Dissection of complex molecular recognition interfaces

Article abstract of DOI:10.1021/ja1084783

The synthesis of a family of zinc porphyrins and pyridine ligands equipped with peripheral H-bonding functionality has provided access to a wide range of closely related supramolecular complexes featuring between zero and four intramolecular H-bonds. An automated UV/vis titration system was used to characterize 120 different complexes, and these data were used to construct a large of number of different chemical double mutant cycles to quantify the intramolecular H-bonding interactions. The results probe the quantitative structure-activity relationship that governs cooperativity in the assembly of complex molecular recognition interfaces. Specifically, variations in the chemical structures of the complexes have allowed us to change the supramolecular architecture, conformational flexibility, geometric complementarity, the number and nature of the H-bond interactions, and the overall stability of the complex. The free energy contributions from individual H-bonds are additive, and there is remarkably little variation with architecture in the effective molarity for the formation of intramolecular interactions. Intramolecular H-bonds are not observed in complexes where they are geometrically impossible, but there are no cases where excellent geometric complementarity leads to very high affinities. Similarly, changes in conformational flexibility seem to have limited impact on the values of effective molarity (EM). The major variation that was found for all of the 48 intramolecular interactions that were examined using double mutant cycles is that the values of EM for intramolecular carboxylate ester-phenol H-bonds (200 mM) are an order of magnitude larger than those found for phosphonate diester-phenol H-bonds (30 mM). The corresponding intermolecular phosphonate diester-phenol H-bonds are 2 orders of magnitude more stable than carboxylate ester-phenol H-bonds, and the large differences in EM may be due to some kind of compensation effect, where the stronger H-bond is harder to make, because it imposes tighter constraints on the geometry of the complex.

Full text of DOI:10.1021/ja1084783

Published on Web 12/21/2010
Dissection of Complex Molecular Recognition Interfaces
Christopher A. Hunter,* Maria Cristina Misuraca, and Simon M. Turega
Department of Chemistry, UniVersity of Sheffield, Sheffield S3 7HF, United Kingdom
Received September 20, 2010; E-mail: c.hunter@sheffield.ac.uk
Abstract: The synthesis of a family of zinc porphyrins and pyridine ligands equipped with peripheral
H-bonding functionality has provided access to a wide range of closely related supramolecular complexes
featuring between zero and four intramolecular H-bonds. An automated UV/vis titration system was used
to characterize 120 different complexes, and these data were used to construct a large of number of different
chemical double mutant cycles to quantify the intramolecular H-bonding interactions. The results probe
the quantitative structure-activity relationship that governs cooperativity in the assembly of complex
molecular recognition interfaces. Specifically, variations in the chemical structures of the complexes have
allowed us to change the supramolecular architecture, conformational flexibility, geometric complementarity,
the number and nature of the H-bond interactions, and the overall stability of the complex. The free energy
contributions from individual H-bonds are additive, and there is remarkably little variation with architecture
in the effective molarity for the formation of intramolecular interactions. Intramolecular H-bonds are not
observed in complexes where they are geometrically impossible, but there are no cases where excellent
geometric complementarity leads to very high affinities. Similarly, changes in conformational flexibility seem
to have limited impact on the values of effective molarity (EM). The major variation that was found for all
of the 48 intramolecular interactions that were examined using double mutant cycles is that the values of
EM for intramolecular carboxylate ester-phenol H-bonds (200 mM) are an order of magnitude larger than
those found for phosphonate diester-phenol H-bonds (30 mM). The corresponding intermolecular
phosphonate diester-phenol H-bonds are 2 orders of magnitude more stable than carboxylate ester-phenol
H-bonds, and the large differences in EM may be due to some kind of compensation effect, where the
stronger H-bond is harder to make, because it imposes tighter constraints on the geometry of the complex.
vation, and cooperativity between multiple interactions.4 For
synthetic supramolecular systems that use strong binding site
Introduction
Self-assembly and self-organization in biological systems are
characterized by complex recognition interfaces between mac-
romolecular surfaces with multiple cooperative intermolecular
interactions.1 Many different factors are recognized to contribute
to the thermodynamic properties of these interfaces: electrostatic
interactions between charges, H-bonds, aromatic interactions,
desolvation, etc. This complexity makes it difficult to disentangle
the relative contributions of individual interaction sites and the
cooperativity involved in assembly of the entire interface.2
The inherent conformational flexibility of biomolecules and the
changes in conformational dynamics that are associated with
complex formation add to the problem.3 Synthetic supramo-
lecular systems provide a controlled environment, where some
of these variables can be systematically studied. This has
provided insight into the magnitude of specific functional group
contacts, the requirements for filling space, the role of desol-
interactions and relatively rigid molecular frameworks, coop-
erativity is usually observed as additive free energy contributions
from the individual interaction sites.5 However, the behavior
of systems that feature multiple weak interactions between
flexible molecules, such as biomolecules, is more complicated.
(2) (a) Williams, D. H.; Bardsley, B. Angew. Chem., Int. Ed. 1999, 38,
1173. (b) Williams, D. H.; Stephens, E.; O’Brien, D. P.; Zhou, M.
Angew. Chem., Int. Ed. 2004, 43, 6596. (c) Baldini, L.; Ballester, P.;
Casnati, A.; Gomila, R. M.; Hunter, C. A.; Sansone, F.; Ungaro, R.
J. Am. Chem. Soc. 2003, 125, 14181. (d) Hughes, A. D.; Anslyn, E. V.
Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 6538. (e) Tobey, S. L.; Anslyn,
E. V. J. Am. Chem. Soc. 2003, 125, 10963. (f) Camara-Campos, A.;
Hunter, C. A.; Tomas, S. Proc. Natl. Acad. Sci. U.S.A. 2006, 103,
3034. (g) Lemonnier, J. F.; Guenee, L.; Bernardinelli, G.; Vigier, J. F.;
Bocquet, B.; Piguet, C. Inorg. Chem. 2010, 49, 1252–1265. (h)
Calderone, C. T.; Williams, D. H. J. Am. Chem. Soc. 2001, 123, 6262.
(i) Shiozawa, H.; Chia, B. C. S.; Davies, N. L.; Zerella, R.; Williams,
D. H. J. Am. Chem. Soc. 2002, 124, 3914. (j) Williams, D. H.; Davies,
N. L.; Zerella, R.; Bardsley, B. J. Am. Chem. Soc. 2004, 126, 2042.
(3) (a) Wang, T. J.; Hoy, J. A.; Lamm, M. H.; Nilsen-Hamilton, M. J. Am.
Chem. Soc. 2009, 131, 14747. (b) Menand, M.; Jabin, I. Chem. 2010,
16, 2159. (c) Koshland, D. E. Proc. Natl. Acad. Sci. U.S.A. 1958, 44,
98. (d) Anderson, H. L.; Hunter, C. A.; Meah, M. N.; Sanders, J. K. M.
J. Am. Chem. Soc. 1990, 112, 5780.
(1) (a) Mulder, A.; Huskens, J.; Reinhoudt, D. N. Org. Biomol. Chem.
2004, 2, 3409. (b) Philp, D.; Stoddart, J. F. Angew. Chem., Int. Ed.
Engl. 1996, 35, 1155. (c) Lehn, J. M. Angew. Chem., Int. Ed. Engl.
1990, 29, 1304. (d) Badjic, J. D.; Nelson, A.; Cantrill, S. J.; Turnbull,
W. B.; Stoddart, J. F. Acc. Chem. Res. 2005, 38, 723. (e) Mammen,
M.; Choi, S. K.; Whitesides, G. M. Angew. Chem., Int. Ed. 1998, 37,
2755. (f) Kay, E. R.; Leigh, D. A.; Zerbetto, F. Angew. Chem., Int.
Ed. 2007, 46, 72. (g) Kolomiets, E.; Lehn, J. M. Chem. Commun.
2005, 1519. (h) Hoffmann, M.; Karnbratt, J.; Chang, M. H.; Herz,
L. M.; Albinsson, B.; Anderson, H. L. Angew. Chem., Int. Ed. 2008,
47, 4993.
(4) (a) Cockroft, S. L.; Hunter, C. A. Chem. Commun. 2006, 3806. (b)
Cockroft, S. L.; Hunter, C. A. Chem Commun. 2009, 3961. (c) Fischer,
F. R.; Schweizer, W. B.; Diederich, F. Chem. Commun. 2008, 4031.
(d) Hof, F.; Scofield, D. M.; Schweizer, W. B.; Diederich, F. Angew.
Chem., Int. Ed. 2004, 43, 5056. (e) Scarso, A.; Trembleau, L.; Rebek,
J. J. Am. Chem. Soc. 2004, 126, 13512.
9
582 J. AM. CHEM. SOC. 2011, 133, 582–594
10.1021/ja1084783  2011 American Chemical Society

