatalytic competence, which are induced by dT in wild-type HSV1TK, could nottake place.

Similar results were found with the double mutant M128F/Y172F. In thismutant the dT binding site is formed by a ‘‘double-F sandwich,’’ a particular

Table 6. Thermodynamic data for dT binding in presence of ATP of wild type HSV1TK and

triad H58/M128/Y172 mutants.

HSV1TK �H (kcalmol�1) KB (105 M�1) �G (kcalmol�1) T�S (kcalmol�1)

Wild type �26.35� 0.47 229� 146 �9.95� 0.40 �16.4� 0.85

M128F �8.14� 0.20 1.04� 0.10 �6.84� 0.08 �1.30� 0.20

M128F/Y172F �3.71� 0.08 0.34� 0.19 �6.14� 0.34 2.43� 0.42

H58L/M128F/Y172F �19.82� 0.13 2.61� 0.06 �7.38� 0.01 �12.44� 0.14

35


ORDER REPRINTS

combination that also occurs in various thymidine kinases. In this case even the sign

for the entropy changed, giving evidence for a favorable preformed binding pocket

with a subsequent reduced induced fit movement. However, the hydrogen bonding

network seems to be impaired as deduced from the almost 700-fold reduced dT

affinity as compared to wild-type HSV1TK.Sequence alignments show that this double-F combination never occurs with a

histidine at position 58 (HSV1TK numbering). Subsequently H58 was replaced by

leucine, giving the triple mutant H58L/M128F/Y172F. With this third mutation

enzymatic activity was regained, but the broad substrate diversity was lost since

activity and kinetic studies as well as HPLC assays did not show any activity of the

triple mutant against non-natural substrates. Although the binding affinity is almost

90-fold reduced, binding enthalpy and entropy adopt similar values as for wild-type

HSV1TK, indicating restored flexibility of the enzyme.

Conclusion

This is the first report providing a comprehensive thermodynamic description

of substrate and cofactor binding to HSVTK1, a representative of the large family

of nucleotide and nucleoside kinases. The results obtained by titration micro-

calorimetry reveal an extreme case of positive heterotropic interaction. Formation

of a binary complex of thymidine with HSV1TK is a stringent prerequisite for

ATP binding. Since the ATP binding site is in fact generated by thymidine binding,

one expects the enzyme to undergo considerable conformational rearrangements.

This has been supported by a semi-empirical analysis of the observed heat capacity

and entropy changes, which were large and negative and indicated burial of

molecular surface to an extent much larger than expected if the substrate binding

sites would pre-exist on the apo-enzyme and no rearrangement would occur (rigid

body binding model). The favorable gain in entropy from water release from buried

surface was overcompensated by a large decrease in conformational entropy. The

findings support the view that substrate binding to HSV1TK leads to a

conformational closing of the substrate binding sites to bring thymidine and ATP

into an orientation appropriate for catalysis. Further mutational studies in the

residue triad H58/M128/Y172 supported this view and could shed light on possible

mechanisms important for substrate diversity. Nevertheless, the details of the

predicted rearrangements have to await a firm structural foundation. The direct

thermodynamic approach using ITC proved to be an invaluable tool for this

study. It made not only possible to characterize HSV1TK interactions with its

substrate and cofactor thermodynamically, but it was also an invaluable tool

to investigate mutants that were not amenable through standard kinetic measure-

ments. The lack of observable activity does not necessarily mean that binding is

prevented, but the association can be altered either subtly or dramatically, thus

leading to catalytic incompetence. ITC clearly is the method of choice for studies of

that kind.



ORDER REPRINTS

Specificity of Citrate Binding to the Histidine Autokinase CitA Receptor

Introduction

The periplasmic receptor domain of the sensor kinase CitA of Klebsiellapneumoniae represents another very interesting biological system that has beeninvestigated by the direct thermodynamic approach (122). In bacteria, regulationof gene expression in response to changing environmental conditions is oftenaccomplished by two-component regulatory systems. In their simplest form theyconsist of two modular proteins, the sensor kinase and the response regulator. Thesensor kinase responds to a certain stimulus by autophosphorylation of a conservedhistidine residue. Subsequently the phosphate group is transferred to a conservedaspartate residue of the response regulator that mediates adaptations in geneexpression or cell behavior. In most cases sensor kinases are transmembrane proteinswith an extracellular N-terminal sensor domain flanked by two transmembranehelices and a cytoplasmic C-terminal autokinase domain that is connected to thesecond transmembrane helix via a linker region of variable length (123). With thisarchitecture bacteria are capable to sense external stimuli and transduce informationto the cytoplasm. Although many two-component systems are known, only in a fewcases have the primary stimulus of the sensor kinase been identified. In part this lackof knowledge is due to the difficulties in isolating and purifying the membrane-bound kinases. The strategy to overcome this problem is the expression of theextracellular N-terminal sensor domain as a separate and soluble protein that allowsstudying its binding properties.

The sensor kinase CitA of Klebsiella pneumoniae and its cognate responseregulator CitB form the paradigm of a subfamily of bacterial two-componentsystems (124). It is involved in the regulation and expression of genes needed forcitrate fermentation. In a previous study it was proposed that citrate was the favoritesignal recognized by CitA (125). Attempts to demonstrate this function with an E.coli in vivo system were not successful. Therefore, the periplasmic domain of CitAwas cloned and expressed as a separate, soluble receptor domain supplemented witha C-terminal His6-tag (CitAPHis). Ligand specificity and thermodynamic bindingproperties were investigated using ITC (122).

Citrate Binding Studies Using ITC

Isothermal titration calorimetry was applied to describe the thermodynamicproperties of citrate and citrate-analog binding to CitAPHis. Citrate binding toCitAPHis was shown to be an exothermic process that could be fit according to asingle-site binding model, yielding a KD value (KD¼ 1/KB) of 5.5 mM, an observedenthalpy change (�Hobs) of�18.25 kcalmol�1. In contrast to citrate, neitherisocitrate nor tricarballylate exhibited demonstrable binding to CitAPHis under thesame conditions. Additional displacement experiments were necessary in order toexclude complete entropically driven binding which would not release significantbinding heat. The results of these experiments showed that the presence of isocitrateand tricarballylate did not alter the binding behavior of citrate towards CitAPHis.

37


ORDER REPRINTS

Therefore it was concluded that both isocitrate and tricarballylate do not bind to the

citrate binding site.

