Thermodynamic and structural effects of conformational constraints in protein-ligand interactions. Entropic paradoxy associated with ligand preorganization

Article abstract of DOI:10.1021/ja904698q

Succinate- and cyclopropane-derived phosphotyrosine (pY) replacements were incorporated into a series of Grb2 SH2 binding ligands wherein the pY+1 residue was varied to determine explicitly how variations in ligand preorganization affect binding energetic

Full text of DOI:10.1021/ja904698q

Published on Web 11/03/2009
Thermodynamic and Structural Effects of Conformational
Constraints in Protein-Ligand Interactions. Entropic
Paradoxy Associated with Ligand Preorganization
John E. DeLorbe, John H. Clements, Martin G. Teresk, Aaron P. Benfield,
Hilary R. Plake, Laura E. Millspaugh, and Stephen F. Martin*
Department of Chemistry and Biochemistry, The Institute of Cellular and Molecular Biology,
and The Texas Institute of Drug and Diagnostic DeVelopment, The UniVersity of Texas,
Austin, Texas 78712
Received June 16, 2009; E-mail: sfmartin@mail.utexas.edu
Abstract: Succinate- and cyclopropane-derived phosphotyrosine (pY) replacements were incorporated into
a series of Grb2 SH2 binding ligands wherein the pY+1 residue was varied to determine explicitly how
variations in ligand preorganization affect binding energetics and structure. The complexes of these ligands
with the Grb2 SH2 domain were examined in a series of thermodynamic and structural investigations using
isothermal titration calorimetry and X-ray crystallography. The binding enthalpies for all ligands were
favorable, and although binding entropies for all ligands having a hydrophobic residue at the pY+1 site
were favorable, binding entropies for those having a hydrophilic residue at this site were unfavorable.
Preorganized ligands generally bound with more favorable Gibbs energies than their flexible controls, but
this increased affinity was the consequence of relatively more favorable binding enthalpies. Unexpectedly,
binding entropies of the constrained ligands were uniformly disfavored relative to their flexible controls,
demonstrating that the widely held belief that ligand preorganization should result in an entropic advantage
is not necessarily true. Crystallographic studies of complexes of several flexible and constrained ligands
having the same amino acid at the pY+1 position revealed extensive similarities, but there were some
notable differences. There are a greater number of direct polar contacts in complexes of the constrained
ligands that correlate qualitatively with their more favorable binding enthalpies and Gibbs energies. There
are more single water-mediated contacts between the domain and the flexible ligand of each pair; although
fixing water molecules at a protein-ligand interface is commonly viewed as entropically unfavorable,
entropies for forming these complexes are favored relative to those of their constrained counterparts.
Crystallographic b-factors in the complexes of constrained ligands are greater than those of their flexible
counterparts, an observation that seems inconsistent with our finding that entropies for forming complexes
of flexible ligands are relatively more favorable. This systematic study highlights the profound challenges
and complexities associated with predicting how structural changes in a ligand will affect enthalpies,
entropies, and structure in protein-ligand interactions.
protein-ligand interactions,2 the ability to predict accurately
the binding affinity of a given ligand and derivatives thereof
Introduction
A major goal in contemporary bioorganic and medicinal
chemistry is the design of small molecules that bind with high
affinities to proteins. Toward this goal, numerous computational
methods and scoring functions have been developed to predict
association constants, and when there are well parametrized
training sets, some of these are reasonably good as qualitative
predictors of protein-ligand affinities; however, none are
reliable as quantitative tools for ordering relative affinities for
a range of different biological systems.1 Indeed, despite the
increasing availability of data relating to the energetics of
for a target protein remains the Holy Grail for medicinal
chemists and those studying molecular recognition in biological
systems.3
The binding affinity, Ka, for a protein-ligand interaction is
reflected by the net change in Gibbs energy, ∆G°, which
comprises enthalpic, ∆H°, and entropic, ∆S°, contributions (i.e.,
∆G° ) ∆H° - T∆S°), that occurs upon complexation according
to the expression ∆G° ) -RT ln Ka. The optimization of
protein-ligand interactions thus requires modifying the structure
of a ligand in a way that results in a more negative binding
enthalpy and a more positive binding entropy.
(1) For lead references on the limitations of computational methods for
predicting binding affinities, see: (a) Warren, G. L.; Andrews, C. W.;
Capelli, A.-M.; Clarke, B.; LaLonde, J.; Lambert, M. H.; Lindvall,
M.; Nevins, N.; Semus, S. F.; Senger, S.; Tedesco, G.; Wall, I. D.;
Woolven, J. M.; Peishoff, C. E.; Head, M. S. J. Med. Chem. 2006,
49, 5912–5931. (b) Klebe, G. Drug DiscoVery Today 2006, 11, 580–
594. (c) Damm, K. L.; Carlson, H. A. J. Am. Chem. Soc. 2007, 129,
8225–8235.
(2) Liu, T.; Lin, Y.; Wen, X.; Jorissen, R. N.; Gilson, M. K. Nucleic Acids
Res. 2007, 35, Database issue D198-D201.
(3) For excellent and comprehensive reviews of energetics in bimolecular
interactions, see: (a) Houk, K. N.; Leach, A. G.; Kim, S. P.; Zhang,
X. Angew. Chem., Int. Ed. 2003, 42, 4872–4897. (b) Zhou, H.-X.;
Gilson, M. K. Chem. ReV. 2009, 109, 4092–4107.
9
16758 J. AM. CHEM. SOC. 2009, 131, 16758–16770
10.1021/ja904698q CCC: $40.75  2009 American Chemical Society

