Understanding binding affinity: A combined isothermal titration calorimetry/molecular dynamics study of the binding of a series of hydrophobically modified benzamidinium chloride inhibitors to trypsin

Article abstract of DOI:10.1021/ja034676g

The binding of a series of p-alkylbenzamidinium chloride inhibitors to the serine proteinase trypsin over a range of temperatures has been studied using isothermal titration (micro)calorimetry and molecular dynamics simulation techniques. The inhibitors h

Full text of DOI:10.1021/ja034676g

Published on Web 08/09/2003
Understanding Binding Affinity: A Combined Isothermal
Titration Calorimetry/Molecular Dynamics Study of the
Binding of a Series of Hydrophobically Modified
Benzamidinium Chloride Inhibitors to Trypsin
Reinskje Talhout,† Alessandra Villa,‡ Alan E. Mark,‡ and Jan B. F. N. Engberts*,†
Contribution from the Physical Organic Chemistry Unit, Stratingh Institute, UniVersity of
Groningen, and the Groningen Biomolecular Sciences and Biotechnology Institute,
Department of Biophysical Chemistry, UniVersity of Groningen,
Nijenborgh 4, 9747 AG Groningen, The Netherlands
Received February 14, 2003; E-mail: J.B.F.N.Engberts@chem.rug.nl
Abstract: The binding of a series of p-alkylbenzamidinium chloride inhibitors to the serine proteinase trypsin
over a range of temperatures has been studied using isothermal titration (micro)calorimetry and molecular
dynamics simulation techniques. The inhibitors have small structural variations at the para position of the
benzamidinium ion. They show small differences in relative binding affinity but large compensating
differences in enthalpy and entropy. Binding affinity decreases with increased branching at the first carbon
but increases with increasing the length of a linear alkyl substituent, suggesting that steric hindrance and
hydrophobic interactions play dominant roles in binding. Structural analysis showed that the backbone of
the enzyme was unaffected by the change of the para substituent. In addition, binding does not correlate
strongly with octanol/water partition data. To further characterize this system, the change in the heat capacity
on binding, the change in solvent-accessible surface area on binding, the effect of inhibitor binding on the
hydration of the active site, the pKa of His57, and interactions within the catalytic triad have been investigated.
Although the changes in inhibitor structure are small, it is demonstrated that simple concepts such as
steric hindrance, hydrophobicity, and buried surface area are insufficient to explain the binding data. Other
factors, such as access to the binding site and the cost of dehydration of the active site, are of equal or
greater importance.
fertilization (plasmin).6,7 They are specific for arginine or lysine,
and many natural peptide inhibitors have an arginine or lysine
Introduction
Knowledge of the factors that determine the binding affinity
of low-molecular-weight compounds to proteins is of funda-
mental scientific interest as well as a prerequisite for structure-
based drug design.1,2 One approach to evaluate the contribution
of a specific part of the inhibitor is to study the influence of
small structural variations on the thermodynamics of binding.3-5
Here we present a combined structural and thermodynamic study
of the binding of p-alkylbenzamidinium chloride inhibitors to
the serine proteinase trypsin. Serine proteinases are important
targets for medicinal chemistry that are associated with a wide
range of biologically important processes, including digestion
(trypsin, elastase), blood coagulation (thrombin, factor Xa), and
that binds in the specificity pocket. Synthetic inhibitors are often
based on analogous groups.8 In particular, benzamidinium, a
structural mimic of arginine and itself a potent inhibitor,9 is the
primary component of many larger inhibitors.4,10,11 The struc-
tures of benzamidinium, and a wide range of benzamidinium
derivatives, complexed with trypsin are known.12,13
Figure 1 shows the benzamidinium ion bound in the pocket
S1 of trypsin, hydrogen-bonded to Asp189, Gly219, Ser190,
and an internal water molecule. In other systems, water
molecules have been shown to complement the interaction
between a ligand and a receptor. The net thermodynamic effect
of such water molecules, however, varies.2 Benzamidinium
† Physical Organic Chemistry Unit, Stratingh Institute.
‡ Groningen Biomolecular Sciences and Biotechnology Institute (GBB),
Department of Biophysical Chemistry.
(6) Proteases, Potential Role in Health and Disease; Ho¨rl, W. H., Heidland,
A., Eds.; Plenum Press: New York, 1984.
(7) Proteinases in Mammalian Cells and Tissues; Barrett, A. J., Ed.; Elsevier:
Amsterdam, 1977.
(1) Bo¨hm, H.-J.; Klebe, G. Angew. Chem., Int. Ed. Engl. 1996, 35, 2588-
2614.
(8) Proteinase Inhibitors; Barrett, A. J., Salvesen, G., Eds.; Elsevier: Am-
sterdam, 1986.
(2) Structure-Based Drug Design: Thermodynamics, Modeling and Strategy;
Ladbury, J. E., Connelly, P. R., Eds.; Springer: Berlin, 1997.
(3) Fersht, A. Structure and Mechanism in Protein Science; Freeman: New
York, 1999.
(9) Mares-Guia, M.; Shaw, E. J. Biol. Chem. 1965, 240, 1579-1585.
(10) Bohm, M.; Stu¨rzebecher, J.; Klebe, G. J. Med. Chem. 1999, 42, 458-477.
(11) Renatus, M.; Bode, W.; Huber, R.; Stu¨rzebecher, J.; Stubbs, M. T. J. Med.
Chem. 1998, 41, 5445-5456.
(4) Dullweber, F.; Stubbs, M. T.; Musil, D.; Stu¨rzebecher, J.; Klebe, G. J.
Mol. Biol. 2001, 313, 593-614.
(5) Morgan, B. P.; Scholtz, J. M.; Ballinger, M. D.; Zipkin, I. D.; Bartlett, P.
A. J. Am. Chem. Soc. 1991, 113, 297-307.
(12) Marquart, M.; Walter, J.; Deisenhofer, J.; Bode, W.; Huber, R. Acta
Crystallogr. 1983, B39, 480-490.
(13) Bode, W.; Schwager, P. J. Mol. Biol. 1975, 98, 693-717.
9
10570
J. AM. CHEM. SOC. 2003, 125, 10570-10579
10.1021/ja034676g CCC: $25.00 © 2003 American Chemical Society

Understanding Binding Affinity
A R T I C L E S
Figure 1. Backbone representation of the benzamidinium-trypsin complex12 (3PTB from the Protein Data Bank) generated using RasMol.14 Benzamidinium
(green) is bound in the S1 binding pocket, close to residues of the catalytic triad, His57 and Ser195. The phenyl ring of the benzamidinium is in van der
Waals contact with residues 214-216 and 190-191. Also shown is the hydrophobic pocket S3/S4, defined by Trp215 and Leu99.
+ 16
Chart 1. Structures of the p-Alkylbenzamidinium Chlorides
Studieda
with Hammett substituent constants (σp
)
ranging between
-0.31 and -0.26. Any significant differences in binding
affinities of the inhibitors cannot be attributed to inductive or
resonance effects of the substituents.
To quantify the effect of varying the substituent, ITC was
used to acquire a complete thermodynamic description of
binding. The advantage of ITC over kinetic techniques is that
it is efficient and noninvasive.17-19 One experiment provides
the binding constant and the enthalpy of binding, from which
the Gibbs energy and entropy of binding are readily calculated.