Investigation of the Probable Citrate Species Binding

Depending on the pH, citrate exists in four different species, i.e., H3-citrate,

H2-citrate�, H-citrate2�, and citrate3� with pK1¼ 3.13, pK2¼ 4.76, and pK3 ¼

6.40 (126). In order to determine the preferred ligand species, the pH-dependency of

the citrate binding constant to CitAPHis was determined in the range of pH 4.0–9.0

(Table 7). The curve obtained from a plot of log KB vs. pH showed a maximum at

about pH 5.7 and decreased above and below this value, indicating that citrate

binding is linked to ionization events (41). The �Hobs and �S values were negative

over the pH range studied, showing that binding of citrate to CitAPHis was driven by

the enthalpy change, whereas the entropy change was opposite.It is very likely that this result can be explained by the assumption that

H-citrate2� is the citrate species responsible for the interaction, as the affinity of

citrate to CitAPHis at different pH values closely reflects the appearance of the

H-citrate2� form at the corresponding pH. As deduced from Table 7, the maximal

affinity was found at pH 5.7 and it decreased above and below this value, similar to

the occurrence of the divalent citrate form. An obvious explanation of this result

is that CitAPHis binds the H-citrate2� species specifically and therefore proton

release has to occur at pH values below pH 5.7 (where the H2-citrate� species

becomes dominant) and proton uptake must take place at pH values above

pH 5.7 (where the citrate3� form becomes prevalent). Assuming that the H-citrate2�

species is recognized by CitAPHis, four functional groups are available for binding

interactions, i.e., one uncharged and two charged carboxyl groups and the hydroxyl

group. Current crystallographic data of macromolecules involved in binding of

uncomplexed citrate clearly indicate that citrate binding is preferred in the extended

conformation with important interactions of all four functional groups (127–130).

Table 7. Thermodynamic parameters of citrate binding to CitAPHis at 25�C in 50mM

sodium phosphate buffer at different pH values as determined by ITC.a

pH n

KB

(104 M�1)

KD

(mM)

�Hobs

(kcalmol�1)

�G

(kcalmol�1)

T�S

(kcalmol�1)

4.0 0.98� 0.01 6.26� 1.33 16.7� 3.6 �24.71� 0.04 �6.53� 0.13 �18.18� 0.17

5.0 0.85� 0.01 38.55� 1.25 2.6� 0.1 �23.37� 0.40 �7.62� 0.02 �15.75� 0.41

6.0 0.80� 0.01 51.65� 2.95 1.9� 0.1 �20.60� 0.19 �7.79� 0.03 �12.81� 0.15

7.0 0.89� 0.01 18.25� 0.75 5.5� 0.2 �18.26� 0.05 �7.17� 0.02 �11.08� 0.02

8.0 0.85� 0.01 9.05� 0.09 11.1� 0.1 �17.52� 0.18 �6.76� 0.01 �10.76� 0.17

9.0 0.83� 0.01 6.47� 0.02 15.5� 0.1 �17.89� 0.04 �6.56� 0.01 �11.33� 0.04

aThe data represent the mean of two independent experiments. Errors are quoted as standard

deviations for the two repeats of each titration.



ORDER REPRINTS

The fact that neither isocitrate nor tricarballylate did bind to CitAPHis showed theessential role of the hydroxyl group at position C3 in the citrate molecule.

Influence of Mg2þon Citrate Binding

The inhibitory influence of Mg2þ ions on citrate binding observed before couldbe confirmed by ITC. As shown by the titration experiments presented in Table 8,increasing Mg2þ concentrations resulted in an exponential decrease of the bindingaffinity. At 20mM Mg2þ, 10-fold reduction of KB was measured. In the context withthe above discussion of the four citrate species at different pH, it became clear thatthe reduced binding affinity of citrate in the presence of Mg2þ ions was due to Mg2þ

complexation by citrate3�. Using a stability constant of 1585M�1(126) for theMg–citrate� complex, the preferred H–citrate2� species is reduced in the presence of20mM Mg2þ at pH 7.0 to less than 0.01%, and nearly all of the citrate (96.2%) iscomplexed as Mg–citrate�. Crystallographic data of metal citrate complexesgenerally show the hydroxyl group and either one or two of the carboxyl groupsbeing involved in metal binding (131). This means that the Mg–citrate� complex hasto dissociate before the functional groups can interact with CitAPHis. This behavioris reflected by the occurrence of additional binding enthalpy in the presence of Mg2þ.In control experiments the additional amount of binding enthalpy (��Hobs) andentropy (��S) corresponded very well to the dissociation heat and entropy of theMg–citrate� complex. In summary, the pH and Mg2þ dependency of citrate bindingto CitAPHis indicated that H–citrate2� was the preferred binding species in a non-restricted, extended conformation and that the inhibitory effect of Mg2þ was not dueto an interaction with CitAPHis, but only due to complex formation with citrate.

Binding Stoichiometry

All data fitted excellently to a single-site binding model. Analysis of all ITCexperiments for the binding stoichiometry revealed values that varied between

Table 8. Thermodynamic parameters of citrate binding to CitAPHis at 25�C in 50mM

sodium phosphate buffer pH 7.0 including different Mg2þ concentrations as determined

by ITC.a

MgCl2(mM) n

KB

(104 M�1)

KD

(mM)

�Hobs

(kcalmol�1)

�G

(kcalmol�1)

T�S

(kcalmol�1)

0 0.89� 0.01 18.25� 0.75 5.5 �18.26� 0.05 �7.17 �11.08

2 0.83� 0.01 7.78� 0.17 12.9 �18.88� 0.02 �6.67 �12.21

10 0.85� 0.05 2.47� 0.20 40.5 �21.30� 0.20 �5.99 �15.31

20 0.89� 0.03 1.74� 0.11 57.5 �21.40� 0.11 �5.78 �15.62

aThe data were obtained from a single experiment for each Mg2þ concentration and the errors

quoted for n, KB, and �H are those for the least-squares fit to the binding isotherms in the

individual titrations.

39


ORDER REPRINTS

0.85 and 0.95 (mean value 0.87� 0.06), which was in agreement with independentbinding assays using [14C]-citrate that typically gave values between 0.6 and 0.8(122). These data clearly show that one molecule of citrate binds per CitAPHis

molecule. Thus deviations from the integer value are most likely due to inaccurateprotein determination and/or to missfolded protein.

