Small molecule inhibitors of PCSK9. SAR investigations of head and amine groups.

Our previous work on the optimization of a new class of small molecule PCSK9 mRNA translation inhibitors focused on empirical optimization of the amide tail region of the lead PF-06446846 (1). This work resulted in compound 3 that showed an improved safety profile. We hypothesized that this improvement was related to diminished binding of 3 to non-translating ribosomes and an apparent improvement in transcript selectivity. Herein, we describe our efforts to further optimize this series of inhibitors through modulation of the heterocyclic head group and the amine fragment. Some of the effort was guided by an emerging cryo electron microscopy structure of the binding mode of 1 in the ribosome. These efforts led to the identification of 15 that was deemed suitable for evaluation in a humanized PCSK9 mouse model and a rat toxicology study. Compound 15 demonstrated a dose dependent reduction of plasma PCSK9 levels. The rat toxicological profile was not improved over that of 1, which precluded 15 from further consideration as a clinical candidate.

TorsionNet: A Deep Neural Network to Rapidly Predict Small-Molecule Torsional Energy Profiles with the Accuracy of Quantum Mechanics.

Fast and accurate assessment of small-molecule dihedral energetics is crucial for molecular design and optimization in medicinal chemistry. Yet, accurate prediction of torsion energy profiles remains challenging as the current molecular mechanics (MM) methods are limited by insufficient coverage of drug-like chemical space and accurate quantum mechanical (QM) methods are too expensive. To address this limitation, we introduce TorsionNet, a deep neural network (DNN) model specifically developed to predict small-molecule torsion energy profiles with QM-level accuracy. We applied active learning to identify nearly 50k fragments (with elements H, C, N, O, F, S, and Cl) that maximized the coverage of our corporate compound library and leveraged massively parallel cloud computing resources for density functional theory (DFT) torsion scans of these fragments, generating a training data set of 1.2 million DFT energies. After training TorsionNet on this data set, we obtain a model that can rapidly predict the torsion energy profile of typical drug-like fragments with DFT-level accuracy. Importantly, our method also provides an uncertainty estimate for the predicted profiles without any additional calculations. In this report, we show that TorsionNet can accurately identify the preferred dihedral geometries observed in crystal structures. Our TorsionNet-based analysis of a diverse set of protein-ligand complexes with measured binding affinity shows a strong association between high ligand strain and low potency. We also present practical applications of TorsionNet that demonstrate how consideration of DNN-based strain energy leads to substantial improvement in existing lead discovery and design workflows. TorsionNet500, a benchmark data set comprising 500 chemically diverse fragments with DFT torsion profiles (12k MM- and DFT-optimized geometries and energies), has been created and is made publicly available.

Crystal structure of human lysosomal acid lipase and its implications in cholesteryl ester storage disease.

Lysosomal acid lipase (LAL) is a serine hydrolase that hydrolyzes cholesteryl ester (CE) and TGs delivered to the lysosomes into free cholesterol and fatty acids. LAL deficiency due to mutations in the LAL gene (LIPA) results in accumulation of TGs and cholesterol esters in various tissues of the body leading to pathological conditions such as Wolman's disease and CE storage disease (CESD). Here, we present the first crystal structure of recombinant human LAL (HLAL) to 2.6 Å resolution in its closed form. The crystal structure was enabled by mutating three of the six potential glycosylation sites. The overall structure of HLAL closely resembles that of the evolutionarily related human gastric lipase (HGL). It consists of a core domain belonging to the classical α/ß hydrolase-fold family with a classical catalytic triad (Ser-153, His-353, Asp-324), an oxyanion hole, and a "cap" domain, which regulates substrate entry to the catalytic site. Most significant structural differences between HLAL and HGL exist at the lid region. Deletion of the short helix, 238NLCFLLC244, at the lid region implied a possible role in regulating the highly hydrophobic substrate binding site from self-oligomerization during interfacial activation. We also performed molecular dynamic simulations of dog gastric lipase (lid-open form) and HLAL to gain insights and speculated a possible role of the human mutant, H274Y, leading to CESD.

Generative Network Complex for the Automated Generation of Drug-like Molecules.