Dissection of Complex Molecular Recognition Interfaces
A R T I C L E S
Practical experience in the fields of both medicinal and
supramolecular chemistry shows that preorganization of the
binding partners improves affinity. The dogma is that confor-
mational restriction of the molecules reduces the entropic cost
of freezing rotors, so that rigid molecules are preferred.
However, in practice, it is very difficult to experimentally
separate this kind of entropic contribution from the observed
free energy of binding, because the entropy change is a property
of the system as a whole. Thus, the enthalpy and entropy of
binding often show large fluctuations that are not expressed in
a change in affinity. Changes in the structure of the solvent or
a macromolecule can lead to very large changes in enthalpy
and entropy that more or less cancel out, and these effects
swamp the small differences that could be caused by confor-
mational restriction. Moreover, the introduction of conforma-
tional restriction into a compound changes more than the number
of torsional degrees of freedom: first, a conformational lock can
fix the compound in a geometry that would otherwise be
inaccessible without a significant energy penalty, that is,
enthalpic strain; second, a conformational lock will prevent the
compound from interacting with alternative sites in an undesired
binding mode; and, third, the conformational lock will usually
involve addition of functional groups that may make direct
contacts with the binding site. Thus, there are many mechanisms
by which the restriction of conformational mobility may improve
binding affinity without the need to invoke an entropic contribu-
tion due to restriction of rotors. Here, we describe a supramo-
lecular model system that we can use as a controlled environ-
ment for dissecting thermodynamic contributions to a cooperative
recognition interface that involves multiple weak interactions
between flexible molecules.
Figure 1. A chemical double mutant cycle designed to measure the free
energy contribution of the intramolecular H-bond in complex A to the overall
stability of the complex (∆∆G° in eq 1).
Figure 2. Stepwise equilibria in the formation of complex A in Figure 1.
K0 is the intermolecular association constant for formation of the zinc-nitrogen
interaction. K1EM is the intramolecular equilibrium constant for formation
of the H-bond.
We have shown previously that a strong zinc porphyrin-
pyridine coordination bond can be used to facilitate cooperative
formation of a weaker intramolecular H-bond. By using refer-
ence compounds (mutants) where one of the interaction sites
has been deleted, it is possible to construct a chemical double
mutant cycle (DMC) to quantify the contribution that the
intramolecular H-bond makes to the overall stability of complex
A in Figure 1 using eq 1.7
Approach
There are two parameters that quantify cooperativity in a
system that makes multiple noncovalent interactions: the allo-
steric interaction parameter, which measures the change in
intrinsic functional group interaction energy; and the effective
molarity, EM, which measures the extent to which intramo-
lecularity enhances the probability of making a functional group
interaction (Figure 1).6 The factors that contribute to the intrinsic
functional group interaction energy are reasonably well-
understood: steric hindrance, secondary electrostatic interactions,
functional group polarization, etc. The relationship between
supramolecular architecture and EM is less obvious. One might
expect that EM will decrease with conformational strain in the
bound state and restriction of conformational flexibility, but the
magnitude of these effects is difficult to predict. We have
therefore designed a set of receptors and ligands with varying
degrees of complementarity and conformational flexibility in
an attempt to map this relationship.
∆∆Go ) ∆GoA - ∆GoB - ∆GoC + ∆GoD
(1)
The DMC removes all allosteric effects (secondary interac-
tions, changes in zinc-nitrogen interaction, etc.), so that the
thermodynamic contribution of the H-bond between two func-
tional groups can be dissected from all of the other factors that
affect the overall stability of complex A. If we consider the
assembly of complex A as a stepwise process, where an
intermolecular coordination bond is formed first, followed by
an intramolecular H-bond, then the cooperativity in the assembly
of this complex can be quantified by the effective molarity for
the intramolecular process (EM in Figure 2). Thus, the free
energy contribution of the intramolecular H-bond to the overall
stability of complex A is given by eq 2.
(5) (a) Drobizhev, M.; Stepanenko, Y.; Rebane, A.; Wilson, C. J.; Screen,
T. E. O.; Anderson, H. L. J. Am. Chem. Soc. 2006, 128, 12432. (b)
Camara-Campos, A.; Hunter, C. A.; Tomas, S. Proc. Natl. Acad. Sci.
U.S.A. 2006, 103, 3034. (c) Hunter, C. A.; Ihekwaba, N.; Misuraca,
M. C.; Segarra-Maset, M. D.; Turega, S. M. Chem. Commun. 2009,
3964.
∆∆Go ) -RT ln(K1EM)
(2)
where K1 is the corresponding intermolecular association
constant for formation of the same H-bond between two
appropriate reference compounds, where no other interactions
are possible.
(6) (a) Fersht, A. R.; Kirby, A. J. J. Am. Chem. Soc. 1967, 89, 4857. (b)
Benedetti, F.; Stirling, C. J. M. Chem. Commun. 1983, 1374. (c)
Sinclair, A. J.; del Amo, V.; Philp, D. Org. Biomol. Chem. 2009, 7,
3308. (d) Chi, X. L.; Guerin, A. J.; Haycock, R. A.; Hunter, C. A.;
Sarson, L. D. Chem. Commun. 1995, 2563. (e) Hunter, C. A.; Tomas,
S. Chem. Biol. 2003, 10, 1023. (f) Cacciapaglia, R.; Di Stefano, S.;
Mandolini, L. Acc. Chem. Res. 2004, 37, 113. (g) Kirby, A. AdV. Phys.
Org. Chem. 1980, 17, 183. (h) Hunter, C. A.; Anderson, H. L. Angew.
Chem., Int. Ed. 2009, 48, 7488.
(7) (a) Cockroft, S. L.; Hunter, C. A. Chem. Soc. ReV. 2007, 36, 172. (b)
Camara-Campos, A.; Musumeci, D.; Hunter, C. A.; Turega, S. J. Am.
Chem. Soc. 2009, 131, 18518.
9
J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011 583