Conclusion

In this study the direct thermodynamic approach turned out to be very powerful.For the first time it was possible to describe the sensory characteristics of theperiplasmic domain of the sensor kinase CitA, and it could be shown that itfunctions as a highly specific citrate receptor that binds citrate in a nonrestricted,extended conformation with a 1:1 stoichiometry. Measurements at different pHvalues shed light on the responsible and preferred ligand species for the receptor. Thedisplacement studies using the citrate analogs isocitrate and tricarballylate gaveimportant hints to the essential role of the hydroxyl group at C3. In addition theinhibitory effect of Mg2þ could be very well observed in the ITC experiments, andthe interpretation of the results clearly indicated that CitAPHis is not inhibited due toa separate receptor-Mg2þ interaction, but only due to complex formation withcitrate. ITC was also used to localize important amino acids involved in citratebinding by studying the influence of point mutations in CitAPHis(132). Just recently,the crystal structure of CitAPHis in complex with citrate has been solved (133), andit shows that the findings learned from the thermodynamic data fit very well tothe structural data.

Protein–Ligand Interactions of Membrane Bound Receptors

In this part of the review we would like to focus on biological systems that aredifficult to address: membrane bound receptors. In general, the direct thermo-dynamic approach is also applicable to such systems as long as ligand binding isaccompanied by a significant heat change. The most severe problems to address arethe quantities of receptor that are needed, and the stability and preparation ofconcentrated receptor solutions. Membrane associated proteins are intrinsicallydifficult to treat and usually need special purification protocols that includedetergents, vesicles and/or liposome containing preparations. At this point it must beemphasized that ITC instrumentation does not at all exclude the use of suchcomplicated or even turbid systems. As the preparation of membrane boundreceptor can be solved in a lot of cases, it remains to provide sufficient amounts ofprotein. With the latest advances in recombinant DNA technology and procaryoticas well as eucaryotic expression systems, it has become possible to prepare significant(i.e., mg) quantities of pure receptors. This means that in the near future moremembrane receptor systems will become amenable to the direct thermodynamicapproach.

To our knowledge the first system to which ITC was applied for investigatingmembrane bound protein–ligand interaction, was the serine receptor of Escherichia



ORDER REPRINTS

coli chemotaxis (134). Two years later followed the thermodynamic work on colicin-N binding to the trimeric membrane bound porins OmpF, OmpC, and Phoe (135).Later in 2003 ITC was used to shed light onto human apo- and holo-transferrinbinding to the Neisseria meningitidis transferrin receptor (136). The next section isdedicated to the most recent example published, in which the interactioncharacteristics for the sensory rhodopsin II of Natronobacterium pharaonis with itscognate transducer are examined using ITC (137). It is a particularly interestingexample for the case of lateral membrane bound receptor interaction. Somemethodological aspects will be discussed since they may serve as starting point forthe development of new projects in this field.

Interaction of Transducer Fragments with

Natronobacterium pharaonis Sensory Rhodopsin II

Introduction

Phototaxis of Natronobacterium pharaonis is mediated by the two sensoryrhodopsins SRI and SRII. They consist of seven transmembrane helices and a retinalchromophore attached to the protein, and they share structural homologies as wellas functional properties. The first photoreceptor SRI enables the bacteria to seeklight conditions optimal for the function of the light-driven ion pumps and to avoidultraviolet light. The second photoreceptor SRII conveys negative phototaxis, whichmight enable the bacteria to evade harmful conditions of high oxygen concentrationsin the presence of light (138). Both receptors are bound to membrane proteins(halobacterial transducers of rhodopsins; HtrI, HtrII) that consist of a transmem-brane two-helical domain and a coiled/coil cytoplasmic domain. SRII binds stronglyto its cognate transducer NpHtrII. Light excitation of the SRII:NpHtrII complexleads to conformational changes in both proteins. The cytoplasmic domain of thetransducer becomes activated and in turn triggers the two-component signallingcascade.

SRII and its cognate transducer NpHtrII form a tight 2:2 complex onmembranes, whereas in micelles it dissociates to a 1:1 homodimer (139). In orderto elucidate the dimerization and to determine the size of the receptor-bindingdomain of the transducer, the dissociation constant (KD) of a series of shortenedtransducers to the receptor has been analyzed by ITC (137).

Methodological Aspects

A major problem to be solved for calorimetric studies remains the availability ofsufficient quantities of the membrane bound receptor SRII. The authors of thisarticle used standard DNA techniques for the cloning. The gene of interest wascloned into a pET plasmid and the receptor was expressed in BL21(DE3) as aC-terminal His6-tagged protein. Expression took place at 37�C in 2YT medium in a30-l fermenter. The receptor was localized in the E. coli membrane. Subsequentlycrude membranes were sedimented by centrifugation and solubilized in buffer

41


ORDER REPRINTS

containing 1.5% n-dodecyl-�-d-maltopyranoside (DDM). Isolation and purificationwas achieved by means of Ni-NTA agarose, but always in presence of the detergent(0.5%). The receptor was purified with a yield of up to 1mg/L of cell culture with apurity of >90%. The various transducer variants are membrane-bound proteins aswell, and they were produced in a similar way affording 1–2mg/L of culture.

For the ITC measurements the receptor was concentrated to 700–1000 mM(18–26mg/mL). The transducer solutions were used at 70–100 mM (1.3–1.8mg/mL).Both solutions were dialyzed at 4�C overnight against degassed buffer containing0.05% (w/v) DDM. The receptor solution (250 mL) was titrated into the solution ofthe transducer placed in the cell (1.4mL). Heat effects due to the detergent wereevaluated by control experiments using detergent-containing buffers (1% DDMin the syringe; 0.1% DDM in the cell chamber). Titrations were carried out at22 and 45�C in order to improve signal-to-noise ratios.

