Artificial intelligence in the prediction of protein-ligand interactions: recent advances and future directions.

New drug production, from target identification to marketing approval, takes over 12 years and can cost around $2.6 billion. Furthermore, the COVID-19 pandemic has unveiled the urgent need for more powerful computational methods for drug discovery. Here, we review the computational approaches to predicting protein-ligand interactions in the context of drug discovery, focusing on methods using artificial intelligence (AI). We begin with a brief introduction to proteins (targets), ligands (e.g. drugs) and their interactions for nonexperts. Next, we review databases that are commonly used in the domain of protein-ligand interactions. Finally, we survey and analyze the machine learning (ML) approaches implemented to predict protein-ligand binding sites, ligand-binding affinity and binding pose (conformation) including both classical ML algorithms and recent deep learning methods. After exploring the correlation between these three aspects of protein-ligand interaction, it has been proposed that they should be studied in unison. We anticipate that our review will aid exploration and development of more accurate ML-based prediction strategies for studying protein-ligand interactions.

Novel Fragment Inhibitors of PYCR1 from Docking-Guided X-ray Crystallography.

The proline biosynthetic enzyme Δ1-pyrroline-5-carboxylate (P5C) reductase 1 (PYCR1) is one of the most consistently upregulated enzymes across multiple cancer types and central to the metabolic rewiring of cancer cells. Herein, we describe a fragment-based, structure-first approach to the discovery of PYCR1 inhibitors. Thirty-seven fragment-like carboxylic acids in the molecular weight range of 143-289 Da were selected from docking and then screened using X-ray crystallography as the primary assay. Strong electron density was observed for eight compounds, corresponding to a crystallographic hit rate of 22%. The fragments are novel compared to existing proline analog inhibitors in that they block both the P5C substrate pocket and the NAD(P)H binding site. Four hits showed inhibition of PYCR1 in kinetic assays, and one has lower apparent IC50 than the current best proline analog inhibitor. These results show proof-of-concept for our inhibitor discovery approach and provide a basis for fragment-to-lead optimization.

Expression and kinetic characterization of PYCR3.

PYCRs are proline biosynthetic enzymes that catalyze the NAD(P)H-dependent reduction of Δ1-pyrroline-5-carboxylate (P5C) to proline in humans. PYCRs - especially PYCR1 - are upregulated in many types of cancers and have been implicated in the altered metabolism of cancer cells. Of the three isoforms of PYCR, PYCR3 remains the least studied due in part to the lack of a robust recombinant expression. Herein, we describe a procedure for the expression of soluble SUMO-PYCR3 in Escherichia coli, purification of the fusion protein, and removal of the SUMO tag. PYCR3 is active with either NADPH or NADH as the coenzyme. Bi-substrate kinetic measurements obtained by varying the concentrations of both L-P5C and NADH, along with product inhibition data for l-proline, suggest a random ordered bi bi mechanism. A panel of 19 proline analogs was screened for inhibition, and the kinetics of competitive inhibition (with L-P5C) were measured for five of the compounds screened, including N-formyl-l-proline, a validated inhibitor of PYCR1. N-formyl-l-proline was found to be ten times more selective for PYCR1 over PYCR3. The SUMO-PYCR3 expression system should be useful for testing the isoform specificity of PYCR1 inhibitors.

Kinetic and Structural Characterization of a Flavin-Dependent Putrescine N-Hydroxylase from Acinetobacter baumannii.