The enthalpic and entropic contributions to the Gibbs energy
of binding were used to infer information regarding the
mechanism of binding. While small structural variations in an
inhibitor often result in only small changes in the Gibbs energy
of binding, the enthalpy and entropy may vary considerably.
Such enthalpy-entropy compensation is common when studying
weak interactions and is important for understanding the nature
of competing interactions.20-22 The temperature dependence of
the enthalpy of binding also gives valuable information on the
origin of the interaction.2
a R ) Me, Et, n-Pr, i-Pr, n-Bu, t-Bu, n-Pent, n-Hex.
derivatives containing large hydrophobic groups often also
occupy the hydrophobic S3/S4 groove, defined by the residues
Trp215 and Leu99.4,11
There have been many studies of the binding of benzami-
dinium derivatives to trypsin. Often, however, only the binding
constant has been considered, potentially leading to the (er-
roneous) conclusion that some structural variations do not affect
binding significantly. Also, the structural variations are often
large, which makes it difficult to infer the precise reason for
changes in binding affinity. In this work we consider the effect
of small systematic changes to the para substituent of benza-
midinium on the thermodynamics of binding to trypsin. Previ-
ously,15 we showed that the electronic properties of a substituent
strongly affect binding. Here we consider the role that steric
and hydrophobic effects play in determining the interaction with
the enzyme.
To systematically probe the role of steric and hydrophobic
effects in binding, we have studied a series of p-alkyl-substituted
benzamidinium chlorides (Chart 1) by means of isothermal
titration calorimetry (ITC) and molecular dynamics (MD)
simulation techniques. The substituents differ in the degree of
branching at the first carbon atom and the length of the carbon
chain. Their electronic properties are, however, very similar,
The enthalpy of binding, the entropy of binding, and the Gibbs
energy are, however, global properties of the system. They can
neither be used to infer the precise nature of an interaction, nor
to provide a detailed structural model. MD computer simulations
have, therefore, been used to link the experimentally measured
(16) Exner, O. The Hammett EquationsThe Present Position. In AdVances in
Linear Free Energy Relationships; Chapman, N. B., Shorter, J., Eds.;
Plenum Press: London, 1972; pp 1-69.
(17) Holdgate, G. A. Biotechniques 2001, 30, 164-184.
(18) Leavitt, S.; Freire, E. Curr. Opin. Struct. Biol. 2001, 11, 560-566.
(19) Applications of Calorimetry in the Biological Sciences; Ladbury, J. E.,
Chowdry, B. Z., Eds.; Wiley: New York, 1998.
(20) Calderone, C. T.; Williams, D. H. J. Am. Chem. Soc. 2001, 123, 6262-
6267.
(14) Rasmol 2.7.2; Herbert J. Bernstein, August 2000.
(21) Cooper, A.; Johnson, C. M.; Lakey, J. H.; No¨llmann, M. Biophys. Chem.
2001, 93, 215-230.
(15) Talhout, R.; Engberts, J. B. F. N. Eur. J. Biochem. 2001, 268, 1554-
1560.
(22) Dunitz, J. D. Chem. Biol. 1995, 2, 709-712.
9
J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003 10571

A R T I C L E S
Talhout et al.
evaluated using a twin range cutoff. Interactions within the shorter-
range cutoff (0.9 nm) were evaluated every step, whereas interactions
within the longer cutoff (1.4 nm) were updated every five steps, together
with the pair list. To correct for the neglect of electrostatic interactions
thermodynamic properties to the microscopic structure of the
system. Specifically, MD simulations have been used to study
the structural changes that accompany binding and to predict
the change in Gibbs energy given a particular structural model.
beyond the 1.4 nm cutoff, a reaction field (RF) correction with ꢀRF
)
78.0 was used. To maintain constant temperature and pressure, a
Berendsen thermostat33 was applied. The protein, inhibitor, and solvent
were independently coupled to a temperature bath (25 °C) with a
coupling time of 0.1 ps. The pressure was held at 1 bar, with a coupling
Experimental Section
Synthesis of Inhibitors. The p-alkylbenzamidinium chlorides were
synthesized using modified literature procedures.23-25 Details of the
synthesis and analysis of the products are given in the Supporting
Information.
time of 0.5 ps. The isothermal compressibility was 4.6 × 10-5 bar-1 33
.
The time step was 0.002 ps. The bond lengths and angle in water were
constrained using the SETTLE algorithm.34 Bond lengths within the
protein were constrained using the LINCS algorithm.35
Isothermal Titration Calorimetry. Trypsin solutions were prepared
as described previously.15 Titration experiments were performed using
either an Omega isothermal titration calorimeter (Microcal, Inc.,
Northampton, MA) coupled to a nanovolt preamplifier in order to
improve the signal-to-noise ratio or an MCS isothermal titration
calorimeter (Microcal, Inc.). Both machines were connected to a water
bath for temperature control. The instruments were calibrated using
standard electrical pulses. Inhibitor solutions (5-10 mM, determined
by weight, depending on the specific inhibitor) were titrated into the
stirred (350 rpm) cell (1.3249 or 1.3496 mL, respectively) containing
a degassed (ca. 10 min) trypsin solution (ca. 0.15 mM) after a stable
baseline (rms noise < 0.0050) was achieved. The injection sequence
consisted of an initial injection of 1 µL to prevent artifacts arising from
the filling of the syringe (not used in data fitting), followed by injections
of 5 µL each at 300-s intervals until saturation was reached. To correct
for the heat of dilution and mixing, blank titrations of inhibitor into
buffer were subtracted from the inhibitor-enzyme titration. Data were
analyzed using Origin software (Microcal, Inc.), assuming a single
binding site.26 This yielded ∆H (enthalpy of binding) and K (binding
constant). Measurements were repeated at least three times; K was
reproducible to within 10%, and ∆H was reproducible to within 5%.
Errors in ∆Cp were within 5%.
Computational Techniques. The starting structure of the benza-
midinium-trypsin complex was taken from the Brookhaven Protein
Data Bank (PDB structure 3PTB).12 The nine inhibitors were placed
in the S1 pocket of the enzyme by superimposing the benzamidinium
ion within the crystallographic complex. The protein and the bound
inhibitor were hydrated in a box containing ∼5700 simple point charge
(SPC)27 water molecules. The protein and inhibitors were described
using the GROMOS96 (43a2) force field,28,29 in which aliphatic
hydrogen atoms are treated as united atoms, together with the carbon
atom to which they are attached. The charges of ionizable groups were
appropriate for pH 7.0. Arg and Lys were protonated. Asp and Glu
were unprotonated. Histidines were neutral. Tautomeric forms were
based on local interactions. For His57, which is close to the binding
pocket, the Nδ1-H tautomer was chosen.
Free Energy Calculations. The difference in Gibbs energy between
two states of a system was determined using the coupling parameter
approach in conjunction with the thermodynamic integration (TI)
formula:36
λB
∂H(λ)
∂λ
∆GA-B
)
dλ
λ
(1)
∫
λA
In this approach, the Hamiltonian H is made a function of a coupling
parameter, λ. The λ-dependence of the Hamiltonian defines a pathway
which connects two states of the system, denoted by A and B. To solve
eq 1, the ensemble average at a number of discrete λ-points was
obtained by performing separate simulations for each chosen λ-point,
and the integral was determined numerically.
