Turning high-throughput structural biology into predictive inhibitor design.

A common challenge in drug design pertains to finding chemical modifications to a ligand that increases its affinity to the target protein. An underutilized advance is the increase in structural biology throughput, which has progressed from an artisanal endeavor to a monthly throughput of hundreds of different ligands against a protein in modern synchrotrons. However, the missing piece is a framework that turns high-throughput crystallography data into predictive models for ligand design. Here, we designed a simple machine learning approach that predicts protein-ligand affinity from experimental structures of diverse ligands against a single protein paired with biochemical measurements. Our key insight is using physics-based energy descriptors to represent protein-ligand complexes and a learning-to-rank approach that infers the relevant differences between binding modes. We ran a high-throughput crystallography campaign against the SARS-CoV-2 main protease (MPro), obtaining parallel measurements of over 200 protein-ligand complexes and their binding activities. This allows us to design one-step library syntheses which improved the potency of two distinct micromolar hits by over 10-fold, arriving at a noncovalent and nonpeptidomimetic inhibitor with 120 nM antiviral efficacy. Crucially, our approach successfully extends ligands to unexplored regions of the binding pocket, executing large and fruitful moves in chemical space with simple chemistry.

Iterative computational design and crystallographic screening identifies potent inhibitors targeting the Nsp3 macrodomain of SARS-CoV-2.

The nonstructural protein 3 (NSP3) of the severe acute respiratory syndrome-coronavirus-2 (SARS-CoV-2) contains a conserved macrodomain enzyme (Mac1) that is critical for pathogenesis and lethality. While small-molecule inhibitors of Mac1 have great therapeutic potential, at the outset of the COVID-19 pandemic, there were no well-validated inhibitors for this protein nor, indeed, the macrodomain enzyme family, making this target a pharmacological orphan. Here, we report the structure-based discovery and development of several different chemical scaffolds exhibiting low- to sub-micromolar affinity for Mac1 through iterations of computer-aided design, structural characterization by ultra-high-resolution protein crystallography, and binding evaluation. Potent scaffolds were designed with in silico fragment linkage and by ultra-large library docking of over 450 million molecules. Both techniques leverage the computational exploration of tangible chemical space and are applicable to other pharmacological orphans. Overall, 160 ligands in 119 different scaffolds were discovered, and 153 Mac1-ligand complex crystal structures were determined, typically to 1 Å resolution or better. Our analyses discovered selective and cell-permeable molecules, unexpected ligand-mediated conformational changes within the active site, and key inhibitor motifs that will template future drug development against Mac1.

Allosteric regulation and crystallographic fragment screening of SARS-CoV-2 NSP15 endoribonuclease.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is the causative agent of coronavirus disease 2019 (COVID-19). The NSP15 endoribonuclease enzyme, known as NendoU, is highly conserved and plays a critical role in the ability of the virus to evade the immune system. NendoU is a promising target for the development of new antiviral drugs. However, the complexity of the enzyme's structure and kinetics, along with the broad range of recognition sequences and lack of structural complexes, hampers the development of inhibitors. Here, we performed enzymatic characterization of NendoU in its monomeric and hexameric form, showing that hexamers are allosteric enzymes with a positive cooperative index, and with no influence of manganese on enzymatic activity. Through combining cryo-electron microscopy at different pHs, X-ray crystallography and biochemical and structural analysis, we showed that NendoU can shift between open and closed forms, which probably correspond to active and inactive states, respectively. We also explored the possibility of NendoU assembling into larger supramolecular structures and proposed a mechanism for allosteric regulation. In addition, we conducted a large fragment screening campaign against NendoU and identified several new allosteric sites that could be targeted for the development of new inhibitors. Overall, our findings provide insights into the complex structure and function of NendoU and offer new opportunities for the development of inhibitors.

Fragment Merging Using a Graph Database Samples Different Catalogue Space than Similarity Search.