Acinetobacter baumannii is a Gram-negative opportunistic pathogen that causes nosocomial infections, especially among immunocompromised individuals. The rise of multidrug resistant strains of A. baumannii has limited the use of standard antibiotics, highlighting a need for new drugs that exploit novel mechanisms of pathogenicity. Disrupting iron acquisition by inhibiting the biosynthesis of iron-chelating molecules (siderophores) secreted by the pathogen is a potential strategy for developing new antibiotics. Here we investigated FbsI, an N-hydroxylating monooxygenase involved in the biosynthesis of fimsbactin A, the major siderophore produced by A. baumannii. FbsI was characterized using steady-state and transient-state kinetics, spectroscopy, X-ray crystallography, and small-angle X-ray scattering. FbsI was found to catalyze the N-hydroxylation of the aliphatic diamines putrescine and cadaverine. Maximum coupling of the reductive and oxidative half-reactions occurs with putrescine, suggesting it is the preferred (in vivo) substrate. FbsI uses both NADPH and NADH as the reducing cofactor with a slight preference for NADPH. The crystal structure of FbsI complexed with NADP+ was determined at 2.2 Å resolution. The structure exhibits the protein fold characteristic of Class B flavin-dependent monooxygenases. FbsI is most similar in 3D structure to the cadaverine N-hydroxylases DesB and DfoA. Small-angle X-ray scattering shows that FbsI forms a tetramer in solution like the N-hydroxylating monooxygenases of the SidA/IucD/PvdA family. A model of putrescine docked into the active site provides insight into substrate recognition. A mechanism for the catalytic cycle is proposed where dehydration of the C4a-hydroxyflavin intermediate is partially rate-limiting, and the hydroxylated putrescine product is released before NADP+.

Structure-affinity relationships of reversible proline analog inhibitors targeting proline dehydrogenase.

Proline dehydrogenase (PRODH) catalyzes the first step of proline catabolism, the FAD-dependent oxidation of L-proline to Δ1-pyrroline-5-carboxylate. PRODH plays a central role in the metabolic rewiring of cancer cells, which has motivated the discovery of inhibitors. Here, we studied the inhibition of PRODH by 18 proline-like compounds to understand the structural and chemical features responsible for the affinity of the best-known inhibitor, S-(-)-tetrahydro-2-furoic acid (1). The compounds were screened, and then six were selected for more thorough kinetic analysis: cyclobutane-1,1-dicarboxylic acid (2), cyclobutanecarboxylic acid (3), cyclopropanecarboxylic acid (4), cyclopentanecarboxylic acid (16), 2-oxobutyric acid (17), and (2S)-oxetane-2-carboxylic acid (18). These compounds are competitive inhibitors with inhibition constants in the range of 1.4-6 mM, compared to 0.3 mM for 1. Crystal structures of PRODH complexed with 2, 3, 4, and 18 were determined. All four inhibitors bind in the proline substrate site, but the orientations of their rings differ from that of 1. The binding of 3 and 18 is accompanied by compression of the active site to enable nonpolar contacts with Leu513. Compound 2 is unique in that the additional carboxylate displaces a structurally conserved water molecule from the active site. Compound 18 also destabilizes the conserved water, but by an unexpected non-steric mechanism. The results are interpreted using a chemical double mutant thermodynamic cycle. This analysis revealed unanticipated synergism between ring size and hydrogen bonding to the conserved water. These structure-affinity relationships provide new information relevant to the development of new inhibitor design strategies targeting PRODH.

Conformational Preferences of Pyridone Adenine Dinucleotides from Molecular Dynamics Simulations.

Pyridone adenine dinucleotides (ox-NADs) are redox inactive derivatives of the enzyme cofactor and substrate nicotinamide adenine dinucleotide (NAD) that have a carbonyl group at the C2, C4, or C6 positions of the nicotinamide ring. These aberrant cofactor analogs accumulate in cells under stress and are potential inhibitors of enzymes that use NAD(H). We studied the conformational landscape of ox-NADs in solution using molecular dynamics simulations. Compared to NAD+ and NADH, 2-ox-NAD and 4-ox-NAD have an enhanced propensity for adopting the anti conformation of the pyridone ribose group, whereas 6-ox-NAD exhibits greater syn potential. Consequently, 2-ox-NAD and 4-ox-NAD have increased preference for folding into compact conformations, whereas 6-ox-NAD is more extended. ox-NADs have distinctive preferences for the orientation of the pyridone amide group, which are driven by intramolecular hydrogen bonding and steric interactions. These conformational preferences are compared to those of protein-bound NAD(H). Our results may help in identifying enzymes targeted by ox-NADs.