The change in Gibbs energy resulting from the changes in the
inhibitor (or protein) was achieved by gradually mutating the atoms of
state A into the atoms of state B. Atoms with no corresponding atoms
in the other state were mutated to “dummy” atoms. A dummy atom
has no nonbonded (Lennard-Jones or electrostatic) interactions with
other atoms. The mutations in this study involved only nonbonded
interactions. The bonded interactions and the masses of the atoms were
not changed, as they make no net contribution to the change in binding
Gibbs energy. The nonbonded interactions between the initial state (A)
and the final state (B) were interpolated using a soft-core potential,37
as implemented in the GROMACS simulation package.31,38
Separate simulations were performed at 18 λ-points from λ ) 0 (state
A) to λ ) 1 (state B). At each λ-point, the system was equilibrated for
50 ps and data were collected for 150 (free inhibitor) or 250 ps
(complex). For complexes with alkyl chains longer than n-propyl, 22
λ points were used with 500 ps sampling. To obtain ∆GA-B, the average,
∂H(λ)/∂λ , at each λ-point (eq 1) was integrated using the trapezoidal
λ
method. The error in ∂H(λ)/∂λ λ was estimated using a block averaging
procedure,39,40 and the errors were integrated to give the total error in
∆GA-B
.
All simulations were performed using the GROMACS (version 3.0)
package30-32 in a periodic triclinic box. Nonbonded interactions were
Results and Discussion
(23) Glock, G. Chem. Ber. 1888, 21, 2650-2659.
Binding Affinity. The thermodynamics of binding of p-
alkylbenzamidinium chlorides (Chart 1) to trypsin were studied
using ITC. Since the solubility of p-n-hexylbenzamidinium
chloride (∼15 mM) was only 3 times higher than the concentra-
tion necessary for the experiments (∼5 mM), solubility problems
(24) Rash, F. H.; Boatman, S.; Hauser, C. R. J. Org. Chem. 1967, 32, 372-
376.
(25) Moss, R. A.; Ma, W.; Merrer, D. C.; Xue, S. Tetrahedron Lett. 1995, 36,
8761-8764.
(26) Wiseman, T.; Williston, S.; Brandts, J. F.; Lin, L. N. Anal. Biochem. 1989,
179, 131-137.
(27) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; Hermans, J.
In Intermolecular Forces; Pullman, B., Ed.; Reidel: Dordrecht, 1981; pp
331-342.
(28) van Gunsteren, W. F.; Billeter, S. R.; Eising, A. A.; Hu¨nenberger, P. H.;
Kru¨ger, P.; Mark, A. E.; Scott, W. R. P.; Tironi, I. G. Biomolecular
Simulation: GROMOS96 Manual and User Guide; BIOMOS b.v.: Zu¨rich,
Groningen, 1996.
(33) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; DiNola, A.;
Haak, J. R. J. Chem. Phys. 1984, 81, 3684-3690.
(34) Miyamoto, S.; Kollman, P. A. J. Comput. Chem. 1992, 13, 952-962.
(35) Hess, B.; Bekker, H.; Berendsen, H. J. C.; Fraaije, J. G. E. M. J. Comput.
Chem. 1997, 18, 1463-1472.
(29) Schuler, L. D.; van Gunsteren, W. F. Mol. Sim. 2000, 25, 301-319.
(30) Berendsen, H. J. C.; van der Spoel, D.; van Drunen, R. Comput. Phys.
Commun. 1995, 91, 43-56.
(31) Lindahl, E.; Hess, B.; van der Spoel, D. J. Mol. Model. 2001, 7, 306-317.
(32) van der Spoel, D.; van Buuren, A. R.; Apol, E.; Meulenhoff, P. J.; Tieleman,
D. P.; Sijbers, A. L. T. M.; Hess, B.; Feenstra, K. A.; Lindahl, E.; van
Drunen, R.; Berendsen, H. J. C. Gromacs User Manual Version 3.0;
Nijenborgh 4, 9747 AG Groningen, The Netherlands, 2001 (http://
www.gromacs.org).
(36) Mark, A. E. In Encyclopedia of Computational Chemistry; von Rague´
Schleyer, P., Ed.; Wiley: New York, 1998; pp 1070-1083.
(37) Beutler, T. C.; Mark, A. E.; van Schaik, R. C.; Geber, P. R.; van Gunsteren,
W. F. Chem. Phys. Lett. 1994, 222, 529-539.
(38) Villa, A.; Mark, A. E. J. Comput. Chem. 2002, 23, 548-553.
(39) Bishop, M.; Frinks, S. J. Chem. Phys. 1987, 87, 3675-3676.
(40) Allen, M. P.; Tildesley, D. J. Computer Simulations of Liquids; Oxford
Science Publications: Oxford, 1987.
9
10572 J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003

Understanding Binding Affinity
A R T I C L E S
Table 1. Thermodynamic Parameters of Binding of
p-Alkylbenzamidinium Chlorides to Trypsin at Several
Temperaturesa
T
K
∆
G
∆
H
T∆S
1
1
1
1
Rb
(
°
C)
(104 M-
)
(kJ mol-
)
(kJ mol-
)
(kJ mol-
)
Hc
20.1
25.1
30.0
37.1
20.1
25.1
30.0
37.1
20.0
25.1
30.0
37.1
20.1
25.2
30.2
37.2
20.0
25.0
30.0
37.0
20.0
25.1
30.0
37.0
6.2 ( 0.5
4.5 ( 0.2
4.1 ( 0.3
3.1 ( 0.3
8.0 ( 0.7
6.9 ( 0.3
5.5 ( 0.2
4.1 ( 0.4
3.4 ( 0.1
2.9 ( 0.2
2.5 ( 0.2
2.1 ( 0.3
3.8 ( 0.3
3.1 ( 0.2
2.5 ( 0.1
2.0 ( 0.2
1.1 ( 0.08
-26.9 ( 0.2 -17.0 ( 0.4
-26.6 ( 0.3 -18.9 ( 0.4
-26.7 ( 0.2 -21.4 ( 1.0
-26.7 ( 0.3 -23.7 ( 1.2
9.9 ( 0.5
7.7 ( 0.6
5.4 ( 0.5
3.0 ( 0.7
Mec
Etd
-27.5 ( 0.3 -16.7 ( 0.4 10.8 ( 0.6
-27.6 ( 0.1 -18.5 ( 0.3
-27.5 ( 0.2 -20.7 ( 0.3
-27.4 ( 0.2 -23.8 ( 0.6
9.1 ( 0.2
6.8 ( 0.3
3.6 ( 0.5
-25.5 ( 0.1 -10.6 ( 0.2 14.9 ( 0.4
-25.4 ( 0.2 -13.9 ( 0.6 11.5 ( 0.7
-25.5 ( 0.2 -17.3 ( 0.5
-25.6 ( 0.3 -22.4 ( 0.8
8.2 ( 0.6
3.2 ( 0.5
n-Pr
i-Prd
n-Bu
t-Bue
-25.7 ( 0.2
-9.7 ( 0.4 16.0 ( 0.5
-25.7 ( 0.2 -12.7 ( 0.5 13.0 ( 0.6
-25.6 ( 0.1 -15.7 ( 0.6
-25.6 ( 0.3 -19.9 ( 0.7
-22.6 ( 0.2
9.9 ( 0.4
5.7 ( 0.8
-4.9 ( 0.2 18 ( 0.4
-7.0 ( 0.4 16 ( 0.5
-9.0 ( 0.4 14 ( 0.6
0.96 ( 0.09 -22.7 ( 0.2
0.84 ( 0.07 -22.8 ( 0.2
0.66 ( 0.05 -22.7 ( 0.2 -12 ( 0.5
11 ( 0.5
4.7 ( 0.3
3.8 ( 0.2
3.4 ( 0.2
2.7 ( 0.1
-
-26.2 ( 0.3
-26.2 ( 0.2
-6.1 ( 0.3 20.1 ( 0.5
-9.9 ( 0.5 16.3 ( 0.7
-26.3 ( 0.2 -13.4 ( 0.6 12.9 ( 0.7
Figure 2. (A) Raw data for titration of p-methylbenzamidinium chloride
into an aqueous trypsin (lower trace) and buffer (upper trace) solution at
25 °C in Tris at pH 8.0. (B) Enthalpogram retrieved from (A), corrected
for the heat of dilution; the line represents the least-squares fit to the single-
site binding model.26
-26.3 ( 0.2 -18.5 ( 0.9
7.8 ( 0.6
-
-
-
n-Pent 20.0
25.0
6.0 ( 0.5
5.8 ( 0.3
5.3 ( 0.5
5.0 ( 0.4
-26.8 ( 0.2
-6.6 ( 0.3 20.3 ( 0.6
-27.2 ( 0.1
-9.9 ( 0.4 17.3 ( 0.3
30.0
-27.4 ( 0.2 -12.8 ( 0.6 14.6 ( 0.7
-27.9 ( 0.2 -17.5 ( 0.8 10.4 ( 0.7
37.0
n-Hex 20.0 14 ( 1
-28.9 ( 0.2
-6.6 ( 0.2 22.2 ( 0.2
are anticipated for alkyl chains longer than n-hexyl. No attempt
was made to study longer chain p-n-alkylbenzamidinium
chlorides.