Fragment merging is a promising approach to progressing fragments directly to on-scale potency: each designed compound incorporates the structural motifs of overlapping fragments in a way that ensures compounds recapitulate multiple high-quality interactions. Searching commercial catalogues provides one useful way to quickly and cheaply identify such merges and circumvents the challenge of synthetic accessibility, provided they can be readily identified. Here, we demonstrate that the Fragment Network, a graph database that provides a novel way to explore the chemical space surrounding fragment hits, is well-suited to this challenge. We use an iteration of the database containing >120 million catalogue compounds to find fragment merges for four crystallographic screening campaigns and contrast the results with a traditional fingerprint-based similarity search. The two approaches identify complementary sets of merges that recapitulate the observed fragment-protein interactions but lie in different regions of chemical space. We further show our methodology is an effective route to achieving on-scale potency by retrospective analyses for two different targets; in analyses of public COVID Moonshot and Mycobacterium tuberculosis EthR inhibitors, potential inhibitors with micromolar IC50 values were identified. This work demonstrates the use of the Fragment Network to increase the yield of fragment merges beyond that of a classical catalogue search.

A Small Step Toward Generalizability: Training a Machine Learning Scoring Function for Structure-Based Virtual Screening.

Over the past few years, many machine learning-based scoring functions for predicting the binding of small molecules to proteins have been developed. Their objective is to approximate the distribution which takes two molecules as input and outputs the energy of their interaction. Only a scoring function that accounts for the interatomic interactions involved in binding can accurately predict binding affinity on unseen molecules. However, many scoring functions make predictions based on data set biases rather than an understanding of the physics of binding. These scoring functions perform well when tested on similar targets to those in the training set but fail to generalize to dissimilar targets. To test what a machine learning-based scoring function has learned, input attribution, a technique for learning which features are important to a model when making a prediction on a particular data point, can be applied. If a model successfully learns something beyond data set biases, attribution should give insight into the important binding interactions that are taking place. We built a machine learning-based scoring function that aimed to avoid the influence of bias via thorough train and test data set filtering and show that it achieves comparable performance on the Comparative Assessment of Scoring Functions, 2016 (CASF-2016) benchmark to other leading methods. We then use the CASF-2016 test set to perform attribution and find that the bonds identified as important by PointVS, unlike those extracted from other scoring functions, have a high correlation with those found by a distance-based interaction profiler. We then show that attribution can be used to extract important binding pharmacophores from a given protein target when supplied with a number of bound structures. We use this information to perform fragment elaboration and see improvements in docking scores compared to using structural information from a traditional, data-based approach. This not only provides definitive proof that the scoring function has learned to identify some important binding interactions but also constitutes the first deep learning-based method for extracting structural information from a target for molecule design.

RHO to the DOCK for GDP disembarking: Structural insights into the DOCK GTPase nucleotide exchange factors.

The human dedicator of cytokinesis (DOCK) family consists of 11 structurally conserved proteins that serve as atypical RHO guanine nucleotide exchange factors (RHO GEFs). These regulatory proteins act as mediators in numerous cellular cascades that promote cytoskeletal remodeling, playing roles in various crucial processes such as differentiation, migration, polarization, and axon growth in neurons. At the molecular level, DOCK DHR2 domains facilitate nucleotide dissociation from small GTPases, a process that is otherwise too slow for rapid spatiotemporal control of cellular signaling. Here, we provide an overview of the biological and structural characteristics for the various DOCK proteins and describe how they differ from other RHO GEFs and between DOCK subfamilies. The expression of the family varies depending on cell or tissue type, and they are consequently implicated in a broad range of disease phenotypes, particularly in the brain. A growing body of available structural information reveals the mechanism by which the catalytic DHR2 domain elicits nucleotide dissociation and also indicates strategies for the discovery and design of high-affinity small-molecule inhibitors. Such compounds could serve as chemical probes to interrogate the cellular function and provide starting points for drug discovery of this important class of enzymes.

Fragment Hotspot Mapping to Identify Selectivity-Determining Regions between Related Proteins.