Entropic Paradoxy Associated with Ligand Preorganization
A R T I C L E S
Obtaining a more favorable enthalpy of binding typically
entails increasing the noncovalent associations that arise from
polar, van der Waals, and charge-dipole interactions.4 This
objective is not easily achieved, however, because desolvation
of polar groups on a ligand is energetically more costly than
desolvating nonpolar groups.5 Further complicating the task is
the difficulty associated with controlling the geometries of
individual polar interactions between the ligand and the protein,
the energies of which are both distance and angle dependent.
A more favorable binding entropy may be achieved by
reducing unfavorable solvation and conformational parameters
associated with the ligand.6,7 One common tactic to enhance
the entropy of solvation involves increasing the hydrophobicity
of the ligand, and although this often leads to an improved
binding affinity, formation of the resultant complex is not
necessarily accompanied by the expected entropy-driven sig-
nature characteristic of the hydrophobic effect.8 For example,
adding methylene groups to increase the hydrophobicity of a
ligand can lead to a less favorable binding entropy and a more
favorable binding enthalpy.9 Indeed, a search of the literature
reveals numerous examples wherein binding of two nonpolar
molecules is characterized by a large enthalpic driving force
rather than an entropic one.10,11
Figure 1. Simple model of the putative energetic effects associated with
ligand preorganization. Cyclization of a flexible ligand limits the degrees
of freedom and reduces the number of conformational isomers in solution,
so the probability that the ligand adopts its biologically active conformation
in solution is enhanced, thereby resulting in a more favorable entropy of
binding. This analysis presumes that the two ligands interact similarly with
the solvent and protein so that the binding enthalpies for the two are
approximately the same.
Constraining a flexible ligand in the three-dimensional shape
it adopts when bound to a receptor, namely its biologically active
conformation, can also result in increased association con-
stants.12-14 This enhanced affinity has been commonly attributed
to the more favorable configurational entropy of binding that is
expected from reducing the dynamic motion of a ligand prior
to its complexation with the protein (i.e., ∆S° < ∆S°′) (Figure
1). An implicit assumption in this reasoning is that solvent and
protein interact in the same way with both the flexible and
constrained ligands so that no significant change in binding
enthalpy is expected (i.e., ∆H° ≈ ∆H°′). However, increases
in potencies accompanying ligand preorganization are often less
than 10-fold, an amount somewhat less than what would be
expected based upon the accepted energetic estimates of 0.7-1.6
kcal mol-1 (i.e., ∼2.3-5.3 eu at 25 °C) for completely restricting
an independent rotor.15 Indeed, the small energetic benefits
observed for constraining rotors in the arena of host-guest
chemistry led Schneider to question whether reducing the
(4) For examples of parametrizing binding enthalpy, see: Luque, I.; Freire,
E. Proteins: Struct., Funct., Genet. 2002, 49, 181–190.
(5) Cabani, S.; Gianni, P.; Mollica, V.; Lepori, L. J. Solution Chem. 1981,
8, 563–595.
(6) (a) Ruben, A. J.; Kiso, Y.; Freire, E. Chem. Biol. Drug Des. 2006,
67, 2–4. (b) Lafont, V.; Armstrong, A. A.; Ohtaka, H.; Kiso, Y.;
Amzel, L. M.; Freire, E. Chem. Biol. Drug Des. 2007, 69, 413–422.
(7) For detailed analyses of intermolecular interactions, see: (a) Gohlke,
H.; Klebe, G. Angew. Chem., Int. Ed. 2002, 41, 2644–2676. (b) Hunter,
C. A. Angew. Chem., Int. Ed. 2004, 43, 5310–5324. (c) Williams,
D. H.; Stephens, E.; O’Brien, D. P.; Zhou, M. Angew. Chem., Int. Ed.
2004, 43, 6596–6616.
(8) For some leading articles on hydrophobic effects associated with
nonpolar and polar groups, see: (a) Bartlett, P. A.; Yusuff, N.; Rico,
A. C.; Lindvall, M. K. J. Am. Chem. Soc. 2002, 124, 3853–3857. (b)
Southall, N. T.; Dill, K. A.; Haymet, A. D. J. J. Phys. Chem. B 2002,
106, 521–533. (c) Kyte, J. Biophys. Chem. 2003, 100, 193–203. (d)
Chalikian, T. V. Biopolymers 2003, 70, 492–496.
(13) For some comparisons wherein the flexible and constrained ligands
have the same number of heavy atoms, see: (a) Widlanski, T.; Bender,
S. L.; Knowles, J. R. J. Am. Chem. Soc. 1989, 111, 2299–2300. (b)
Morgan, B. P.; Holland, D. R.; Matthews, B. W.; Bartlett, P. A. J. Am.
Chem. Soc. 1994, 116, 3251–3260. (c) Ettmayer, P.; France, D.;
Gounarides, J.; Jarosinski, M.; Martin, M. S.; Rondeau, J. M.; Sabio,
M.; Topiol, S.; Weidmann, B.; Zurini, M.; Bair, K. W. J. Med. Chem.
1999, 42, 971–980. (d) Hansen, K. K.; Grosch, B.; Greiveldinger-
Poenaru, S.; Bartlett, P. A. J. Org. Chem. 2003, 68, 8465–8470. (f)
Tsantrizos, Y. S.; Bolger, G.; Bonneau, P.; Cameron, D. R.; Goudreau,
N.; Kukolj, G.; LaPlante, S. R.; Llinas-Brunet, M.; Nar, H.; Lamarre,
D. Angew. Chem., Int. Ed. 2003, 42, 1355–1360.
(9) Malham, R.; Johnstone, S.; Bingham, R. J.; Barratt, E.; Phillips,
S. E. V.; Laughton, C. A.; Homans, S. W. J. Am. Chem. Soc. 2005,
127, 17061–17067.
(10) For a review of so-called nonclassical hydrophobic effects, see: Meyer,
E. A.; Castellano, R. K.; Diederich, F. Angew. Chem., Int. Ed. 2003,
42, 1210–1250.
(11) For some leading references to enthalpy driven hydrophobic associa-
tions, see: (a) Chervenak, M. C.; Toone, E. J. J. Am. Chem. Soc. 1994,
116, 10533–10539. (b) Carey, C.; Cheng, Y.-K.; Rossky, P. J. Chem.
Phys. 2000, 258, 415–425. (c) Bingham, R. J.; Findlay, J. H. B. C.;
Hsieh, S.-Y.; Kalverda, A. P.; Kjellberg, A.; Perazzolo, C.; Phillips,
S. E. V.; Seshadri, K.; Trinh, C. H.; Turnbull, W. B.; Bodenhausen,
G.; Homans, S. W. J. Am. Chem. Soc. 2004, 126, 1675–1681. (d)
Homans, S. W. Drug DiscoVery Today 2007, 12, 534–539.
(12) For reviews, see:(a) Kessler, H. Angew. Chem., Int. Ed. Engl. 1982,
21, 512–523. (b) Cram, D. J. Angew. Chem., Int. Ed. Engl. 1986, 25,
1039–1057. (c) Bo¨hm, H.-J.; Klebe, G. Angew. Chem., Int. Ed. 1996,
35, 2588–2614. (d) Hanessian, S.; McNaughton-Smith, G.; Lombart,
H.-G.; Lubell, W. D. Tetrahedron 1997, 53, 12789–l2854. (e) Mann,
A. In The Practice of Medicinal Chemistry, 2nd ed.; Wermuth, C. G.,
Ed.; Academic Press: London, U.K., 2003; pp 233-250. (f) Nakanishi,
H.; Kahn, M. Design of Peptidomimetics. In The Practice of Medicinal
Chemistry, 2nd ed.; Wermuth, C. G., Ed.; Academic Press: London,
U.K., 2003; pp 477-500. (g) Loughlin, W. A.; Tyndall, J. D. A.;
Glenn, M. P.; Fairlie, D. P. Chem. ReV. 2004, 104, 6085–6118.
(14) For some comparisons wherein the flexible and constrained ligands
differ by one or two heavy atoms, see: (a) Meyer, J. H.; Bartlett, P. A.
J. Am. Chem. Soc. 1998, 120, 4600–4609. (b) Smith, W. W.; Bartlett,
P. A. J. Am. Chem. Soc. 1998, 120, 4622–4628. (c) Marquis, R. W.;
et al. J. Med. Chem. 2001, 44, 1380–1395. (d) Dekker, F. J.; de Mol,
N. J.; Fischer, M. J. E.; Kemmink, J.; Liskamp, R. M. J. Org. Biomol.
Chem. 2003, 1, 3297–3303. (e) Nam, N.-H.; Ye, G.; Sun, G.; Parang,
K. J. Med. Chem. 2004, 47, 3131–3141. (f) Ghosh, A. K.; Swanson,
L. M.; Cho, H.; Leshchenko, S.; Hussain, K. A.; Kay, S.; Walters,
D. E.; Koh, Y.; Mitsuya, H. J. Med. Chem. 2005, 48, 3576–3585.
(15) For leading references to energetic contributions to binding of freezing
rotors, see: (a) Gerhard, U.; Searle, M. S.; Williams, D. H. Bioorg.
Med. Chem. Lett. 1993, 3, 803–808. (b) Khan, A. R.; Parrish, J. C.;
Fraser, M. E.; Smith, W. W.; Bartlett, P. A.; James, M. N. G.
Biochemistry 1998, 37, 16839–16845. (c) Hossain, M. A.; Schneider,
H.-J. Chem.sEur. J. 1999, 5, 1284–1290. (d) See also in: Schneider,
H.-J. In Protein-Ligand Interactions: From Molecular Recognition to
Drug Design; Bo¨hm, H.-J., Schneider, G., Eds.; Wiley-VCH: Wein-
heim, 2003; pp 21-50.
9
J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009 16759

A R T I C L E S
DeLorbe et al.
number of single bonds in synthetic hosts was actually a viable
strategy to enhance the binding affinity.16
of preorganizing pseudopeptide ligands by incorporating cy-
clopropane rings as conformational constraints in flexible
controls. Toward this objective, the phosphotyrosine residue of
Ac-pTyr-Glu-Glu-Ile-OH (Ac-pY-E-E-I-OH) (1), a tet-
rapeptide sequence present in peptides that bind to the Src SH2
domain,24 was replaced with substituted succinyl and cyclo-
propane dicarboxyl moieties leading to the pseudopeptides 2
and 3.18 Cyclization of 2 to give the cyclopropane-derived ligand
3 completely restricts two rotors, and the three rotors associated
with the carbon-carbon single bonds joining the two carbonyl
and the aryl substituents on the cyclopropane ring are partially
restricted. Significantly, compounds 2 and 3 have the same
number and type of heavy atoms, identical functional groups,
and the same number of hydrogen bond donors and acceptors;
they differ by only two hydrogen atoms arising from formation
or scission of a carbon-carbon single bond as shown in 2 and
3. Six related pseudopeptides were also prepared in which each
Glu residue in 2 and 3 was replaced with an Asp residue, and
the Ile residue was substituted with a Val residue.25
Notwithstanding some notable exceptions,13a,15b ligand pre-
organization has not provided the magnitude of increased
binding affinity that would be expected from the number
of rotors being restricted; however, understanding the origin(s)
of this shortcoming is problematic. For example, the lack of
structural information for complexes of flexible and constrained
ligand pairs makes it difficult to determine whether the two
ligands interact similarly with the protein. Furthermore, because
of the dynamic nature of protein-ligand interactions, it is not
easy to assess the extent to which individual rotors of a flexible
molecule are restricted upon binding. A related complicating
factor is that ligands typically bind to proteins in conformations
that are higher in energy than their global minima in solution,17
and the effects of differential conformational strain energies
upon the relative binding affinities of related compounds are
poorly understood. The challenges associated with evaluating
the specific energetic consequences of ligand preorganization
are further exacerbated by the common failure to compare the
affinity of a constrained molecule with an appropriate flexible
control having the same number and type of heavy atoms, the
same functionality, and the same number of hydrogen bond
donors and acceptors.13 Another impediment to understanding
how ligand preorganization affects binding affinity is that
potencies are typically reported as Ki’s or IC50’s, and the specific
contributions to ∆S° and ∆H° of binding are rarely determined.
Exceptions to this generalization are found in reports from our
group,18,19 as well as in work disclosed by Spaller, who has
conducted a detailed analysis of thermodynamic parameters for
the binding of a series of linear and macrocyclic ligands to a
PDZ3 domain.20 These and other studies further reveal how
favorable changes in binding enthalpies or entropies are
attenuated by enthalpy-entropy compensation, an ubiquitous
phenomenon in bimolecular interactions.21,22
The thermodynamic parameters for complexation of each of
these ligands with recombinant Src SH2 domain were deter-
mined by isothermal titration calorimetry (ITC). The constrained
ligand 3 and its derivatives containing Asp or Val replacements
at the pY+1 - pY+3 positions invariably bound with more
favorable entropies than their flexible counterparts, and the
observed average entropic advantage of ∼7.5 eu correlated
modestly with the total number of rotors restricted.15,25 However,
this expected entropic advantage was always balanced by an
enthalpic penalty that eventuated in comparable binding affinities
for each ligand in a flexible/constrained pair. Because of this
balancing enthalpy-entropy compensation, no net energetic
advantage attended ligand preorganization. We determined the
structure of the complex of 3 with the domain, and comparing
this structure with that of an 11-mer peptide containing the pTyr-
Our first venture toward correlating ligand preorganization
and energetics in protein-ligand interactions involved introduc-
ing cyclopropane-derived amino acid replacements into analogs
of known enzyme inhibitors.23Although these studies led to the
discovery of potent inhibitors of a number of enzymes, we, like
others, found that the benefits of ligand preorganization fell short
of expectations. These results inspired us to initiate studies that
would explicitly elucidate the energetic and structural effects
(16) Eblinger, F.; Schneider, H.-J. Angew. Chem., Int. Ed. 1998, 37, 826–
829.
(17) Perola, E.; Charifson, P. S. J. Med. Chem. 2004, 47, 2499–2510.
(18) Davidson, J. P.; Lubman, O.; Rose, T.; Waksman, G.; Martin, S. F.
J. Am. Chem. Soc. 2002, 124, 205–215.
(19) For a preliminary account of some of the results presented in this
account, see: Benfield, A. P.; Teresk, M. G.; Plake, H. R.; DeLorbe,
J. E.; Millspaugh, L. E.; Martin, S. F. Angew. Chem., Int. Ed. 2006,
45, 6830–6835.
(20) (a) Li, T.; Saro, D.; Spaller, M. R. Bioorg. Med. Chem. Lett. 2004,
14, 1385–1388. (b) Udugamasooriya, D. G.; Spaller, M. R. Biopoly-
mers 2008, 8, 653–667, and references therein.
(23) For a review, see: (a) Reichelt, A.; Martin, S. F. Acc. Chem. Res.
2006, 39, 433–442. See also: (b) Martin, S. F.; Austin, R. E.; Oalmann,
C. J.; Baker, W. R.; Condon, S. L.; deLara, E.; Rosenberg, S. H.;
Spina, K. P.; Stein, H. H.; Cohen, J.; Kleinert, H. D. J. Med. Chem.
1992, 35, 1710–1721. (c) Martin, S. F.; Dorsey, G. O.; Gane, T.;
Hillier, M. C.; Kessler, H.; Baur, M.; Matha, B.; Erickson, J. W.; Bhat,
T. N.; Munshi, S.; Gulnick, S. V.; Topol, I. A. J. Med. Chem. 1998,
41, 1581–1597. (d) Martin, S. F.; Oalmann, C. J.; Liras, S. Tetrahedron
1993, 49, 3521–3532. (e) Reichelt, A.; Gaul, C.; Frey, R.; Kennedy,
A.; Martin, S. F. J. Org. Chem. 2002, 67, 4062–4075.
(21) For reviews, see: (a) Sharp, K. Protein Sci. 2001, 10, 661–667. (b)
Liu, L.; Guo, Q.-X. Chem. ReV. 2001, 101, 673–695.
(22) For some leading references of enthalpy-entropy compensation, see:
(a) Gilli, P.; Ferretti, V.; Gillli, G.; Borea, P. A. J. Phys. Chem. 1994,
98, 1515–1518. (b) Dunitz, J. D. Chem. Biol. 1995, 2, 709–712. (c)
Calderone, C. T.; Williams, D. H. J. Am. Chem. Soc. 2001, 123, 6262–
6267. (d) Cooper, A.; Johnson, C. M.; Lakey, J. H.; No¨llmann, M.
Biophys. Chem. 2001, 93, 215–230. (e) Talhout, R.; Villa, A.; Mark,
A. E.; Engberts, J. B. F. N. J. Am. Chem. Soc. 2003, 125, 10570–
10579. (f) Williams, D. H.; Stephens, E.; O’Brien, D. P.; Zhou, M.
Angew. Chem., Int. Ed. 2004, 43, 6596–6616. (g) Krishnamurthy,
V. M.; Bohall, B. R.; Semetey, V.; Whitesides, G. M. J. Am. Chem.
Soc. 2006, 128, 5802–5812.
(24) For reviews of SH2 domains, see: (a) Cody, W. L.; Lin, Z.; Panek,
R. L.; Rose, D. W.; Rubin, J. R. Curr. Pharm. Des. 2000, 6, 59–98.
(b) Bradshaw, J. M.; Waksman, G. AdV. Protein Chem. 2002, 61, 161–
210. (c) Machida, K.; Mayer, B. J. Biochim. Biophys. Acta 2005, 1747,
1–25.
(25) Davidson, J. P. Ph.D. Dissertation, The University of Texas, 2001.
9
16760 J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009

