Elizabeth A. Amin

Zn Coordination Chemistry: Development of Benchmark Suites for Geometries, Dipole Moments, and Bond Dissociation Energies and Their Use To Test and Validate Density Functionals and Molecular Orbital Theory

We present nonrelativistic and relativistic benchmark databases (obtained by coupled cluster calculations) of 10 Zn-ligand bond distances, 8 dipole moments, and 12 bond dissociation energies in Zn coordination compounds with O, S, NH3, H2O, OH, SCH3, and H ligands. These are used to test the predictions of 39 density functionals, Hartree-Fock theory, and seven more approximate molecular orbital theories. In the nonrelativisitic case, the M05-2X, B97-2, and mPW1PW functionals emerge as the most accurate ones for this test data, with unitless balanced mean unsigned errors (BMUEs) of 0.33, 0.38, and 0.43, respectively. The best local functionals (i.e., functionals with no Hartree-Fock exchange) are M06-L and τ-HCTH with BMUEs of 0.54 and 0.60, respectively. The popular B3LYP functional has a BMUE of 0.51, only slightly better than the value of 0.54 for the best local functional, which is less expensive. Hartree-Fock theory itself has a BMUE of 1.22. The M05-2X functional has a mean unsigned error of 0.008 Å for bond lengths, 0.19 D for dipole moments, and 4.30 kcal/mol for bond energies. The X3LYP functional has a smaller mean unsigned error (0.007 Å) for bond lengths but has mean unsigned errors of 0.43 D for dipole moments and 5.6 kcal/mol for bond energies. The M06-2X functional has a smaller mean unsigned error (3.3 kcal/mol) for bond energies but has mean unsigned errors of 0.017 Å for bond lengths and 0.37 D for dipole moments. The best of the semiempirical molecular orbital theories are PM3 and PM6, with BMUEs of 1.96 and 2.02, respectively. The ten most accurate functionals from the nonrelativistic benchmark analysis are then tested in relativistic calculations against new benchmarks obtained with coupled-cluster calculations and a relativistic effective core potential, resulting in M05-2X (BMUE = 0.895), PW6B95 (BMUE = 0.90), and B97-2 (BMUE = 0.93) as the top three functionals. We find significant relativistic effects (∼0.01 Å in bond lengths, ∼0.2 D in dipole moments, and ∼4 kcal/mol in Zn-ligand bond energies) that cannot be neglected for accurate modeling, but the same density functionals that do well in all-electron nonrelativistic calculations do well with relativistic effective core potentials. Although most tests are carried out with augmented polarized triple-ζ basis sets, we also carried out some tests with an augmented polarized double-ζ basis set, and we found, on average, that with the smaller basis set DFT has no loss in accuracy for dipole moments and only ∼10% less accurate bond lengths.

Energies, Geometries, and Charge Distributions of Zn Molecules, Clusters, and Biocenters from Coupled Cluster, Density Functional, and Neglect of Diatomic Differential Overlap Models

We present benchmark databases of Zn-ligand bond distances, bond angles, dipole moments, and bond dissociation energies for Zn-containing small molecules and Zn coordination compounds with H, CH3, C2H5, NH3, O, OH, H2O, F, Cl, S, and SCH3 ligands. The test set also includes clusters with Zn-Zn bonds. In addition, we calculated dipole moments and binding energies for Zn centers in coordination environments taken from zinc metalloenzyme X-ray structures, representing both structural and catalytic zinc centers. The benchmark values are based on relativistic-core coupled cluster calculations. These benchmark calculations are used to test the predictions of four density functionals, namely B3LYP and the more recently developed M05-2X, M06, and M06-2X levels of theory, and six semiempirical methods, including neglect of diatomic differential overlap (NDDO) calculations incorporating the new PM3 parameter set for Zn called ZnB, developed by Brothers and co-workers, and the recent PM6 parametrization of Stewart. We found that the best DFT method to reproduce dipole moments and dissociation energies of our Zn compound database is M05-2X, which is consistent with a previous study employing a much smaller and less diverse database and a much larger set of density functionals. Here we show that M05-2X geometries and single-point coupled cluster calculations with M05-2X geometries can also be used as benchmarks for larger compounds, where coupled cluster optimization is impractical, and in particular we use this strategy to extend the geometry, binding energy, and dipole moment databases to additional molecules, and we extend the tests involving crystal-site coordination compounds to two additional proteins. We find that the most predictive NDDO methods for our training set are PM3 and MNDO/d. Notably, we also find large errors in B3LYP for the coordination compounds based on experimental X-ray geometries.

