- Email: [email protected]
Toward the computer-aided design of metal ion sequestering agents
Journal of Alloys and Compounds 374 (2004) 416–419
Toward the computer-aided design of metal ion sequestering agents Benjamin P. Hay a,∗ , Timothy K. Firman a , Gregg J. Lumetta a , Brian M. Rapko a , Priscilla A. Garza a , Sergei I. Sinkov a , James E. Hutchison b , Bevin W. Parks b , Robert D. Gilbertson b , Timothy J.R. Weakley b a
Pacific Northwest National Laboratory, Mail Stop, P.O. Box 999, Richland, WA 99352, USA b Department of Chemistry, University of Oregon, Eugene, OR 97403, USA
Abstract The concepts embodied in de novo structure-based drug design are being adapted for the computer-aided design of metal ion sequestering agents. This adaptation requires the development of methods for (i) generating candidate structures and (ii) evaluating and prioritizing these structures with respect to their binding affinity for a specific guest. This article summarizes recent progress in this area that includes the creation of a new computer software program, called HostDesigner, that can generate and evaluate millions of new molecular structures per minute on a desktop personal computer. Several methods for evaluating the degree of binding site organization in a host structure are presented. An example is provided to demonstrate how these methods have been used to identify ligand architectures that provide enhanced metal ion binding affinity. © 2003 Elsevier B.V. All rights reserved. Keywords: Computer-aided design; Molecular modeling; Metal ion; Chelate; Ligand
1. Introduction Organic ligands with a high degree of metal ion recognition provide the foundation for the development of sensors, separating agents, improved analytical techniques, homogeneous catalysis, imaging agents, encapsulated radionuclides for use in cancer treatments, therapeutic agents for the treatment of metal intoxication, and models to study enzyme function. As a result, there has been large effort expended to design effective and selective receptors for targeted metal ion species. Simply put, ligand design is the process of choosing a set of binding sites and then choosing the connecting geometric structure that ties them together. Although there are good criteria for selecting the number and type of binding sites [1–3], the choice of structure remains a challenge. The deliberate design of ligand architectures by assembling sets of disconnected binding sites in three dimensions is not a trivial task. Until recently, we could only generate trial structures by hand with a graphical user interface, an extremely time-consuming process. ∗ Corresponding author. Tel.: +1-509-372-6239; fax: +1-509-372-6328. E-mail address: [email protected] (B.P. Hay).
0925-8388/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.jallcom.2003.11.049
Often, it is not readily obvious which linkage structures might be best used to connect the binding sites to obtain a host cavity that is organized for coordinaton to a targeted metal ion guest. To address the problem of how to identify new host molecules that recognize and bind strongly to specific metal ion guests, we have adopted a computational approach pioneered by the pharmaceutical industry. Drug designers have developed methods to address the inverse of this problem, in other words, how to identify molecular structures (guests) that complement the binding site of a protein (host) [4–6]. These approaches include de novo structure-based design strategies that couple molecule building algorithms with scoring functions used to prioritize the candidate structures. The building algorithms assemble guest molecule structures that can physically interact with a known protein structure from pieces which are either atoms or larger, chemically reasonable fragments. The ability to generate large numbers of potential guest structures necessitates the use of simple scoring functions to prioritize the output. To this end, methods have been developed to estimate the binding free energy by summing free energy increments for hydrogen bond interactions, ionic interactions, lipophilic interactions, the number of rotatable bonds in the guest molecule, etc. [7,8].
B.P. Hay et al. / Journal of Alloys and Compounds 374 (2004) 416–419
The computer programs that have been developed to perform de novo structure-based drug design are, in general, not applicable for the design of host molecules. These programs require input of the atomic coordinates of a protein binding site, are highly specialized to address protein–organic interactions, and do not contain scoring functions to address the interactions that occur in other types of host–guest systems. To apply the powerful concepts embodied in de novo structure-based drug design to the design of metal ion sequestering agents, it is necessary to develop methods for (i) the generation of candidate structures and (ii) evaluating and prioritizing these structures with respect to their binding affinity for a specific guest. This article summarizes recent progress that we have made in these areas.