Entropic Paradoxy Associated with Ligand Preorganization
A R T I C L E S
Glu-Glu-Ile sequence revealed that all interatomic distances
between the domain and the ligand in each of the two structures
were in close agreement.18 Although this structural study
suggested that the cyclopropane-derived pY replacement in 3
is a viable rigid mimic of the bound conformation of phospho-
tyrosine in Src SH2 binding ligands, we were unable to identify
the origin of the enthalpic penalty associated with preorganiz-
ing 2.
between the pY residues of 1 and 3 and the Src SH2
domain.13c,26 It thus occurred to us that the flexible and
constrained pY mimics found in 2 and 3, respectively, might
again serve as efficacious replacements of the pY residue in
derivatives of 4. Accordingly, we targeted the series of flexible
ligands 5-10, in which a benzyl succinyl moiety serves as a
flexible pY replacement, and the corresponding series of
constrained ligands 11-16, in which a cyclopropane ring is a
rigid mimic of the pY residue. Any flexible/constrained ligand
pair would have enabled us to evaluate the energetic effects of
preorganization; however, we reasoned that a broader study
To our knowledge, these key experiments were the first to
demonstrate explicitly that introducing a conformational con-
straint into a flexible ligand resulted in the expected entropic
advantage for forming protein-ligand complexes. It was
nevertheless perplexing that preorganizing flexible Src SH2
binding ligands did not afford analogs having higher affinities,
despite their having more favorable binding entropies. To further
explore the effects of preorganization upon energetics in
protein-ligand interactions, we extended the scope of our
inquiry to the binding of phosphotyrosine-containing peptide
analogs to the SH2 domain of growth receptor binding protein
2 (Grb2), a 25 kDa cytosolic adapter protein that participates
in the Ras signal transduction pathway.24 The binding of pY
residues on receptor tyrosine kinases (RTKs) to the SH2 domain
of Grb2 leads to Ras activation and consequent cell growth and
differentiation. There has thus been considerable interest in
identifying compounds that selectively and potently inhibit
binding of the Grb2 SH2 domain to RTKs as a strategy to
modulate Ras signaling and to discover potential anticancer
agents.13c,26,27 The SH2 domain of Grb2 recognizes and binds
pY peptides containing the amino acid sequence pTyr-Xaa-Asn
(pYXN). Although the pTyr and the Asn residues are essential
for high affinity binding to the Grb2 SH2 domain, there is
considerable flexibility in the nature of the residue at the pY+1
position; it is usually hydrophobic, but Gln, Glu, and Lys
residues are found at this position in potent Grb2 SH2 binding
ligands.28
wherein the polar and functional nature of the side chain of the
pY+1 residue was varied might provide additional insights
relative to how changing the hydrophobicity and charge of a
ligand might affect binding energetics.
The syntheses of fpYVN (5) and cpYVN (11) by coupling
the known acids 17 and 18,18 respectively, with a Val-Asn-
NH2 dipeptide, followed by hydrogenolysis of the O-benzyl
groups, have been described,29 and compounds 6-10 and
12-16 were prepared analogously (eqs 1 and 2).
Results and Discussion
Ligand Design and Synthesis. Given the known structure-
activity relationships of Grb2 SH2 binding ligands, our initial
studies were focused upon derivatives of Ac-pTyr-Xaa-Asn-
NH2 (4).19 Inspection of crystal structures of complexes of the
Grb2 SH2 domain with different phosphotyrosine-containing
peptides revealed that the interactions between the pY residue
and the Grb2 SH2 domain were closely similar to those observed
(26) For some structural studies, see: (a) Rahuel, J.; Gay, B.; Erdmann,
D.; Strauss, A.; Garcia-Echeverria, C.; Furet, P.; Caravatti, G.; Fretz,
H.; Schoepfer, J.; Gru¨tter, M. G. Nat. Struct. Biol. 1996, 3, 586–589.
(b) Ogura, K.; Tsuchiya, S.; Terasawa, H.; Yuzawa, S.; Hatanaka,
H.; Mandiyan, V.; Schlessinger, J.; Inagaki, F. J. Biomol. NMR 1997,
10, 273–278. (c) Rahuel, J.; Garcia-Echeverria, C.; Furet, P.; Strauss,
A.; Caravatti, G.; Fretz, H.; Schoepfer, J.; Gay, B. J. Mol. Biol. 1998,
279, 1013–1022. (d) Nioche, P.; Liu, W.-Q.; Broutin, I.; Charbonnier,
F.; Latreille, M.-T.; Vidal, M.; Roques, B.; Garbay, C.; Ducruix, A.
J. Mol. Biol. 2002, 315, 1167–1177. (e) Ogura, K.; Shiga, T.; Yokochi,
M.; Yuzawa, S.; Burke, T. R., Jr.; Inagaki, F. J. Biomol. NMR 2008,
42, 197–207.
Thermodynamic Properties of Flexible and Constrained
Ligands. The thermodynamic parameters (Ka, ∆G°, ∆H°, ∆S°)
for binding of 5 and 11 to the Grb2 SH2 domain were first
determined by titration studies using ITC.30 Briefly, a solution
(27) For a review, see: (a) Fretz, H.; Furet, P.; Garcia-Echeverria, C.;
Rahuel, J.; Schoepfer, J. Curr. Pharm. Des. 2000, 6, 1777–1796. See
also: (b) McNemar, C.; Snow, M. E.; Windsor, W. T.; Prongay, A.;
Mui, P.; Zhang, R.; Durkin, J.; Le, H. V.; Weber, P. C. Biochemistry
1997, 36, 10006–10014. (c) Garcia-Echeverria, C.; Furet, P.; Gay, B.;
Fretz, H.; Rahuel, J.; Schoepfer, J.; Caravatti, G. J. Med. Chem. 1998,
41, 1741–1744. (d) Liu, W.-Q.; Vidal, M.; Olszowy, C.; Million, E.;
Lenoir, C.; Dhoˆtel, H.; Garbay, C. J. Med. Chem. 2004, 47, 1223–
1233.
(29) (a) Davidson, J. P.; Martin, S. F. Tetrahedron Lett. 2000, 41, 9459–
9564. (b) Plake, H. R.; Sundberg, T. B.; Woodward, A. R.; Martin,
S. F. Tetrahedron Lett. 2003, 44, 1571–1574.
(28) Kessels, H. W. H. G.; Ward, A. C.; Schumacher, T. N. M. Proc. Natl.
Acad. Sci. U.S.A. 2002, 99, 8524–8529.
(30) For a review of the use of ITC, see: Ladbury, J. E. Thermochim. Acta
2001, 380, 209–215.
9
J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009 16761