A R T I C L E S
Hunter et al.
Figure 3. Porphyrin receptors, P1a-P4a (R ) OH) and P1b-P4b (R )
OMe).
If the association constant for the intermolecular H-bond,
K1, can be measured using reference compounds, then the
DMC can be used to determine the value of EM for the
intramolecular interaction in complex A using eq 2. The
system illustrated in Figure 1 therefore provides an ideal
platform for exploring the relationship between the structures
of the groups linking the two interaction sites on both the
porphyrin receptor and the ligand and the magnitude of the
EM.8 The use of a phenol as the H-bond donor (R ) 3.8)
and a phosphonate diester as the H-bond acceptor (ꢀ ) 8.9)
ensures that the H-bond interaction is sufficiently strong to
allow measurement of K1 in nonpolar solvents.9 Figures 3
and 4 illustrate a range of readily accessible metalloporphy-
rins and ligands that can be combined in a combinatorial
manner to construct a large number of DMCs for different
supramolecular architectures and thereby begin to characterize
the relationship between chemical structure and EM.
Figure 4. Ligand structures, L1a-L6a (R ) CH2PO(OEt)2), L1b-L6b
(R ) Et), and L4c-L6c.
Scheme 1. Synthesis of Zinc Porphyrin Receptors P1a and P1b
Results and Discussion
Porphyrin Receptor Synthesis. Free base porphyrins 1 and 2
were prepared according to literature procedures.10 Metalation
with zinc acetate provided receptors P1a and P1b (Scheme 1).
Two different routes were investigated for the synthesis of the
other receptor architectures. In route I (Scheme 2), Suzuki
couplings between tetrabromoporphyrin 3 and 2-, 3-, or 4-hy-
droxyphenylboronic acid only proceeded in good yield for the
3-isomer, 4. The 2- and 4-isomers, 5 and 6, can be obtained by
route I, but in very low yield. The alternative route (II) shown
in Scheme 3 proved to be superior. Suzuki coupling of
3-bromobenzaldehyde with 2-, 3-, or 4-methoxyphenylboronic
(8) Chekmeneva, E.; Hunter, C. A.; Packer, M. J.; Turega, S. M. J. Am.
Chem. Soc. 2008, 130, 17718.
(9) (a) Abraham, M. H.; Platts, J. A. J. Org. Chem. 2001, 66, 3484. (b)
Hunter, C. A. Angew. Chem., Int. Ed. 2004, 43, 5310.
9
584 J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011

Dissection of Complex Molecular Recognition Interfaces
A R T I C L E S
Scheme 2. Route I for Synthesis of the Zinc Porphyrin Receptors
Scheme 3. Route II for Synthesis of the Zinc Porphyrin Receptors
acid gave the corresponding biphenyl aldehydes 7, 8,10 and 9.
The aldehydes were converted to the corresponding free-base
porphyrins 10, 11, and 12 using the Lindsey method.11 The
tetramethoxyporphyrins 10 and 11 were deprotected with BBr3
to give the corresponding tetrahydroxyporphyrins (Scheme 3).
Metalation of the free base porphyrins with zinc acetate gave
the corresponding porphyrin receptors P2a, P3a, P4a (Scheme
2), P2b, P3b, and P4b (Scheme 3).
Ligand Synthesis. Some of the ligands, ethyl nicotinate, ethyl
isonicotinate, 3-picoline, 4-picoline, and 3,5-lutidine (corre-
sponding to L2b, L1b, L4c, L5c, and L6c, respectively), were
commercially available. Ligands L2a, L3a, and L3b were
synthesized according to previously published procedures.8 The
remaining ligands were synthesized from the corresponding
carboxylic acids. The dicarboxylic acid, 15, is the only acid
that is not commercially available, and this was prepared using
the route shown in Scheme 4. Sonogashira coupling of propargyl
alcohol with 3,5-dibromopyridine gave diol 13, which was
reduced using a Pd/C CatCart in a H-cube apparatus to provide
14.12 Diol 14 was oxidized to the corresponding diacid with
Jones reagent, but this compound proved easier to isolate as
the diethyl ester after refluxing with acidic ethanol. The diester,
which corresponds to one of the control ligands, L6b, was then
hydrolyzed to give the diacid 15. Carboxylic acids 16 and 17
were coupled with diethyl hydroxymethane phosphonate using
N,N′-dicyclohexylcarbodiimide (DCC) and 4-dimethylaminopy-
ridine (DMAP) to yield ligands L1a and L5a. To convert
carboxylic acids 18 and 15 into ligands L4a and L6a, diethyl
hydroxymethane phosphonate was first tosylated to give 19,13
(12) Mor, M.; Rivara, S.; Lodola, A.; Plazzi, P. V.; Tarzia, G.; Duranti,
A.; Tontini, A.; Piersanti, G.; Kathuria, S.; Piomelli, D. J. Med. Chem.
2004, 47, 4998.
(10) (a) Crossley, M. J.; Field, L. D.; Forster, A. J.; Harding, M. M.;
Sternhell, S. J. Am. Chem. Soc. 1987, 109, 341. (b) Wiehe, A.; Shaker,
Y. M.; Brandt, J. C.; Mebs, S.; Senge, M. O. Tetrahedron 2005, 61,
5535.
(13) (a) Desai, B.; Kappe, C. O. J. Comb. Chem. 2005, 7, 641. (b) Saaby,
S.; Knudsen, K. R.; Ladlow, M.; Ley, S. V. Chem. Commun. 2005,
2909. (c) Thales Nanotechnology Inc. website: http://www.thalesnano.
com.
(11) Lindsey, J. S.; Schreiman, I. C.; Hsu, H. C.; Kearney, P. C.;
Marguerettaz, A. M. J. Org. Chem. 1987, 52, 827.
9
J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011 585