Selectivity is a crucial property in small molecule development. Binding site comparisons within a protein family are a key piece of information when aiming to modulate the selectivity profile of a compound. Binding site differences can be exploited to confer selectivity for a specific target, while shared areas can provide insights into polypharmacology. As the quantity of structural data grows, automated methods are needed to process, summarize, and present these data to users. We present a computational method that provides quantitative and data-driven summaries of the available binding site information from an ensemble of structures of the same protein. The resulting ensemble maps identify the key interactions important for ligand binding in the ensemble. The comparison of ensemble maps of related proteins enables the identification of selectivity-determining regions within a protein family. We applied the method to three examples from the well-researched human bromodomain and kinase families, demonstrating that the method is able to identify selectivity-determining regions that have been used to introduce selectivity in past drug discovery campaigns. We then illustrate how the resulting maps can be used to automate comparisons across a target protein family.

SAMPL7 protein-ligand challenge: A community-wide evaluation of computational methods against fragment screening and pose-prediction.

A novel crystallographic fragment screening data set was generated and used in the SAMPL7 challenge for protein-ligands. The SAMPL challenges prospectively assess the predictive power of methods involved in computer-aided drug design. Application of various methods to fragment molecules are now widely used in the search for new drugs. However, there is little in the way of systematic validation specifically for fragment-based approaches. We have performed a large crystallographic high-throughput fragment screen against the therapeutically relevant second bromodomain of the Pleckstrin-homology domain interacting protein (PHIP2) that revealed 52 different fragments bound across 4 distinct sites, 47 of which were bound to the pharmacologically relevant acetylated lysine (Kac) binding site. These data were used to assess computational screening, binding pose prediction and follow-up enumeration. All submissions performed randomly for screening. Pose prediction success rates (defined as less than 2 Å root mean squared deviation against heavy atom crystal positions) ranged between 0 and 25% and only a very few follow-up compounds were deemed viable candidates from a medicinal-chemistry perspective based on a common molecular descriptors analysis. The tight deadlines imposed during the challenge led to a small number of submissions suggesting that the accuracy of rapidly responsive workflows remains limited. In addition, the application of these methods to reproduce crystallographic fragment data still appears to be very challenging. The results show that there is room for improvement in the development of computational tools particularly when applied to fragment-based drug design.

Crystallographic approach to fragment-based hit discovery against Schistosoma mansoni purine nucleoside phosphorylase.

Several Schistosoma species cause Schistosomiasis, an endemic disease in 78 countries that is ranked second amongst the parasitic diseases in terms of its socioeconomic impact and human health importance. The drug recommended for treatment by the WHO is praziquantel (PZQ), but there are concerns associated with PZQ, such as the lack of information about its exact mechanism of action, its high price, its effectiveness - which is limited to the parasite's adult form - and reports of resistance. The parasites lack the de novo purine pathway, rendering them dependent on the purine salvage pathway or host purine bases for nucleotide synthesis. Thus, the Schistosoma purine salvage pathway is an attractive target for the development of necessary and selective new drugs. In this study, the purine nucleotide phosphorylase II (PNP2), a new isoform of PNP1, was submitted to a high-throughput fragment-based hit discovery using a crystallographic screening strategy. PNP2 was crystallized and crystals were soaked with 827 fragments, a subset of the Maybridge 1000 library. X-ray diffraction data was collected and structures were solved. Out of 827-screened fragments we have obtained a total of 19 fragments that show binding to PNP2. Fourteen of these fragments bind to the active site of PNP2, while five were observed in three other sites. Here we present the first fragment screening against PNP2.

Data Set Augmentation Allows Deep Learning-Based Virtual Screening to Better Generalize to Unseen Target Classes and Highlight Important Binding Interactions.

Current deep learning methods for structure-based virtual screening take the structures of both the protein and the ligand as input but make little or no use of the protein structure when predicting ligand binding. Here, we show how a relatively simple method of data set augmentation forces such deep learning methods to take into account information from the protein. Models trained in this way are more generalizable (make better predictions on protein/ligand complexes from a different distribution to the training data). They also assign more meaningful importance to the protein and ligand atoms involved in binding. Overall, our results show that data set augmentation can help deep learning-based virtual screening to learn physical interactions rather than data set biases.