Current drug discovery is expensive and time-consuming. It remains a challenging task to create a wide variety of novel compounds that not only have desirable pharmacological properties but also are cheaply available to low-income people. In this work, we develop a generative network complex (GNC) to generate new drug-like molecules based on the multiproperty optimization via the gradient descent in the latent space of an autoencoder. In our GNC, both multiple chemical properties and similarity scores are optimized to generate drug-like molecules with desired chemical properties. To further validate the reliability of the predictions, these molecules are reevaluated and screened by independent 2D fingerprint-based predictors to come up with a few hundreds of new drug candidates. As a demonstration, we apply our GNC to generate a large number of new BACE1 inhibitors, as well as thousands of novel alternative drug candidates for eight existing market drugs, including Ceritinib, Ribociclib, Acalabrutinib, Idelalisib, Dabrafenib, Macimorelin, Enzalutamide, and Panobinostat.

Are 2D fingerprints still valuable for drug discovery?

Recently, molecular fingerprints extracted from three-dimensional (3D) structures using advanced mathematics, such as algebraic topology, differential geometry, and graph theory have been paired with efficient machine learning, especially deep learning algorithms to outperform other methods in drug discovery applications and competitions. This raises the question of whether classical 2D fingerprints are still valuable in computer-aided drug discovery. This work considers 23 datasets associated with four typical problems, namely protein-ligand binding, toxicity, solubility and partition coefficient to assess the performance of eight 2D fingerprints. Advanced machine learning algorithms including random forest, gradient boosted decision tree, single-task deep neural network and multitask deep neural network are employed to construct efficient 2D-fingerprint based models. Additionally, appropriate consensus models are built to further enhance the performance of 2D-fingerprint-based methods. It is demonstrated that 2D-fingerprint-based models perform as well as the state-of-the-art 3D structure-based models for the predictions of toxicity, solubility, partition coefficient and protein-ligand binding affinity based on only ligand information. However, 3D structure-based models outperform 2D fingerprint-based methods in complex-based protein-ligand binding affinity predictions.

Comprehensive Assessment of Torsional Strain in Crystal Structures of Small Molecules and Protein-Ligand Complexes using ab Initio Calculations.

The energetics of rotation around single bonds (torsions) is a key determinant of the three-dimensional shape that druglike molecules adopt in solution, the solid state, and in different biological environments, which in turn defines their unique physical and pharmacological properties. Therefore, accurate characterization of torsion angle preference and energetics is essential for the success of computational drug discovery and design. Here, we analyze torsional strain in crystal structures of druglike molecules in Cambridge structure database (CSD) and bioactive ligand conformations in protein data bank (PDB), expressing the total strain energy as a sum of strain energy from constituent rotatable bonds. We utilized cloud computing to generate torsion scan profiles of a very large collection of chemically diverse neutral fragments at DFT(B3LYP)/6-31G*//6-31G** or DFT(B3LYP)/6-31+G*//6-31+G** (for sulfur-containing molecule). With the data generated from these ab initio calculations, we performed rigorous analysis of strain due to deviation of observed torsion angles relative to their ideal gas-phase geometries. Contrary to the previous studies based on molecular mechanics, we find that in the crystalline state, molecules generally adopt low-strain conformations, with median per-torsion strain energy in CSD and PDB under one-tenth and one-third of a kcal/mol, respectively. However, for a small fraction (<5%) of motifs, external effects such as steric hindrance and hydrogen bonds result in strain penalty exceeding 2.5 kcal/mol. We find that due to poor quality of PDB structures in general, bioactive structures tend to have higher torsional strain compared to small-molecule crystal conformations. However, in the absence of structural fitting artifacts in PDB structures, protein-induced strain in bioactive conformations is quantitatively similar to those due to the packing forces in small-molecule crystal structures. This analysis allows us to establish strain energy thresholds to help identify biologically relevant conformers in a given ensemble. The work presented here is the most comprehensive study to date that demonstrates the utility and feasibility of gas-phase quantum mechanics (QM) calculations to study conformational preference and energetics of drug-size molecules. Potential applications of this study in computational lead discovery and structure-based design are discussed.

Receptor-Mediated Delivery of CRISPR-Cas9 Endonuclease for Cell-Type-Specific Gene Editing.