A R T I C L E S
Hunter et al.
Scheme 4. Synthesis of Dicarboxylic Acid 15
of the spectroscopic data. A UV/vis plate reader equipped
with internal syringes was programmed to titrate ligand into
receptor with acquisition of the UV/vis spectrum after each
addition. A potential disadvantage of this system is that the
plate is open to the atmosphere, but the use of a nonpolar
high boiling point solvent, toluene, minimizes potential
problems due to solvent evaporation and absorption of
atmospheric water. The association constants obtained using
the plate reader agree very well with those obtained from
conventional manual titrations: representative titration data
for automated and manual experiments on the P1a · L1b
complex are shown in Figure 5b and c, respectively. There
is a well-defined isosbestic point (Figure 5a), and the titration
data fit well to a 1:1 isotherm (Figure 5b and c). The
association constants for automated and manual experiments
are identical within experimental error: the average value over
three different experiments is 2600 ( 500 M-1 for the
automated experiment and 2200 ( 100 M-1 for the manual
titration. The errors tend to be somewhat higher for automated
experiments, but this is balanced against the speed advantage,
which allows collection of a large number of data points for
each titration and multiple repetitions of each titration. The
lower limit on the porphyrin concentration that is required
to obtain sufficient signal-to-noise to ensure good titration
data is about 3 µM. This places an upper limit of about 107
M-1 on the association constant that can be reliably measured
by the automated experiment. For complexes that are close
to this limit, the association constants that are reported here
were obtained by manual titration experiments, due to the
greater sensitivity of the conventional UV/vis absorption
spectrometer.
Scheme 5. Synthesis of Ligands
The association constants for all 120 porphyrin-ligand
combinations are collected in Table 1. In most cases, the
data fit well to a simple 1:1 binding isotherm, but there are
four complexes that gave very different isotherms: P1a · L1a,
P3a · L1a, P4a · L1a, and P4a · L2a. Figure 5e shows the
titration data for the P1a · L1a complex. The isotherm is
clearly biphasic, indicating the presence of at least three
different species in equilibrium: free porphyrin, the 1:1
complex, and a higher stoichiometry complex that is formed
at high ligand concentrations. The titration data were fit to a
number of different isotherms, and although a number of
different models fit the data reasonably well, the best fit was
obtained using a model that includes formation of a 2:5
complex in addition to the 1:1 complex (Figure 5e, see the
Supporting Information for details). 1H NMR titrations failed
to provide further insight into the stoichiometry of the higher
order complex formed at high ligand concentrations due to
broadening of the signals. However, the reason that it is only
these four systems that form higher order complexes will
become clear based on the DMC analysis in the next section
of this Article. In short, all four of the complexes that exhibit
this anomalous behavior are systems where intramolecular
phosphonate-phenol H-bonds are difficult to make. Thus,
the only way in which the H-bonding potential of these
groups can be satisfied is through intermolecular interactions
between two porphyrin complexes. Dimerization of the
porphyrin complexes establishes a template for binding
multiple bridging ligands via pyridine-phenol and
phosphonate-phenol H-bonds (Figure 6). For ligands that
have only one phosphonate group, the 2:5 complex is the
lowest stoichiometry complex that satisfies the principle of
maximum site occupancy, where all of the H-bond and
which underwent nucleophilic substitution under basic condi-
tions to give the esters in high yield. Carboxylic acids 17 and
18 were esterified in acidic EtOH to give control ligands L5b
and L4b (Scheme 5).
High Throughput Titration Analysis of Binding. Interaction
of the ligands with the porphyrin receptors was analyzed
using UV/vis titration experiments. Complexation of the
pyridine nitrogen of a ligand to the zinc center of a porphyrin
causes a change of approximately 10 nm in the UV/vis
absorption maximum of the Soret band (Figure 5a), and this
change was used to determine association constants. How-
ever, the large number of porphyrin-ligand combinations
encouraged us to develop an automated approach to collection
9
586 J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011

Dissection of Complex Molecular Recognition Interfaces
A R T I C L E S
Figure 5. UV/vis titration data in toluene at 298 K. (a) Data for manual titration of P1a with L1b and (b) the corresponding fit of the absorbance at 420,
425, 430, and 440 nm to a 1:1 binding isotherm. (c) Data for automated titration of P1a with L1b showing the fit of the absorbance at 420, 425, 430, and
440 nm to a 1:1 binding isotherm. (d) Data for manual titration of P1a with L1a and (e) the corresponding fit of the absorbance at 420, 425, 430, and 440
nm to a 1:1 plus 2:5 binding isotherm.
Table 1. Association Constants (M-1) for Formation of 1:1 Complexes in Toluene at 298 K (Percentage Errors in Brackets)a
porphyrin
ligand
P1a
P2a
P3a
P4a
P1b
P2b
P3b
P4b
L1a 4.0 × 103 (30%)b 1.9 × 104 (10%) 1.5 × 104 (10%)b 8.2 × 103 (7%)b 2.9 × 103 (7%) 2.0 × 103 (3%) 2.4 × 103 (8%) 2.2 × 103 (1%)
L2a 2.4 × 105 (25%) 2.5 × 104 (8%) 9.5 × 104 (30%) 1.6 × 104 (20%)b 9.2 × 103 (1%) 7.4 × 103 (3%) 4.6 × 103 (40%) 3.3 × 103 (6%)
L3a 6.5 × 106 (15%) 7.4 × 104 (1%) 9.1 × 105 (3%)
L4a 3.8 × 105 (20%) 2.0 × 105 (1%) 1.4 × 106 (2%)
L5a 2.7 × 105 (20%) 2.1 × 105 (20%) 4.8 × 105 (1%)
6.4 × 104 (9%)
5.9 × 105 (1%)
7.6 × 103 (1%) 2.3 × 103 (9%) 9.1 × 103 (20%) 5.4 × 103 (4%)
1.4 × 104 (1%) 1.2 × 104 (2%) 1.9 × 104 (10%) 1.5 × 104 (1%)
1.2 × 105 (20%) 5.1 × 103 (4%) 4.1 × 103 (2%) 7.1 × 103 (1%) 5.5 × 103 (4%)
L6a 8.9 × 106 (3%)c 8.3 × 106 (20%) 9.3 × 106 (20%)c 2.2 × 106 (10%) 6.8 × 103 (10%) 4.4 × 103 (7%) 8.2 × 103 (6%) 7.9 × 103 (3%)
L1b 2.6 × 103 (8%)
L2b 3.6 × 103 (8%)
L3b 3.2 × 103 (1%)
3.9 × 103 (5%) 6.3 × 103 (2%)
5.1 × 103 (2%) 8.5 × 103 (1%)
4.3 × 103 (5%) 6.9 × 103 (9%)
3.5 × 103 (4%)
5.1 × 103 (6%)
3.0 × 103 (10%) 2.8 × 103 (7%) 3.8 × 103 (8%) 3.4 × 103 (6%)
3.9 × 103 (20%) 3.3 × 103 (6%) 6.2 × 103 (30%) 5.4 × 103 (4%)
1.3 × 104 (8%) 1.1 × 104 (30%) 1.9 × 104 (20%) 1.6 × 104 (30%)
7.0 × 103 (30%) 5.0 × 103 (10%) 8.6 × 103 (2%) 8.0 × 103 (10%)
5.1 × 103 (40%) 3.8 × 103 (20%) 2.6 × 103 (10%) 6.0 × 103 (20%) 4.5 × 103 (20%)
L4b 1.6 × 104 (30%) 1.9 × 104 (20%) 3.8 × 104 (20%) 2.5 × 104 (9%)
L5b 2.1 × 104 (20%) 2.7 × 104 (4%) 5.1 × 104 (2%)
1.2 × 104 (4%)
L6b 2.7 × 104 (20%) 4.3 × 104 (8%) 1.2 × 105 (9%)
7.3 × 103 (30%) 4.4 × 103 (2%) 2.7 × 103 (3%) 5.7 × 103 (1%) 4.9 × 103 (20%)
1.1 × 104 (20%) 9.1 × 103 (1%) 1.6 × 104 (4%) 1.4 × 104 (7%)
L4c 1.2 × 104 (8%)
1.5 × 104 (5%) 2.6 × 104 (20%) 2.1 × 104 (5%)
L5c 6.7 × 103 (20%) 8.0 × 103 (4%) 1.7 × 104 (2%)
1.1 × 104 (10%) 7.1 × 103 (6%) 5.4 × 103 (20%) 9.1 × 103 (8%) 6.7 × 103 (6%)
1.5 × 104 (10%) 9.5 × 103 (1%) 7.0 × 103 (1%) 1.3 × 104 (2%) 8.5 × 103 (20%)
L6c 7.5 × 103 (9%)
1.0 × 104 (3%) 1.7 × 104 (2%)
a Errors are quoted as twice the standard deviation of the values from multiple repeats of the same titration (3-12 repeats). The errors estimated in this way range
from 1% to 40%, with an average value of 10% over the entire data set. We therefore assume that (10% is the real error in the experiment, and this value is used as
the lower limit on the error in K in all subsequent analysis. b Fitted to a 1:1 plus 2:5 isotherm. c Measured by manual titration.
9
J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011 587