A R T I C L E S
DeLorbe et al.
Table 1. Thermodynamic Parameters for Complex Formation between the Grb2 SH2 Domain and Pseudopeptides 5-16a
Ka
(M-1
∆G°
∆H°
∆S°
-T∆S°
ligand
)
(kcal mol-1
)
(kcal mol-1
)
(cal mol-1 K-1
)
(kcal mol-1
)
5 (fpYVN)
11 (cpYVN)
6 (fpYIN)
12 (cpYIN)
7 (fpYLN)
13 (cpYLN)
8 (fpYQN)
14 (cpYQN)
9 (fpYEN)
15 (cpYEN)
(4.5 ( 0.12) × 105
(2.8 ( 0.10) × 106
(4.0 ( 0.15) × 105
(2.1 ( 0.08) × 106
(1.7 ( 0.06) × 105
(7.1 ( 0.27) × 105
(5.6 ( 0.15) × 105
(1.2 ( 0.06) × 106
(3.0 ( 0.08) × 105
(3.6 ( 0.10) × 105
-7.7 ( 0.02
-8.8 ( 0.02
-7.7 ( 0.02
-8.6 ( 0.02
-7.1 ( 0.02
-8.0 ( 0.02
-7.8 ( 0.02
-8.3 ( 0.01
-7.5 ( 0.02
-7.6 ( 0.02
-5.4 ( 0.14
-7.9 ( 0.29
-5.5 ( 0.20
-8.3 ( 0.30
-4.6 ( 0.17
-6.0 ( 0.22
-8.7 ( 0.23
-9.8 ( 0.20
-8.8 ( 0.23
-10.3 ( 0.27
7.9 ( 0.22
3.0 ( 0.30
7.4 ( 0.30
1.3 ( 0.30
8.6 ( 0.30
6.6 ( 0.30
-2.8 ( 0.22
-5.2 ( 0.18
-4.3 ( 0.22
-9.0 ( 0.22
-2.4 ( 0.07
-0.9 ( 0.09
-2.2 ( 0.09
-0.4 ( 0.09
-2.6 ( 0.09
-2.0 ( 0.09
0.8 ( 0.07
1.5 ( 0.05
1.3 ( 0.07
2.7 ( 0.07
10 (fpYKN)
16 (cpYKN)
(9.8 ( 0.23) × 104
-6.8 ( 0.02
-7.8 ( 0.02
-7.7 ( 0.20
-9.2 ( 0.24
-3.0 ( 0.21
-4.6 ( 0.22
0.9 ( 0.07
1.4 ( 0.07
(5.5 ( 0.15) × 105
a ITC experiments were conducted at 25 °C in duplicate in 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid (HEPES, 50 mM) with NaCl (150
mM) at pH 7.45 ( 0.05.31 A value of 5.2% for the error in ligand concentration was used, and the error arising therefrom was propagated
accordingly.22g
of monomeric Grb2 SH2 domain in buffer (50 mM HEPES,
150 mM NaCl, pH 7.45 ( 0.05) at 25 °C was titrated with
ligand in the same buffer, and the raw data were processed to
give Ka and ∆H°. The ∆G° term was calculated indirectly by
applying the modified Arrhenius equation, ∆G° ) -RT ln Ka,
and ∆S° was calculated using the Gibbs relationship ∆G° )
∆H° - T∆S°. These experiments revealed that the constrained
ligand 11 bound approximately 6-fold better than its flexible
counterpart 5 (Table 1). This increased affinity is qualitatively
in accord with the net thermodynamic benefit that would be
expected from ligand preorganization; however, the entropy of
binding for the constrained pseudopeptide 11 was 4.9 eu less
favorable than that for its flexible counterpart 5. Hence, the
increased binding affinity of 11 relative to 5, which corresponds
to a ∆∆G° of 1.1 kcal mol-1, arose from an enthalpic advantage
Figure 2. Graphical representation of the thermodynamic parameters for
complex formation between the Grb2 SH2 domain and pseudopeptides 5-16
as determined by ITC.
of 2.5 kcal mol-1 and not the more favorable entropy of binding
that is commonly associated with restricting rotors.15
Because the results of these studies were inconsistent with
the conventional wisdom regarding the putative energetic,
especially entropic, effects of ligand preorganization, the
thermodynamic parameters for the binding of 6-10 and 12-16
to the Grb2 SH2 domain were determined by ITC to ascertain
whether other flexible/constrained ligand pairs exhibited similar
thermodynamic behavior; these results are collected in Table 1
entropies were found to be either favorable or unfavorable. For
example, the ∆S° term for 5-7 and 11-13, each of which has
a hydrophobic residue at the pY+1 site, is favorable, whereas
the ∆S° term for 8-10 and 14-16, wherein this side chain is
-
either polar (i.e., -CONH2) or charged at pH ∼7.5 (i.e., -CO2
or -NH3+), is unfavorable.
and Figure 2. Each of the constrained pseudopeptides 12-16
The stunning finding that preorganization is not necessarily
accompanied by the anticipated favorable change in entropy but
bound with ∆G°’s that were more favorable than those of their
respective flexible controls 6-10 by 0.1-1.0 kcal mol-1
.
rather by an enthalpic advantage prompted us to query why.
The overall energetics of complex formation can be broadly
partitioned into the thermodynamics associated with (1) proton
transfer and desolvation of the ligand and the binding site of
the protein; (2) new nonbonded interactions, including polar
and van der Waals, that form between the ligand and the pro-
tein; and (3) conformational and dynamic changes of both ligand
and protein. Each of these factors was probed to some degree
in the following experimental studies.
Consistent with observations for the flexible/constrained ligand
pair 5 and 11, the increased affinities for the preorganized
ligands 12-16 were uniformly the consequence of binding
enthalpies that were more favorable by 1.1-2.8 kcal mol-1
.
Without exception the binding entropies for the preorganized
ligands 12-16 were unfavorable relative to their flexible
counterparts 6-10 by 1.6-6.1 eu or 0.5-1.8 kcal mol-1 at
25 °C.
Protein-ligand binding of all ligands 5-16 was primarily
enthalpy driven as binding enthalpies contributed more to the
overall energetics of complex formation than binding entropies.
Depending upon the side chain at the pY+1 position, binding
Effect of Proton Transfer on Binding Energetics. Protein
binding of ligands having ionizable groups such as a phosphate
group, which plays a significant role in the overall binding
energetics of 5-16, may be accompanied by proton transfer.
Values determined for ∆H° will then comprise not only an
enthalpy term for binding, ∆H°bind, but also an enthalpy term
that is associated with proton exchange. The contribution from
this term is dependent upon the heat of ionization of the buffer,
(31) The values of the thermodynamic parameters reported herein for 5-
7 and 11-13 differ slightly from those reported in ref 18 because
ITC experiments were conducted using a different protocol for drying
ligands that led to n values closer to 1.0.
9
16762 J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009