CRISPR-Cas RNA-guided endonucleases hold great promise for disrupting or correcting genomic sequences through site-specific DNA cleavage and repair. However, the lack of methods for cell- and tissue-selective delivery currently limits both research and clinical uses of these enzymes. We report the design and in vitro evaluation of S. pyogenes Cas9 proteins harboring asialoglycoprotein receptor ligands (ASGPrL). In particular, we demonstrate that the resulting ribonucleoproteins (Cas9-ASGPrL RNP) can be engineered to be preferentially internalized into cells expressing the corresponding receptor on their surface. Uptake of such fluorescently labeled proteins in liver-derived cell lines HEPG2 (ASGPr+) and SKHEP (control; diminished ASGPr) was studied by live cell imaging and demonstrates increased accumulation of Cas9-ASGPrL RNP in HEPG2 cells as a result of effective ASGPr-mediated endocytosis. When uptake occurred in the presence of a peptide with endosomolytic properties, we observed receptor-facilitated and cell-type specific gene editing that did not rely on electroporation or the use of transfection reagents. Overall, these in vitro results validate the receptor-mediated delivery of genome-editing enzymes as an approach for cell-selective gene editing and provide a framework for future potential applications to hepatoselective gene editing in vivo.

Induction of human cytochrome P450 3A4 by the irreversible myeloperoxidase inactivator PF-06282999 is mediated by the pregnane X receptor.

1. 2-(6-(5-Chloro-2-methoxyphenyl)-4-oxo-2-thioxo-3,4-dihydropyrimidin-1(2H)-yl) acetamide (PF-06282999) is a member of the thiouracil class of irreversible inactivators of human myeloperoxidase enzyme and a candidate for the treatment of cardiovascular disease. PF-06282999 is an inducer of CYP3A4 mRNA and midazolam-1'-hydroxylase activity in human hepatocytes, which is consistent with PF-06282999-dose dependent decreases in mean maximal plasma concentrations (Cmax) and area under the plasma concentration time curve (AUC) of midazolam in humans following 14-day treatment with PF-06282999. 2. In the present study, the biochemical mechanism(s) of CYP3A4 induction by PF-06282999 was studied. Incubations in reporter cells indicated that PF-06282999 selectively activated human pregnane X receptor (PXR). Treatment of human HepaRG cells with PF-06282999 led to ∼14-fold induction in CYP3A4 mRNA and 5-fold increase in midazolam-1'-hydroxylase activity, which was nullified in PXR-knock out HepaRG cells. TaqMan® gene expression analysis of human hepatocytes treated with PF-06282999 and the prototypical PXR agonist rifampin demonstrated increases in mRNA for CYP3A4 and related CYPs that are regulated by PXR. 3. Docking studies using a published human PXR crystal structure provided insights into the molecular basis for PXR activation by PF-06282999. Implementation of PXR transactivation assays in a follow-on discovery campaign should aid in the identification of back-up compounds devoid of PXR activation and CYP3A4 induction liability.

Efficient Liver Targeting by Polyvalent Display of a Compact Ligand for the Asialoglycoprotein Receptor.

A compact and stable bicyclic bridged ketal was developed as a ligand for the asialoglycoprotein receptor (ASGPR). This compound showed excellent ligand efficiency, and the molecular details of binding were revealed by the first X-ray crystal structures of ligand-bound ASGPR. This analogue was used to make potent di- and trivalent binders of ASGPR. Extensive characterization of the function of these compounds showed rapid ASGPR-dependent cellular uptake in vitro and high levels of liver/plasma selectivity in vivo. Assessment of the biodistribution in rodents of a prototypical Alexa647-labeled trivalent conjugate showed selective hepatocyte targeting with no detectable distribution in nonparenchymal cells. This molecule also exhibited increased ASGPR-directed hepatocellular uptake and prolonged retention compared to a similar GalNAc derived trimer conjugate. Selective release in the liver of a passively permeable small-molecule cargo was achieved by retro-Diels-Alder cleavage of an oxanorbornadiene linkage, presumably upon encountering intracellular thiol. Therefore, the multicomponent construct described here represents a highly efficient delivery vehicle to hepatocytes.

Development of a selective activity-based probe for glycosylated LIPA.

Loss of LIPA activity leads to diseases such as Wolman's Disease and Cholesterol Ester Storage Disease. While it is possible to measure defects in LIPA protein levels, it is difficult to directly measure LIPA activity in cells. In order to measure LIPA activity directly we developed a LIPA specific activity based probe. LIPA is heavily glycosylated although it is unclear how glycosylation affects LIPA activity or function. Our probe is specific for a glycosylated form of LIPA in cells, although it labels purified LIPA regardless of glycosylation.

Expression and functional characterization of human lysosomal acid lipase gene (LIPA) mutation responsible for cholesteryl ester storage disease (CESD) phenotype.