Design, synthesis and evaluation of analogs of initiation factor 4E (eIF4E) cap-binding antagonist Bn7-GMP

Aberrant regulation of cap-dependent translation has been frequently observed in the development of cancer. Association of the cap-binding protein eIF4E with N(7)-methylated guanosine capped mRNA is the rate limiting step governing translation initiation; and therefore represents an attractive process for cancer drug discovery. Previously, replacement of the 7-Me group of the Me(7)-guanosine monophosphate with a benzyl group has been found to increase binding affinity to eIF4E. Recent X-ray crystallographic studies have revealed that the cap-binding pocket undergoes a unique structural change in order to accommodate the benzyl group. To explore the structure-activity relationships governing the affinity of N(7)-benzylated guanosine monophosphate (Bn(7)-GMP) for eIF4E, we virtually screened a library of 80 Bn(7)-GMP analogs utilizing CombiGlide as implemented in Schrodinger. A subset library of substituted Bn(7)-GMP analogs was synthesized and their dissociation constants (K(d)) were determined. Due to the poor correlation between docking/scoring results and experimental binding affinities, three-dimensional quantitative structure-activity relationship (3D-QSAR) calculations were performed. Two highly predictive and self-consistent CoMFA (comparative molecular field analysis) and CoMSIA (comparative molecular similarity indices analysis) models were derived and optimized. These models may be useful for the future design of eIF4E cap-binding antagonists.

Tactical Approaches to Interconverting GPCR Agonists and Antagonists.

There are many reported examples of small structural modifications to GPCR-targeted ligands leading to major changes in their functional activity, converting agonists into antagonists or vice versa. These shifts in functional activity are often accompanied by negligible changes in binding affinity. The current perspective focuses on outlining and analyzing various approaches that have been used to interconvert GPCR agonists, partial agonists, and antagonists in order to achieve the intended functional activity at a GPCR of therapeutic interest. An improved understanding of specific structural modifications that are likely to alter the functional activity of a GPCR ligand may be of use to researchers designing GPCR-targeted drugs and/or probe compounds, specifically in cases where a particular ligand exhibits good potency but not the preferred functional activity at the GPCR of choice.

Development of a Comprehensive, Validated Pharmacophore Hypothesis for Anthrax Toxin Lethal Factor (LF) Inhibitors Using Genetic Algorithms, Pareto Scoring, and Structural Biology

Anthrax is an acute infectious disease caused by the spore-forming bacterium Bacillus anthracis. The anthrax toxin lethal factor (LF), an 89-kDa zinc hydrolase secreted by the bacilli, is the toxin component chiefly responsible for pathogenesis and has been a popular target for rational and structure-based drug design. Although hundreds of small-molecule compounds have been designed to target the LF active site, relatively few reported inhibitors have exhibited activity in cell-based assays, and no LF inhibitor is currently available to treat or prevent anthrax. This study presents a new pharmacophore map assembly, validated by experiment, designed to rapidly identify and prioritize promising LF inhibitor scaffolds from virtual compound libraries. The new hypothesis incorporates structural information from all five available LF enzyme-inhibitor complexes deposited in the Protein Data Bank (PDB) and is the first LF pharmacophore map reported to date that includes features representing interactions involving all three key subsites of the LF catalytic binding region. In a wide-ranging validation study on all 546 compounds for which published LF biological activity data exist, this model displayed strong selectivity toward nanomolar-level LF inhibitors, successfully identifying 72.1% of existing nanomolar-level compounds in an unbiased test set, while rejecting 100% of weakly active (>100 μM) compounds. In addition to its capabilities as a database searching tool, this comprehensive model points to a number of key design principles and previously unidentified ligand-receptor interactions that are likely to influence compound potency.

Assessment and Validation of the Electrostatically Embedded Many-Body Expansion for Metal-Ligand Bonding.

The electrostatically embedded many-body method has been very successful for calculating cohesive energies and relative conformational energies of clusters, and here we extend it to calculate bond breaking energies for metal-ligand bonds in inorganic coordination chemistry. We find that, on average, the electrostatically embedded pairwise additive method is able to predict bond energies yielded by conventional full-system calculations done at the same level of theory to within 2.5 kcal/mol and that the electrostatically embedded three-body method consistently yields energies within 1.0 kcal/mol of the full-system calculations.

Methemoglobin Formation and Characterization of Hemoglobin Adducts of Carcinogenic Aromatic Amines and Heterocyclic Aromatic Amines.

Arylamines (AAs) and heterocyclic aromatic amines (HAAs) are structurally related carcinogens formed during the combustion of tobacco or cooking of meat. They undergo cytochrome P450 mediated N-hydroxylation to form metabolites which bind to DNA and lead to mutations. The N-hydroxylated metabolites of many AAs also can undergo a co-oxidation reaction with oxy-hemolgobin (HbO2) to form methemoglobin (met-Hb) and the arylnitroso intermediates, which react with the β-Cys(93) chain of Hb to form Hb-arylsulfinamide adducts. The biochemistry of arylamine metabolism has been exploited to biomonitor certain AAs through their Hb arylsulfinamide adducts in humans. We examined the reactivity of HbO2 with the N-hydroxylated metabolites of 4-aminobiphenyl (ABP, HONH-ABP), aniline (ANL, HONH-ANL), and the HAAs 2-amino-9H-pyrido[2,3-b]indole (AαC, HONH-AαC), 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP, HONH-PhIP), and 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx, HONH-MeIQx). HONH-ABP, HO-ANL, and HONH-AαC induced methemoglobinemia and formed Hb sulfinamide adducts. However, HONH-MeIQx and HONH-PhIP did not react with the oxy-heme complex, and met-Hb formation and chemical modification of the β-Cys(93) residue were negligible. Molecular modeling studies showed that the distances between the H-ON-AA or H-ON-HAA substrates and the oxy-heme complex of HbO2 were too far away to induce methemoglobinemia. Different conformational changes in flexible helical and loop regions around the heme pocket induced by the H-ON-AA or H-ON-HAAs may explain the different proclivities of these chemicals to induce methemoglobinemia. Hb-Cys(93β) sulfinamide and sulfonamide adducts of ABP, ANL, and AαC were identified, by Orbitrap MS, following the proteolysis of Hb with trypsin, Glu-C, or Lys-C. Hb sulfinamide and sulfonamide adducts of ABP were identified in the blood of mice exposed to ABP, by Orbitrap MS. This is the first report of the identification of intact Hb sulfinamide adducts of carcinogenic AAs in vivo. The high reactivity of HONH-AαC with HbO2 suggests that the Hb sulfinamide adduct of AαC may be a promising biomarker of exposure to this HAA in humans.