2. Generating host structures We have developed a de novo structure-based design software, HostDesigner, that is specifically created for the discovery of host molecules for metal ion guests [9]. HostDesigner generates and evaluates millions of candidate structures in minutes on a desktop personal computer, and rapidly identifies three-dimensional architectures that position binding sites to provide an optimal interaction with a metal ion. The molecule building algorithms combine user-input host fragments with linking fragments taken from a database. The user-input host fragments define the optimal geometry for one or more binding sites interacting with a guest. The current linking fragment database used by HostDesigner contains over 10 000 structures composed of Csp3 , Csp2 , and hydrogen atoms. The fragments consist of all connectivities and conformers that can be made from 0 to 6 carbons, excluding three- and four-membered rings. The fragments also include all dimethylated five- and sixmembered rings, and selected bicyclic structures. When using these fragments to build host molecules, all possible connectivities, stereochemistries, and conformations are constructed, which generates large numbers of structures. These structures are prioritized based on how well the host binding sites converge at the metal ion guest. Cartesian coordinates for the top candidates are output to a file for subsequent viewing. Recent improvements to the HostDesigner software, which is available at no cost from the website http://www. hostdesigner.emsl.pnl.gov, allow (i) the treatment of multiatom guests, (ii) structural variation within the input fragments, and (iii) use of empirical methods to estimate the relative conformational energy for each output structure [10].
3. Evaluating host structures Deliberate ligand design requires a knowledge of how structure impacts reactivity and the ability to distinguish a
417
good architecture from a poor architecture. Extensive research effort has been expended toward understanding how host structure influences metal ion binding with the goal of discovering more effective and more selective metal ion receptors [1–3,11,12]. It is clear from these studies that certain properties are needed in order to achieve significant increases in binding affinity or ion recognition. These host properties include (i) the presence of multiple binding sites [1–3], (ii) the ability to adopt a conformation in which all binding sites are positioned to structurally complement the metal ion [13], and (iii) a limited degree of conformational freedom [14]. It has long been recognized that increased binding affinities are obtained when a collection of donor atoms is structurally constrained to the binding conformation, in other words, when the host is preorganized [15]. There are various approaches to evaluate the degree of complementarity and preorganization offered by a host structure. The HostDesigner software, which can examine millions of structures in a single run, requires fast methods for accomplishing these evaluations. This is achieved by using geometric factors to prioritize structures on the basis of how well their binding sites correspond to those of the guest [9]. In addition, conformational energy increments, based on potential surfaces for simple hydrocarbon analogs, are used to estimate the conformational stability of the host structures, allowing them to be further prioritized on the basis of their degree of preorganization [10]. Although approximate in nature, these methods provide a rapid means of selecting a list of the best candidates from a large group of potential structures. This short list can then be re-prioritized using more accurate evaluation methods. With increasing application to coordination compounds [13,16,17], molecular mechanics (MM) models provide an ideal, computationally efficient tool to evaluate the degree to which a ligand is structurally organized for metal complexation [13,18–20]. Conformational analyses yield the stable conformers of the ligand both in the uncomplexed state and in the metal complex. MM models partition the steric energy into stretching, bending, torsion, and non-bonded (Van der Waals, electrostatics, hydrogen bonding) interactions. The process of parameterizing these models requires knowledge of the geometries and potential energy surfaces for each individual interaction, which are precisely the criteria needed to evaluate metal ion complementarity. Examination of the steric energy components from calculations on metal ion complexes provides a way to quantify the effect of (i) cavity size mismatch (strain in M–L bonds), (ii) poor donor group orientation (strain in M–L–X and M–L–X–X angles), and, for metals that exhibit preferences for distinct polyhedra, (iii) topographical mismatch (strain in L–M–L angles) where M: metal, L: donor atom, and X: any other atom [13]. Work we have done on the design of improved diamide sequestering agents demonstrates how these methods have been used to identify ligand architectures that provide enhanced metal ion binding affinity before they are prepared and tested. Conventional malonamide ligands, in which two
418
B.P. Hay et al. / Journal of Alloys and Compounds 374 (2004) 416–419
Fig. 2. MM3 optimized geometries for the only stable conformer of the transoid form, 2, and the lowest energy conformer of the cisoid form, 3, with atom types and oxygen vectors as defined in Fig. 1.
