Deciphering Antifungal Drug Resistance in Pneumocystis jirovecii DHFR with Molecular Dynamics and Machine Learning.

Drug resistance impacts the effectiveness of many new therapeutics. Mutations in the therapeutic target confer resistance; however, deciphering which mutations, often remote from the enzyme active site, drive resistance is challenging. In a series of Pneumocystis jirovecii dihydrofolate reductase variants, we elucidate which interactions are key bellwethers to confer resistance to trimethoprim using homology modeling, molecular dynamics, and machine learning. Six molecular features involving mainly residues that did not vary were the best indicators of resistance.

Deciphering Complex Mechanisms of Resistance and Loss of Potency through Coupled Molecular Dynamics and Machine Learning.

Drug resistance threatens many critical therapeutics through mutations in the drug target. The molecular mechanisms by which combinations of mutations, especially those remote from the active site, alter drug binding to confer resistance are poorly understood and thus difficult to counteract. A machine learning strategy was developed that coupled parallel molecular dynamics simulations with experimental potency to identify specific conserved mechanisms underlying resistance. Physical features were extracted from the simulations, analyzed, and integrated into one consistent and interpretable elastic network model. To rigorously test this strategy, HIV-1 protease variants with diverse mutations were used, with potencies ranging from picomolar to micromolar to the drug darunavir. Feature reduction resulted in a model with four specific features that predicts for both the training and test sets inhibitor binding free energy within 1 kcal/mol of the experimental value over this entire range of potency. These predictive features are physically interpretable, as they vary specifically with affinity and diagonally transverse across the protease homodimer. This physics-based strategy of parallel molecular dynamics and machine learning captures mechanisms by which complex combinations of mutations confer resistance and identify critical features that serve as bellwethers of affinity, which will be critical in future drug design.

Drug Design Strategies to Avoid Resistance in Direct-Acting Antivirals and Beyond.

Drug resistance is prevalent across many diseases, rendering therapies ineffective with severe financial and health consequences. Rather than accepting resistance after the fact, proactive strategies need to be incorporated into the drug design and development process to minimize the impact of drug resistance. These strategies can be derived from our experience with viral disease targets where multiple generations of drugs had to be developed to combat resistance and avoid antiviral failure. Significant efforts including experimental and computational structural biology, medicinal chemistry, and machine learning have focused on understanding the mechanisms and structural basis of resistance against direct-acting antiviral (DAA) drugs. Integrated methods show promise for being predictive of resistance and potency. In this review, we give an overview of this research for human immunodeficiency virus type 1, hepatitis C virus, and influenza virus and the lessons learned from resistance mechanisms of DAAs. These lessons translate into rational strategies to avoid resistance in drug design, which can be generalized and applied beyond viral targets. While resistance may not be completely avoidable, rational drug design can and should incorporate strategies at the outset of drug development to decrease the prevalence of drug resistance.

Optimizing the refinement of merohedrally twinned P61 HIV-1 protease-inhibitor cocrystal structures.

Twinning is a crystal-growth anomaly in which protein monomers exist in different orientations but are related in a specific way, causing diffraction reflections to overlap. Twinning imposes additional symmetry on the data, often leading to the assignment of a higher symmetry space group. Specifically, in merohedral twinning, reflections from each monomer overlap and require a twin law to model unique structural data from overlapping reflections. Neglecting twinning in the crystallographic analysis of quasi-rotationally symmetric homo-oligomeric protein structures can mask the degree of structural non-identity between monomers. In particular, any deviations from perfect symmetry will be lost if higher than appropriate symmetry is applied during crystallographic analysis. Such cases warrant choosing between the highest symmetry space group possible or determining whether the monomers have distinguishable structural asymmetries and thus require a lower symmetry space group and a twin law. Using hexagonal cocrystals of HIV-1 protease, a C2-symmetric homodimer whose symmetry is broken by bound ligand, it is shown that both assigning a lower symmetry space group and applying a twin law during refinement are critical to achieving a structural model that more accurately fits the electron density. By re-analyzing three recently published HIV-1 protease structures, improvements in nearly every crystallographic metric are demonstrated. Most importantly, a procedure is demonstrated where the inhibitor can be reliably modeled in a single orientation. This protocol may be applicable to many other homo-oligomers in the PDB.