A R T I C L E S
Hunter et al.
Figure 8. DMC for evaluating the thermodynamic contribution of
phosphonate diester-phenol H-bonds to the stability of a porphyrin-ligand
complex.
Table 2. Free Energy Contributions from Phosphonate
Diester-Phenol H-Bonds (∆∆G°/kJ mol-1) at 298 K in Toluene,
Determined Using the Chemical Double Mutant Cycle in Figure 8a
Figure 6. Exploded view of possible intermolecular interactions in the
2:5 complexes formed by ligands that cannot make intramolecular H-bonds
in the 1:1 complex (P1a·L1a, P3a·L1a, P4a·L1a, and P4a·L2a).
a Complexes that make weak H-bonds are shaded (cf., Figure 9).
Errors are (1 kJ mol-1
.
systems are the values for the 1:1 complexes that are formed
at low ligand concentrations.
Thermodynamics of Phosphonate Diester-Phenol H-Bond-
ing. The results in Table 1 are illustrated graphically in Figure
7. The complexes are grouped into families according to the
four components of the DMC. The blue region shows the results
for complexes formed between the porphyrins that contain
phenol H-bond donors (P1a-P4a) and the ligands that contain
phosphonate H-bond acceptors (L1a-L6a). In general, these
complexes are more stable than the other complexes that do
not have the potential to form phosphonate-phenol H-bonds.
However, within this family, there is considerable variation in
stability due to differences in the number and the geometry of
the H-bond interactions. Some of the blue complexes have a
stability similar to that of the control complexes, indicating that
no intramolecular phosphonate-phenol H-bonds are formed,
whereas in the most extreme case an increase of 3 orders of
magnitude in stability is observed. The green region corresponds
to the single mutants where there are no H-bond donors on the
porphyrin. The yellow region corresponds to the single mutants
where the phosphonate groups have been removed from the
ligand. The red region corresponds to the double mutants. The
association constants are similar in the green and red regions,
but there are some control complexes in the yellow region that
show enhanced stability. Comparison between the complexes
formed with ligands L5b and L6b and the complexes formed
with ligands L5c and L6c suggests that it is interactions with
Figure 7. Association constants (log K/M-1) for formation of 1:1 complexes
(phosphonate ligand-hydroxyporphyrin complexes in blue, phosphonate
ligand-methoxyporphyrincomplexesingreen,controlligand-hydroxyporphyrin
complexes in yellow, and control ligand-methoxyporphyrin complexes in
red), and the DMC used to quantify the phosphonate-phenol (A ·D) H-bond
interaction in porphyrin-ligand (P ·L) complexes.
coordination sites on all of the molecules are fully bound.
The association constants quoted in Table 1 for these four
9
588 J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011

Dissection of Complex Molecular Recognition Interfaces
A R T I C L E S
Figure 9. Illustration of possible geometries of different porphyrin-ligand combinations. The shaded areas correspond to systems where the contribution
of a phosphonate diester-phenol H-bond to the overall stability of the complex is low (∆∆G > -6 kJ mol-1): there are obvious geometric limitations on
the formation of H-bonds in most of these systems. The four dotted boxes indicate systems that make 2:5 complexes. Ligands L3a and L6a have the same
geometry as ligands L2a and L4a, respectively, but with a second side chain indicated by the dotted lines.
the ester carbonyl H-bond acceptors that are responsible for the
enhanced stability of these yellow complexes. Thus, there is
more than one kind of H-bond present, and the contributions
of these multiple interactions have to be disentangled. The
double mutant cycle approach is specifically designed to factor
out the thermodynamic contributions of individual interactions.
Figure 8 shows how the effects of ester-phenol H-bonding
interactions can be removed to quantify the phosphonate-phenol
H-bonds in complexes that make multiple H-bonds.
In complex A in Figure 8, there is the possibility of making
both phosphonate diester-phenol and ester-phenol H-bonds.
The overall stability of the complex reflects the contributions
from metal-ligand binding, both types of H-bond, and any other
secondary interactions that might be present. However, com-
parison of the stability of complex D with complex B provides
a measure of the thermodynamic contribution of the ester-phenol
H-bond. Comparison of complex D with complex C provides a
measure of the contributions due to changes in the porphyrin-
ligand coordination interaction. Thus, the DMC allows us to
dissect the free energy contribution due to the phosphonate
diester-phenol H-bond from all of the other interactions present
in this system. This analysis assumes that all free energy
contributions are additive. In other words, the presence of the
phosphonate diester-phenol H-bond in complex A does not
affect the stability of the ester-phenol H-bond as compared to
the corresponding ester-phenol H-bond in complex B. This is
an assumption that is difficult to verify experimentally, but there
is experimental evidence that cooperative H-bond networks of
this type behave in an additive manner with respect to free
energy.5c Deviations from additivity are likely to be second-
order effects as compared to the primary interactions, and if
there are major secondary effects, they will show up as
anomalies in the DMC analysis.
Table 2 summarizes the DMC results for all 24 complexes
that can make phosphonate diester-phenol H-bonds. Ligands
L3a and L6a both have two phosphonate diester groups and
therefore have the potential to make two H-bonds, whereas the
other ligands, L1a, L2a, L4a, and L5a, can only make one.
Table 2 shows that the free energy contribution is approximately
double for the ligands that have two H-bonding groups, L3a
and L6a, when compared to the corresponding ligands that have
one H-bonding group, L2a and L5a, respectively. This is rather
good evidence for the additivity of the H-bond free energy
contributions in these systems. There is considerable variation
in the contributions due to H-bonding from one complex to
another, and these differences can be rationalized to some extent
on the basis of the geometries of the porphyrin-ligand
combinations (Figure 9). For complexes where the phenol and
phosphonate diester groups can achieve close proximity, there
is a free energy contribution of -6 to -9 kJ mol-1 due to the
formation of each H-bond. For the complexes where the
H-bonding partners cannot get close to each other without major
distortion, either in the metal-ligand geometry or in the
configuration of the ester groups, the free energy contributions
due to H-bonding are significantly weaker. It appears that for
an optimal geometry the phosphonate diester-phenol H-bond
9
J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011 589