Fig. 1. Structural reorganization of a simple malonamide ligand, 1, during the chelation of a lanthanide metal ion. Vectors attached to each oxygen atom show the optimal direction of approach for a lanthanide metal ion.
amides are connected by a methylene link, are poorly organized for complexation. Fig. 1 compares the metal free conformations of N,N,N ,N -tetramethylmalonamide [21], 1, with a lanthanide–coordinated form of this ligand [22]. In the most stable form, the two carbonyl groups point in opposite directions such that it is not possible for the two oxygen atoms to simultaneously contact the same metal ion. Rotation about one of the two C–C bonds yields a second form, 3.0 kcal/mol higher in energy, which is predisposed for chelation. However, this metal binding conformer fails to provide a complementary array of binding sites. The geometries of simple amides coordinated with lanthanide metal ions show the presence of a distinct oxygen donor directionality in which the metal ion lies in the plane of the amide moiety with a C=O–M angle of 143 ± 1◦ [23]. This directionality is illustrated in Fig. 1 by attaching a vector to each oxygen atom. In a complementary chelate architecture, the two vectors would intersect at a point where the metal ion is located. In the binding form of malonamide, however, these vectors diverge. The use of such vectors provides a convenient way to visualize the degree of complementarity that is offered by a host structure. Attaching vectors of the appropriate length to each binding site also provides a geometric basis for quantitatively evaluating how well the host fits the guest. For example, when molecular fragments consisting of single binding sites are connected by HostDesigner, the software uses the distance between
the ends of binding site vectors to prioritize the candidates [9]. Metal ion coordination causes further structural change to the binding form of malonamide that involves rotation about both C–C bonds. MM calculations on an isolated Eu3+ -1 chelate ring show that the structural change results in an additional 2.5 kcal/mol increase in ligand strain, giving a total strain energy of 5.5 kcal/mol on going from the low energy form to the bound form [24]. The foregoing analysis suggested that metal binding affinity afforded by two amide binding sites could be increased significantly if the binding sites were conformationally constrained in a complementary geometry. After examination of a variety of possible bicyclic scaffolds, we discovered that the desired structural attributes are attained in the stereoisomers 2 and 3 illustrated in Fig. 2. In contrast to 1, both of these architectures have convergent binding site vectors and they both chelate Eu3+ with the development of less than 0.1 kcal/mol of ligand strain [24]. Synthesis and testing have shown that the preorganized architecture afforded by 3 exhibits the anticipated increase in metal ion binding affinity. We have described the preparation and extraction characteristics for a hydrophobic derivative of 3, in which the N-methyl groups have been replaced with n-octyl substituents [24,25]. In liquid-liquid extraction experiments the designed sequestering agent exhibits a dramatic 106 –107 enhancement in the distribution coefficients for trivalent lanthanides and actinides when compared with conventional tetraalkylmalonamide extractants. For all ligands examined in these extraction experiments, the stoichiometry of the extracted species is approximately three ligands to one metal ion revealing that the bicyclic diamide gives an average enhancement of at least 102 per ligand over the acyclic diamides. Recent measurements of aqueous formation constants suggest that this enhancement in metal ion extraction can be attributed directly to the increased metal
B.P. Hay et al. / Journal of Alloys and Compounds 374 (2004) 416–419
ion binding affinity of the bicyclic architecture. Spectrophotometric titrations of Am3+ in 1.0 M nitric acid at 23 ◦ C yield log K values for the formation of 1:1 complexes, −0.17 for 1 and 1.81 for 3, that confirm the bicyclic architecture binds this metal two orders of magnitude more strongly than the acyclic architecture, even under acidic aqueous conditions. Acknowledgements HostDesigner was developed at Pacific Northwest National Laboratory (PNNL) with support from the Division of Chemical Sciences, Office of Science, US Department of Energy (DOE). The diamide research, performed at PNNL and at the University of Oregon (UO), was sponsored by the DOE Environmental Management Science Program (EMSP–54679, EMSP–73759), the National Science Foundation (DUE-0088986 and CHE-023563) and the University of Oregon. PNNL is managed and operated under DOE contract DE–AC06–76RLO–1830 by Battelle Memorial Institute. References [1] R.D. Hancock, A.E. Martell, Chem. Rev. 89 (1989) 1875. [2] A.E. Martell, R.D. Hancock, in: J.P. Fackler (Ed.), Metal Complexes in Aqueous Solution, Plenum Press, New York, 1996. [3] H.-J. Schneider, A. Yatsimirsky, in: Principles and Methods in Supramolecular Chemistry, Wiley, New York, 2000.