Molecular and Structural Mechanism of Pan-Genotypic HCV NS3/4A Protease Inhibition by Glecaprevir.

Hepatitis C virus, causative agent of chronic viral hepatitis, infects 71 million people worldwide and is divided into seven genotypes and multiple subtypes with sequence identities between 68 to 82%. While older generation direct-acting antivirals had varying effectiveness against different genotypes, the newest NS3/4A protease inhibitors including glecaprevir (GLE) have pan-genotypic activity. The structural basis for pan-genotypic inhibition and effects of polymorphisms on inhibitor potency were not well-known due to lack of crystal structures of GLE-bound NS3/4A or genotypes other than 1. In this study, we determined the crystal structures of NS3/4A from genotypes 1a, 3a, 4a, and 5a in complex with GLE. Comparison with the highly similar grazoprevir indicated the mechanism of GLE's drastic improvement in potency. We found that, while GLE is highly potent against wild-type NS3/4A of all genotypes, specific resistance-associated substitutions (RASs) confer orders of magnitude loss in inhibition. Our crystal structures reveal molecular mechanisms behind pan-genotypic activity of GLE, including potency loss due to RASs at D168. Our structures permit for the first time analysis of changes due to polymorphisms among genotypes, providing insights into design principles that can aid future drug development and potentially can be extended to other proteins.

Molecular Determinants of Epistasis in HIV-1 Protease: Elucidating the Interdependence of L89V and L90M Mutations in Resistance.

Protease inhibitors have the highest potency among antiviral therapies against HIV-1 infections, yet the virus can evolve resistance. Darunavir (DRV), currently the most potent Food and Drug Administration-approved protease inhibitor, retains potency against single-site mutations. However, complex combinations of mutations can confer resistance to DRV. While the interdependence between mutations within HIV-1 protease is key for inhibitor potency, the molecular mechanisms that underlie this control remain largely unknown. In this study, we investigated the interdependence between the L89V and L90M mutations and their effects on DRV binding. These two mutations have been reported to be positively correlated with one another in HIV-1 patient-derived protease isolates, with the presence of one mutation making the probability of the occurrence of the second mutation more likely. The focus of our investigation is a patient-derived isolate, with 24 mutations that we call "KY"; this variant includes the L89V and L90M mutations. Three additional KY variants with back-mutations, KY(V89L), KY(M90L), and the KY(V89L/M90L) double mutation, were used to experimentally assess the individual and combined effects of these mutations on DRV inhibition and substrate processing. The enzymatic assays revealed that the KY(V89L) variant, with methionine at residue 90, is highly resistant, but its catalytic function is compromised. When a leucine to valine mutation at residue 89 is present simultaneously with the L90M mutation, a rescue of catalytic efficiency is observed. Molecular dynamics simulations of these DRV-bound protease variants reveal how the L90M mutation induces structural changes throughout the enzyme that undermine the binding interactions.

Target-Specific Prediction of Ligand Affinity with Structure-Based Interaction Fingerprints.

Discovery and optimization of small molecule inhibitors as therapeutic drugs have immensely benefited from rational structure-based drug design. With recent advances in high-resolution structure determination, computational power, and machine learning methodology, it is becoming more tractable to elucidate the structural basis of drug potency. However, the applicability of machine learning models to drug design is limited by the interpretability of the resulting models in terms of feature importance. Here, we take advantage of the large number of available inhibitor-bound HIV-1 protease structures and associated potencies to evaluate inhibitor diversity and machine learning models to predict ligand affinity. First, using a hierarchical clustering approach, we grouped HIV-1 protease inhibitors and identified distinct core structures. Explicit features including protein-ligand interactions were extracted from high-resolution cocrystal structures as 3D-based fingerprints. We found that a gradient boosting machine learning model with this explicit feature attribution can predict binding affinity with high accuracy. Finally, Shapley values were derived to explain local feature importance. We found specific van der Waals (vdW) interactions of key protein residues are pivotal for the predicted potency. Protein-specific and interpretable prediction models can guide the optimization of many small molecule drugs for improved potency.