Evidence for Proline Catabolic Enzymes in the Metabolism of Thiazolidine Carboxylates.

Thiazolidine carboxylates such as thiazolidine-4-carboxylate (T4C) and thiazolidine-2-carboxylate (T2C) are naturally occurring sulfur analogues of proline. These compounds have been observed to have both beneficial and toxic effects in cells. Given that proline dehydrogenase has been proposed to be a key enzyme in the oxidative metabolism of thioprolines, we characterized T4C and T2C as substrates of proline catabolic enzymes using proline utilization A (PutA), which is a bifunctional enzyme with proline dehydrogenase (PRODH) and l-glutamate-γ-semialdehyde dehydrogenase (GSALDH) activities. PutA is shown here to catalyze the FAD-dependent PRODH oxidation of both T4C and T2C with catalytic efficiencies significantly higher than with proline. Stopped-flow experiments also demonstrate that l-T4C and l-T2C reduce PutA-bound FAD at rates faster than proline. Unlike proline, however, oxidation of T4C and T2C does not generate a substrate for NAD+-dependent GSALDH. Instead, PutA/PRODH oxidation of T4C leads to cysteine formation, whereas oxidation of T2C generates an apparently stable Δ4-thiazoline-2-carboxylate species. Our results provide new insights into the metabolism of T2C and T4C.

Biochemical Characterization of the Two-Component Flavin-Dependent Monooxygenase Involved in Valanimycin Biosynthesis.

The flavin reductase (FRED) and isobutylamine N-hydroxylase (IBAH) from Streptomyces viridifaciens constitute a two-component, flavin-dependent monooxygenase system that catalyzes the first step in valanimycin biosynthesis. FRED is an oxidoreductase that provides the reduced flavin to IBAH, which then catalyzes the hydroxylation of isobutylamine (IBA) to isobutylhydroxylamine (IBHA). In this work, we used several complementary methods to investigate FAD binding, steady-state and rapid reaction kinetics, and enzyme-enzyme interactions in the FRED:IBAH system. The affinity of FRED for FADox is higher than its affinity for FADred, consistent with its function as a flavin reductase. Conversely, IBAH binds FADred more tightly than FADox, consistent with its role as a monooxygenase. FRED exhibits a strong preference (28-fold) for NADPH over NADH as the electron source for FAD reduction. Isothermal titration calorimetry was used to study the association of FRED and IBAH. In the presence of FAD, either oxidized or reduced, FRED and IBAH associate with a dissociation constant of 7-8 µM. No interaction was observed in the absence of FAD. These results are consistent with the formation of a protein-protein complex for direct transfer of reduced flavin from the reductase to the monooxygenase in this two-component system.

Impaired folate binding of serine hydroxymethyltransferase 8 from soybean underlies resistance to the soybean cyst nematode.

Management of the agricultural pathogen soybean cyst nematode (SCN) relies on the use of SCN-resistant soybean cultivars, a strategy that has been failing in recent years. An underutilized source of resistance in the soybean genotype Peking is linked to two polymorphisms in serine hydroxy-methyltransferase 8 (SHMT8). SHMT is a pyridoxal 5'-phosphate-dependent enzyme that converts l-serine and (6S)-tetrahydrofolate to glycine and 5,10-methylenetetrahydrofolate. Here, we determined five crystal structures of the 1884-residue SHMT8 tetramers from the SCN-susceptible cultivar (cv.) Essex and the SCN-resistant cv. Forrest (whose resistance is derived from the SHMT8 polymorphisms in Peking); the crystal structures were determined in complex with various ligands at 1.4-2.35 Å resolutions. We find that the two Forrest-specific polymorphic substitutions (P130R and N358Y) impact the mobility of a loop near the entrance of the (6S)-tetrahydrofolate-binding site. Ligand-binding and kinetic studies indicate severely reduced affinity for folate and dramatically impaired enzyme activity in Forrest SHMT8. These findings imply widespread effects on folate metabolism in soybean cv. Forrest that have implications for combating the widespread increase in virulent SCN.