419
[4] I.D. Kuntz, E.C. Meng, B.K. Shoichet, Acct. Chem. Res. 27 (1994) 117. [5] T.P. Lybrand, Curr. Opin. Struct. Biol. 5 (1995) 224. [6] H.-J. Böhm, Prog. Biophys. Mol. Biol. 66 (1996) 197. [7] A. Murcko, M.A. Murcko, J. Med. Chem. 38 (1995) 4953. [8] M.D. Eldridge, C.W. Murray, T.R. Auton, G.V. Paolini, R.P. Mee, J. Comput. Aided Mol. Des. 11 (1997) 425. [9] B.P. Hay, T.K. Firman, Inorg. Chem. 41 (2002) 5502. [10] B.P. Hay, T.K. Firman, in: HostDesigner User’s Manual, Pacific Northwest National Laboratory, Richland, Washington, 2003, http://www.hostdesigner.emsl.pnl.gov. [11] A.E. Martell, R.D. Smith, in: Critical Stability Constants, vol. 1–6, Plenum Press, New York, pp. 1974–1988. [12] R.M. Izatt, K. Pawlak, J.S. Bradshaw, R.L. Bruening, Chem. Rev. 95 (1995) 2529. [13] B.P. Hay, R.D. Hancock, Coord. Chem. Rev. 212 (2001) 61. [14] D.H. Busch, K. Farmery, V. Goedken, V. Katovic, A.C. Melnyk, C.R. Sperati, N. Tokel, Adv. Chem. Ser. 100 (1971) 44. [15] D.J. Cram, T. Kaneda, R.C. Helgeson, S.B. Brown, C.B. Knobler, E. Maverick, K.N. Trueblood, J. Am. Chem. Soc. 107 (1985) 3645. [16] B.P. Hay, Coord. Chem. Rev. 126 (1993) 177. [17] P. Comba, T.W. Hambley, in: Molecular Modeling of Inorganic Compounds, VHC, New York, 1995. [18] B.P. Hay, J.R. Rustad, J. Am. Chem. Soc. 116 (1994) 6316. [19] B.P. Hay, D. Zhang, J.R. Rustad, Inorg. Chem. 35 (1996) 2650. [20] B.P. Hay, D.A. Dixon, R. Vargas, J. Garza, K.N. Raymond, Inorg. Chem. 40 (2001) 3922. [21] G. Sandrone, D.A. Dixon, B.P. Hay, J. Phys. Chem. 103 (1999) 3554. [22] E.E. Castellano, R.W. Becker, Acta Crystallogr. Sect. B 37 (1981) 61. [23] O. Clement, B.M. Rapko, B.P. Hay, Coord. Chem. Rev. 170 (1998) 203. [24] G.J. Lumetta, B.M. Rapko, P.A. Garza, B.P. Hay, R.D. Gilbertson, T.J.R. Weakley, J.E. Hutchison, J. Am. Chem. Soc. 124 (2002) 5644. [25] G.J. Lumetta, B.M. Rapko, B.P. Hay, P.A. Garza, J.E. Hutchison, R.D. Gilbertson, Solv. Extr. Ion Exch. 21 (2003) 29.