Ribosomal oxygenases are structurally conserved from prokaryotes to humans.

2-Oxoglutarate (2OG)-dependent oxygenases have important roles in the regulation of gene expression via demethylation of N-methylated chromatin components and in the hydroxylation of transcription factors and splicing factor proteins. Recently, 2OG-dependent oxygenases that catalyse hydroxylation of transfer RNA and ribosomal proteins have been shown to be important in translation relating to cellular growth, TH17-cell differentiation and translational accuracy. The finding that ribosomal oxygenases (ROXs) occur in organisms ranging from prokaryotes to humans raises questions as to their structural and evolutionary relationships. In Escherichia coli, YcfD catalyses arginine hydroxylation in the ribosomal protein L16; in humans, MYC-induced nuclear antigen (MINA53; also known as MINA) and nucleolar protein 66 (NO66) catalyse histidine hydroxylation in the ribosomal proteins RPL27A and RPL8, respectively. The functional assignments of ROXs open therapeutic possibilities via either ROX inhibition or targeting of differentially modified ribosomes. Despite differences in the residue and protein selectivities of prokaryotic and eukaryotic ROXs, comparison of the crystal structures of E. coli YcfD and Rhodothermus marinus YcfD with those of human MINA53 and NO66 reveals highly conserved folds and novel dimerization modes defining a new structural subfamily of 2OG-dependent oxygenases. ROX structures with and without their substrates support their functional assignments as hydroxylases but not demethylases, and reveal how the subfamily has evolved to catalyse the hydroxylation of different residue side chains of ribosomal proteins. Comparison of ROX crystal structures with those of other JmjC-domain-containing hydroxylases, including the hypoxia-inducible factor asparaginyl hydroxylase FIH and histone N(ε)-methyl lysine demethylases, identifies branch points in 2OG-dependent oxygenase evolution and distinguishes between JmjC-containing hydroxylases and demethylases catalysing modifications of translational and transcriptional machinery. The structures reveal that new protein hydroxylation activities can evolve by changing the coordination position from which the iron-bound substrate-oxidizing species reacts. This coordination flexibility has probably contributed to the evolution of the wide range of reactions catalysed by oxygenases.

Targeting the Small GTPase Superfamily through Their Regulatory Proteins.

The Ras superfamily of small GTPases are guanine-nucleotide-dependent switches essential for numerous cellular processes. Mutations or dysregulation of these proteins are associated with many diseases, but unsuccessful attempts to target the small GTPases directly have resulted in them being classed as "undruggable". The GTP-dependent signaling of these proteins is controlled by their regulators; guanine nucleotide exchange factors (GEFs), GTPase activating proteins (GAPs), and in the Rho and Rab subfamilies, guanine nucleotide dissociation inhibitors (GDIs). This review covers the recent small molecule and biologics strategies to target the small GTPases through their regulators. It seeks to critically re-evaluate recent chemical biology practice, such as the presence of PAINs motifs and the cell-based readout using compounds that are weakly potent or of unknown specificity. It highlights the vast scope of potential approaches for targeting the small GTPases in the future through their regulatory proteins.

An allosteric binding site of the α7 nicotinic acetylcholine receptor revealed in a humanized acetylcholine-binding protein.