A R T I C L E S
Hunter et al.
contributes -9 kJ mol-1, and this is eroded by strain in the
other complexes. There do not appear to be any significant
contributions due to differences in conformational flexibility.
Apart from the complexes that are highly strained, there is
surprisingly little difference in the free energy contribution due
to H-bonding from one architecture to another.
Thermodynamics of Ester-Phenol H-Bonding. As high-
lighted in Figure 7, there is evidence of increased stability in
some of the control complexes due to the formation of
ester-phenol H-bonds. By using a set of additional control
ligands, L4c-L6c, it is possible to construct a different double
mutant cycle to measure the thermodynamic contribution of
these interactions to the overall stabilities of the complexes
(Figure 10). In ligands L1b-L3b, the ester group is directly
conjugated to the pyridine nitrogen, and so removal of the ester
also has a significant effect on the metal-ligand interaction.
However, this difference is accounted for in the double mutant
cycle, because complexes A and C have the ester group and
complexes B and D do not, so the change in the metal-ligand
interaction cancels out.
Figure 10. DMC for evaluating the thermodynamic contribution of
ester-phenol H-bonds to the stability of a porphyrin-ligand complex.
Table 3 summarizes the results for all 24 complexes that can
make ester-phenol H-bonds. It is clear that only ligands L5b
and L6b make H-bonds. Moreover, in all cases, the free energy
contributions due to H-bonding for ligand L6b, which has two
ester groups, are approximately double that for ligand L5b,
which has only one ester group. This again indicates that free
energy increments due to H-bond interactions are additive in
this system. The results can again be rationalized on the basis
of the geometries of the complexes, as illustrated in Figure 11.
For the systems that do not make H-bonds, the ester and phenol
groups are too remote to interact without considerable strain.
The exception is the P2a·L2b/L3b complex, which appears to
be well-suited to the formation of H-bonds. However, the flat
ChemDraw representation hides the fact that the dihedral angle
around the biphenyl bond is closer to 90° than 0° in this system,
due to the ortho-hydroxyl group. This appears to be sufficient
to hold the two groups apart in this relatively rigid complex.
Each ester-phenol H-bond contributes a constant increment
of -3 kJ mol-1 to the overall stability of the complex in these
systems. This behavior is quite different from that observed for
the phosphonate diester H-bonds. With the esters, it appears
that H-bonding is an all or nothing process that depends on
geometric complementarity. For the phosphonate diesters, the
H-bonding interactions are significantly stronger, and so it is
possible to trade off some of the binding energy in exchange
for conformational strain to make more demanding contacts.
Structural Evidence for H-Bonding. The analysis above relies
on the use of DMCs to determine thermodynamic contributions
associated with particular functional group combinations. In
essence, a more stable complex implies the presence of H-bond
interactions. The implications of this analysis are that some of
the complexes make as many as four H-bonds. For example,
P3a·L6a is the most stable complex with an association constant
of nearly 107 M-1, and the DMC analysis indicates that both of
the ester groups and both of the phosphonate diester groups
make H-bonds with the porphyrin phenols. The schematic
diagrams in Figures 9 and 11 show that this is geometrically
feasible, and it is possible to construct full three-dimensional
molecular models that confirm this (Figure 12).
Table 3. Free Energy Contributions from Ester-Phenol H-Bonds
(∆∆G°/kJ mol-1) at 298 K in Toluene, Determined Using the
Chemical Double Mutant Cycle in Figure 10a
a Complexes that do not make H-bonds are shaded (cf., Figure 11).
Errors are (1 kJ mol-1
.
spectra were broad in deuterated toluene, but for some com-
plexes, 1:1 mixtures of porphyrin and ligand gave well-resolved
spectra in deuterated tetrachloroethane. The association constants
are sufficiently high that the complexes are almost fully bound
at a concentration of 1 mM. The phenol OH signals were
generally shifted by +2-3 ppm relative to the corresponding
signals in the spectra of the free porphyrin, indicative of the
presence of H-bonding interactions. However, signal broadening
and the sensivity of this signal to residual water precluded
quantitative analysis.
The signals due to the porphyrin pyrrole protons proved
more informative. Rotation around the porphyrin-meso-
1
phenyl bond is slow on the H NMR time scale at room
temperature, leading to atropoisomers for all four receptors.
For the free porphyrins, the presence of several atropisomers
had no impact on the spectrum, and the pyrrole protons were
observed as one singlet around 9 ppm (Figure 13a). Com-
plexation of a simple ligand that does not form H-bonds led
to a change of -0.1 ppm in the pyrrole signal, but did not
affect the multiplicity (Figure 13b). Ligands that form one
H-bond with one of the porphyrin meso-phenol substituents
gave very similar spectra (Figure 13c). However, ligands that
form more than one H-bond according to the DMC analysis
above showed multiplicity in the pyrrole signal (Figure
13d-g). Ligands that make more than one H-bond have
different affinities for different atropisomers, so they can be
resolved in slow exchange in the NMR spectrum. Thus,
To obtain direct experimental evidence for the presence of
1
H-bonding interactions, we used H NMR spectroscopy. The
(14) Barral, K.; Priet, S.; Sire, J.; Neyts, J.; Balzarini, J.; Canard, B.;
1
Alvarez, K. J. Med. Chem. 2006, 49, 7799.
complexes that form two H-bonds give the characteristic H
9
590 J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011

Dissection of Complex Molecular Recognition Interfaces
A R T I C L E S
Figure 11. Illustration of possible geometries of different porphyrin-ligand combinations. The shaded areas correspond to systems that do not make
ester-phenol H-bonds: there are obvious geometric limitations on the formation of H-bonds in most cases. Ligands L3b and L6b have the same geometry
as ligands L2b and L4b, respectively, with a second side chain indicated by the dotted lines.
-RT ln Kcalc ) ∆Go ) -(R-RS)(ꢀ-ꢀS) + 6 kJ mol-1
NMR spectra shown in Figure 13d and e. Complexes that
form four H-bonds according to the DMC analysis also give
(3)
rise to characteristic but different spectra (Figure 13f and
g): the multiplicity in the pyrrole region is replaced by a
single broad signal. In these cases, only one of the atropi-
somers can make four simultaneous H-bonds to the ligand,
and so the atropisomer population is shifted to the symmetric
R,R,R,R-isomer, resulting in a simpler spectrum. Although
somewhat indirect, these observations corroborate the ther-
modynamic analysis from the DMCs above.
Comparison of inter- and intramolecular H-bonding is compli-
cated by the fact that the fully bound state may not be fully
populated in systems that make multiple weak interactions. Figure
15 illustrates three different bound states that can be populated for
ligands that make one coordination interaction and one H-bond.
Because K0 . K1, state 1, where the porphyrin-ligand bond is
broken, can be ignored, because it will not be significantly
populated as compared to state 3, where the H-bond is broken.
The population of partially bound state 3 depends on the product
K1EM. If K1EM . 1, then the partially bound state will not be
populated to any significant extent, and the overall stability of the
complex will simply reflect the sum of the free energy contributions
of the individual intermolecular contacts. However, if K1EM ≈ 1,
then the fully and partially bound states (2 and 3 in Figure 15)
will both be populated, and the thermodynamic properties of the
complex will be a population weighted average of the properties
of the different species (eq 4).
Partially Bound States. The DMC analysis allows us to
separate the thermodynamic contributions from individual
H-bond interactions to the overall stability of the complexes.
These contributions are made up of two different components,
the intrinsic strength of the H-bond interaction and the
effective molarity (EM) for formation of the intramolecular
contact. These two factors can be separated by comparing
the intramolecular interactions, which we have measured
above, with the corresponding intermolecular interactions
between model compounds: phenol, 20, phosphonate diester,
21, and ester, 22 (Figure 14). The association constants for
1
Kobs ) K0(1 + K1EM)
(4)
the 20 · 21 and 20 · 22 complexes were determined using H
NMR titrations in toluene, and the results are listed in Table
4. The value for the phenol-ester complex is at the limit of
reliable experimental measurement, but the values of as-
sociation constants for simple complexes of this type can
also be estimated using eq 3.7b Table 4 shows that the
experimental results are in good agreement with the values
expected on the basis of H-bond parameters that have been
determined independently for these functional groups.
For the porphyrin-ligand complexes that show phenol-ester
H-bonding interactions, the increase in the stability of the
complex is -3 kJ mol-1 per H-bond. A change in free energy
of -3 kJ mol-1 corresponds to a value of K1EM of 2 in eq 4,
which means that the partially bound state is 30% populated,
and we must therefore analyze the populations of states in more
detail. As the number of different intermolecular interactions
9
J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011 591