Results and Discussion

Binding affinities of the four shortened transducer fragments (NpHtrII-82,NpHtrII-101, NpHtrII-114, NpHtrII-157) to the receptor SRII were determinedusing ITC. It was necessary to measure the two longer fragments (NpHtrII-114,NpHtrII-157) at elevated temperature (45�C) in order to detect sufficient bindingheat. Shorter fragments aggregated at higher temperature and were thereforemeasured at 22�C. Both of the longer transducer fragments displayed no detectableheat effect at lower temperature. The thermodynamic parameters of fragmentassociation to the receptor are shown in Table 9. The dissociation constants forNpHtrII-157 and NpHtrII-114 were almost identical in the range of 200 nM, and thebinding heat was almost identical with�4.3 kcalmol�1. However, the KD ofNpHtrII-101 was increased by almost two orders of magnitude, indicating that thesequence between 101 and 114 is important for the receptor–transducer interaction.The shortest fragment (NpHtrII-82) was inactive in the binding assay. All threeNpHtrII fragments exhibited a stoichiometry of 1:1 within experimental error.

It must be pointed out that the data do not necessarily mirror the situation innative membranes and the association constants and �H measured might representminimal values. For example from the well-documented glycophorin A dimerization

Table 9. Thermodynamic parameters of the association of NpHtrII fragments to NpSRII.a

NpHtrII T (K)

KB

(106 M�1)

KD

(nM)

�Hobs

(kcalmol�1)

�G

(kcalmol�1)

�S

(cal K�1mol�1)

157 318 6.2 160 �4.28 �9.88 �17.62

114 318 4.32 240 �4.20 �9.65 �17.0

101 295 0.1 104 �1.39 �6.74 �18.20

82 295 <0.01 >105 — — —

aThe measurements were performed at indicated temperatures in degassed buffer containing

150mM NaCl, 10mM Tris/HCl (pH 8), and 0.05% DDM.



ORDER REPRINTS

in membranes it could be shown that �G of monomer association was influencedby the type and concentration of the detergent used, strongly influencing thethermodynamics of membrane–protein interactions. Despite these restrictions thedata provide important results on membrane–protein interactions and shed light onthe structure and size of the receptor-binding domain.

Conclusion

Complex formation of NpSRII and its transducer NpHtrII is a special case inthe sense that the association proceeds laterally along the membrane to form afunctional complex. Such lateral complex formation has only been elucidated in afew examples. To date there is only rare data available that were primarily gained byusing sedimentation equilibrium and/or analytical ultracentrifugation (140). Ofgeneral interest are those experiments that provide information about thethermodynamics and kinetics of the transducer binding. This can be accomplishedby using a direct thermodynamic approach. As these data are difficult to determinefor membrane proteins, the present results provide a new approach to study theseinteractions using a direct thermodynamic approach by means of ITC. This methodis rarely used so far for the study of membrane protein assembly due to the difficultnature of the membrane protein family.

To our knowledge this work presents the first calorimetric approach to gaininformation about the strength and energetics of lateral binding processes. This is ofgeneral interest because it could be shown that ITC is a suitable method to studydirectly membrane protein/protein interactions (137). It will contribute to ourunderstanding about this class of proteins, although the experiments need to beperformed in a micelle environment.

FUTURE DEVELOPMENTS

As outlined in the text and also shown with the examples, the ITC is a powerfultool with high informational content that is applicable to various biological and non-biological systems. A major drawback of first generation instrumentation has beenovercome by markedly increasing the sensitivity in a way that it has now reached aphysical limit. Significant further improvements with respect to sensitivity do notseem possible in the near future (MicroCal, personal communication). Anotherpotential approach to decrease sample size may be miniaturization. Integratedcircuit microcalorimeters are being built and may allow the application of ITC tomicro-titre plate formats in the future.

Modern screening processes for drug candidates would profit from developmentof high-throughput instruments. This would be a major breakthrough in thedetection of compounds in pharmaceutical chemistry. The current throughput onlyallows measuring a limited number of promising candidates from a series whichcould lead to missing candidates with unusual and unexpected binding character-istics worth being further investigated. New calorimetric methods using suchtechnology are now being developed, and with this new methodology the

43


ORDER REPRINTS

calorimetric approach will undoubtedly play a dominant role in high-throughput

screening processes in the future.

REFERENCES

1. Lodwig TH, Smeaton WA. The ice calorimeter of Lavoisier and Laplace and

some of its critics. Ann Sci 1974; 31:1–18.2. Wintermeyer-Weppler U. Zur Geschichte der Entwicklung der Physikalischen

Chemie. Dissertation, Frankfurt am Main, 1974.3. Daumas M. Les appareils D’experimentation de lavoisier. In: Tenney LD, ed.

Chymia. Vol. 3. Philadelphia: University of Pennsylvania Press, 1950:43–62.4. Guerlac H. Chemistry as a branch of physics: Laplace’s collaboration with

Lavoisier. In: McCormmach R, ed. Historical Studies in the Physical Sciences.

Vol. 7. Princeton, 1976:193–276.5. Hemminger W, Hohne GWH. Grundlagen Der Kalorimetrie. Weinheim:

Verlag Chemie, 1979.6. Pledge HT. Science Since 1500: A Short History of Mathematics, Physics,

Chemistry, Biology. London: His Majesty’s Stationery Office, 1939.7. Cobb C, Goldwhite H. Creations of Fire. New York: Plenum Press, 1995.8. Hudson J. The History of Chemistry. London: Macmillan, 1992.9. Partington JR. A Short History of Chemistry. 3rd ed. New York: Dover

Publications, 1989.10. Oscarson JL, Izatt RM. Determination of thermodynamic properties. In:

Rossiter BW, Baetzold RC, eds. Physical Methods of Chemistry. Vol. 2. 2nd ed.

New York: John Wiley & Sons, 1992.11. Wadso I. Microcalorimetry of aqueous and biological systems. In: Marsh KN,

O’Hare PAG, eds. Solution Calorimetry: Experimental Thermodynamics.

Oxford: Blackwell Scientific Publications, 1994:267–302.12. Wadso I. Microcalorimetry and its application in biological sciences. In: Pain

RH, Smith BJ, eds. New Techniques in Biophysics and Cell Biology. Vol. 2.

London: John Whiley & Sons, 1975:85–126.13. Privalov PL, Potekhin SA. Scanning microcalorimetry in studying temperature-

induced changes in proteins. Method Enzymol 1986; 131:4–51.14. Privalov PL. Thermodynamic problems of protein structure. Ann Rev Biophys

& Biophys Chem 1989; 18:47–69.15. Sturtevant M. Biochemical applications of differential scanning calorimetry.