Toward the computer-aided design of metal ion sequestering agents Benjamin P. Hay a,∗ , Timothy K. Firman a , Gregg J. Lumetta a , Brian M. Rapko a , Priscilla A. Garza a , Sergei I. Sinkov a , James E. Hutchison b , Bevin W. Parks b , Robert D. Gilbertson b , Timothy J.R. Weakley b a
Pacific Northwest National Laboratory, Mail Stop, P.O. Box 999, Richland, WA 99352, USA b Department of Chemistry, University of Oregon, Eugene, OR 97403, USA
Abstract The concepts embodied in de novo structure-based drug design are being adapted for the computer-aided design of metal ion sequestering agents. This adaptation requires the development of methods for (i) generating candidate structures and (ii) evaluating and prioritizing these structures with respect to their binding affinity for a specific guest. This article summarizes recent progress in this area that includes the creation of a new computer software program, called HostDesigner, that can generate and evaluate millions of new molecular structures per minute on a desktop personal computer. Several methods for evaluating the degree of binding site organization in a host structure are presented. An example is provided to demonstrate how these methods have been used to identify ligand architectures that provide enhanced metal ion binding affinity. © 2003 Elsevier B.V. All rights reserved. Keywords: Computer-aided design; Molecular modeling; Metal ion; Chelate; Ligand
1. Introduction Organic ligands with a high degree of metal ion recognition provide the foundation for the development of sensors, separating agents, improved analytical techniques, homogeneous catalysis, imaging agents, encapsulated radionuclides for use in cancer treatments, therapeutic agents for the treatment of metal intoxication, and models to study enzyme function. As a result, there has been large effort expended to design effective and selective receptors for targeted metal ion species. Simply put, ligand design is the process of choosing a set of binding sites and then choosing the connecting geometric structure that ties them together. Although there are good criteria for selecting the number and type of binding sites [1–3], the choice of structure remains a challenge. The deliberate design of ligand architectures by assembling sets of disconnected binding sites in three dimensions is not a trivial task. Until recently, we could only generate trial structures by hand with a graphical user interface, an extremely time-consuming process. ∗ Corresponding author. Tel.: +1-509-372-6239; fax: +1-509-372-6328. E-mail address: [email protected] (B.P. Hay).
0925-8388/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.jallcom.2003.11.049
Often, it is not readily obvious which linkage structures might be best used to connect the binding sites to obtain a host cavity that is organized for coordinaton to a targeted metal ion guest. To address the problem of how to identify new host molecules that recognize and bind strongly to specific metal ion guests, we have adopted a computational approach pioneered by the pharmaceutical industry. Drug designers have developed methods to address the inverse of this problem, in other words, how to identify molecular structures (guests) that complement the binding site of a protein (host) [4–6]. These approaches include de novo structure-based design strategies that couple molecule building algorithms with scoring functions used to prioritize the candidate structures. The building algorithms assemble guest molecule structures that can physically interact with a known protein structure from pieces which are either atoms or larger, chemically reasonable fragments. The ability to generate large numbers of potential guest structures necessitates the use of simple scoring functions to prioritize the output. To this end, methods have been developed to estimate the binding free energy by summing free energy increments for hydrogen bond interactions, ionic interactions, lipophilic interactions, the number of rotatable bonds in the guest molecule, etc. [7,8].
B.P. Hay et al. / Journal of Alloys and Compounds 374 (2004) 416–419
The computer programs that have been developed to perform de novo structure-based drug design are, in general, not applicable for the design of host molecules. These programs require input of the atomic coordinates of a protein binding site, are highly specialized to address protein–organic interactions, and do not contain scoring functions to address the interactions that occur in other types of host–guest systems. To apply the powerful concepts embodied in de novo structure-based drug design to the design of metal ion sequestering agents, it is necessary to develop methods for (i) the generation of candidate structures and (ii) evaluating and prioritizing these structures with respect to their binding affinity for a specific guest. This article summarizes recent progress that we have made in these areas.