Lysosomal acid lipase (LAL) is a serine hydrolase which hydrolyzes cholesteryl ester and triglycerides delivered to the lysosomes into free cholesterol and free fatty acids. Mutations in the LAL gene (LIPA) result in accumulation of triglycerides and cholesterol esters in various tissues of the body, leading to pathological conditions such as Wolman's disease (WD) and cholesteryl ester storage disease (CESD). CESD patients homozygous for His295Tyr (H295Y) mutation have less than 5% of normal LAL activity. To shed light on the molecular basis for this loss-of-function phenotype, we have generated the recombinant H295Y enzyme and studied its biophysical and biochemical properties. No significant differences were observed in the expression levels or glycosylation patterns between the mutant and the wild type LAL. However, the H295Y mutant displayed only residual enzymatic activity (<5%) compared to the wild type. While wild type LAL is mostly a monomer at pH 5.0, the vast majority H295Y exists as a high molecular soluble aggregate. Besides, the H295Y mutant has a 20°C lower melting temperature compared to the wild type. Transient expression studies in WD fibroblasts showed that mutation of His295 to other amino acids resulted in a significant loss of enzymatic activity. A homology model of LAL revealed that His295 is located on an α-helix of the cap domain and could be important for tethering it to its core domain. The observed loss-of-function phenotype in CESD patients might arise from a combination of protein destabilization and the shift to a non-functional soluble aggregate.

Discovery of spirocyclic-diamine inhibitors of mammalian acetyl CoA-carboxylase.

A novel series of spirocyclic-diamine based, isoform non-selective inhibitors of acetyl-CoA carboxylase (ACC) is described. These spirodiamine derivatives were discovered by design of a library to mimic the structural rigidity and hydrogen-bonding pattern observed in the co-crystal structure of spirochromanone inhibitor I. The lead compound 3.5.1 inhibited de novo lipogenesis in rat hepatocytes, with an IC50 of 0.30 µM.

Toward a unified model of passive drug permeation II: the physiochemical determinants of unbound tissue distribution with applications to the design of hepatoselective glucokinase activators.

In this work, we leverage a mathematical model of the underlying physiochemical properties of tissues and physicochemical properties of molecules to support the development of hepatoselective glucokinase activators. Passive distribution is modeled via a Fick-Nernst-Planck approach, using in vitro experimental data to estimate the permeability of both ionized and neutral species. The model accounts for pH and electrochemical potential across cellular membranes, ionization according to Henderson-Hasselbalch, passive permeation of the neutral species using Fick's law, and passive permeation of the ionized species using the Nernst-Planck equation. The mathematical model of the physiochemical system allows derivation of a single set of parameters governing the distribution of drug molecules across multiple conditions both in vitro and in vivo. A case study using this approach in the development of hepatoselective glucokinase activators via organic anion-transporting polypeptide-mediated hepatic uptake and impaired passive distribution to the pancreas is described. The results for these molecules indicate the permeability penalty of the ionized form is offset by its relative abundance, leading to passive pancreatic exclusion according to the Nernst-Planck extension of Fickian passive permeation. Generally, this model serves as a useful construct for drug discovery scientists to understand subcellular exposure of acids or bases using specific physiochemical properties.

Identification of a novel conformationally constrained glucagon receptor antagonist.

Identification of orally active, small molecule antagonists of the glucagon receptor represents a novel treatment paradigm for the management of type 2 diabetes mellitus. The present work discloses novel glucagon receptor antagonists, identified via conformational constraint of current existing literature antagonists. Optimization of lipophilic ligand efficiency (LLE or LipE) culminated in enantiomers (+)-trans-26 and (-)-trans-27 which exhibit good physicochemical and in vitro drug metabolism profiles. In vivo, significant pharmacokinetic differences were noted with the two enantiomers, which were primarily driven through differences in clearance rates. Enantioselective oxidation by cytochrome P450 was ruled out as a causative factor for pharmacokinetic differences.

The design and synthesis of a potent glucagon receptor antagonist with favorable physicochemical and pharmacokinetic properties as a candidate for the treatment of type 2 diabetes mellitus.