A R T I C L E S
Hunter et al.
Figure 14. Model compounds used to quantify intermolecular H-bond
interactions.
Table 4. Association Constants (M-1) for the Formation of
1
H-Bonded Complexes in Toluene-d8 at 298 K Measured by H
a
NMR Titrations (Kexpt) and Estimated Using Eq 3 (Kcalc
)
complex
R
ꢀ
RS
ꢀS
Kexpt
Kcalc
20·21
20·22
3.8
3.8
8.9
5.3
1.0
1.0
2.2
2.2
140 ( 10
3 ( 1
180
3
a H-bond parameters, R and ꢀ, from ref 9.
The population of the fully bound state, Pb, is given by eq
6.
Figure 12. Three-dimensional structure of the P3a·L6a complex energy
minimized using AM1.17 There are four H-bonds (OH-O distances <2.2
Å) involving both of the ester groups and both of the phosphonate diester
groups of the ligand.
N
K0σij...N
K EM
i i
∏
i
Pb )
(6)
Kobs
The populations of the partially bound states, where only
some of the intramolecular H-bond are made, can be
determined in a similar way.
The approach is illustrated below for complexes that can make
a single intramolecular H-bond, and full details for the more
complex systems can be found in the Supporting Information.
For a single intramolecular H-bond, the observed association
constant is given by the sum of the equilibrium constants for
the fully bound and one partially bound state (eq 7).
Kobs ) K0(1 + σ1K1EM1)
(7)
where K0 is the intermolecular association constant for formation
of the zinc-nitrogen interaction, K1 is the intermolecular association
constant for formation of the H-bonding interaction, EM1 is the
effective molarity for the intramolecular interaction, and σ1 is the
statistical factor that describes the degeneracy of the fully bound
state. In the complexes discussed here, the porphyrins all have four
identical H-bond donor sites, so σ1 ) 4.
1
Figure 13. Pyrrole region of the H NMR spectra of (a) [P3a] ) 1 mM
with no ligand, (b) [P3a] ) [L4b] ) 1 mM (no H-bonds), (c) [P3a] )
[L5b] ) 1 mM (one H-bond), (d) [P3a] ) [L6b] ) 1 mM (two H-bonds),
(e) [P4a] ) [L6a] ) 1 mM (two H-bonds), (f) [P1a] ) [L6a] ) 1 mM
(four H-bonds), and (g) [P3a] ) [L6a] ) 1 mM (four H-bonds) in deuterated
tetrachloroethane.
increases in the more complex systems, the number of partially
bound states that are possible increases, and this affects any
attempt to deconvolute the thermodynamic contributions of
individual interaction sites.
In general, for complexes that make N intramolecular
H-bonds, the observed association constant is given by the sum
of the equilibrium constants for all of the bound states (eq 5).
The population of the fully bound state, Pb, is given by
σ1K1EM1
Pb )
(8)
1 + σ1K1EM1
and the population of the partially bound state, Pf, where the
intramolecular H-bond is not made is given by
N
1
Kobs ) K (1 +
σ K EM +
σ K EM K EM + ...
Pf )
(9)
∑
∑
0
i
i
i
ij
i
i
j
j
1 + σ1K1EM1
i
i,j
N
Note that the key parameter that defines the behavior of
the system is the product σ1K1EM1. When σ1K1EM1 , 1, the
H-bonded state is not populated. In the limit of σ1K1EM1 .
1, the partially bound state is not populated, and eq 7 reduces
to Kobs ) σ1K0K1EM1, which is the basis for the double mutant
cycle analysis above. In general, the value of ∆∆G0 obtained
from the DMC provides a measure of the ratio Kobs/K0
corrected for any secondary interactions and substituent
effects and so can be used to determine effective molarities
(eq 10).
+ σij...N
K EM ) (5)
i i
∏
i
where K0 is the intermolecular association constant for
formation of the zinc-nitrogen interaction, Ki is the inter-
molecular association constant for formation of the relevant
H-bonding interaction, EMi is the effective molarity for
formation of the intramolecular interaction, and σi is the
statistical factor that describes the degeneracy of the partially
bound state.15
9
592 J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011

Dissection of Complex Molecular Recognition Interfaces
A R T I C L E S
Figure 15. Different bound states that are possible in a 1:1 complex between a porphyrin and a ligand that can make one H-bond and one coordination
interaction. K0 is the intermolecular association constant for porphyrin-ligand coordination. K1 is the intermolecular association constant for the H-bonded
complex. EM is the effective molarity for the intramolecular interaction.
Table 5. Effective Molarities (EM/mM) for Intramolecular
Phosphonate Diester-Phenol H-Bonds Measured at 298 K in
Toluenea
Kobs
K0
e-∆∆G°/RT
)
) 1 + σ1K1EM1
(10)
Effective Molarities. The DMC free energies in Tables 2 and
3 were used in conjunction with the intermolecular association
constants in Table 4 and the partially bound states analysis in
the Supporting Information to determine effective molarities for
the intramolecular H-bonds observed in these systems. The
results are summarized in Tables 5 and 6. The variations in the
values of EM with both ligand and receptor structure are
surprisingly small. In most cases, the intramolecular phosphonate
diester-phenol H-bonds have an EM of about 30 mM. In cases
where the geometry is particularly unfavorable, this values drops
by an order of magnitude, and in two cases, H-bonding cannot
be detected (shaded boxes in Table 5). Neither differences in
ligand side chain flexibility nor the formation of additional
H-bonds with the ester linkers, which might be expected to
organize the ligand side-chains, seem to have a major effect on
EM in these systems.
a Complexes that make weak or no H-bonds are shaded (cf., Figure
9). Errors are (40%. b No interaction detected.
Table 6. Effective Molarities (EM/mM) for Intramolecular
Ester-Phenol H-Bonds Measured at 298 K in Toluenea
The EM is an order of magnitude higher for the intramolecular
ester-phenol H-bonds, 200 mM. The ester and phosphonate
diester H-bond groups are located at different positions on the
ligand framework, which means that the EM values cannot be
directly compared. However, we have explored a large number
of different geometries, so the results suggest that there is an
intrinsic difference in the ability of these two functional groups
to form intramolecular interactions. One possibility is that the
increased steric crowding around the phosphonate group inter-
feres with the formation of strong intramolecular interactions.
In general, the values of σKEM for intramolecular phospho-
nate diester H-bonds are around 10, which means that these
interactions are 90% bound, and partially bound states do not
play a significant role. The value of σKEM drops in the more
strained complexes highlighted by the shaded boxes in Table
5, so partially bound states are populated. These are the cases
in which the formation of the higher order 2:5 complexes
complicates analysis of the titration data. For intramolecular
ester-phenol H-bonds, the value of σKEM is around 2, so these
interactions are never more than 60% populated.
a Complexes that do not make H-bonds are shaded (cf., Figure 11).
Errors are (50%. b No interaction detected.
effective molarities for the intramolecular processes. Exploitation
of an automated UV/vis titration system has enabled straight-
forward characterization of 120 different porphyrin-ligand
combinations, and these data have been used to construct a large
of number of different DMCs to examine a variety of closely
related intramolecular H-bonds in different supramolecular
architectures. The number of intramolecular H-bonds formed
varies from zero to four depending on geometric complemen-
tarity and the number of functional groups on the ligand, but,
in all cases, the free energy contribution from each functional
group contact could be dissected using appropriate DMCs.
Two types of H-bond were detected: between phosphonate
diester H-bond acceptors on the ligand and phenol H-bond
donors on the porphyrin, and between carboxylate ester accep-
tors on the ligand and phenol donors on the porphyrin. As
expected from the properties of simple intermolecular H-bonds
measured for model compounds, the intramolecular phosphonate
diester H-bonds contribute significantly more to the stability of
the complexes (-6 to -9 kJ mol-1) than do the carboxylate
ester H-bonds (-3 kJ mol-1). However, the two classes of
interaction show qualitatively very different behavior. The weak
carboxylate ester H-bonds show all-or-nothing behavior: when
there is good geometric complementarity between the ligand
and receptor, a H-bond is observed, and when there is not, there
Conclusions
The metalloporphyrin-ligand architecture provides an ideal
platform for conducting a systematic survey of the relationship
between chemical structure and cooperativity at relatively
complex molecular recognition interfaces. In this Article, we
have developed an approach to dissecting cooperative networks
of noncovalent interactions based on chemical double mutant
cycles. Systematic deletion of functional groups allows quan-
tification of the free energy contributions of individual functional
group contacts, and these have been characterized in terms of
9
J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011 593