Trapping conformational states of a flavin-dependent N-monooxygenase in crystallo reveals protein and flavin dynamics.

The siderophore biosynthetic enzyme A (SidA) ornithine hydroxylase from Aspergillus fumigatus is a fungal disease drug target involved in the production of hydroxamate-containing siderophores, which are used by the pathogen to sequester iron. SidA is an N-monooxygenase that catalyzes the NADPH-dependent hydroxylation of l-ornithine through a multistep oxidative mechanism, utilizing a C4a-hydroperoxyflavin intermediate. Here we present four new crystal structures of SidA in various redox and ligation states, including the first structure of oxidized SidA without NADP(H) or l-ornithine bound (resting state). The resting state structure reveals a new out active site conformation characterized by large rotations of the FAD isoalloxazine around the C1-'C2' and N10-C1' bonds, coupled to a 10-Å movement of the Tyr-loop. Additional structures show that either flavin reduction or the binding of NADP(H) is sufficient to drive the FAD to the in conformation. The structures also reveal protein conformational changes associated with the binding of NADP(H) and l-ornithine. Some of these residues were probed using site-directed mutagenesis. Docking was used to explore the active site of the out conformation. These calculations identified two potential ligand-binding sites. Altogether, our results provide new information about conformational dynamics in flavin-dependent monooxygenases. Understanding the different active site conformations that appear during the catalytic cycle may allow fine-tuning of inhibitor discovery efforts.

In crystallo screening for proline analog inhibitors of the proline cycle enzyme PYCR1.

Pyrroline-5-carboxylate reductase 1 (PYCR1) catalyzes the biosynthetic half-reaction of the proline cycle by reducing Δ1-pyrroline-5-carboxylate (P5C) to proline through the oxidation of NAD(P)H. Many cancers alter their proline metabolism by up-regulating the proline cycle and proline biosynthesis, and knockdowns of PYCR1 lead to decreased cell proliferation. Thus, evidence is growing for PYCR1 as a potential cancer therapy target. Inhibitors of cancer targets are useful as chemical probes for studying cancer mechanisms and starting compounds for drug discovery; however, there is a notable lack of validated inhibitors for PYCR1. To fill this gap, we performed a small-scale focused screen of proline analogs using X-ray crystallography. Five inhibitors of human PYCR1 were discovered: l-tetrahydro-2-furoic acid, cyclopentanecarboxylate, l-thiazolidine-4-carboxylate, l-thiazolidine-2-carboxylate, and N-formyl l-proline (NFLP). The most potent inhibitor was NFLP, which had a competitive (with P5C) inhibition constant of 100 µm The structure of PYCR1 complexed with NFLP shows that inhibitor binding is accompanied by conformational changes in the active site, including the translation of an α-helix by 1 Å. These changes are unique to NFLP and enable additional hydrogen bonds with the enzyme. NFLP was also shown to phenocopy the PYCR1 knockdown in MCF10A H-RASV12 breast cancer cells by inhibiting de novo proline biosynthesis and impairing spheroidal growth. In summary, we generated the first validated chemical probe of PYCR1 and demonstrated proof-of-concept for screening proline analogs to discover inhibitors of the proline cycle.

Structure and function of a flavin-dependent S-monooxygenase from garlic (Allium sativum).