Total Synthesis and Evaluation of C26-Hydroxyepothilone D Derivatives for Photoaffinity Labeling of β-Tubulin

Three photoaffinity labeled derivatives of epothilone D were prepared by total synthesis, using efficient novel asymmetric synthesis methods for the preparation of two important synthetic building blocks. The key step for the asymmetric synthesis of (S,E)-3-(tert-butyldimethylsilyloxy)-4-methyl-5-(2-methylthiazol-4-yl)pent-4-enal involved a ketone reduction with (R)-Me-CBS-oxazaborolidine. For the synthesis of (5S)-5,7-di[(tert-butyldimethylsilyl)oxy]-4,4-dimethylheptan-3-one an asymmetric Noyori reduction of a beta-ketoester was employed. The C26 hydroxyepothilone D derivative was constructed following a well-established total synthesis strategy and the photoaffinity labels were attached to the C26 hydroxyl group. The photoaffinity analogues were tested in a tubulin assembly assay and for cytotoxicity against MCF-7 and HCT-116 cancer cell lines. The 3- and 4-azidobenzoic acid analogues were found to be as active as epothilone B in a tubulin assembly assay, but demonstrated significantly reduced cellular cytotoxicity compared to epothilone B. The benzophenone analogue was inactive in both assays. Docking and scoring studies were conducted that suggested that the azide analogues can bind to the epothilone binding site, but that the benzophenone analogue undergoes a sterically driven ligand rearrangement that interrupts all hydrogen bonding and therefore protein binding. Photoaffinity labeling studies with the 3-azidobenzoic acid derivative did not identify any covalently labeled peptide fragments, suggesting that the phenylazido side chain was predominantly solvent-exposed in the bound conformation.

Highly predictive support vector machine (SVM) models for anthrax toxin lethal factor (LF) inhibitors.

Anthrax is a highly lethal, acute infectious disease caused by the rod-shaped, Gram-positive bacterium Bacillus anthracis. The anthrax toxin lethal factor (LF), a zinc metalloprotease secreted by the bacilli, plays a key role in anthrax pathogenesis and is chiefly responsible for anthrax-related toxemia and host death, partly via inactivation of mitogen-activated protein kinase kinase (MAPKK) enzymes and consequent disruption of key cellular signaling pathways. Antibiotics such as fluoroquinolones are capable of clearing the bacilli but have no effect on LF-mediated toxemia; LF itself therefore remains the preferred target for toxin inactivation. However, currently no LF inhibitor is available on the market as a therapeutic, partly due to the insufficiency of existing LF inhibitor scaffolds in terms of efficacy, selectivity, and toxicity. In the current work, we present novel support vector machine (SVM) models with high prediction accuracy that are designed to rapidly identify potential novel, structurally diverse LF inhibitor chemical matter from compound libraries. These SVM models were trained and validated using 508 compounds with published LF biological activity data and 847 inactive compounds deposited in the Pub Chem BioAssay database. One model, M1, demonstrated particularly favorable selectivity toward highly active compounds by correctly predicting 39 (95.12%) out of 41 nanomolar-level LF inhibitors, 46 (93.88%) out of 49 inactives, and 844 (99.65%) out of 847 Pub Chem inactives in external, unbiased test sets. These models are expected to facilitate the prediction of LF inhibitory activity for existing molecules, as well as identification of novel potential LF inhibitors from large datasets.

Ligand‐induced expansion of the S1′ site in the anthrax toxin lethal factor

The Bacillus anthracis lethal factor (LF) is one component of a tripartite exotoxin partly responsible for persistent anthrax cytotoxicity after initial bacterial infection. Inhibitors of the zinc metalloproteinase have been investigated as potential therapeutic agents, but LF is a challenging target because inhibitors lack sufficient selectivity or possess poor pharmaceutical properties. These structural studies reveal an alternate conformation of the enzyme, induced upon binding of specific inhibitors, that opens a previously unobserved deep pocket termed S1′∗ which might afford new opportunities to design selective inhibitors that target this subsite.