A R T I C L E S
Hunter et al.
known concentration (3-7 µM) in spectroscopic grade solvent. A
10 mL solution of ligand (40-5000 µM) was prepared using
spectroscopic grade solvent (i.e., without any host present). 150
µL of porphyrin solution was added to each well of a 96 well
Hellma quartz plate, and the UV/vis absorbance was recorded at
five wavelengths using a BMG FLUOstar Omega plate reader. The
plate reader was thermostatted at 298 K for all measurements.
Aliquots of ligand solution (3, 6, or 10 µL) were added successively
to each well containing the porphyrin solution, and the UV/vis
absorbance was recorded after each addition. Changes in the
absorbance of the Soret band of the porphyrin were fit to a 1:1
binding isotherm in Microsoft Excel to obtain the association
constant. Each titration was repeated at least three times, and the
experimental error is quoted as twice the standard deviation at a
precision of one significant figure.
is no H-bond. The stronger phosphonate diester H-bonds make
intramolecular contacts even in cases where there is a significant
geometrical mismatch between the ligand and the receptor. In
these cases, the associated strain leads to a decrease in the free
energy contribution to the overall stability of the complex as
compared to the -6 to -9 kJ mol-1 observed for the more
complementary architectures. Thus, H-bonds to phosphonate
diesters were observed for 22 of the 24 systems examined,
whereas H-bonds to carboxylate esters were observed in only
6 of the 24 geometries studied.
The magnitude of the cooperativity involved in the formation
of the intramolecular interactions was quantified using the
effective molarity (EM). There is remarkably little variation in
the value of EM with supramolecular architecture, conforma-
tional flexibility, geometric complementarity, and number of
interactions. In the phosphonate diester series, the EM is around
30 mM for 16 of the 24 systems studied, but it drops to 5 mM
when there is significant geometric strain. In the carboxylate
ester series, the EM is around 200 mM in all 6 systems that
make intramolecular H-bonds. The large difference between the
two sets of EM values is striking, given that both series cover
a range of different architectures. There are two possible
explanations: the phosphonate diester groups are very bulky and
steric constraints may prevent them from reaching optimal
geometries in the confined spaces on the faces of the receptors;
or there may be some kind of compensation effect, where the
stronger H-bond constrains the interacting groups more tightly
leading to a lower EM. Extending this survey to a wider range
of functional groups would provide further insight into the origin
of this effect.
Manual UV/Visible Absorption Titrations. UV/vis titrations
were carried out by preparing a 10 mL sample of porphyrin at
known concentration (0.5-7 µM) in spectroscopic grade solvent.
Two milliliters of this solution was removed, and a UV/vis spectrum
was recorded using a Peltier thermostat set at 298 K. A 2 mL
solution of ligand (40-3000 µM) was prepared using the porphyrin
stock solution, so that the concentration of porphyrin remained
constant throughout the titration. Aliquots of ligand solution were
added successively to the cell containing the porphyrin solution,
and the UV/vis spectrum was recorded after each addition. Changes
in the absorbance of the Soret band of the porphyrin were fit to a
1:1 binding isotherm in Microsoft Excel to obtain the association
constant. Each titration was repeated at least three times, and the
experimental error is quoted as twice the standard deviation at a
precision of one significant figure.
1H NMR Titrations. NMR titrations were carried out by
preparing a 2 mL sample of host at known concentration (7-270
The analysis in this Article is based on the assumption that
free energy contributions from different functional group
interactions are additive, and the self-consistency of the results
obtained supports this assumption. For example, the free energy
contribution due to H-bonding in ligands that have two
symmetrical side arms is almost exactly double that for the
corresponding ligands, which have only one side arm in all
cases. This suggests that simple additive schemes, such as the
Free-Wilson analysis,16 have some validity even in complex
systems where multiple cooperative interactions are present.
1
mM). Next, 0.6 mL of this solution was removed, and a H NMR
spectrum was recorded. A 1 mL solution of guest (70-2600 mM)
was prepared using the host solution, so that the concentration of
host remained constant throughout the titration. Aliquots of guest
solution were added successively to the NMR tube containing the
host, and the NMR spectrum was recorded after each addition.
Changes in chemical shifts were analyzed by using the appropriate
binding isotherms in Microsoft Excel. Each titration was repeated
at least three times, and the experimental error is quoted as twice
the standard deviation at a precision of one significant figure.
Experimental Section
Acknowledgment. We thank the EPSRC for funding and Dr.
M. Fisher for assistance with protocols for the BMG Labtech plate
reader.
Automated UV/Visible Absorption Titrations. UV/vis titrations
were carried out by preparing a 10 mL sample of porphyrin at
(15) (a) Bailey, W. F.; Monahan, A. S. J. Chem. Educ. 1978, 55, 489–
493. (b) Ercolani, G.; Piguet, C.; Borkovec, M.; Hamacek, J. J. Phys.
Chem. B 2007, 111, 12195–12203.
Supporting Information Available: Synthetic procedures and
spectroscopic data, protocol for operation of the UV/vis plate
reader, details of the fitting of the titration data to different
binding isotherms, analysis of partially bound states for systems
that make more than one H-bond, and tables of K EM values
and occupancies for intramolecular H-bonds and coordinates
of the structure shown in Figure 12. This material is available
free of charge via the Internet at http://pubs.acs.org.
(16) (a) Free, S. M.; Wilson, J. W. J. Med. Chem. 1964, 7, 395. (b)
Andrews, P. R.; Craik, D. J.; Martin, J. L. J. Med. Chem. 1984, 27,
1648–1657. (c) Tomic, S.; Nilsson, L.; Wade, R. C. J. Med. Chem.
2000, 43, 1780–1792. (d) Gohlke, H.; Klebe, G. Angew. Chem., Int.
Ed. 2002, 41, 2645–2676.
(17) (a) Conformations were analyzed using an MMFFs MCMM search
using Macomodel, and the lowest energy conformation was minimized
at a semi-empirical AM1 level using Spartan 06. (b) Macromodel
(Version 7.1); Schro¨dinger, Inc.: Portland, OR, 2000, running on a
Linux cluster. (c) Spartan 06; Wavefunction, Inc.: Irvine, CA, running
on a Windows PC.
JA1084783
9
594 J. AM. CHEM. SOC. VOL. 133, NO. 3, 2011