Ann Rev Phys Chem 1987; 38:463–488.16. Freire E. Differential scanning calorimetry. Method Mol Biol 1995;

40:191–218.17. Chen A, Wadso I. Simultaneous determination of delta G, delta H and delta S

by an automatic microcalorimetric titration technique: application to protein

ligand binding. J Biochem Biophys Meth 1982; 6:307–316.18. Freire E, Mayorga OL, Straume M. Isothermal titration calorimetry. Anal

Chem 1990; 62:950A–959A.



ORDER REPRINTS

19. Wiseman T, Williston S, Brandts JF, Lin LN. Rapid measurement of binding

constants and heats of binding using a new titration calorimeter. Anal Biochem

1989; 179:131–137.20. MicroCal LLC. VP-ITC Microcalorimeter, User’s Manual. MAU130030

(Rev. A). ed. Northampton: MicroCal LLC, 2003.21. Bains G, Freire E. Calorimetric determination of cooperative interactions in

high affinity binding processes. Anal Biochem 1991; 192:203–206.22. Bhatnagar RS, Gordon JI. Thermodynamic studies of myristoyl-CoA:protein

N-myristoyltransferase using isothermal titration calorimetry. Method

Enzymol 1995; 250:467–486.23. Christensen JJ, Hansen LD, Izatt RM. Handbook of Proton Ionization

Heats and Related Thermodynamic Quantities. New York: John Wiley &

Sons, 1976.24. Fukada H, Takahashi K. Enthalpy and heat capacity changes for the proton

dissociation of various buffer components in 0.1m potassium chloride. Proteins

1998; 33:159–166.25. Jelesarov I, Bosshard HR. Thermodynamics of ferredoxin binding to

ferredoxin:NADPþ reductase and the role of water at the complex interface.

Biochemistry 1994; 33:13321–13328.26. Cooper A, Johnson CM. Isothermal titration microcalorimetry. Method Mol

Biol 1994; 22:137–150.27. Sigurskjold BW, Altman E, Bundle DR. Sensitive titration microcalorimetric

study of the binding of salmonella O-antigenic oligosaccharides by a

monoclonal antibody. Eur J Biochem 1991; 197:239–246.28. Bundle DR, Sigurskjold BW. Determination of accurate thermodynamics of

binding by titration microcalorimetry. Method Enzymol 1994; 247:288–305.29. Cooper A, Johnson CM. Introduction to microcalorimetry and biomolecular

energetics. Method Mol Biol 1994; 22:109–124.30. Fisher HF, Singh N. Calorimetric methods for interpreting protein-ligand

interactions. Method Enzymol 1995; 259:194–221.31. Wymann J, Gill SJ. Binding and Linkage: Functional Chemistry of Biological

Macromolecules. Mill Valley: University Science Books, 1990.32. Connors KA. Binding Constants: The Measurement of Molecular Complex

Stability. New York: John Wiley & Sons, 1987.33. Di Cera E. Thermodynamic Theory of Site-Specific Binding Processes in

Biological Macromolecules. Cambridge: Cambridge University Press, 1995.34. Gopal B, Swaminathan CP, Bhattacharya S, Bhattacharya A, Murthy MR,

Surolia A. Thermodynamics of metal ion binding and denaturation of a

calcium binding protein from Entamoeba histolytica. Biochemistry 1997;

36:10910–10916.35. Hyre DE, Spicer LD. Thermodynamic evaluation of binding interactions in the

methionine repressor system of Escherichia coli using isothermal titration

calorimetry. Biochemistry 1995; 34:3212–3221.36. Eisenstein E, Yu HD, Schwarz FP. Cooperative binding of the feedback

modifiers isoleucine and valine to biosynthetic threonine deaminase from

Escherichia coli. J Biol Chem 1994; 269:29423–29429.

45


ORDER REPRINTS

37. Bruzzese FJ, Connelly PR. Allosteric properties of inosine monophosphatedehydrogenase revealed through the thermodynamics of binding of inosine50-monophosphate and mycophenolic acid. Temperature dependent heatcapacity of binding as a signature of ligand-coupled conformational equilibria.Biochemistry 1997; 36:10428–10438.

38. Ferrari ME, Lohman TM. Apparent heat capacity change accompanying anonspecific protein–DNA interaction Escherichia coli SSB tetramer binding tooligodeoxyadenylates. Biochemistry 1994; 33:12896–12910.

39. Doyle ML, Louie G, Dal Monte PR, Sokoloski TD. Tight binding affinitiesdetermined from thermodynamic linkage to protons by titration calorimetry.Method Enzymol 1995; 259:183–194.

40. Baker BM, Murphy KP. Evaluation of linked protonation effects in proteinbinding reactions using isothermal titration calorimetry. Biophys J 1996;71:2049–2055.

41. Bradshaw JM, Waksman G. Calorimetric investigation of proton linkage bymonitoring both the enthalpy and association constant of binding: applicationto the interaction of the Src SH2 domain with a high-affinity tyrosylphosphopeptide. Biochemistry 1998; 37:15400–15407.

42. Doyle ML, Hensley P. Tight ligand binding affinities determined fromthermodynamic linkage to temperature by titration calorimetry. MethodEnzymol 1998; 295:88–99.

43. Li J, Swanson RV, Simon MI, Weis RM. The response regulators CheB andCheY exhibit competitive binding to the kinase CheA. Biochemistry 1995;34:14626–14636.

44. Hu DD, Eftink MR. Thermodynamic studies of the interaction of trpaporepressor with tryptophan analogs. Biophys Chem 1994; 49:233–239.

45. Khalifah RG, Zhang F, Parr JS, Rowe ES. Thermodynamics of binding of theCO2-competitive inhibitor imidazole and related compounds to humancarbonic anhydrase I: an isothermal titration calorimetry approach to studyingweak binding by displacement with strong inhibitors. Biochemistry 1993;32:3058–3066.

46. Sigurskjold BW, Berland CR, Svensson B. Thermodynamics of inhibitorbinding to the catalytic site of glucoamylase from Aspergillus niger determinedby displacement titration calorimetry. Biochemistry 1994; 33:10191–10199.

47. Sigurskjold BW. Exact analysis of competition ligand binding by displacementisothermal titration calorimetry. Anal Biochem 2000; 277:260–266.

48. Aqvist J, Medina C, Samuelsson JE. A new method for predicting bindingaffinity in computer-aided drug design. Protein Eng 1994; 7:385–391.