A novel and potent small molecule glucagon receptor antagonist for the treatment of diabetes mellitus is reported. This candidate, (S)-3-[4-(1-{3,5-dimethyl-4-[4-(trifluoromethyl)-1H-pyrazol-1-yl]phenoxy}butyl)benzamido]propanoic acid, has lower molecular weight and lipophilicity than historical glucagon receptor antagonists, resulting in excellent selectivity in broad-panel screening, lower cytotoxicity, and excellent overall in vivo safety in early pre-clinical testing. Additionally, it displays low in vivo clearance and excellent oral bioavailability in both rats and dogs. In a rat glucagon challenge model, it was shown to reduce the glucagon-elicited glucose excursion in a dose-dependent manner and at a concentration consistent with its rat in vitro potency. Its properties make it an excellent candidate for further investigation.

Pyrimidone-based series of glucokinase activators with alternative donor-acceptor motif.

Glucokinase activators are a class of experimental agents under investigation as a therapy for Type 2 diabetes mellitus. An X-ray crystal structure of a modestly potent agent revealed the potential to substitute the common heterocyclic amide donor-acceptor motif for a pyridone moiety. We have successfully demonstrated that both pyridone and pyrimidone heterocycles can be used as a potent donor-acceptor substituent. Several sub-micromolar analogs that possess the desired partial activator profile were synthesized and characterized. Unfortunately, the most potent activators suffered from sub-optimal pharmacokinetic properties. Nonetheless, these donor-acceptor motifs may find utility in other glucokinase activator series or beyond.

Identification of potent, selective, CNS-targeted inverse agonists of the ghrelin receptor.

The optimization for selectivity and central receptor occupancy for a series of spirocyclic azetidine-piperidine inverse agonists of the ghrelin receptor is described. Decreased mAChR muscarinic M2 binding was achieved by use of a chiral indane in place of a substituted benzylic group. Compounds with desirable balance of human in vitro clearance and ex vivo central receptor occupancy were discovered by incorporation of heterocycles. Specifically, heteroaryl rings with nitrogen(s) vicinal to the indane linkage provided the most attractive overall properties.

Structure-guided inhibitor design for human acetyl-coenzyme A carboxylase by interspecies active site conversion.

Inhibition of acetyl-CoA carboxylases (ACCs), a crucial enzyme for fatty acid metabolism, has been shown to promote fatty acid oxidation and reduce body fat in animal models. Therefore, ACCs are attractive targets for structure-based inhibitor design, particularly the carboxyltransferase (CT) domain, which is the primary site for inhibitor interaction. We have cloned, expressed, and purified the CT domain of human ACC2 using baculovirus-mediated insect cell expression system. However, attempts to crystallize the human ACC2 CT domain have not been successful in our hands. Hence, we have been using the available crystal structure of yeast CT domain to design human ACC inhibitors. Unfortunately, as the selectivity of the lead series has increased against the full-length human enzyme, the potency against the yeast enzyme has decreased significantly. This loss of potency against the yeast enzyme correlated with a complete lack of binding of the human-specific compounds to crystals of the yeast CT domain. Here, we address this problem by converting nine key active site residues of the yeast CT domain to the corresponding human residues. The resulting humanized yeast ACC-CT (yCT-H9) protein exhibits biochemical and biophysical properties closer to the human CT domain and binding to human specific compounds. We report high resolution crystal structures of yCT-H9 complexed with inhibitors that show a preference for the human CT domain. These structures offer insights that explain the species selectivity of ACC inhibitors and may guide future drug design programs.

A novel series of glucagon receptor antagonists with reduced molecular weight and lipophilicity.

A novel series of glucagon receptor antagonists has been discovered. These pyrazole ethers and aminopyrazoles have lower molecular weight and increased polarity such that the molecules fall into better drug-like property space. This work has culminated in compounds 44 and 50 that were shown to have good pharmacokinetic attributes in dog, in contrast to rats, in which clearance was high; and compound 49, which demonstrated a dose-dependent reduction in glucose excursion in a rat glucagon challenge experiment.

The design and synthesis of indazole and pyrazolopyridine based glucokinase activators for the treatment of type 2 diabetes mellitus.

Glucokinase activators represent a promising potential treatment for patients with Type 2 diabetes. Herein, we report the identification and optimization of a series of novel indazole and pyrazolopyridine based activators leading to the identification of 4-(6-(azetidine-1-carbonyl)-5-fluoropyridin-3-yloxy)-2-ethyl-N-(5-methylpyrazin-2-yl)-2H-indazole-6-carboxamide (42) as a potent activator with favorable preclinical pharmacokinetic properties and in vivo efficacy.