2. Generating host structures We have developed a de novo structure-based design software, HostDesigner, that is specifically created for the discovery of host molecules for metal ion guests [9]. HostDesigner generates and evaluates millions of candidate structures in minutes on a desktop personal computer, and rapidly identifies three-dimensional architectures that position binding sites to provide an optimal interaction with a metal ion. The molecule building algorithms combine user-input host fragments with linking fragments taken from a database. The user-input host fragments define the optimal geometry for one or more binding sites interacting with a guest. The current linking fragment database used by HostDesigner contains over 10 000 structures composed of Csp3 , Csp2 , and hydrogen atoms. The fragments consist of all connectivities and conformers that can be made from 0 to 6 carbons, excluding three- and four-membered rings. The fragments also include all dimethylated five- and sixmembered rings, and selected bicyclic structures. When using these fragments to build host molecules, all possible connectivities, stereochemistries, and conformations are constructed, which generates large numbers of structures. These structures are prioritized based on how well the host binding sites converge at the metal ion guest. Cartesian coordinates for the top candidates are output to a file for subsequent viewing. Recent improvements to the HostDesigner software, which is available at no cost from the website http://www. hostdesigner.emsl.pnl.gov, allow (i) the treatment of multiatom guests, (ii) structural variation within the input fragments, and (iii) use of empirical methods to estimate the relative conformational energy for each output structure [10].
3. Evaluating host structures Deliberate ligand design requires a knowledge of how structure impacts reactivity and the ability to distinguish a
417
good architecture from a poor architecture. Extensive research effort has been expended toward understanding how host structure influences metal ion binding with the goal of discovering more effective and more selective metal ion receptors [1–3,11,12]. It is clear from these studies that certain properties are needed in order to achieve significant increases in binding affinity or ion recognition. These host properties include (i) the presence of multiple binding sites [1–3], (ii) the ability to adopt a conformation in which all binding sites are positioned to structurally complement the metal ion [13], and (iii) a limited degree of conformational freedom [14]. It has long been recognized that increased binding affinities are obtained when a collection of donor atoms is structurally constrained to the binding conformation, in other words, when the host is preorganized [15]. There are various approaches to evaluate the degree of complementarity and preorganization offered by a host structure. The HostDesigner software, which can examine millions of structures in a single run, requires fast methods for accomplishing these evaluations. This is achieved by using geometric factors to prioritize structures on the basis of how well their binding sites correspond to those of the guest [9]. In addition, conformational energy increments, based on potential surfaces for simple hydrocarbon analogs, are used to estimate the conformational stability of the host structures, allowing them to be further prioritized on the basis of their degree of preorganization [10]. Although approximate in nature, these methods provide a rapid means of selecting a list of the best candidates from a large group of potential structures. This short list can then be re-prioritized using more accurate evaluation methods. With increasing application to coordination compounds [13,16,17], molecular mechanics (MM) models provide an ideal, computationally efficient tool to evaluate the degree to which a ligand is structurally organized for metal complexation [13,18–20]. Conformational analyses yield the stable conformers of the ligand both in the uncomplexed state and in the metal complex. MM models partition the steric energy into stretching, bending, torsion, and non-bonded (Van der Waals, electrostatics, hydrogen bonding) interactions. The process of parameterizing these models requires knowledge of the geometries and potential energy surfaces for each individual interaction, which are precisely the criteria needed to evaluate metal ion complementarity. Examination of the steric energy components from calculations on metal ion complexes provides a way to quantify the effect of (i) cavity size mismatch (strain in M–L bonds), (ii) poor donor group orientation (strain in M–L–X and M–L–X–X angles), and, for metals that exhibit preferences for distinct polyhedra, (iii) topographical mismatch (strain in L–M–L angles) where M: metal, L: donor atom, and X: any other atom [13]. Work we have done on the design of improved diamide sequestering agents demonstrates how these methods have been used to identify ligand architectures that provide enhanced metal ion binding affinity before they are prepared and tested. Conventional malonamide ligands, in which two
418
B.P. Hay et al. / Journal of Alloys and Compounds 374 (2004) 416–419
Fig. 2. MM3 optimized geometries for the only stable conformer of the transoid form, 2, and the lowest energy conformer of the cisoid form, 3, with atom types and oxygen vectors as defined in Fig. 1.