49. Cummings MD, Hart TN, Read RJ. Atomic solvation parameters in theanalysis of protein–protein docking results. Protein Sci 1995; 4:2087–2099.

50. Horton N, Lewis M. Calculation of the free energy of association for proteincomplexes. Protein Sci 1992; 1:169–181.

51. Krystek S, Stouch T, Novotny J. Affinity and specificity of serineendopeptidase-protein inhibitor interactions: empirical free energy calculationsbased on x-ray crystallographic structures. J Mol Biol 1993; 234:661–679.

52. Searle MS, Williams DH, Gerhard U. Partitioning of free energy contri-butions in the estimation of binding constants: residual motions and



ORDER REPRINTS

consequences for amide–amide hydrogen bond strengths. J Amer Chem Soc

1992; 114:10697–10704.53. Wallqvist A, Jernigan RL, Covell DG. A preference-based free-energy

parameterization of enzyme-inhibitor binding. Applications to HIV-1-protease

inhibitor design. Protein Sci 1995; 4:1881–1903.54. Williams DH, Cox JPL, Doig AJ, Gardner M, Gerhard U, Kaye PT, Lal AR,

Nicholls IA, Salter CJ, Mitchell RC. Toward the semiquantitative estimation

of binding constants: guides for peptide–petpide binding in aqueous solution.

J Amer Chem Soc 1991; 113:7020–7030.55. Williams DH, Searle MS, Mackay JP, Gerhard U, Maplestone RA. Toward an

estimation of binding constants in aqueous solution: studies of associations of

vancomycin group antibiotics. Proc Natl Acad Sci USA 1993; 90:1172–1178.56. Eftink MR, Anusiem AC, Biltonen RL. Enthalpy–entropy compensation and

heat capacity changes for protein–ligand interactions: general thermodynamic

models and data for the binding of nucleotides to ribonuclease A. Biochemistry

1983; 22:3884–3896.57. Lumry R, Rajender S. Enthalpy–entropy compensation phenomena in water

solutions of proteins and small molecules: a ubiquitous property of water.

Biopolymers 1970; 9:1125–1227.58. Chervenak MC, Toone EJ. A direct measure of the contribution of solvent

reorganization to the enthalpy of ligand binding. J Amer Chem Soc 1994;

116:10533–10539.59. Connelly PR, Thomson JA, Fitzgibbon MJ, Bruzzese FJ. Probing hydration

contributions to the thermodynamics of ligand binding by proteins. Enthalpy

and heat capacity changes of tacrolimus and rapamycin binding to FK506

binding protein in D2O and H2O. Biochemistry 1993; 32:5583–5590.60. Ladbury JE. Just add water! The effect of water on the specificity of protein–

ligand binding sites and its potential application to drug design. Chem Biol

1996; 3:973–980.61. Holdgate GA, Tunnicliffe A, Ward WH, Weston SA, Rosenbrock G, Barth PT,

Taylor IW, Pauptit RA, Timms D. The entropic penalty of ordered water

accounts for weaker binding of the antibiotic novobiocin to a resistant mutant

of DNA gyrase: a thermodynamic and crystallographic study. Biochemistry

1997; 36:9663–9673.62. Bhat TN, Bentley GA, Boulot G, Greene MI, Tello D, Dall’Acqua W,

Souchon H, Schwarz FP, Mariuzza RA, Poljak RJ. Bound water molecules

and conformational stabilization help mediate an antigen–antibody associa-

tion. Proc Natl Acad Sci USA 1994; 91:1089–1093.63. Xavier KA, Shick KA, Smith-Gil SJ, Willson RC. Involvement of water

molecules in the association of monoclonal antibody HyHEL-5 with bobwhite

quail lysozyme. Biophys J 1997; 73:2116–2125.64. Goldbaum FA, Schwarz FP, Eisenstein E, Cauerhff A, Mariuzza RA,

Poljak RJ. The effect of water activity on the association constant and the

enthalpy of reaction between lysozyme and the specific antibodies D1.3 and

D44.1. J Mol Recognit 1996; 9:6–12.

47


ORDER REPRINTS

65. Kornblatt JA, Kornblatt MJ, Hoa GH, Mauk AG. Responses of two protein–protein complexes to solvent stress: does water play a role at the interface?Biophys J 1993; 65:1059–1065.

66. Robinson CR, Sligar SG. Molecular recognition mediated by bound water: amechanism for star activity of the restriction endonuclease EcoRI. J Mol Biol.1993; 234:302–306.

67. Pearce KH Jr, Ultsch MH, Kelley RF, de Vos AM, Wells JA. Structural andmutational analysis of affinity-inert contact residues at the growth hormone-receptor interface. Biochemistry 1996; 35:10300–10307.

68. Connelly PR, Aldape RA, Bruzzese FJ, Chambers SP, Fitzgibbon MJ, FlemingMA, Itoh S, Livingston DJ, Navia MA, Thomson JA, Wilson KP. Enthalpy ofhydrogen bond formation in a protein–ligand binding reaction. Proc Natl AcadSci USA 1994; 91:1964–1968.

69. Frisch C, Schreiber G, Johnson CM, Fersht AR. Thermodynamics of theinteraction of barnase and barstar: changes in free energy versus changes inenthalpy on mutation. J Mol Biol 1997; 267:696–706.

70. Mark AE, van Gunsteren WF. Decomposition of the free energy of a system interms of specific interactions. Implications for theoretical and experimentalstudies. J Mol Biol 1994; 240:167–176.

71. Brady GP, Sharp KA. Decomposition of interaction free energies in proteinsand other complex systems. J Mol Biol 1995; 254:77–85.

72. Boresch S, Karplus M. The meaning of component analysis: decomposition ofthe free energy in terms of specific interactions. J Mol Biol 1995; 254:801–807.

73. Murphy KP, Xie D, Garcia KC, Amzel LM, Freire E. Structural energeticsof peptide recognition: angiotensin II/antibody binding. Proteins 1993;15:113–120.

74. Murphy KP, Xie D, Thompson KS, Amzel LM, Freire E. Entropy in biologicalbinding processes: estimation of translational entropy loss. Proteins 1994;18:63–67.

75. Murphy KP, Freire E, Paterson Y. Configurational effects in antibody–antigeninteractions studied by microcalorimetry. Proteins 1995; 21:83–90.

76. Jelesarov I, Bosshard HR. Isothermal titration calorimetry and differentialscanning calorimetry as complementary tools to investigate the energetics ofbiomolecular recognition. J Mol Recognit 1999; 12:3–18.