Entropic Paradoxy Associated with Ligand Preorganization
A R T I C L E S
∂(∆Ho)
∂T
∂(∆So)
∂T
∆H°ion, and the number of protons, n, transferred according to
eq 3.32,33
∆C )
) T
(4)
(
)
(
)
p
p
p
∆H° ) ∆H° + n∆H°
(3)
bind
ion
The ∆Cps for 5 and 11, each of which binds with a favorable
entropy, and 8 and 14, each of which binds with an unfavorable
entropy, were determined by obtaining binding enthalpies for
these ligands at three different temperatures within the range
15-35 °C. This study was designed to ascertain how ∆Cps
might vary depending upon whether binding entropies are
positive or negative. The values of ∆Cp for 5 and 11 derived
from plots of ∆H° versus temperature, T, were found to be
-158.5 ( 8.6 and -167.5 ( 6.2 cal mol-1 K-1, respectively,
whereas the corresponding values for 8 and 14 were -133.0 (
15.1 and -125.8 ( 29.5 cal mol-1 K-1 (Figure 4). The negative
values of these ∆Cps are in accord with the burial of nonpolar
surfaces upon binding, and the less negative values for 8 and
14 relative to 5 and 11 are qualitatively consistent with the
presence of a more polar side chain at the pY+1 position. More
important from the perspective of evaluating the energetic effects
of preorganization, the values of ∆Cp for each ligand in a
flexible/constrained pair are the same within experimental error,
suggesting that differences in the binding free energies of pairs
of flexible and constrained ligands haVing the same pY+1
residue do not appear to arise from desolVation or hydrophobic
effects.
Although the thermodynamic parameter ∆Cp provides some
information relative to desolvation effects in protein-ligand
interactions, it does not give a direct measure of differences in
ligand hydrophobicity. Such information may be obtained by
comparing solvation free energies,36 but the nonvolatile nature
of 5-16 precluded a direct experimental determination of
solvation free energy. The contribution of ligand desolvation
to binding can also be estimated by transferring the ligand from
an aqueous buffer to an organic solvent such as octanol.37
However, owing to the presence of the ionized phosphate group,
none of the ligands of this study were sufficiently soluble in
octanol at pH 7.5, the pH at which ITC experiments were
performed, to determine the partition coefficient.
To ascertain the extent to which the binding enthalpies
included contributions from proton transfers, we determined
∆H° for the pair of flexible and constrained ligands 5 and 11
in PIPES, HEPES, and imidazole buffers, which have corre-
27b
sponding heats of ionization of 2.7, 5.0, and 8.8 kcal mol-1
.
The slopes of the plots of ∆H° values against ∆H°ion for 5 and
11 in these buffers reveal that the number of protons transferred
upon complexation was 0.27 and 0.20 for 5 and 11, respectively
(Figure 3). This corresponds to a contribution to ∆H° from
proton transfer processes of ∼1.0-1.4 kcal mol-1 in HEPES,
the buffer used in the ITC studies. Because the number of proton
equivalents transferred is nearly the same for 5 and 11, the
contribution from ∆H°ion to ∆H° of binding for each is within
experimental error of being identical. Consequently, the en-
hanced ∆H° for binding associated with preorganizing 5 does
not appear to arise from any significant difference in proton
transfer processes. We posit that this conclusion holds for the
other flexible/constrained ligand pairs in this study, although
the number of protons transferred might be different owing to
the presence of acidic and basic groups on the side chains
of the Glu- and Lys-derived ligands 9 and 15 and 10 and 16,
respectively.
Effects of Enthalpy-Entropy Compensation in Binding
Energetics. Examination of the data in Table 1 reveals that ∆G°s
for forming complexes between the Grb2 SH2 domain and the
constrained ligands 11-16 are more favorable than for their
respective flexible controls 5-10 by up to 1.1 kcal mol-1. The
variation in binding enthalpy for a pair of flexible and
constrained ligands is as large as 2.8 kcal mol-1, but this
enthalpic advantage is partially offset by an unfavorable binding
entropy of as much as 1.8 kcal mol-1. Hence, enthalpy-entropy
compensation for forming these complexes of corresponding
Figure 3. Dependence of the measured ∆H° of binding of 5 (b) and 11
(9) at pH 7.5 and 25 °C upon the heat of ionization of the buffer: PIPES,
∆H°ion ) 2.7 kcal mol-1; HEPES, ∆H°ion ) 5.0 kcal mol-1; imidazole,
∆H°ion ) 8.8 kcal mol-1. The slope of each line is proportional to the
contribution of proton transfer to ∆H° of binding.
(32) For example, see: (a) Baker, B. M.; Murphy, K. P. Biophys. J. 1996,
71, 2049–2055. (b) Fukuda, H.; Takahashi, K. Proteins: Struct., Funct.,
Genet. 1998, 33, 159–166. (c) Bradshaw, J. M.; Waksman, G.
Biochemistry 1998, 37, 15400–15407.
Temperature Dependence of Binding Thermodynamics. Tem-
perature dependent studies of ∆H° for bimolecular interactions
give the heat capacity changes, ∆Cp, for complex formation as
defined in eq 4. The significance of this parameter has been
ascribed to a number of phenomena,34 but in biological systems,
∆Cp serves primarily as a barometer of the hydrophobic effect.
A negative value of ∆Cp signifies the burial of nonpolar surfaces,
whereas a positive value is attributed to the burial of polar
surfaces.35 Hence, in protein-ligand interactions, a comparison
of ∆Cps for the binding of closely related ligands to a protein
provides an indication of whether desolvation effects might be
involved in differential binding affinities.
(33) For example, see: (a) Bradshaw, J. M.; Waksman, G. Biochemistry
1998, 37, 15400–15407.
(34) (a) Sturtevant, J. M. Proc. Natl. Acad. Sci. U.S.A. 1977, 74, 2236–
2240. (b) Spolar, R. S.; Ha, J. H.; Record, M. T., Jr. Proc. Natl. Acad.
Sci. U.S.A. 1989, 86, 8382–8385.
(35) For a review, see: Southall, N. T.; Dill, K. A.; Haymet, A. D. J. J.
Phys. Chem. B 2002, 106, 521–533.
(36) For tables of experimental solvation Gibbs energies and computional
methods for their calculation, see: (a) Wang, J.; Wang, W.; Huo, S.;
Lee, M.; Kollman, P. A. J. Phys. Chem. B 2001, 105, 5055–5067. (b)
Chuman, H.; Mori, A.; Tanaka, H. Anal. Sci. 2002, 18, 1015–1020.
(c) Gallicchio, E.; Zhang, L. Y.; Levy, R. M. J. Comput. Chem. 2002,
23, 517–529.
9
J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009 16763

A R T I C L E S
DeLorbe et al.
Figure 4. ∆H° of ligand binding as a function of temperature. Measurements were performed in 50 mM HEPES buffer (150 mM NaCl, pH 7.45 ( 0.05)
at three different temperatures. At least two measurements were performed at each temperature. ∆Cp values for the binding of each ligand were obtained
from the slope of the plots with the error in ∆Cp being the standard error in the slope. (a) Plot for the flexible ligand 5 (b) and the constrained ligand 11
(9). (b) Plot for the flexible ligand 8 (b) and the constrained ligand 14 (9).
Structural Studies of Complexes of Flexible and Constrained
Ligands with the Grb2 SH2 Domain. In the next stage of the
inquiry, we undertook X-ray crystallographic investigations of
complexes of flexible and constrained ligands having the same
amino acid residue at the pY+1 position. We recognize,
however, that interpreting such structural data is subject to the
caveats that differences in resolution necessarily lead to
uncertainties in structure and contact distances and angles and
that the static interactions observed in the solid state may not
accurately reflect the dynamic interactions in solution. Neverthe-
less, such studies are commonly relied upon to ascertain whether
structurally related ligands bind similarly to proteins.39 By
comparing the binding modes and interactions for each member
of different flexible/constrained ligand pairs, we hoped, perhaps
naively, to identify structural differences that might be correlated
with differential binding energetics.
Figure 5. Correlation between ∆H° and ∆S° of binding as obtained from
ITC measurements for flexible ligands (b) and constrained ligands (9).
Suitable crystals for X-ray analysis were obtained for
complexes of the Grb2 SH2 domain with 5, 6, 8, 11, 12, and
14.40 Data for the complexes of 5 and 11 were collected to 1.7
and 1.9 Å resolution, respectively, and the structures were solved
flexible and constrained ligands is not balancing, and relatiVe
by molecular replacement using a known structure.26d Data for
changes in ∆H° dominate those in ∆S°.
the complexes with 6, 8, 12, and 14 were collected to 1.7, 1.8,
2.0, and 1.7 Å resolution, respectively, and the structures were
solved similarly. There was one complex in the asymmetric unit
in the structures of 5 and 6, but there were two complexes in
The range of ∆G° values for the set of ligands 5-16 was
2.0 kcal mol-1, whereas the values of ∆H° varied by 5.7 kcal
mol-1 and T∆S° differed by 5.3 kcal mol-1. It follows directly
from the Gibbs relationship, ∆G° ) ∆H° - T∆S°, that when
the asymmetric unit for 8, 11, 12, and 14. Having multiple copies
variations in ∆G° are small, a change in ∆H° will be offset by
of a protein-ligand complex in the asymmetric unit is useful
a compensating change in ∆S°. A plot of ∆H° versus ∆S° will
because it affords independent snapshots of approximately
then be linear, and the slope of the line thus obtained is the
isoenergetic structures, thereby providing a benchmark for
compensation temperature, Tc. Based upon such a plot for
the binding of 5-16 (Figure 5), Tc was found to be 296 K. The
interpreting the importance of any observed structural differences.
Comparing the structures of the complexes of 5, 6, and 8
test of Krug (eq 5),38 where σ is the standard error in the slope
with those of their corresponding cyclopropane-derived ligands
obtained from linear regression analysis and T is the temperature
at which the titration experiments were conducted, was applied
11, 12, and 14 reveals a number of similarities as well as some
to the data to ascertain whether the difference between Tc and
T was statistically significant at a 95% confidence level.21a
(37) For example, see: (a) Radzicka, A.; Wolfenden, R. Biochemistry 1988,
27, 1664–1670. (b) Wimley, W. C.; Creamer, T. P.; White, S. H.
Because application of this relationship gives a range for T of
Biochemistry 1996, 35, 5109–5124.
(38) Krug, R.; Hunter, W.; Grieger, R. Nature 1976, 261, 566–567.
(39) A number of structurally similar ligands are known to bind differently
to the same protein. See: Bostro¨m, J.; Hogner, A.; Schmitt, S. J. Med.
Chem. 2006, 49, 6716–6725.
Tc - 1.81σ < T < Tc + 1.81σ (5)
(40) Complexes of the Grb2 SH2 domain with ligands 6, 8, 12, and 14
were deposited into the Research Collaboratory for Structural Bioin-
formatics (RCSB) Protein Data Bank under entries 3IN8, 3IMD, 3IMJ,
and 3IN7, respectively. Complexes with ligands 5 and 11 were
deposited previously under entries 3C7I and 2HUW, respectively.
243-350 K for experiments that were performed at 298 K, the
observed enthalpy-entropy compensation for 5-16 appears to
be a consequence of the Gibbs free energy relationship rather
than some extra thermodynamic phenomenon.
9
16764 J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009