Fig. 1. Structural reorganization of a simple malonamide ligand, 1, during the chelation of a lanthanide metal ion. Vectors attached to each oxygen atom show the optimal direction of approach for a lanthanide metal ion.
amides are connected by a methylene link, are poorly organized for complexation. Fig. 1 compares the metal free conformations of N,N,N ,N -tetramethylmalonamide [21], 1, with a lanthanide–coordinated form of this ligand [22]. In the most stable form, the two carbonyl groups point in opposite directions such that it is not possible for the two oxygen atoms to simultaneously contact the same metal ion. Rotation about one of the two C–C bonds yields a second form, 3.0 kcal/mol higher in energy, which is predisposed for chelation. However, this metal binding conformer fails to provide a complementary array of binding sites. The geometries of simple amides coordinated with lanthanide metal ions show the presence of a distinct oxygen donor directionality in which the metal ion lies in the plane of the amide moiety with a C=O–M angle of 143 ± 1◦ [23]. This directionality is illustrated in Fig. 1 by attaching a vector to each oxygen atom. In a complementary chelate architecture, the two vectors would intersect at a point where the metal ion is located. In the binding form of malonamide, however, these vectors diverge. The use of such vectors provides a convenient way to visualize the degree of complementarity that is offered by a host structure. Attaching vectors of the appropriate length to each binding site also provides a geometric basis for quantitatively evaluating how well the host fits the guest. For example, when molecular fragments consisting of single binding sites are connected by HostDesigner, the software uses the distance between
the ends of binding site vectors to prioritize the candidates [9]. Metal ion coordination causes further structural change to the binding form of malonamide that involves rotation about both C–C bonds. MM calculations on an isolated Eu3+ -1 chelate ring show that the structural change results in an additional 2.5 kcal/mol increase in ligand strain, giving a total strain energy of 5.5 kcal/mol on going from the low energy form to the bound form [24]. The foregoing analysis suggested that metal binding affinity afforded by two amide binding sites could be increased significantly if the binding sites were conformationally constrained in a complementary geometry. After examination of a variety of possible bicyclic scaffolds, we discovered that the desired structural attributes are attained in the stereoisomers 2 and 3 illustrated in Fig. 2. In contrast to 1, both of these architectures have convergent binding site vectors and they both chelate Eu3+ with the development of less than 0.1 kcal/mol of ligand strain [24]. Synthesis and testing have shown that the preorganized architecture afforded by 3 exhibits the anticipated increase in metal ion binding affinity. We have described the preparation and extraction characteristics for a hydrophobic derivative of 3, in which the N-methyl groups have been replaced with n-octyl substituents [24,25]. In liquid-liquid extraction experiments the designed sequestering agent exhibits a dramatic 106 –107 enhancement in the distribution coefficients for trivalent lanthanides and actinides when compared with conventional tetraalkylmalonamide extractants. For all ligands examined in these extraction experiments, the stoichiometry of the extracted species is approximately three ligands to one metal ion revealing that the bicyclic diamide gives an average enhancement of at least 102 per ligand over the acyclic diamides. Recent measurements of aqueous formation constants suggest that this enhancement in metal ion extraction can be attributed directly to the increased metal
B.P. Hay et al. / Journal of Alloys and Compounds 374 (2004) 416–419
ion binding affinity of the bicyclic architecture. Spectrophotometric titrations of Am3+ in 1.0 M nitric acid at 23 ◦ C yield log K values for the formation of 1:1 complexes, −0.17 for 1 and 1.81 for 3, that confirm the bicyclic architecture binds this metal two orders of magnitude more strongly than the acyclic architecture, even under acidic aqueous conditions. Acknowledgements HostDesigner was developed at Pacific Northwest National Laboratory (PNNL) with support from the Division of Chemical Sciences, Office of Science, US Department of Energy (DOE). The diamide research, performed at PNNL and at the University of Oregon (UO), was sponsored by the DOE Environmental Management Science Program (EMSP–54679, EMSP–73759), the National Science Foundation (DUE-0088986 and CHE-023563) and the University of Oregon. PNNL is managed and operated under DOE contract DE–AC06–76RLO–1830 by Battelle Memorial Institute. References [1] R.D. Hancock, A.E. Martell, Chem. Rev. 89 (1989) 1875. [2] A.E. Martell, R.D. Hancock, in: J.P. Fackler (Ed.), Metal Complexes in Aqueous Solution, Plenum Press, New York, 1996. [3] H.-J. Schneider, A. Yatsimirsky, in: Principles and Methods in Supramolecular Chemistry, Wiley, New York, 2000.