77. Murphy KP, Bhakuni V, Xie D, Freire E. Molecular basis of cooperativity inprotein folding. III. Structural identification of cooperative folding units andfolding intermediates. J Mol Biol 1992; 227:293–306.

78. Gomez J, Hilser VJ, Xie D, Freire E. The heat capacity of proteins. Proteins1995; 22:404–412.

79. Gomez J, Freire E. Thermodynamic mapping of the inhibitor site of theaspartic protease endothiapepsin. J Mol Biol 1995; 252:337–350.

80. Murphy KP, Freire E. Thermodynamics of structural stability and cooperativefolding behavior in proteins. Adv Protein Chem 1992; 43:313–361.

81. Spolar RS, Livingstone JR, Record MT Jr. Use of liquid hydrocarbon andamide transfer data to estimate contributions to thermodynamic functions ofprotein folding from the removal of nonpolar and polar surface from water.Biochemistry 1992; 31:3947–3955.



ORDER REPRINTS

82. Baker BM, Murphy KP. Prediction of binding energetics from structure using

empirical parameterization. Method Enzymol 1998; 295:294–315.83. Luque I, Freire E. Structure-based prediction of binding affinities and

molecular design of peptide ligands. Method Enzymol 1998; 295:100–127.84. Dunitz JD. Win some, lose some: enthalpy–entropy compensation in weak

intermolecular interactions. Chem Biol 1995; 2:709–712.85. Gilli P, Ferretti V, Gilli G, Borea PA. Enthalpy–entropy compensation in

drug receptor binding. J Phys Chem 1994; 98:1515–1518.86. Doyle ML. Characterization of binding interactions by isothermal titration

calorimetry. Curr Opin Biotechnol 1997; 8:31–35.87. Makhatadze GI, Privalov PL. Contribution of hydration to protein folding

thermodynamics I. The enthalpy of hydration. J Mol Biol 1993; 232:639–659.88. Privalov PL, Gill SJ. Stability of protein structure and hydrophobic

interaction. Adv Protein Chem 1988; 39:191–234.89. Privalov PL, Makhatadze GI. Contribution of hydration to protein folding

thermodynamics. II. The entropy and Gibbs energy of hydration. J Mol Biol

1993; 232:660–679.90. Spolar RS, Ha JH, Record MT Jr. Hydrophobic effect in protein folding and

other noncovalent processes involving proteins. Proc Natl Acad Sci USA

1989; 86:8382–8385.91. Xie D, Freire E. Molecular basis of cooperativity in protein folding. V.

Thermodynamic and structural conditions for the stabilization of compact

denatured states. Proteins 1994; 19:291–301.92. Janin J, Chothia C. The structure of protein–protein recognition sites. J Biol

Chem 1990; 265:16027–16030.93. Burrows SD, Doyle ML, Murphy KP, Franklin SG, White JR, Brooks I,

McNulty DE, Scott MO, Knutson JR, Porter D, Young PR, Hensley P.

Determination of the monomer-dimer equilibrium of interleukin-8 reveals

it is a monomer at physiological concentrations. Biochemistry 1994;

33:12741–12745.94. Baker BM, Murphy KP. Dissecting the energetics of a protein–protein

interaction: the binding of ovomucoid third domain to elastase. J Mol Biol

1997; 268:557–569.95. Lee B, Richards FM. The interpretation of protein structures: estimation of

static accessibility. J Mol Biol 1971; 55:379–400.96. Hubbard SJ, Thornton JM. NACCESS Computer Program. 2.1.1 ed.

London: Department of Biochemisry & Molecular Biology, University

College, 1996.97. Luque I, Freire E. Structural parameterization of the binding enthalpy of

small ligands. Proteins 2002; 49:181–190.98. Murphy KP, Gill SJ. Solid model compounds and the thermodynamics of

protein unfolding. J Mol Biol 1991; 222:699–709.99. Murphy KP, Privalov PL, Gill SJ. Common features of protein unfolding and

dissolution of hydrophobic compounds. Science 1990; 247:559–561.100. Baldwin RL. Temperature dependence of the hydrophobic interaction in

protein folding. Proc Natl Acad Sci USA 1986; 83:8069–8072.

49


ORDER REPRINTS

101. Finkelstein AV, Janin J. The price of lost freedom: entropy of bimolecularcomplex formation. Protein Eng 1989; 3:1–3.

102. Tamura A, Privalov PL. The entropy cost of protein association. J Mol Biol1997; 273:1048–1060.

103. Amzel LM. Loss of translational entropy in binding, folding, and catalysis.Proteins 1997; 28:144–149.

104. Kauzmann W. Some factors in the interpretation of protein denaturation. AdvProtein Chem 1959; 14:1–63.

105. Gilson MK, Given JA, Bush BL, McCammon JA. The statistical-thermo-dynamic basis for computation of binding affinities: a critical review. BiophysJ 1997; 72:1047–1069.

106. Holtzer A. The ‘‘cratic correction’’ and related fallacies. [published erratumappears in biopolymers 1996; 39:753]. Biopolymers 1995; 35:595–602.

107. Murphy KP, Freire E. Thermodynamic strategies for rational protein anddrug design. Pharm Biotechnol 1995; 7:219–241.

108. Creamer TP, Rose GD. Side-chain entropy opposes alpha-helix formation butrationalizes experimentally determined helix-forming propensities. Proc NatlAcad Sci USA 1992; 89:5937–5941.

109. Creamer TP, Rose GD. Alpha-helix-forming propensities in peptides andproteins. Proteins 1994; 19:85–97.

110. D’Aquino JA, Gomez J, Hilser VJ, Lee KH, Amzel LM, Freire E. Themagnitude of the backbone conformational entropy change in protein folding.Proteins 1996; 25:143–156.

111. Lee KH, Xie D, Freire E, Amzel LM. Estimation of changes in side chainconfigurational entropy in binding and folding: general methods andapplication to helix formation. Proteins 1994; 20:68–84.

112. Bardi JS, Luque I, Freire E. Structure-based thermodynamic analysis ofHIV-1 protease inhibitors. Biochemistry 1997; 36:6588–6596.

113. Wild K, Bohner T, Aubry A, Folkers G, Schulz GE. The three-dimensionalstructure of thymidine kinase from Herpes simplex virus type 1. FEBS Lett1995; 368:289–292.