Picomolar to Micromolar: Elucidating the Role of Distal Mutations in HIV-1 Protease in Conferring Drug Resistance.

Drug resistance continues to be a growing global problem. The efficacy of small molecule inhibitors is threatened by pools of genetic diversity in all systems, including antibacterials, antifungals, cancer therapeutics, and antivirals. Resistant variants often include combinations of active site mutations and distal "secondary" mutations, which are thought to compensate for losses in enzymatic activity. HIV-1 protease is the ideal model system to investigate these combinations and underlying molecular mechanisms of resistance. Darunavir (DRV) binds wild-type (WT) HIV-1 protease with a potency of <5 pM, but we have identified a protease variant that loses potency to DRV 150 000-fold, with 11 mutations in and outside the active site. To elucidate the roles of these mutations in DRV resistance, we used a multidisciplinary approach, combining enzymatic assays, crystallography, and molecular dynamics simulations. Analysis of protease variants with 1, 2, 4, 8, 9, 10, and 11 mutations showed that the primary active site mutations caused ∼50-fold loss in potency (2 mutations), while distal mutations outside the active site further decreased DRV potency from 13 nM (8 mutations) to 0.76 µM (11 mutations). Crystal structures and simulations revealed that distal mutations induce subtle changes that are dynamically propagated through the protease. Our results reveal that changes remote from the active site directly and dramatically impact the potency of the inhibitor. Moreover, we find interdependent effects of mutations in conferring high levels of resistance. These mechanisms of resistance are likely applicable to many other quickly evolving drug targets, and the insights may have implications for the design of more robust inhibitors.

Structural Analysis of the Active Site and DNA Binding of Human Cytidine Deaminase APOBEC3B.

APOBEC3 (A3) proteins, a family of human cytidine deaminases, protect the host from endogenous retro-elements and exogenous viral infections by introducing hypermutations. However, overexpressed A3s can modify genomic DNA to promote tumorigenesis, especially A3B. Despite their overall similarity, A3 proteins have distinct deamination activity. Recently determined A3 structures have revealed the molecular determinants of nucleotide specificity and DNA binding. However, for A3B, the structural basis for regulation of deamination activity and the role of active site loops in coordinating DNA had remained unknown. Using advanced molecular modeling followed by experimental mutational analysis and dynamics simulations, we investigated the molecular mechanism of DNA binding by A3B-CTD. We modeled fully native A3B-DNA structure, and we identified Arg211 in loop 1 as the gatekeeper coordinating DNA and critical residue for nucleotide specificity. We also identified a unique autoinhibited conformation in A3B-CTD that restricts access and binding of DNA to the active site. Our results reveal the structural basis for DNA binding and relatively lower catalytic activity of A3B and provide opportunities for rational design of specific inhibitors to benefit cancer therapeutics.

Structural Adaptation of Darunavir Analogues against Primary Mutations in HIV-1 Protease.

HIV-1 protease is one of the prime targets of agents used in antiretroviral therapy against HIV. However, under selective pressure of protease inhibitors, primary mutations at the active site weaken inhibitor binding to confer resistance. Darunavir (DRV) is the most potent HIV-1 protease inhibitor in clinic; resistance is limited, as DRV fits well within the substrate envelope. Nevertheless, resistance is observed due to hydrophobic changes at residues including I50, V82, and I84 that line the S1/S1' pocket within the active site. Through enzyme inhibition assays and a series of 12 crystal structures, we interrogated susceptibility of DRV and two potent analogues to primary S1' mutations. The analogues had modifications at the hydrophobic P1' moiety compared to DRV to better occupy the unexploited space in the S1' pocket where the primary mutations were located. Considerable losses of potency were observed against protease variants with I84V and I50V mutations for all three inhibitors. The crystal structures revealed an unexpected conformational change in the flap region of I50V protease bound to the analogue with the largest P1' moiety, indicating interdependency between the S1' subsite and the flap region. Collective analysis of protease-inhibitor interactions in the crystal structures using principle component analysis was able to distinguish inhibitor identity and relative potency solely based on van der Waals contacts. Our results reveal the complexity of the interplay between inhibitor P1' moiety and S1' mutations and validate principle component analyses as a useful tool for distinguishing resistance and inhibitor potency.