Entropic Paradoxy Associated with Ligand Preorganization
A R T I C L E S
Figure 7. Overlay of residues SerꢀB7 to AspꢀC1, which includes the BC-
loop (GluBC1-ProBC4), of the Grb2 SH2 domain complexed with flexible
and constrained ligands showing water molecules involved in hydrogen bond
network between the N-terminal carbonyl oxygen atom, the amide moiety
in the Gln side chain, and HisꢀD4. Residues are represented as ribbons,
and the bound ligands as sticks. (a) The set of flexible ligands 5 (orange),
6 (purple), and 8 (green). (b) The set of constrained ligands 11 (orange),
12 (purple), and 14 (green).
loops relative to the backbone of the domains and the oxygen
atoms of the phosphate groups (cf. Figure 7b); the relative
relationship between the BC loops and the phosphate groups
of the flexible ligands are virtually identical (cf. Figure 7a).
Inspection of Figure 6b reveals that the backbone atoms in the
BC loops in both complexes in the asymmetric unit of 11 are
packed closer to the phosphate moiety than in the complex with
5. On the other hand, there is a notable difference in the
relationship between the phosphate groups and the BC loops in
the two complexes in the asymmetric unit of 12, and the spatial
relationship in one of these is similar to that found in the
complex of its flexible counterpart 6 (Figure 6c). The orienta-
tions of the phosphate groups relative to the backbone atoms
of the BC loops in the two complexes of 8 and in the two
complexes of its constrained analog 14 are similar (Figure 6d).
In the context of making structural comparisons, it is noteworthy
that Variations in the dispositions of the backbone atoms in the
BC loops relatiVe to the phosphate groups of flexible ligands
and their constrained counterparts are generally comparable
to differences found in multiple copies of the same complex in
some of the asymmetric units, especially those of 12. Accord-
ingly, it is difficult to assess the extent to which these differences
in relative orientations are energetically significant.
Structural alignments of the pYVN analogs 5 and 11 (Figure
6b), the pYIN analogs 6 and 12 (Figure 6c), and the pYQN
analogs 8 and 14 (Figure 6d) reveal that the pY+1 and pY+2
residues in each are virtually identical. Indeed, the positions of
all atoms in these residues align with rmsd values of less than
0.24 Å, a value that is within the rmsd cutoff of 0.25 Å
commonly used to define “identical” structures found by
molecular dynamics calculations.41 These superimpositions do
reveal, however, significant differences in the positions of atoms
in the flexible and constrained Ac-pY replacements. Namely,
atoms in these replacements in the complexes of 5 and 11, 6
and 12, and 8 and 14 align with average rmsd values of 0.88,
0.77, and 1.08 Å, respectively, with the greatest variations being
in the relative positions of the bridging phosphate oxygen atoms
and the atoms of the N-acetyl moieties. Most notably, the
Figure 6. Overlay of Grb2 SH2 domain complexed with flexible and
constrained ligands. Loops are labeled, and residues are represented as
ribbons and the bound ligands as sticks. In images (b)-(d), only residues
SerꢀB7 to AspꢀC1 of the domain, which includes the BC loop (GluBC1-
ProBC4), are shown for all complexes in the asymmetric unit. (a) Overlay
of complex of 5 (orange) and one complex in the asymmetric unit of 11
(purple), showing the complete domain. (b) Overlay of Grb2 SH2 in complex
with 5 (orange) and 11 (purple). (c) Overlay of Grb2 SH2 in complex with
6 (orange) and 12 (purple). (d) Overlay of Grb2 SH2 in complex with 8
(orange) and 14 (purple) showing water molecules involved in hydrogen
bond network between the N-terminal carbonyl oxygen atom, the amide
moiety in the Gln side chain, and HisꢀD4.
differences (Figure 6). Flexible and constrained ligands bind to
the domain in a ꢀ-turn-like conformation featuring an intramo-
lecular hydrogen bond between the pY+1 carbonyl oxygen atom
and the C-terminal amide. The alignment of the complex of 5
(orange) and one of the complexes in the asymmetric unit of
11 (purple) (Figure 6a) shows that the helical and the ꢀ-strand
secondary structural elements of the domain in the two
complexes are virtually identical; however, there are some
notable differences in the loop regions, especially in the BC
loop. A comparison of the structures of other complexes of
flexible and constrained ligand pairs reveals similar features.
For example, the backbone atoms of the Grb2 SH2 domain in
its complexes with 5, 6, and 8 align with the backbone atoms
in the corresponding complexes with 11, 12, and 14 with a root-
mean-square deviation (rmsd) of 0.37-0.56 Å. The rmsd values
for these alignments are comparable to those obtained from
aligning the backbone atoms in each of the two complexes in
the asymmetric unit for 8 (0.07 Å), 11 (0.29 Å), 12 (0.42 Å),
and 14 (0.42 Å). Hence, excepting some Variations in the flexible
loop regions, there are no significant conformational differences
for the backbone atoms of the Grb2 SH2 domain in its complexes
with flexible and constrained ligands.
The aforementioned dissimilarities in the orientations of the
backbone atoms of the residues in the BC loop (GluBC1-
ProBC4) in complexes of flexible and constrained ligand pairs
are apparent in the overlays in Figure 6b-d. These differences
do not result from variations in the conformation of the backbone
atoms in the BC loops in the complexes of the constrained
ligands. Rather they arise from variable orientations of these
(41) See: Macromodel, version 9.0; Schro¨dinger, LLC: New York, NY,
2005.
9
J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009 16765

A R T I C L E S
DeLorbe et al.
in the ligands 5 and 19 align with a rmsd of 0.3 Å. There are
slight variations in the pairwise polar interactions and contact
distances between the ligands and the domain in these two
complexes; however, because these differences are similar to
those observed in structures of other complexes of phosphoty-
rosine-containing ligands bound to the Grb2 SH2 domain,13c,26
attaching significance to these dissimilarities is difficult. The
backbone atoms of the domain in the two complexes of 11 and
the complex of 20 align with a rmsd of 0.35 Å, and the
corresponding atoms in 11 and 20 align with a rmsd of 0.6 Å.
Despite these differences, the pairwise polar interactions and
contact distances between the BC loop of the domain and the
phosphate groups of 11 and 20 in these complexes are closely
comparable. This analysis suggests that the succinyl- and
cyclopropane-deriVed pY replacements in 5 and 11 are not
uniquely responsible for any significant structural Variations
in their respectiVe complexes with the Grb2 SH2 domain. The
pY surrogates in 5 and 11 thus appear to be good mimics of
the corresponding pY residues in the linear peptide 19 and the
macrocycle 20.
That there are discernible structural differences in complexes
of corresponding pairs of flexible and constrained ligands is
not surprising, as Bostro¨m has shown that like molecules often
do not bind to and interact with a given target in a similar
manner.39 This, however, begs the question of whether there
are any dissimilarities in the interactions between the domain
and pairs of flexible and constrained ligands that may be
correlated with the observed differences in binding enthalpies
or entropies. To address this issue, the polar contacts, which
may be categorized as direct or single water-mediated, between
the domain and flexible and constrained ligands having the same
residue at the pY+1 position were compared using contact
diagrams such as those in Figure 9 (see Supporting Information
for contact diagrams for 6, 12, 8, and 14). The only direct
contacts considered in this analysis are those wherein the
distance between an electronegative atom on the ligand and an
electronegative atom on the protein is within 2.5-3.4 Å. The
same criteria were applied to contacts that are mediated by a
single water molecule; protein-ligand interactions that are
mediated by more than one water molecule are not considered.
Because of differences in the resolution of the crystallographic
data, we made no attempt to further distinguish polar interactions
based upon their measured contact distances or angles.
As highlighted in Figure 9a, there are seven direct contacts
between the domain and the Ac-pY replacement that are
conserved in all complexes of flexible ligands, whereas there
are eight direct contacts involving the Ac-pY replacement that
are conserved in all complexes of constrained ligands (Figure
9c). The contact between SerꢀC3 and the bridging phosphate
oxygen atom that is conserved in the complexes of the
constrained ligands is not found in any of the complexes of
their flexible counterparts owing to the different orientation of
this atom in these latter complexes. There are two direct contacts
between the N-terminal carbonyl oxygen atom and the side chain
of ArgRA2 that are conserved in the complexes of 5, 6, and all
of the constrained ligands. On the other hand, the N-terminal
carbonyl oxygen atom of 8 makes one contact with the side
chain of ArgRA2 and a second contact with a water molecule
involved in a hydrogen bond network with the Gln side chain
of the ligand. There are two conserved, single water-mediated
interactions involving W1 and W2, the domain, and the
phosphate group in the flexible ligands 5, 6, and 8. In one of
these, W1 participates in interactions with the backbone N-H
Figure 8. Complexes of ligands with the Grb2 SH2 domain showing
relationship between the ligands and residues SerꢀB7-AspꢀC1, which
includes the BC-loop (GluBC1-ProBC4), of the domain. (a) Overlay of
structures of complexes of 5 (orange) with the linear nonapeptide 19
(purple),26d showing only the Ser-pTyr-Val-Asn segment of 19 for clarity.
(b) Alignment of one complex in the asymmetric unit of 11 (orange) with
the structure of the complex of macrocycle 20 (purple),13c showing only
the portion of 20 that is purple for clarity. (c) Structure of 20.
N-terminal amide of 8 is oriented so that the carbonyl oxygen
atom can participate in a water-mediated hydrogen bond network
with the side chain of the glutamine residue at the pY+1 site
(Figure 6d). Because the corresponding carbon-carbon bond
in 14 is incorporated in the cyclopropane ring, this conformation
is not accessible to 14.
A comparison of structural alignments for all complexes
reveals that the set of flexible ligands 5, 6, and 8 (Figure 7a)
bind and interact with the Grb2 SH2 domain with more
consistency than does the set of their corresponding constrained
ligands 11, 12, and 14 (Figure 7b). Excepting the orientation
of the N-terminal amide group in 8, all of the corresponding
atoms in the flexible ligands align with rmsd values e0.12 Å,
and the maximum displacement of the backbone atoms of the
BC loops is 0.3 Å. Although the corresponding atoms in the
pY+1 and pY+2 residues of the constrained ligands 11, 12
and 14 also align closely (rmsd values e0.16 Å), superimposi-
tion of the atoms of the cyclopropane-derived Ac-pY replace-
ment is not quite as good (rmsd values e0.26 Å) because of
variable orientations of the N-acetyl group. The maximum
displacement of the backbone atoms of the BC loops in these
complexes is 1.8 Å, suggesting that the BC loop can readily
adopt a number of orientations so the amino acid side chains
can achieve optimal interactions with the phosphate group.
We then compared the structures of the complexes of the
Grb2 SH2 domain with 5 and the linear nonapeptide Ala-Pro-
Ser-pTyr-Val-Asn-Val-Gln-Asn (19)26d (Figure 8a) and of the
domain with 11 and the macrocycle 2013c (Figure 8b). The
backbone atoms of the domain in the complexes of 5 and 19
align with a rmsd of 0.5 Å, whereas the corresponding atoms
9
16766 J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009