Nicotinic acetylcholine receptors (nAChRs) belong to the family of pentameric ligand-gated ion channels and mediate fast excitatory transmission in the central and peripheral nervous systems. Among the different existing receptor subtypes, the homomeric α7 nAChR has attracted considerable attention because of its possible implication in several neurological and psychiatric disorders, including cognitive decline associated with Alzheimer's disease or schizophrenia. Allosteric modulators of ligand-gated ion channels are of particular interest as therapeutic agents, as they modulate receptor activity without affecting normal fluctuations of synaptic neurotransmitter release. Here, we used X-ray crystallography and surface plasmon resonance spectroscopy of α7-acetylcholine-binding protein (AChBP), a humanized chimera of a snail AChBP, which has 71% sequence similarity with the extracellular ligand-binding domain of the human α7 nAChR, to investigate the structural determinants of allosteric modulation. We extended previous observations that an allosteric site located in the vestibule of the receptor offers an attractive target for receptor modulation. We introduced seven additional humanizing mutations in the vestibule-located binding site of AChBP to improve its suitability as a model for studying allosteric binding. Using a fragment-based screening approach, we uncovered an allosteric binding site located near the ß8-ß9 loop, which critically contributes to coupling ligand binding to channel opening in human α7 nAChR. This work expands our understanding of the topology of allosteric binding sites in AChBP and, by extrapolation, in the human α7 nAChR as determined by electrophysiology measurements. Our insights pave the way for drug design strategies targeting nAChRs involved in ion channel-mediated disorders.

Rapid Covalent-Probe Discovery by Electrophile-Fragment Screening.

Covalent probes can display unmatched potency, selectivity, and duration of action; however, their discovery is challenging. In principle, fragments that can irreversibly bind their target can overcome the low affinity that limits reversible fragment screening, but such electrophilic fragments were considered nonselective and were rarely screened. We hypothesized that mild electrophiles might overcome the selectivity challenge and constructed a library of 993 mildly electrophilic fragments. We characterized this library by a new high-throughput thiol-reactivity assay and screened them against 10 cysteine-containing proteins. Highly reactive and promiscuous fragments were rare and could be easily eliminated. In contrast, we found hits for most targets. Combining our approach with high-throughput crystallography allowed rapid progression to potent and selective probes for two enzymes, the deubiquitinase OTUB2 and the pyrophosphatase NUDT7. No inhibitors were previously known for either. This study highlights the potential of electrophile-fragment screening as a practical and efficient tool for covalent-ligand discovery.

Construction of a Shape-Diverse Fragment Set: Design, Synthesis and Screen against Aurora-A Kinase.

Historically, chemists have explored chemical space in a highly uneven and unsystematic manner. As an example, the shape diversity of existing fragment sets does not generally reflect that of all theoretically possible fragments. To assess experimentally the added value of increased three dimensionality, a shape-diverse fragment set was designed and collated. The set was assembled by both using commercially available fragments and harnessing unified synthetic approaches to sp3 -rich molecular scaffolds. The resulting set of 80 fragments was highly three-dimensional, and its shape diversity was significantly enriched by twenty synthesised fragments. The fragment set was screened by high-throughput protein crystallography against Aurora-A kinase, revealing four hits that targeted the binding site of allosteric regulators. In the longer term, it is envisaged that the fragment set could be screened against a range of functionally diverse proteins, allowing the added value of more shape-diverse screening collections to be more fully assessed.

Structure-guided fragment-based drug discovery at the synchrotron: screening binding sites and correlations with hotspot mapping.

Structure-guided drug discovery emerged in the 1970s and 1980s, stimulated by the three-dimensional structures of protein targets that became available, mainly through X-ray crystal structure analysis, assisted by the development of synchrotron radiation sources. Structures of known drugs or inhibitors were used to guide the development of leads. The growth of high-throughput screening during the late 1980s and the early 1990s in the pharmaceutical industry of chemical libraries of hundreds of thousands of compounds of molecular weight of approximately 500 Da was impressive but still explored only a tiny fraction of the chemical space of the predicted 1040 drug-like compounds. The use of fragments with molecular weights less than 300 Da in drug discovery not only decreased the chemical space needing exploration but also increased promiscuity in binding targets. Here we discuss advances in X-ray fragment screening and the challenge of identifying sites where fragments not only bind but can be chemically elaborated while retaining their positions and binding modes. We first describe the analysis of fragment binding using conventional X-ray difference Fourier techniques, with Mycobacterium abscessus SAICAR synthetase (PurC) as an example. We observe that all fragments occupy positions predicted by computational hotspot mapping. We compare this with fragment screening at Diamond Synchrotron Light Source XChem facility using PanDDA software, which identifies many more fragment hits, only some of which bind to the predicted hotspots. Many low occupancy sites identified may not support elaboration to give adequate ligand affinity, although they will likely be useful in drug discovery as 'warm spots' for guiding elaboration of fragments bound at hotspots. We discuss implications of these observations for fragment screening at the synchrotron sources. This article is part of the theme issue 'Fifty years of synchrotron science: achievements and opportunities'.