25.0 13 ( 0.6
-29.2 ( 0.1 -10.6 ( 0.2 18.6 ( 0.3
-29.4 ( 0.2 -15.1 ( 0.4 14.3 ( 0.2
30.0 12 ( 0.7
37.0 10 ( 0.6
-29.7 ( 0.2 -21.0 ( 0.6
8.7 ( 0.5
Figure 2A shows the raw data of a typical binding experiment,
the titration of p-methylbenzamidinium chloride to trypsin in
Tris buffer at 25 °C. Binding is strongly exothermic. Note that
the dilution peaks are small, endothermic, and equal in size.
This indicates that, at this concentration, no aggregation of the
inhibitor in water occurs. Figure 2B shows the corresponding
enthalpogram. Similar observations were made for all other
inhibitors. A binding isotherm (see Experimental Section) has
been fitted to the data to obtain values for K and ∆H. From
these values, ∆G and T∆S can be calculated using
a Titrations performed in 50 mM Tris, 10 mM CaCl2 buffer at pH 8.0.
b Substituent at para position. c Data taken from ref 15. d Since these
compounds contain some NH4Cl, control experiments were performed with
10 wt % NH4Cl added to a benzamidinium chloride solution at the usual
concentration. Within error, the thermodynamic parameters were identical.
Therefore, the concentration of these inhibitors, based on weight, can be
safely corrected for the amount of NH4Cl present. e When p-tert-butylben-
zamidinium chloride was titrated into trypsin, only peaks of dilution were
observed at 25 °C. Therefore, binding of this compound is negligible, and
it is therefore not considered an inhibitor.
∆G ) -RT ln K ) ∆H - T∆S
(2)
Titration experiments were performed at 20, 25, 30, and 37
°C in Tris buffer. The resulting thermodynamic parameters are
listed in Table 1. Benzamidinium chloride is also listed for
comparison but is not considered a member of this series due
to different electronic para-substituent effects. Changes in the
alkyl substituent at the para position lead to systematic differ-
ences in binding affinities (Table 1). At all temperatures studied,
p-n-hexylbenzamidinium chloride (the best inhibitor) binds ∼14
times more tightly than p-isopropylbenzamidinium chloride (the
worst).
In Figure 3, the change in Gibbs energy (∆∆G) at 25 °C
relative to p-methylbenzamidinium is plotted. Figure 3A shows
the effect of increasing the branching at the first position of the
chain, and Figure 3B shows the effect of increasing the length
of the chain. This relative order of binding was found for all
temperatures studied, except for a reversal of n-Pent and H at
20 °C and n-Pent and Me at 37 °C.
Figure 3. Experimental (b) and calculated (0) relative Gibbs energies of
binding to trypsin of p-alkylbenzamidinium ions with respect to the
p-methylbenzamidinium ion at 25 °C. The difference in Gibbs energy of
octanol/water partitioning41 (4) is also reported. (A) Values for the methyl-,
ethyl-, isopropyl-, and tert-butylbenzamidinium ions, showing the effect
of branching at the first carbon. (B) Values for the methyl-, ethyl-, n-propyl-,
n-butyl-, n-pentyl-, and n-hexylbenzamidinium ions, showing the effect of
extending the aliphatic chain.
for at least 15 ns (5 ns equilibration, 10 ns sampling). For alkyl
chains longer than n-propyl, the simulations were 25 ns (5 ns
equilibration; 20 ns sampling). In all cases, the structures
The structure and dynamics of the complexes were also
investigated using MD simulations. Each complex was simulated
9
J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003 10573

A R T I C L E S
Talhout et al.
Table 3. Experimental and Calculated Relative Binding Gibbs
Table 2. RMS Positional Deviations (nm)a of the Backbone
Atoms, the Backbone Atoms within 1.4 nm of the Inhibitor, the
Phenyl Ring, and Some Residues with Respect to the
Crystallographic Structure for the p-Alkylbenzamidinium-Trypsin
Complexes Studied
Energies, ∆∆GAfB, at 25 °C for All Mutations Investigated
1
∆∆GA B (kJ mol-
)
f
mutation
A f RB
R
exptl
calcd
Asp189,
H f Me
Me f Et
Et f n-Pr
n-Pr f n-Bu
n-Bu f n-Pent
n-Pent f n-Hex
Et f i-Pr
-1.0 ( 0.2
2.1 ( 0.1
-0.3 ( 0.3
-0.5 ( 0.3
-1.0 ( 0.1
-2.0 ( 0.1
2.7 ( 0.3
-1.3 ( 0.8
0.9 ( 0.9
backbone
(1.4 nm)
phenyl
ring
Ser190,
Gly219
Rb
backbone
-2.5 ( 0.9
-1.6 ( 1.0
-2.4 ( 0.7
-1.8 ( 0.9
1.1 ( 0.9
H
0.14
0.17
0.13
0.20
0.15
0.15
0.14
0.18
0.16
0.12
0.14
0.12
0.17
0.11
0.12
0.12
0.14
0.13
0.15
0.10
0.08
0.01
0.11
0.09
0.13
0.11
0.11
0.13
0.11
0.11
0.18
0.17
0.12
0.16
0.11
0.15
Me
Et
n-Pr
i-Pr
n-Bu
t-Bu
n-Pent
n-Hex
Et f t-Bu
-4.4 ( 1.1
Gibbs energies of binding calculated from the simulations show
good agreement with the experimental trends. The exception is
tert-butylbenzamidinium, which was predicted to have a binding
constant similar to that of n-butylbenzamidinium but did not
bind in ITC experiments. Note that, in the simulations, the
inhibitor is grown directly inside the S1 pocket. If the tert-butyl
inhibitor were unable to enter the S1 pocket, this would not be
reflected in the calculations. To test if the binding of tert-
butylbenzamidinium was on a time scale not detectable using
the standard ITC approach, a competitive binding experiment
was also performed. Trypsin solutions with and without tert-
butylbenzamidinium chloride (∼1 mM) were stored for one
week at 7 °C. The solution containing tert-butylbenzamidinium
and the control were then titrated with benzamidinium chloride.