Entropic Paradoxy Associated with Ligand Preorganization
A R T I C L E S
Figure 9. Polar interactions in the range of 2.5-3.4 Å in complexes of Grb2 SH2 domain with 5 and one complex in the asymmetric unit of 11. All labile
hydrogen atoms have been omitted for clarity except those on protein backbone nitrogen atoms. Only those water molecules that mediate a contact between
the domain and the ligand are shown, and these are numbered so that water molecules that are conserved in at least two complexes have the same number.
Solid lines in (a) and (b) indicate those polar contacts that are conserved for all complexes of flexible ligands, and solid lines in (c) and (d) indicate those
polar contacts that are conserved for all complexes of constrained ligands. Dotted lines represent contacts that are not conserved within the set of flexible
or constrained ligands. (a) Interactions between pY replacement of 5 and domain. (b) Interactions between VN of 5 and domain. (c) Interactions between
pY replacement of 11 and domain. (d) Interactions between VN of 11 and domain.
of GluBC1 and one of the nonbridging phosphate oxygen atoms
of the Ac-pY replacement, and in the other, W2 interacts with
the side chain hydroxyl group of SerBC2 and the bridging
phosphate oxygen atom of the ligand (cf. Figure 9a). There are
no water-mediated, protein-ligand interactions in the Ac-pY
binding pocket that are conserved in the set of constrained
ligands.
mediated protein-ligand interactions in this binding pocket that
are conserved in either set of complexes.
A summary of direct and single water-mediated contacts for
all complexes is presented in Table 2. An inventory of the
interactions for ligands having the same amino acid at the pY+1
site reveals that there are generally more direct contacts between
the domain and the constrained member of a pair, whereas there
are more single water-mediated contacts involving its flexible
counterpart. For each pair of ligands, the larger number of direct
contacts correlates qualitatively with the more favorable binding
enthalpies and Gibbs energies that are observed, although there
is no such correlation for the total number of polar contacts. A
There are four direct contacts between the domain and the
pY+1-N subunit of the ligands that are conserved in all
complexes (cf. Figure 9b and 9d). There are more single water-
mediated contacts in the pY+1-N binding pocket in the
complexes of the flexible ligands, but there are no water-
9
J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009 16767

A R T I C L E S
DeLorbe et al.
Table 2. Direct and Single Water-Mediated Polar Protein-Ligand Contacts As Determined by X-Ray Crystallographya
a See Supporting Information for contact diagrams of 6, 8, 12 and 14.b One water molecule makes a contact with the backbone nitrogen atom of
HisꢀD4 and two contacts with the ligand, one with the N-Ac moiety of the pY replacement and the other with the Gln side chain; each are counted
independently.
closer examination of the summary shown in Table 2, however,
underscores the difficulties associated with trying to correlate
polar protein-ligand contacts with binding energetics because
the Variations in the number of direct and water mediated
contacts inVolVing similar ligands may be comparable to
differences found in multiple copies of a complex in an
asymmetric unit as exemplified by analyzing 12a and 12b
(Table 2).
complexes is the orientation and packing of the side chain of
LysꢀD6, which may form a van der Waals contact either with
the aromatic ring of the pY residue or with the C-terminal amide
moiety of the ligand. These differences would seem to be
approximately offsetting. There are a number of other dis-
similarities in the van der Waals contacts in the structures of
flexible and constrained ligands, but these generally do not
appear significant relative to the variations that occur in different
complexes in the same asymmetric unit. Based upon this
analysis, the Van der Waals contacts in complexes of flexible
and constrained ligands seem comparable and cannot be easily
correlated with any differences in binding enthalpies.
Crystallographic b-factors have been used to probe changes
in protein flexibility in protein-ligand complexes.45 This
practice is, however, subject to numerous caveats because these
thermal parameters are influenced by crystal packing, resolution,
temperature of data acquisition, methods used to solve the
structure, and a number of other effects that are not directly
related to changes in protein flexibility.46 Atomic b-factors also
reflect a static, solid-state picture of nonbonded interactions that
are much more dynamic in solution owing to greater thermal
fluctuations in atomic positions. These issues notwithstanding,
comparisons of b-factors can provide a sense of the relative
magnitudes of motion in the solid state, if all of the b-factors in
the data sets are adjusted by a constant value for the data set so
the lowest value in each set corresponds to the lowest value
in the data set having the highest overall displacement.46a
It is evident that the interfacial interactions involving water
molecules vary for the flexible and constrained ligands of a given
pair, but assessing the detailed role of water in the energetics
of protein-ligand interactions is a daunting challenge.42 Al-
though such a discussion is beyond the scope of the present
work, several points are worth brief mention. Fixing a water
molecule at a protein-ligand interface is generally regarded as
being entropically unfavorable.43 Because there are more single-
water mediated contacts in the complexes of the flexible ligands,
it is somewhat perplexing that the binding entropies for forming
these complexes are more favorable. Other contributions to
binding entropy must overcome the entropic penalty associated
with these bound water molecules. It has also been shown that
the release of water molecules that are not optimally hydrogen
bonded can have favorable enthalpic consequences.11 This
finding raises the interesting question of whether the more
favorable enthalpies observed for forming complexes of the
preorganized ligands, which have fewer water molecules bound
at the protein-ligand interface, might arise from the release of
such water molecules from the domain. Unfortunately, the Grb2
SH2 domain invariably crystallizes as a domain swapped dimer
in the absence of ligand,44 so the details of the hydration of the
ligand binding site of the uncomplexed domain and the fate of
these water molecules upon binding are unknown.
Applying this protocol to the crystallographic data sets reveals
that the average adjusted b-factors for all atoms in residues
59-151 of the Grb2 SH2 domain in its complexes with the
flexible ligands 5, 6, and 8 are 3.7-6.6 Å2 less than those in
the corresponding set of complexes of the constrained ligands
11, 12, and 14 (Figure 10). Although there are differences in
the magnitudes of these b-factors throughout the complexes,
the most notable variations are in the loop regions, especially
in the four residues in the BC loop, which is involved in
The van der Waals contacts in complexes of the Grb2 SH2
domain with the flexible and constrained ligands were then
examined. Perhaps the most notable variation in the different
(42) For some leading, general references, see: (a) Ladbury, John E. Chem.
Biol. 1996, 3, 973–980. (b) Levy, Y.; Onuchic, J. N. Annu. ReV.
Biophys. Biomol. Struct. 2006, 35, 389–415. (c) Li, Z.; Lazaridis, T.
Phys. Chem. Chem. Phys. 2007, 9, 573-581. See also ref 11d.
(43) The maximal value for transferring a water molecule from bulk water
(45) For a recent discussion of using b-factors to compare protein flexibility
in protein-ligand complexes, see: Yang, C.-Y.; Wang, R.; Wang, S.
J. Med. Chem. 2005, 48, 5648–5650.
to a protein binding site has been suggested to be ∼2 kcal mol-1
.
(46) For leading references, see: (a) Ringe, D.; Petsko, G. A. Methods
Enzymol. 1986, 131, 389–433. (b) Bhalla, J.; Storchan, G. B.;
MacCarthy, C. M.; Uversky, V. N.; Tcherkasskay, O. Mol. Cell.
Proteomics 2006, 5, 1212–1223.
See: Dunitz, J. D. Science 1994, 264, 670.
(44) Benfield, A. P.; Whiddon, B. B.; Clements, J. H.; Martin, S. F. Arch.
Biochem. Biophys. 2007, 462, 47–53.
9
16768 J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009