Allicin is a component of the characteristic smell and flavor of garlic (Allium sativum). A flavin-containing monooxygenase (FMO) produced by A. sativum (AsFMO) was previously proposed to oxidize S-allyl-l-cysteine (SAC) to alliin, an allicin precursor. Here, we present a kinetic and structural characterization of AsFMO that suggests a possible contradiction to this proposal. Results of steady-state kinetic analyses revealed that AsFMO exhibited negligible activity with SAC; however, the enzyme was highly active with l-cysteine, N-acetyl-l-cysteine, and allyl mercaptan. We found that allyl mercaptan with NADPH was the preferred substrate-cofactor combination. Rapid-reaction kinetic analyses showed that NADPH binds tightly (KD of ∼2 µm) to AsFMO and that the hydride transfer occurs with pro-R stereospecificity. We detected the formation of a long-wavelength band when AsFMO was reduced by NADPH, probably representing the formation of a charge-transfer complex. In the absence of substrate, the reduced enzyme, in complex with NADP+, reacted with oxygen and formed an intermediate with a spectrum characteristic of C4a-hydroperoxyflavin, which decays several orders of magnitude more slowly than the kcat The presence of substrate enhanced C4a-hydroperoxyflavin formation and, upon hydroxylation, oxidation occurred with a rate constant similar to the kcat The structure of AsFMO complexed with FAD at 2.08-Å resolution features two domains for binding of FAD and NADPH, representative of class B flavin monooxygenases. These biochemical and structural results are consistent with AsFMO being an S-monooxygenase involved in allicin biosynthesis through direct formation of sulfenic acid and not SAC oxidation.

Structural analysis of prolines and hydroxyprolines binding to the l-glutamate-γ-semialdehyde dehydrogenase active site of bifunctional proline utilization A.

Proline utilization A (PutA) proteins are bifunctional proline catabolic enzymes that catalyze the 4-electron oxidation of l-proline to l-glutamate using spatially-separated proline dehydrogenase and l-glutamate-γ-semialdehyde dehydrogenase (GSALDH, a.k.a. ALDH4A1) active sites. The observation that l-proline inhibits both the GSALDH activity of PutA and monofunctional GSALDHs motivated us to study the inhibition of PutA by proline stereoisomers and analogs. Here we report five high-resolution crystal structures of PutA with the following ligands bound in the GSALDH active site: d-proline, trans-4-hydroxy-d-proline, cis-4-hydroxy-d-proline, l-proline, and trans-4-hydroxy-l-proline. Three of the structures are of ternary complexes of the enzyme with an inhibitor and either NAD+ or NADH. To our knowledge, the NADH complex is the first for any GSALDH. The structures reveal a conserved mode of recognition of the inhibitor carboxylate, which results in the pyrrolidine rings of the d- and l-isomers having different orientations and different hydrogen bonding environments. Activity assays show that the compounds are weak inhibitors with millimolar inhibition constants. Curiously, although the inhibitors occupy the aldehyde binding site, kinetic measurements show the inhibition is uncompetitive. Uncompetitive inhibition may involve proline binding to a remote site or to the enzyme-NADH complex. Together, the structural and kinetic data expand our understanding of how proline-like molecules interact with GSALDH, reveal insight into the relationship between stereochemistry and inhibitor affinity, and demonstrate the pitfalls of inferring the mechanism of inhibition from crystal structures alone.

Probing the function of a ligand-modulated dynamic tunnel in bifunctional proline utilization A (PutA).