The results were identical within the experimental error,
indicating no binding of tert-butylbenzamidinium.
a The RMS deviations were calculated for each frame after a least-squares
fit on the backbone. b Substituent at para position.
remained close to that of the benzamidinium-trypsin complex12
on which the starting configurations were based. Table 2 shows
the average root-mean-square (RMS) deviation of the complexes
with respect to the crystal structure 3PTB. The RMS deviation
was evaluated after performing a least-squares best fit on all
backbone atoms. The RMS deviations of the backbone atoms
with respect to the X-ray structure were between 0.13 and 0.20
nm (Table 2). The position of the phenyl ring of the inhibitor
was the same as that observed in the crystal structure 3PTB for
all compounds. No obvious strain is present in any of the
structures.
For comparison to the ITC results, the relative Gibbs energies
of binding of the inhibitors to trypsin at 25 °C were calculated.
The relative binding Gibbs energy (∆∆GAfB) of inhibitor B
with respect to inhibitor A was evaluated using the following
thermodynamic cycle:
For comparison to the ITC results and the free energy
calculations, the relative Gibbs energies estimated from octanol/
water partition constants41 of the isolated substituent groups (i.e.,
going from benzamidinium to p-methylbenzamidinium is com-
pared to going from methane to ethane, etc.) are also indicated
in Figure 3. Excluding tert-butylbenzamidinium, increasing the
number of carbons at the first position of the substituent results
in a systematic increase in the relative Gibbs energy of binding
both experimentally and in the free energy calculations. Increas-
ing the length of the chain initially results in a small rise and
then in a slight systematic decrease in the relative Gibbs energy
of binding. The absolute magnitude of the changes in Gibbs
energy of binding is, however, small when compared to the
corresponding estimates from octanol/water partition constants.
Experimentally, the binding constant decreases systematically
as the steric bulk at the first carbon of the substituent is
increased: KMe > KEt > Ki-Pr > Kt-Bu. One explanation might
be that increasing the steric bulk of the substituents leads to a
progressive distortion of the binding pocket and its surroundings.
However, from Table 2, it is evident that the structure of the
backbone of the protein is relatively unaffected. This is also in
line with crystallographic studies of trypsin complexes.11-13
Even tert-butylbenzamidinium, which does not bind experi-
mentally, could be accommodated in the structure with no major
distortion of the backbone. Interestingly, the larger backbone
deviations were observed for complexes of inhibitors with an
odd number of carbon atoms in the alkyl chain (methyl,
n-propyl, n-pentyl), although the significance of this is hard to
determine, given the small sample size. A rotation of Trp215
(at the beginning of pocket S3/S4) was observed in the
where ∆GA and ∆GB are the binding Gibbs energy of inhibitor
A and B, respectively, and ∆GA-B is the work required to mutate
inhibitor A into B in trypsin (∆GA-B(tryp.)) and in water
(∆GA-B(aq.)). As the Gibbs energy is a state function, the
change in Gibbs energy for any closed cycle is zero. Thus, the
difference ∆∆GAfB may be obtained either by calculating ∆GA
and ∆GB, which corresponds to simulating the process of
complexation, or by the computationally more efficient approach
of determining ∆GA-B(tryp.) and ∆GA-B(aq.). This corresponds
to the (nonphysical) mutation of A into B in the complex and
free in solution. The change in binding Gibbs energy was
estimated using the TI formula (see Experimental Section). In
Table 3, the experimental and calculated relative binding Gibbs
energies at 25 °C are reported for all the transformations
investigated.
The estimates from the free energy calculations of the changes
in Gibbs energy relative to p-methylbenzamidinium are shown
in Figure 3, along with the experimental results. Overall, relative
(41) Sangster, J. Octanol-Water Partition Coefficients: Fundamentals and
Physical Chemistry; Wiley: Chichester, 1997.
9
10574 J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003

Understanding Binding Affinity
A R T I C L E S
Figure 4. (A) Entropy and (B) enthalpy of binding of p-n-alkylbenzami-
dinium chlorides versus the number of carbons in the substituent at 20, 25,
30, and 37 °C in Tris at pH 8.0. Values and errors are listed in Table 1.
Figure 5. Average structure of the binding site of the p-n-butylbenzami-
dinium-trypsin complex at 25 °C. Green, the average structure of the
inhibitor; blue and red, configurations illustrating the range of movement
of the inhibitor within the binding site. These correspond to the extreme
values of the first eigenvector for the motion of the inhibitor obtained from
principle component analysis (PCA).
simulations of the p-n-pentylbenzamidinium-trypsin complex,
and smaller motions of this group were observed in the p-n-
propylbenzamidinium complex (results not shown). No major
systematic distortions in the S1 pocket associated with the
elongation of the tail were observed. In the case of the sterically
demanding alkyl groups on the para position of the inhibitor,
small distortions of the S1 pocket were observed involving
Asp189, Ser190, and Gly219 (Table 2). However, as the largest
distortion is only 0.18 nm (n-propylbenzamidinium), it is
doubtful if any are significant.
ingly, when the chain has an odd number of carbons, the
terminal methyl group of the tail points away from the pocket.
To analyze the dynamics of the system, a principal component
analysis (PCA)46-48 of the motions of the inhibitors inside the
S1 pocket was performed. All inhibitors show some motion
within the pocket. For the p-methyl-, p-ethyl-, p-n-propyl-,
p-isopropyl-, p-tert-butyl-, and p-n-butylbenzamidinium com-
plexes, the average structure corresponds to the most populated
conformation. This is illustrated in Figure 5 for the p-n-
butylbenzamidinium complex, which shows the largest motion
within this subset of inhibitors. The red and blue structures in
Figure 5 correspond to the extremes of the motion. The green
structure corresponds to the average (most populated) structure.
In contrast, the chains of p-n-pentyl- and p-n-hexylbenzami-
dinium move between two distinct conformations. Figure 6
shows the population distributions at 25 and 37 °C for the p-n-
hexylbenzamidinium complex. The blue and red structures
correspond to the predominate conformations at each temper-
ature. The average structure is again green. At 25 °C, the tail is
predominately in the S3/S4 pocket. At 37 °C, the two confor-
mations are equally populated. Figure 6C,D shows projections
of the first eigenvectors at 25 and 37 °C, respectively. Clearly,
the distribution at 37 °C is wider than that at 25 °C. This
suggests that the entropy of the bound inhibitor is higher at 37
than at 25 °C. The overall change in entropy on binding is,
however, lower at 37 than at 25 °C, reflecting the fact that the
entropy is dominated by the displaced solvent and not by the
bound inhibitor.
The binding constant initially falls and then systematically
increases upon elongating the chain on the para position: Kn-Hex
> Kn-Pent > Kn-Bu > Kn-Pr > KEt. KMe falls between Kn-Hex
and Kn-Pent, which may be due to this substituent being less
sterically demanding. In Figure 4, ∆H and T∆S are plotted
against the number of carbon atoms in the linear alkyl chain.