419
[4] I.D. Kuntz, E.C. Meng, B.K. Shoichet, Acct. Chem. Res. 27 (1994) 117. [5] T.P. Lybrand, Curr. Opin. Struct. Biol. 5 (1995) 224. [6] H.-J. Böhm, Prog. Biophys. Mol. Biol. 66 (1996) 197. [7] A. Murcko, M.A. Murcko, J. Med. Chem. 38 (1995) 4953. [8] M.D. Eldridge, C.W. Murray, T.R. Auton, G.V. Paolini, R.P. Mee, J. Comput. Aided Mol. Des. 11 (1997) 425. [9] B.P. Hay, T.K. Firman, Inorg. Chem. 41 (2002) 5502. [10] B.P. Hay, T.K. Firman, in: HostDesigner User’s Manual, Pacific Northwest National Laboratory, Richland, Washington, 2003, http://www.hostdesigner.emsl.pnl.gov. [11] A.E. Martell, R.D. Smith, in: Critical Stability Constants, vol. 1–6, Plenum Press, New York, pp. 1974–1988. [12] R.M. Izatt, K. Pawlak, J.S. Bradshaw, R.L. Bruening, Chem. Rev. 95 (1995) 2529. [13] B.P. Hay, R.D. Hancock, Coord. Chem. Rev. 212 (2001) 61. [14] D.H. Busch, K. Farmery, V. Goedken, V. Katovic, A.C. Melnyk, C.R. Sperati, N. Tokel, Adv. Chem. Ser. 100 (1971) 44. [15] D.J. Cram, T. Kaneda, R.C. Helgeson, S.B. Brown, C.B. Knobler, E. Maverick, K.N. Trueblood, J. Am. Chem. Soc. 107 (1985) 3645. [16] B.P. Hay, Coord. Chem. Rev. 126 (1993) 177. [17] P. Comba, T.W. Hambley, in: Molecular Modeling of Inorganic Compounds, VHC, New York, 1995. [18] B.P. Hay, J.R. Rustad, J. Am. Chem. Soc. 116 (1994) 6316. [19] B.P. Hay, D. Zhang, J.R. Rustad, Inorg. Chem. 35 (1996) 2650. [20] B.P. Hay, D.A. Dixon, R. Vargas, J. Garza, K.N. Raymond, Inorg. Chem. 40 (2001) 3922. [21] G. Sandrone, D.A. Dixon, B.P. Hay, J. Phys. Chem. 103 (1999) 3554. [22] E.E. Castellano, R.W. Becker, Acta Crystallogr. Sect. B 37 (1981) 61. [23] O. Clement, B.M. Rapko, B.P. Hay, Coord. Chem. Rev. 170 (1998) 203. [24] G.J. Lumetta, B.M. Rapko, P.A. Garza, B.P. Hay, R.D. Gilbertson, T.J.R. Weakley, J.E. Hutchison, J. Am. Chem. Soc. 124 (2002) 5644. [25] G.J. Lumetta, B.M. Rapko, B.P. Hay, P.A. Garza, J.E. Hutchison, R.D. Gilbertson, Solv. Extr. Ion Exch. 21 (2003) 29.