Structure activity relationship studies on rhodanines and derived enethiol inhibitors of metallo-ß-lactamases.

Metallo-ß-lactamases (MBLs) enable bacterial resistance to almost all classes of ß-lactam antibiotics. We report studies on enethiol containing MBL inhibitors, which were prepared by rhodanine hydrolysis. The enethiols inhibit MBLs from different subclasses. Crystallographic analyses reveal that the enethiol sulphur displaces the di-Zn(II) ion bridging 'hydrolytic' water. In some, but not all, cases biophysical analyses provide evidence that rhodanine/enethiol inhibition involves formation of a ternary MBL enethiol rhodanine complex. The results demonstrate how low molecular weight active site Zn(II) chelating compounds can inhibit a range of clinically relevant MBLs and provide additional evidence for the potential of rhodanines to be hydrolysed to potent inhibitors of MBL protein fold and, maybe, other metallo-enzymes, perhaps contributing to the complex biological effects of rhodanines. The results imply that any medicinal chemistry studies employing rhodanines (and related scaffolds) as inhibitors should as a matter of course include testing of their hydrolysis products.

Structure and inhibitor specificity of the PCTAIRE-family kinase CDK16.

CDK16 (also known as PCTAIRE1 or PCTK1) is an atypical member of the cyclin-dependent kinase (CDK) family that has emerged as a key regulator of neurite outgrowth, vesicle trafficking and cancer cell proliferation. CDK16 is activated through binding to cyclin Y via a phosphorylation-dependent 14-3-3 interaction and has a unique consensus substrate phosphorylation motif compared with conventional CDKs. To elucidate the structure and inhibitor-binding properties of this atypical CDK, we screened the CDK16 kinase domain against different inhibitor libraries and determined the co-structures of identified hits. We discovered that the ATP-binding pocket of CDK16 can accommodate both type I and type II kinase inhibitors. The most potent CDK16 inhibitors revealed by cell-free and cell-based assays were the multitargeted cancer drugs dabrafenib and rebastinib. An inactive DFG-out binding conformation was confirmed by the first crystal structures of CDK16 in separate complexes with the inhibitors indirubin E804 and rebastinib, respectively. The structures revealed considerable conformational plasticity, suggesting that the isolated CDK16 kinase domain was relatively unstable in the absence of a cyclin partner. The unusual structural features and chemical scaffolds identified here hold promise for the development of more selective CDK16 inhibitors and provide opportunity to better characterise the role of CDK16 and its related CDK family members in various physiological and pathological contexts.

Human RECQ1 helicase-driven DNA unwinding, annealing, and branch migration: insights from DNA complex structures.

RecQ helicases are a widely conserved family of ATP-dependent motors with diverse roles in nearly every aspect of bacterial and eukaryotic genome maintenance. However, the physical mechanisms by which RecQ helicases recognize and process specific DNA replication and repair intermediates are largely unknown. Here, we solved crystal structures of the human RECQ1 helicase in complexes with tailed-duplex DNA and ssDNA. The structures map the interactions of the ssDNA tail and the branch point along the helicase and Zn-binding domains, which, together with reported structures of other helicases, define the catalytic stages of helicase action. We also identify a strand-separating pin, which (uniquely in RECQ1) is buttressed by the protein dimer interface. A duplex DNA-binding surface on the C-terminal domain is shown to play a role in DNA unwinding, strand annealing, and Holliday junction (HJ) branch migration. We have combined EM and analytical ultracentrifugation approaches to show that RECQ1 can form what appears to be a flat, homotetrameric complex and propose that RECQ1 tetramers are involved in HJ recognition. This tetrameric arrangement suggests a platform for coordinated activity at the advancing and receding duplexes of an HJ during branch migration.