For all of the compounds except p-n-hexylbenzamidinium, ∆H
becomes less favorable and T∆S becomes more favorable upon
extending the chain at all temperatures studied. This is
characteristic of hydrophobic interactions.42-45 The behavior of
p-n-hexylbenzamidinium is somewhat different. At low tem-
peratures, there is little to no increase in ∆H on adding an extra
methylene group to p-n-pentylbenzamidinium. At 30 and 37
°C, however, the trend observed for the smaller substituents
reverses: adding an extra methylene group to p-n-pentylben-
zamidinium results in a significant decrease in ∆H and an
unfavorable change in T∆S.
In Figure 5, the structure of the binding site of the protein
together with the p-n-butylbenzamidinium ion is illustrated. The
alkyl tail of the inhibitor is oriented toward the hydrophobic
pocket S3/S4 defined by residues Trp215 and Leu99. Interest-
(42) Lee, B. Biophys. Chem. 1994, 51, 271-278.
(43) Blokzijl, W.; Engberts, J. B. F. N. Angew. Chem., Int. Ed. Engl. 1993, 32,
1545-1579.
(44) Graziano, G. Phys. Chem. Chem. Phys 1999, 1, 3567-3576.
(45) Southall, N. T.; Dill, K. A.; Haymet, A. D. J. J. Phys. Chem. B 2002, 106,
2812.
(46) Karplus, M.; Kushick, J. N. Macromolecules 1981, 14, 325-332.
(47) Garc´ıa, A. E. Phys. ReV. Lett. 1992, 68, 2696-2699.
(48) Amedei, A.; Linssen, A. B. M.; Berendsen, H. J. C. Struct., Funct. Gen.
1993, 17, 412-425.
9
J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003 10575

A R T I C L E S
Talhout et al.
Figure 6. (Upper plots) The average structure of the binding site of the p-n-hexylbenzamidinium-trypsin complex at (left, A) 25 and (right, B) 37 °C.
Green, the average structure of the inhibitor; blue and red, configurations corresponding to the values of the first eigenvector with maximum probability.
(Lower plots) Graphs showing the distribution along the first eigenvector obtained from a PCA of the motion of the inhibitor within the binding site at (left,
C) 25 and (right, D) 37 °C.
trypsin. ∆Cp, the heat capacity change upon binding, is given
by
∂∆H/∂T ) ∆Cp
(4)
From Figure 7, it is clear that, in the temperature range studied,
∆H is linearly dependent on T. This is also true for the other
p-alkylbenzamidinium chlorides. For each inhibitor, ∆Cp was
estimated from the slope of a least-squares line of the best fit
to the ∆H data. The results are listed in Table 4.
Figure 8 shows the dependence of ∆Cp on the number of
carbon atoms in the linear alkyl chain. Overall, ∆Cp becomes
more negative with increasing chain length. Interestingly,
inhibitors with an odd number of carbon atoms in R have a
less negative (or those with an even number have a more
negative) ∆Cp than expected, leading to a zigzag pattern. A
similar pattern is evident in the simulations. Odd-numbered
chains show slightly larger RMS backbone positional deviations
(Table 2) and have the terminal carbons projecting away from
the protein. However, this is not reflected in the change in
solvent-accessible surface area of the inhibitor on binding (see
below).
Figure 7. ∆G, ∆H, and -T∆S of binding of p-methylbenzamidinium
chloride at 20, 25, 30, and 37 °C in Tris at pH 8.0. Values and errors are
listed in Table 1.
Heat Capacity Change upon Binding. The hydrophobic
properties of the inhibitors can also be inferred from the
temperature dependence of the thermodynamic parameters.
Figure 7 depicts the temperature dependence of ∆G, ∆H, and
T∆S of the binding of p-methylbenzamidinium chloride to
9
10576 J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003

Understanding Binding Affinity
A R T I C L E S
Table 4. ∆Cp, TS and TH, and the Solvent-Accessible Surfacea of
the Inhibitor before (SAS(aq.)) and after (SAS(tryps.)) Binding, for
Binding of p-Alkylbenzamidinium Chlorides to Trypsinb
in the solvent-accessible surface area of the inhibitor on binding
is plotted in Figure 8. For the longer alkyl tails (n-Pent and
n-Hex), there is little or no change in SAS(tryps.) (Table 4).
This suggests that the longer alkyl tails are partially buried
(presumably in the S3/S4 groove) and not exposed to water.
A negative ∆Cp means that the net thermodynamic driving
force for association will shift from being entropic to enthalpic
with increasing temperature. Temperatures TS and TH can be
defined, at which the entropic and the enthalpic contributions
to binding are zero. TS and TH were calculated from a linear fit
to the temperature dependence of T∆S and ∆H and are listed
in Table 4.
∆
Cp
Ts
C)
TH
C)
SAS(tryps.)
(nm2)
SAS(aq.)
(nm2)
1
Rc
(J mol-1 K-
)
(
°
(
°
Hd
-400 ( 20
-420 ( 12
-693 ( 12
-598 ( 2
-419 ( 4
-728 ( 4
-632 ( 10
-849 ( 10
44
46
42
47
63
48
55
48
-23
-19
0.40
0.35
0.48
0.68
0.36
0.87
0.87
0.92
2.78
2.83
3.19
3.46
3.37
3.70
3.90
4.15
Med
Ete
n-Pr
4.8
3.9
8.4
12
9.6
12
i-Pre
n-Bu
n-Pent
n-Hex
a The solvent-accessible surface was computed numerically49 on the basis
of an average structure. The atomic radii used were 0.16 nm for carbon,
0.13 nm for oxygen, 0.14 nm for nitrogen, 0.20 nm for sulfur, and 0.10 nm
for hydrogen atoms. The atomic radius of the solvent was 0.14 nm.
b Titrations performed in 50 mM Tris, 10 mM CaCl2 buffer at pH 8.0.
c Substituent at para position. d Data taken from ref 15. e Since these
compounds contain some NH4Cl, control experiments were performed with
10 wt % NH4Cl added to a benzamidinium chloride solution at the usual
concentration. Within error, the thermodynamic parameters were identical.
Therefore, the concentrations of these inhibitors, based on weight, can be
safely corrected for the amount of NH4Cl present.
From Figure 7, it can be seen that ∆G is practically
independent of temperature. This is a result of a compensation
between ∆H and T∆S, which are both linearly dependent on
temperature. Such behavior is characteristic of processes with
42,44,45,55
a large, negative ∆Cp
and is a consequence of the
vanishing of the hydrophobic hydration of the inhibitor. This
enthalpy-entropy compensation arises from the relative mag-
nitudes of ∆S and ∆Cp.42,54 The change in the enthalpy and the
entropy upon variation of the temperature are given respectively
by eq 4 and the following:
∂T∆S/∂T ) ∆Cp + ∆S
(5)
When |∆Cp| . |∆S| at all temperatures, the temperature
dependence of ∆S is approximately equal to ∆Cp. This means
that the temperature dependences of both ∆H and ∆S are equal
to ∆Cp, which results in ∆G being relatively temperature-
independent. The enthalpy-entropy compensation is similar for
almost all of the p-alkylbenzamidinium chlorides studied. Only
in the case of the n-pentyl and n-hexyl substituents was complete
compensation not observed. In these cases, δ∆H/δT has a larger
negative value than δT∆S/δT and leads to a decrease in ∆G
upon increasing temperature. This could be due to the tail being
more dynamic at higher temperatures (Figure 6), which results
in the overall entropy change upon binding being more favorable
at higher temperatures than expected.