In many bacteria, the reactions of proline catabolism are catalyzed by the bifunctional enzyme known as proline utilization A (PutA). PutA catalyzes the two-step oxidation of l-proline to l-glutamate using distinct proline dehydrogenase (PRODH) and l-glutamate-γ-semialdehyde dehydrogenase (GSALDH) active sites, which are separated by over 40 Å and connected by a complex tunnel system. The tunnel system consists of a main tunnel that connects the two active sites and functions in substrate channeling, plus six ancillary tunnels whose functions are unknown. Here we used tunnel-blocking mutagenesis to probe the role of a dynamic ancillary tunnel (tunnel 2a) whose shape is modulated by ligand binding to the PRODH active site. The 1.90 Å resolution crystal structure of Geobacter sulfurreducens PutA variant A206W verified that the side chain of Trp206 cleanly blocks tunnel 2a without perturbing the surrounding structure. Steady-state kinetic measurements indicate the mutation impaired PRODH activity without affecting the GSALDH activity. Single-turnover experiments corroborated a severe impairment of PRODH activity with flavin reduction decreased by nearly 600-fold in A206W relative to wild-type. Substrate channeling is also significantly impacted as A206W exhibited a 3000-fold lower catalytic efficiency in coupled PRODH-GSALDH activity assays, which measure NADH formation as a function of proline. The structure suggests that Trp206 inhibits binding of the substrate l-proline by preventing the formation of a conserved glutamate-arginine ion pair and closure of the PRODH active site. Our data are consistent with tunnel 2a serving as an open space through which the glutamate of the ion pair travels during the opening and closing of the active site in response to binding l-proline. These results confirm the essentiality of the conserved ion pair in binding l-proline and support the hypothesis that the ion pair functions as a gate that controls access to the PRODH active site.

Disease variants of human Δ1-pyrroline-5-carboxylate reductase 2 (PYCR2).

Pyrroline-5-carboxylate reductase (PYCR in humans) catalyzes the final step of l-proline biosynthesis by catalyzing the reduction of L-Δ1-pyrroline-5-carboxylate (L-P5C) to l-proline using NAD(P)H as the hydride donor. In humans, three isoforms PYCR1, PYCR2, and PYCR3 are known. Recent genome-wide association and clinical studies have revealed that homozygous mutations in human PYCR2 lead to postnatal microcephaly and hypomyelination, including hypomyelinating leukodystrophy type 10. To uncover biochemical and structural insights into human PYCR2, we characterized the steady-state kinetics of the wild-type enzyme along with two protein variants, Arg119Cys and Arg251Cys, that were previously identified in patients with microcephaly and hypomyelination. Kinetic measurements with PYCR2 suggest a sequential binding mechanism with L-P5C binding before NAD(P)H and NAD(P)+ releasing before L-Pro. Both disease-related variants are catalytically impaired. Depending on whether NADPH or NADH was used, the catalytic efficiency of the R119C protein variant was 40 or 366 times lower than that of the wild-type enzyme, while the catalytic efficiency of the R251C protein variant was 7 or 26 times lower than that of the wild-type enzyme. In addition, thermostability and circular dichroism measurements suggest that the R251C protein variant has a pronounced folding defect. These results are consistent with the involvement of Arg119Cys and Arg251Cys in disease pathology.

Structure, biochemistry, and gene expression patterns of the proline biosynthetic enzyme pyrroline-5-carboxylate reductase (PYCR), an emerging cancer therapy target.

Proline metabolism features prominently in the unique metabolism of cancer cells. Proline biosynthetic genes are consistently upregulated in multiple cancers, while the proline catabolic enzyme proline dehydrogenase has dual, context-dependent pro-cancer and pro-apoptotic functions. Furthermore, the cycling of proline and Δ1-pyrroline-5-carboxylate through the proline cycle impacts cellular growth and death pathways by maintaining redox homeostasis between the cytosol and mitochondria. Here we focus on the last enzyme of proline biosynthesis, Δ1-pyrroline-5-carboxylate reductase, known as PYCR in humans. PYCR catalyzes the NAD(P)H-dependent reduction of Δ1-pyrroline-5-carboxylate to proline and forms the reductive half of the proline metabolic cycle. We review the research on the three-dimensional structure, biochemistry, inhibition, and cancer biology of PYCR. To provide a global view of PYCR gene upregulation in cancer, we mined RNA transcript databases to analyze differential gene expression in 28 cancer types. This analysis revealed strong, widespread upregulation of PYCR genes, especially PYCR1. Altogether, the research over the past 20 years makes a compelling case for PYCR as a cancer therapy target. We conclude with a discussion of some of the major challenges for the field, including developing isoform-specific inhibitors, elucidating the function of the long C-terminus of PYCR1/2, and characterizing the interactome of PYCR.