Entropic Paradoxy Associated with Ligand Preorganization
A R T I C L E S
of 11 (cpYVN1) and in one complex of 12 (cpYIN1) in their
respective asymmetric units is similar, yet the relative orienta-
tions of the BC loops in each of these complexes differ
markedly. There are also significant dissimilarities in the crystal
packing for each of the two complexes in the asymmetric unit
of 14 (cpYQN1 and cpYQN2), but the structures of the two
complexes are comparable. Moreover, the crystal packing in
these two complexes is different from either of the two
complexes in the asymmetric unit of 8 (fpYQN1 and fpYQN2),
even though the structures of all of these complexes, with the
notable exception of the orientation of the N-terminal acetyl
group in 8, are similar. Based upon these observations, it is not
possible to correlate differences in crystal packing with Varia-
tions in contact distances, b-factors, or other structural details.
Figure 10. Adjusted atomic b-factors in complexes of the Grb2 SH2
domain (residues 59-151) with flexible and constrained ligands. Backbone
atoms of residues are color-coded from blue to green to red for atoms having
values of 20, 30, and 40 Å2. (a) Complexes of flexible ligands 5, 6, and 8.
(b) Complexes of constrained ligands 11, 12, and 14.
Summary and Conclusions
There is ample evidence that ligand preorganization may
provide compounds having improved target selectivity,48 bio-
availability,49 and higher binding affinity.13,14 The present
investigations also demonstrate that constraining ligands in their
biologically active conformations can lead to more potent
analogs. The higher affinities observed for the conformationally
constrained ligands 11-16 arise, however, because they benefit
from more favorable binding enthalpies than their respective
flexible controls 5-10. Indeed, binding entropies for these
preorganized ligands are uniformly disfavored relative to their
flexible counterparts. That this surprising finding is not simply
a consequence of the peptide replacements used in these studies
is evident from our previous work with Src SH2 domain binding
ligands where we found that compounds having cyclopropane-
derived pY replacements did exhibit more favorable binding
entropies than their flexible succinate analogs. Moreover, Spaller
has recently found that preorganization of a linear peptide by
macrocyclization can lead to increased affinity for binding to a
PDZ3 domain because a relative enthalpic advantage dominates
a corresponding entropic penalty.20b Based upon several inde-
pendent studies, it is thus eVident that the widely held assumption
that ligand preorganization should be accompanied by an
entropic adVantage is not always true.50
phosphate binding, and in the seven residues in the BG loop
(SerBG1-GlnBG7), which makes single water-mediated contacts
with the pY+1 residue. Indeed, the average b-factor for all
backbone atoms in the BC loops in the complexes of 5, 6, and
8 is approximately the same as the domain average, whereas
the average b-factor for the same backbone atoms in the
complexes of 11, 12, and 14 is nearly one standard deviation
above the domain average.
It is known, but not widely appreciated, that forming a
protein-ligand complex may result in either an overall decrease
or increase in protein flexibility and dynamics.22f,47 Such
changes in flexibility will be accompanied by changes in
nonbonded interactions and order throughout the complex that
will eventuate in enthalpic and entropic consequences. It is
difficult, however, to evaluate how such changes in dynamics
will affect the relative magnitudes of the compensating enthal-
pies and entropies of binding. For example, one might infer
from the preceding analysis of b-factors that the thermal
motions, at least in the solid state, in the complexes of the
constrained ligands are greater than those in the complexes of
their more flexible counterparts. Considering the contributions
of the protein to the thermodynamic binding parameters, one
might then predict that the entropies, especially the configura-
tional entropies, for forming complexes of the constrained
ligands would be more favorable than those for their flexible
analogs, a prediction that is opposite the experimental observation.
Crystal packing in the different complexes of the Grb2 SH2
domain with the flexible ligands 5, 6, and 8 is virtually identical,
whereas the crystal packing in the complexes of the constrained
ligands 11, 12, and 14 varies from one to another and from
their flexible counterparts. Despite these differences, however,
variations in crystal packing do not seem to correlate with
structural changes. For example, crystal packing in one complex
A search for the origin of this unexpected discovery has,
however, not yet led to concrete answers. We found that
variations in binding enthalpies and entropies for the flexible
and constrained ligands did not appear to arise from differences
in either proton exchange or desolvation phenomena.
The Gibbs energy relationship accounts for the observed
enthalpy-entropy compensation that accompanies preorgani-
zation and changes in the nature of the pY+1 residue.
Crystallographic studies of complexes of flexible and constrained
(48) For a review, see: (a) Tyndall, J. D. A.; Pfeiffer, B.; Abbenante, G.;
Fairlie, D. P. Chem. ReV. 2005, 105, 793–826. For some selected
examples, see: (b) Kawai, M.; Horikawa, Y.; Ishihara, T.; Shimamoto,
K.; Ohfune, Y. Eur. J. Pharmacol. 1992, 211, 195–202. (c) Liao, S.;
Alfaro-Lopez, J.; Shenderovich, M. D.; Hosohata, K.; Lin, J.; Li, X.;
Stropova, D.; Davis, P.; Jernigan, K. A.; Porreca, F.; Yamamura, H. I.;
Hruby, V. J. J. Med. Chem. 1998, 41, 4767–4776. (d) Suich, D. J.;
Mousa, S. A.; Singh, G.; Liapakis, G.; Reisine, T.; DeGrado, W. F.
Bioorg. Med. Chem. 2000, 8, 2229–2241. (e) Ying, J.; Gu, X.; Cai,
M.; Dedek, M.; Vagner, J.; Trivedi, D. B.; Hruby, V. J. J. Med. Chem.
2006, 49, 6888–6896.
(47) For some leading references, see: (a) Anderegg, R. J.; Wagner, D. S.
J. Am. Chem. Soc. 1995, 117, 1374–1377. (b) Kay, L. E.; Muhandiram,
D. R.; Wolf, G.; Shoelson, S. E.; Forman-Kay, J. D. Nat. Struct. Biol.
1998, 5, 156–163. (c) Engen, J. R.; Gmeiner, W. H.; Smithgall, T. E.;
Smith, D. L. Biochemistry 1999, 38, 8926–8935. (d) Wang, F.; Miles,
R. W.; Kicska, G.; Nieves, E.; Schramm, V. L.; Angelettin, R. H.
Protein Sci. 2000, 9, 1660–1668. (e) Stone, M. J. Acc. Chem. Res.
2001, 34, 379–388. (f) Wang, F.; Shi, W.; Nieves, E.; Angeletti, R. H.;
Schramm, V. L.; Grubmeyer, C. Biochemistry 2001, 40, 8043–8054.
(g) Ferreon, J. C.; Hilser, V. J. Protein Sci. 2003, 12, 982–996. (h) de
Mol, N. J.; Catalina, M. I.; Fischer, M. J. E.; Broutin, I.; Maier, C. S.;
Heck, A. J. R. Biochim. Biophys. Acta 2004, 1700, 53–64. (i)
MacRaild, C.; Daranas, H.; Bronowska, A.; Homans, S. W. J. Mol.
Biol. 2007, 368, 822–832.
(49) Veber, D. F.; Johnson, S. R.; Cheng, H.-Y.; Smith, B. R.; Ward, K. W.;
Kopple, K. D. J. Med. Chem. 2002, 45, 2615–2623.
(50) For a recent computational study of the energetic consequences
associated with the flexibility of hosts and guests, see: Moghaddam,
S.; Inoue, Y.; Gilson, M. K. J. Am. Chem. Soc. 2009, 109, 4012–
4021.
9
J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009 16769

A R T I C L E S
DeLorbe et al.
ligands having the same pY+1 residue reveal widespread
similarities, and many of the observed differences are compa-
rable to those seen in multiple copies of the same complex in
some asymmetric units. The more notable dissimilarities in the
complexes of a flexible/constrained pair are found in the number
and type of polar interactions between the BC loop and the Ac-
pY replacements in the ligands. Specifically, the constrained
ligand of a given pair makes more direct contacts with the
domain, so there is a qualitative correlation between the total
number of direct protein-ligand contacts and the relative
binding enthalpies and Gibbs energies for each ligand pair.
Conversely, the more flexible ligand of a pair makes more single
water-mediated contacts with the domain. Inasmuch as the
entropies for forming complexes of the flexible ligands are more
favorable, this finding seems incompatible with the prevailing
view that there is an entropic cost associated with fixing water
molecules at a protein-ligand interface. There are no significant
differences in the van der Waals contacts for a given flexible/
constrained ligand pair. An analysis of crystallographic b-factors
suggests that thermal motions in complexes of the constrained
ligands are generally greater than those in the corresponding
complexes of their flexible controls. If these motions in the solid
state actually reflect more disorder, and hence more favorable
configurational entropies, in the complexes of constrained
ligands, one might anticipate that the binding entropies for their
formation would be greater than those for their flexible
counterparts; however, this prediction is inconsistent with our
results.
These systematic and extensive studies clearly demonstrate
our lack of understanding of energetics in protein-ligand
interactions, even in biological systems that are well-character-
ized by thermodynamic and structural studies. They also
underscore the difficulty of correlating the number or type of
protein-ligand contacts or the number of water molecules at
the protein-ligand interface with specific contributions to
binding enthalpies and entropies. One shortcoming of many
studies of protein-ligand interactions is that they focus upon
the direct interactions between the protein and the ligand,
perhaps including regions of the protein proximal to the ligand
binding site; the effects of ligand binding upon protein dynamics
and structure distal to the binding site are rarely considered.
Indeed, there is a profound lack of experimental data that
correlates the energetic effects associated with the net confor-
mational and dynamic changes that occur in the protein and
the ligand upon complexation.
We believe that future models of protein-ligand interactions
must include explicit consideration of the enthalpic and entropic
contributions arising from changes in nonbonded interactions
and order that occur upon complex formation and how these
changes vary as a function of ligand structure. Ligands may
bind to proteins with high affinity in conformations that are
energetically higher than their global minima in solution, so
the relationship between the conformational strain energy of a
ligand in its bound conformation and its affinity must be better
understood. Moreover, the detailed interactions of the protein,
ligand, and the protein-ligand complex with water molecules
as well as the effects of solvent reorganization must be
incorporated in the analysis. Finally, to optimize protein-ligand
interactions, means must be found to modify the structure of a
ligand in a way that minimizes enthalpy-entropy compensation.
We are beginning to address these and other difficult questions
in ongoing studies of energetics and structure in protein-ligand
interactions that will include the use of NMR and computational
methods. These results will be reported in due course.
Acknowledgment. We thank the National Institutes of Health
(GM 84965), the National Science Foundation (CHE 0750329),
The Robert A. Welch Foundation (F-652), the Norman Hackerman
Advanced Research Program, The Texas Institute for Drug and
Diagnostic Development through a Welch Foundation Grant
(#H-F-0032) for their generous support of this research. J.E.D.
thanks the ACS Division of Medicinal Chemistry and Pfizer Inc.
as well as the Dorothy B. Banks Foundation for predoctoral
fellowships. We are also thankful to Dr. Stuart Black (Schering-
Plough Research Institute) for the Grb2 SH2 expression vector and
helpful discussions and to Dr. James Davidson for helpful discus-
sions. We are also grateful to Brahm Carlson, David Cramer, Greg
Frayne, Monica Lotz, Nicole Martin, James Myslinski, Naveed
Nosrati, Nicole Pulliam, Adithya Rangarajan, and Benjamin Whid-
don for their help in numerous aspects of these investigations.
Supporting Information Available: Complete ref 14c, ex-
perimental procedures and characterization data for all new
compounds, methods and materials for ITC and X-ray crystal-
lographic experiments, plots of ITC data, and contact diagrams
for complexes of 6, 8, 12, and 14. This material is available
free of charge via the Internet at http://pubs.acs.org.
JA904698Q
9
16770 J. AM. CHEM. SOC. VOL. 131, NO. 46, 2009