Dehydration of the Active Site. The binding of several
natural inhibitors is known to induce an acidic shift in the pKa
of His57 N2.56,57 This pKa shift is attributed to the cost of
burying a charged His57 in an apolar environment, which is
only partly compensated by a stronger Ser195-His57 hydrogen
bond in the complex.56,57 To examine whether a change in the
protonation state of His57 N2 also occurs upon binding of the
benzamidinium-based inhibitors to trypsin, we investigated the
dependence of binding of benzamidinium chloride and p-n-
hexylbenzamidinium chloride on the buffer used. When proto-
nation/deprotonation effects are coupled to binding, ∆H depends
not only on the intrinsic enthalpy of binding ∆Hint but also on
the enthalpy change associated with the ionization of the
buffer.17,18 This can be expressed as
Figure 8. Heat capacity change (∆Cp, y-axis left-hand side) and change
in the solvent accessible surface (∆SAS, y-axis right-hand side) upon binding
of p-n-alkylbenzamidinium ions to trypsin versus the number of carbons in
the substituent in Tris at pH 8.0. Values and errors are listed in Table 4.
Thermodynamic data from various sources, i.e., the transfer
of compounds from a nonpolar phase to water,50,51 protein
folding,50-52 and the binding of different ligands to proteins,52-54
suggest that a negative value of ∆Cp often correlates with the
burial of nonpolar (hydrophobic) surface area. The burial of a
polar surface results in an increase of ∆Cp. Although the precise
origin of this hydrophobic effect is still debated, a common
interpretation is that hydrophobic surfaces induce the tangential
orientation (relative to the surface) of the water O-H bonds in
the first hydration shell42-45,55 and that this ordered water is
released on burial of the surface.
The marked decrease in ∆Cp with increasing chain length
suggests that the burial of a hydrophobic surface is a major
factor in binding. In Table 4, the values of the solvent-accessible
surface of the inhibitor calculated free in solution (SAS(aq.))
and bound to the protein (SAS(tryps.)) are given. The net change
∆H ) ∆Hint + NH+∆Hion
(6)
(49) Eisenhaber, F.; Lijnzaad, P.; Argos, P.; Sander, C.; Scharf, M. J. Comput.
Chem. 1995, 16, 273-284.
+
where NH is the number of protons from the buffer taken up
by the enzyme and ∆Hion is the ionization enthalpy of the buffer.
(50) Spolar, R. S.; Livingstone, J. R.; Record, M. T. Biochemistry 1992, 31,
3947-3955.
+
NH is equal to the change in the protonation state of residues
(51) Murphy, K. P.; Privalov, P. L.; Gill, S. J. Science 1990, 247, 559-561.
(52) Sturtevant, J. M. Proc. Natl. Acad. Sci. U.S.A. 1977, 74, 2236-2240.
(53) Spolar, R. S.; Record, M. T. Science 1994, 263, 777-784.
(54) Ha, J. H.; Spolar, R. S.; Record, M. T. J. Mol. Biol. 1989, 209, 801-816.
(55) Sharp, K. A.; Madan, B. J. Phys. Chem. B 1997, 101, 4343-4348.
(56) Baker, B. M.; Murphy, K. P. J. Mol. Biol. 1997, 268, 557-569.
(57) Antonini, E.; Ascenzi, P.; Bolognesi, M.; Gatti, G.; Guarneri, M.; Menegatti,
E. J. Mol. Biol. 1983, 165, 543-558.
9
J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003 10577

A R T I C L E S
Talhout et al.
Table 5. Thermodynamic Parameters for Binding of
p-Alkylbenzamidinium Chlorides to Trypsin at 25 °Ca
K
∆
G
∆
H
T∆S
1
1
1
1
Rb (pH)
buffer
(104 M-
)
(kJ mol-
)
(kJ mol-
)
(kJ mol-
)
H (8.0)
Hepes
Bicine
Tricine
5.3 ( 0.3 -26.9 ( 0.1 -17.9 ( 0.1 9.0 ( 0.2
6.3 ( 0.1 -27.4 ( 0.1 -16.0 ( 0.8 11.4 ( 0.9
4.6 ( 0.4 -26.6 ( 0.2 -16.8 ( 0.5 9.8 ( 0.6
Glycylglycine 5.0 ( 0.4 -26.8 ( 0.2 -16.5 ( 0.2 10.3 ( 0.3
Tris
Pipes
Hepes
Bicine
Tricine
4.5 ( 0.2 -26.6 ( 0.1 -18.9 ( 0.4 7.7 ( 0.5
4.6 ( 0.2 -26.6 ( 0.1 -13.2 ( 0.3 13.4 ( 0.4
4.9 ( 0.3 -26.8 ( 0.2 -15.3 ( 0.4 11.5 ( 0.5
4.3 ( 0.3 -26.5 ( 0.2 -15.7 ( 0.6 10.8 ( 0.6
3.8 ( 0.3 -26.1 ( 0.2 -16.5 ( 0.8 9.6 (0.9
15 ( 1 -29.5 ( 0.2 -2.9 ( 0.2 26.6 ( 0.4
14 ( 1 -29.4 ( 0.2 -6.2 ( 0.3 23.2 ( 0.6
13 ( 1 -29.2 ( 0.2 -6.6 ( 0.3 22.6 ( 0.5
12 ( 1 -29.1 ( 0.2 -8.4 ( 0.4 20.7 ( 0.3
H (7.4)
n-Hex (7.4) Pipes
Hepes
Bicine
Trycine
Figure 10. Distribution functions of the distance (nanometers) between
the oxygen OG of Ser195 and the nitrogen N2 of His57 in the complexes
of trypsin with p-methyl-, p-isopropyl-, and p-n-hexylbenzamidinium ions.
a Titrations performed in 50 mM concentrations of the different buffers
(10 mM CaCl2) at pH 8.0 or 7.4. b Substituent at the para position.
For p-n-hexylbenzamidinium chloride at pH 7.4, the values of
∆Hint and NH are -0.3(0.96) kJ mol-1 and -0.26(0.04),
+
respectively. Since at pH 7.4 the intrinsic enthalpy of binding
is close to zero, the driving force for binding is almost entirely
entropic. It is evident that more protons are transferred upon
binding for p-n-hexylbenzamidinium chloride than for benza-
midinium chloride. Using eq 7, the change in the fraction of
protonated His57 N2 for a serine proteinase upon binding of a
protein inhibitor at pH 7.4 was estimated to be -0.28.59 This
decrease in the fraction of protonated His57 N2 corresponds to
60
+
a transfer of -0.28 proton from the buffer to the enzyme (NH ).
This is (within error) equal to that reported for the binding of
p-n-hexylbenzamidinium chloride to trypsin and larger than that
reported for benzamidinium chloride (see above). Since the
decrease of the fraction of protonated His57 reflects a dehydra-
tion penalty, these data indicate that His57 is more shielded
from water upon binding of p-n-hexylbenzamidinium chloride
than upon binding of benzamidinium chloride.