114. Wild K, Bohner T, Folkers G, Schulz GE. The structures of thymidine kinasefrom Herpes simplex virus type 1 in complex with substrates and a substrateanalogue. Protein Sci 1997; 6:2097–2106.

115. Champness JN, Bennett MS, Wien F, Visse R, Summers WC, Herdewijn P,de Clerq E, Ostrowski T, Jarvest RL, Sanderson MR. Exploring the active siteof Herpes simplex virus type-1 thymidine kinase by x-ray crystallography ofcomplexes with aciclovir and other ligands. Proteins 1998; 32:350–361.

116. Brown DG, Visse R, Sandhu G, Davies A, Rizkallah PJ, Melitz C, SummersWC, Sanderson MR. Crystal structures of the thymidine kinase from Herpessimplex virus type-1 in complex with deoxythymidine and ganciclovir. NatureStruct Biol 1995; 2:876–881.

117. Bennett MS, Wien F, Champness JN, Batuwangala T, Rutherford T,Summers WC, Sun H, Wright G, Sanderson MR. Structure to 1.9 Aresolution of a complex with Herpes simplex virus type-1 thymidine kinaseof a novel, non-substrate inhibitor: X-ray crystallographic comparison withbinding of aciclovir. FEBS Lett 1999; 443:121–125.



ORDER REPRINTS

118. Perozzo R, Jelesarov I, Bosshard HR, Folkers G, Scapozza L. Compulsoryorder of substrate binding to Herpes simplex virus type 1 thymidine kinase.A calorimetric study. J Biol Chem 2000; 275:16139–16145.

119. Pilger BD, Perozzo R, Alber F, Wurth C, Folkers G, Scapozza L. Substratediversity of Herpes simplex virus thymidine kinase—impact of the kinematicsof the enzyme. J Biol Chem 1999; 274:31967–31973.

120. Schlauderer GJ, Proba K, Schulz GE. Structure of a mutant adenylate kinaseligated with an ATP-analogue showing domain closure over ATP. J Mol Biol1996; 256:223–227.

121. Muller CW, Schlauderer GJ, Reinstein J, Schulz GE. Adenylate kinasemotions during catalysis: an energetic counterweight balancing substratebinding. Structure 1996; 4:147–156.

122. Kaspar S, Perozzo R, Reinelt S, Meyer M, Pfister K, Scapozza L, Bott M. Theperiplasmic domain of the histidine autokinase CitA functions as a highlyspecific citrate receptor. Mol Microbiol 1999; 33:858–872.

123. Stock AM, Robinson VL, Goudreau PN. Two-component signal transduc-tion. Annu Rev Biochem 2000; 69:183–215.

124. Meyer M. Regulation of the Citrate Fermentation Genes in Klebsiellapneumoniae: Functional Analysis of the Two-Component System CitA/CitBand of the cAMP Receptor Protein. Dissertation, Zurich, 1999.

125. Bott M, Meyer M, Dimroth P. Regulation of anaerobic citrate metabolism inKlebsiella pneumoniae. Mol Microbiol 1995; 18:533–546.

126. Sillen LG, Martell AE. Stability Constants of Metal-ion Complexes. 2nd ed.London: The Chemical Society, 1964.

127. Wilson JE, Chin A. Chelation of divalent cations by A.TP, studied by titrationcalorimetry. Anal Biochem 1991; 193:16–19.

128. Remington S, Wiegand G, Huber R. Crystallographic refinement and atomicmodels of two different forms of citrate synthase at 2.7 and 1.7 A resolution.J Mol Biol 1982; 158:111–152.

129. Glusker JP. Structural aspects of citrate biochemistry. Curr Top Cell Regul1992; 33:169–184.

130. Russell RJ, Ferguson JM, Hough DW, Danson MJ, Taylor GL. The crystalstructure of citrate synthase from the hyperthermophilic Archaeon Pyrococcusfuriosus at 1.9 A resolution. Biochemistry 1997; 36:9983–9994.

131. Glusker JP. Citrate conformation and chelation: enzymatic implications.Acc Chem Res 1980; 13:345–352.

132. Gerharz T, Reinelt S, Kaspar S, Scapozza L, Bott M. Identification of basicamino acid residues important for citrate binding by the periplasmic receptordomain of the sensor kinase CitA. Biochemistry 2003; 42:5917–5924.

133. Reinelt S, Hofmann E, Gerharz T, Bott M, Madden DR. The structure of theperiplasmic ligand-binding domain of the sensor kinase CitA reveals the firstextracellular PAS domain. J Biol Chem 2003; 278:39189–39196.

134. Lin LN, Li J, Brandts JF, Weis RM. The serine receptor of bacterialchemotaxis exhibits half-site saturation for serine binding. Biochemistry 1994;33:6564–6570.

135. Evans LJ, Cooper A, Lakey JH. Direct measurement of the association of aprotein with a family of membrane receptors. J Mol Biol 1996; 255:559–563.

51


ORDER REPRINTS

136. Krell T, Renauld-Mongenie G, Nicolai MC, Fraysse S, Chevalier M,Berard Y, Oakhill J, Evans RW, Gorringe A, Lissolo L. Insight intothe structure and function of the transferrin receptor from Neisseriameningitidis using microcalorimetric techniques. J Biol Chem 2003;278:14712–14722.

137. Hippler-Mreyen S, Klare JP, Wegener AA, Seidel R, Herrmann C, Schmies G,Nagel G, Bamberg E, Engelhard M. Probing the sensory rhodopsin II bindingdomain of its cognate transducer by calorimetry and electrophysiology. J MolBiol 2003; 330:1203–1213.

138. Spudich JL. Variations on a molecular switch: transport and sensorysignalling by archaeal rhodopsins. Mol Microbiol 1998; 28:1051–1058.

139. Wegener AA, Klare JP, Engelhard M, Steinhoff HJ. Structural insights intothe early steps of receptor-transducer signal transfer in archaeal phototaxis.EMBO J 2001; 20:5312–5319.

140. Fleming KG. Probing stability of helical transmembrane proteins. MethodEnzymol 2000; 323:63–77.

141. Presnell SR. ACCESS Computer Program. 3.1 ed. Seattle: ZymoGenetics, 1994.



thermodynamics of protein ligand interactions: history,...

Documents