Molecular Mechanism of Resistance in a Clinically Significant Double-Mutant Variant of HCV NS3/4A Protease.

Despite significant progress in hepatitis C virus (HCV) protease inhibitor (PI) drug design, resistance remains a problem causing treatment failure. Double-substitution variants, notably Y56H/D168A, have emerged in patients who fail therapy with a PI-containing regimen. The resistance conferred by Asp168 substitutions has been well characterized and avoided in newer inhibitors. However, an additional mutation at Tyr56 confers resistance to even the most robust inhibitors. Here, we elucidate the molecular mechanisms of resistance for the Y56H/D168A variant against grazoprevir (and four analogs), paritaprevir, and danoprevir through inhibition assays, co-crystal structures, and molecular dynamics simulations. The PIs' susceptibility to Y56H/D168A varies, with those stacking on the catalytic His57 losing the most potency. For such inhibitors, the Y56H substitution disrupts favorable stacking interactions with the neighboring catalytic His57. This indirect mechanism of resistance threatens to cause multi-PI failure as all HCV PIs in clinical development rely on interactions with the catalytic triad.

Hydration Structure and Dynamics of Inhibitor-Bound HIV-1 Protease.

Water is essential in many biological processes, and the hydration structure plays a critical role in facilitating protein folding, dynamics, and ligand binding. A variety of biophysical spectroscopic techniques have been used to probe the water solvating proteins, often complemented with molecular dynamics (MD) simulations to resolve the spatial and dynamic features of the hydration shell, but comparing relative water structure is challenging. In this study 1 µs MD simulations were performed to identify and characterize hydration sites around HIV-1 protease bound to an inhibitor, darunavir (DRV). The water density, hydration site occupancy, extent and anisotropy of fluctuations, coordinated water molecules, and hydrogen bonds were characterized and compared to the properties of bulk water. The water density of the principal hydration shell was found to be higher than bulk, dependent on the topology and physiochemical identity of the biomolecular surface. The dynamics of water molecules occupying principal hydration sites was highly dependent on the number of water-water interactions and inversely correlated with hydrogen bonds to the protein-inhibitor complex. While many waters were conserved following the symmetry of homodimeric HIV protease, the asymmetry induced by DRV resulted in asymmetric lower-occupancy hydration sites at the concave surface of the active site. Key interactions between water molecules and the protease, that stabilize the protein in the inhibited form, were altered in a drug resistant variant of the protease indicating that modulation of solvent-solute interactions might play a key role in conveying drug resistance. Our analysis provides insights into the interplay between an enzyme inhibitor complex and the hydration shell and has implications in elucidating water structure in a variety of biological processes and applications including ligand binding, inhibitor design, and resistance.

Interdependence of Inhibitor Recognition in HIV-1 Protease.

Molecular recognition is a highly interdependent process. Subsite couplings within the active site of proteases are most often revealed through conditional amino acid preferences in substrate recognition. However, the potential effect of these couplings on inhibition and thus inhibitor design is largely unexplored. The present study examines the interdependency of subsites in HIV-1 protease using a focused library of protease inhibitors, to aid in future inhibitor design. Previously a series of darunavir (DRV) analogs was designed to systematically probe the S1' and S2' subsites. Co-crystal structures of these analogs with HIV-1 protease provide the ideal opportunity to probe subsite interdependency. All-atom molecular dynamics simulations starting from these structures were performed and systematically analyzed in terms of atomic fluctuations, intermolecular interactions, and water structure. These analyses reveal that the S1' subsite highly influences other subsites: the extension of the hydrophobic P1' moiety results in 1) reduced van der Waals contacts in the P2' subsite, 2) more variability in the hydrogen bond frequencies with catalytic residues and the flap water, and 3) changes in the occupancy of conserved water sites both proximal and distal to the active site. In addition, one of the monomers in this homodimeric enzyme has atomic fluctuations more highly correlated with DRV than the other monomer. These relationships intricately link the HIV-1 protease subsites and are critical to understanding molecular recognition and inhibitor binding. More broadly, the interdependency of subsite recognition within an active site requires consideration in the selection of chemical moieties in drug design; this strategy is in contrast to what is traditionally done with independent optimization of chemical moieties of an inhibitor.