Figure 9. Enthalpy of binding of benzamidinium chloride at pH 8.0 and
7.4 and of p-n-hexylbenzamidinium chloride at pH 7.4 versus the ionization
enthalpy58 of the buffer. See Table 5.
involved in the binding process. At a given pH, the change in
the fraction of protonated His57 N2 can be calculated by using
the Henderson-Hasselbalch equation:
To understand the origin of this pKa shift, the average number
of water molecules in the active site around the OG of Ser195
and the N2 of His57 in the simulations was examined. Upon
increasing the number of carbon atoms in the p-alkyl substituent,
the two residues (in particular His57) become progressively less
accessible to water. The more sterically demanding substituents
shield the active site from water (results not shown). For these
inhibitors, the penalty for dehydration of the active site on
binding will be larger. Exclusion of water from the active side
promotes the formation of a hydrogen bond between the OG
of Ser195 and the N2 of His57. Figure 10 shows the distribution
of this distance between the OG of Ser195 and the N2 of His57
for the complexes of trypsin with methyl-, isopropyl-, and
n-hexylbenzamidinium ions. The distributions have two
maxima: one at 2.8 Å (a H-bonding distance) and the other
around 4.0 Å. In general, the larger the substituent, the more
probable is a hydrogen bond between the two residues.
[His57]
[His57+]
pH ) pKa + log
(7)
where [His57] is the concentration of unprotonated His57 N2
and [His57+] is the concentration of protonated His57 N2.
Table 5 lists the thermodynamic parameters of binding of
benzamidinium chloride to trypsin at 25 °C in several buffers
at pH 8.0 and of benzamidinium chloride and p-n-hexylbenza-
midinium chloride at pH 7.4. Figure 9 shows the dependence
of the enthalpy of binding on the ionization enthalpy of the
buffer for benzamidinium chloride at pH 8.0 and 7.4 and for
p-n-hexylbenzamidinium chloride at pH 7.4, both at 25 °C. For
benzamidinium chloride at pH 8.0, the linear regression yields
an abscissa, equal to ∆Hint, of -16.3(1.9) kJ mol-1 and a slope,
+
equal to NH , of -0.03(0.05) indicating that, within error, no
protons are transferred upon binding. By contrast, for benz-
+
amidinium chloride at pH 7.4, the values of ∆Hint and NH are
(58) Christensen, J. J.; Hansen, L. D.; Izatt, R. M. Handbook of Proton Ionization
Heats and Related Thermodynamic Quantities; Wiley: New York, 1976.
(59) At pH 7.4, the fraction of protonated His57 N2 amounts to 0.285 for pKa
) 7 and 0.004 for pKa ) 5.
-11.6(0.61) kJ mol-1 and -0.16(0.03), respectively. Thus, at
pH 7.4, protons are transferred upon binding, and the binding
constant decreases due to a less favorable ∆Hint that is partly
compensated by a more favorable entropic contribution.
(60) Baker and Murphy56 report a pKa shift of His57 from 6.7 in the free elastase
to 5.2 in the ovomucoid third domain-elastase complex. Using their data
+
at pH 7, an experimental value of around -0.29 for NH is found, while a
calculated value of -0.32 is found using the Henderson-Hasselbalch
equation, indicating that this approach leads to good agreement with
experiment.
It is interesting to look at the influence of an alkyl chain at
+
the para position on the thermodynamics of binding and NH
.
9
10578 J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003

Understanding Binding Affinity
A R T I C L E S
As the substituent binds into the hydrophobic pocket S3/S4,
one might expect that binding would strongly correlate with
octanol/water partition constants.41 From Figure 3 it is clear this
is not the case. The fact that inhibitor binding is associated with
the dehydration of the active site, the pKa shift of His57, and
the changes in the interaction between Ser195 and His57 goes
some way to explain this apparent discrepancy. The effect of
dehydration of the protein should, however, be reflected in the
free energy calculations. The GROMOS96 force field accurately
reproduces the cyclohexane/water partition properties of small
linear and branched aliphatic compounds.38 However, the force
field is known to underestimate the Gibbs energy of hydration
of analogues of the more polar R-amino acids, including Trp.
Thus, we would expect the calculations to correctly reproduce
the hydrophobic contributions to binding but to underestimate
the cost of dehydration of the protein, leading to an overestima-
tion of the experimental binding Gibbs energies, as is seen. Thus,
the free energy calculations reinforce the idea that dehydration
is an important determining factor for binding. Note that the
degree of dehydration of the catalytic triad is influenced by both
the length of the substituent and the degree of branching at the
first carbon (Figure 10). Thus, for the ethyl and isopropyl
substituents, for example, one cannot distinguish steric effects
due to clashes with the protein from the increased dehydration
of the active site.
However, the above picture is overly simplistic. Simulation
studies suggest that all of the inhibitors, including p-tert-
butylbenzamidinium, can be accommodated within the binding
pocket without obvious strain. Free energy calculations predict
that the binding of p-tert-butylbenzamidinium is favorable,
suggesting that the failure of p-tert-butylbenzamidinium to bind
may be due to an inability to enter the binding pocket rather
than steric restrictions. Comparison with octanol/water partition
data also suggests that the changes in binding Gibbs energy on
extending the chain length of the substituent are small compared
to that expected for a purely hydrophobic interaction. In fact,
the binding constants for these inhibitors are remarkably similar,
indicating that there is a penalty for binding which must be
considered. Calculation of the pKa shift of His57 shows that
the alkyl substituents shield the active site from water. The larger
and more sterically demanding a substituent is, the less water
can be found around Ser195 OG and His57 N2. This dehydra-
tion, which mechanistically is associated with the activation of
the catalytic triad, also incurs a thermodynamic cost and results
in a lower affinity.
Overall, the study shows that, even in the case of a simple
series of substituents, the factors that determine the binding
affinity can be highly involved. Superficially, the series of
compounds investigated display only small differences in
binding affinity (except for p-tert-butylbenzamidinium, which
does not bind). Detailed thermodynamic analysis showed,
however, that this masks large compensating differences in
enthalpy and entropy. Simulation studies in turn showed that
simple concepts such as steric hindrance and hydrophobic
interactions based on buried surface area were insufficient to
fully explain the results. Secondary effects, such as an ability
to access the binding site and the cost of dehydration of the
active site, are of equal or even greater importance. The present
study serves to underline the danger of using empirical models
based on the structural properties of an inhibitor alone to
rationalize binding behavior, even in the simplest cases.
Conclusions
Understanding molecular recognition requires knowledge of
the thermodynamic properties of the system, together with
insight into the structure and dynamics of the complex of
interest. Here, ITC experiments and MD simulations have been
used to study the binding of a series of p-alkylbenzamidinium
chloride inhibitors to trypsin, to understand the influence of
small structural variations in the alkyl substituent on binding.
The steric and hydrophobic properties of the p-alkyl sub-
stituent are clearly correlated with binding affinity. Steric bulk
on the first carbon decreases the binding affinity. Experimen-
tally, the binding of p-tert-butylbenzamidinium was found to
be negligible. The presence of a hydrophobic linear alkyl tail
increases the binding affinity. Increasing the length of the p-n-
alkyl tail (n ) 2-6) resulted in a progressive (favorable)
increase in the entropy of binding. This was partially compen-
sated by an unfavorable increase in the enthalpy of binding.
The change in heat capacity upon binding becomes more
negative with increasing tail length, suggesting that binding is
dominated by hydrophobic interactions.
Acknowledgment. Thorsten Stafforst is gratefully acknowl-
edged for synthesizing p-ethylbenzamidinium chloride.
Supporting Information Available: Synthesis and analysis
of p-alkylbenzamidinium chlorides (PDF). This material is
available free of charge via the Internet at http://pubs.acs.org.
JA034676G
9
J. AM. CHEM. SOC. VOL. 125, NO. 35, 2003 10579