Kinetics of human pyrroline-5-carboxylate reductase in L-thioproline metabolism.

L-Thioproline (L-thiazolidine-4-carboxylate, L-T4C) is a cyclic sulfur-containing analog of L-proline found in multiple kingdoms of life. The oxidation of L-T4C leads to L-cysteine formation in bacteria, plants, mammals, and protozoa. The conversion of L-T4C to L-Cys in bacterial cell lysates has been attributed to proline dehydrogenase and L-Δ1-pyrroline-5-carboxylate (P5C) reductase (PYCR) enzymes but detailed kinetic studies have not been conducted. Here, we characterize the dehydrogenase activity of human PYCR isozymes 1 and 2 with L-T4C using NAD(P)+ as the hydride acceptor. Both PYCRs exhibit significant L-T4C dehydrogenase activity; however, PYCR2 displays nearly tenfold higher catalytic efficiency (136 M-1 s-1) than PYCR1 (13.7 M-1 s-1). Interestingly, no activity was observed with either L-Pro or the analog DL-thiazolidine-2-carboxylate, indicating that the sulfur at the 4-position is critical for PYCRs to utilize L-T4C as a substrate. Inhibition kinetics show that L-Pro is a competitive inhibitor of PYCR1 [Formula: see text] with respect to L-T4C, consistent with these ligands occupying the same binding site. We also confirm by mass spectrometry that L-T4C oxidation by PYCRs leads to cysteine product formation. Our results suggest a new enzyme function for human PYCRs in the metabolism of L-T4C.

N-Propargylglycine: a unique suicide inhibitor of proline dehydrogenase with anticancer activity and brain-enhancing mitohormesis properties.

Proline dehydrogenase (PRODH) is a mitochondrial inner membrane flavoprotein critical for cancer cell survival under stress conditions and newly recognized as a potential target for cancer drug development. Reversible (competitive) and irreversible (suicide) inhibitors of PRODH have been shown in vivo to inhibit cancer cell growth with excellent host tolerance. Surprisingly, the PRODH suicide inhibitor N-propargylglycine (N-PPG) also induces rapid decay of PRODH with concordant upregulation of mitochondrial chaperones (HSP-60, GRP-75) and the inner membrane protease YME1L1, signifying activation of the mitochondrial unfolded protein response (UPRmt) independent of anticancer activity. The present study was undertaken to address two aims: (i) use PRODH overexpressing human cancer cells (ZR-75-1) to confirm the UPRmt inducing properties of N-PPG relative to another equipotent irreversible PRODH inhibitor, thiazolidine-2-carboxylate (T2C); and (ii) employ biochemical and transcriptomic approaches to determine if orally administered N-PPG can penetrate the blood-brain barrier, essential for its future use as a brain cancer therapeutic, and also potentially protect normal brain tissue by inducing mitohormesis. Oral daily treatments of N-PPG produced a dose-dependent decline in brain mitochondrial PRODH protein without detectable impairment in mouse health; furthermore, mice repeatedly dosed with 50 mg/kg N-PPG showed increased brain expression of the mitohormesis associated protease, YME1L1. Whole brain transcriptome (RNAseq) analyses of these mice revealed significant gene set enrichment in N-PPG stimulated neural processes (FDR p < 0.05). Given this in vivo evidence of brain bioavailability and neural mitohormesis induction, N-PPG appears to be unique among anticancer agents and should be evaluated for repurposing as a pharmaceutical capable of mitigating the proteotoxic mechanisms driving neurodegenerative disorders.

DeepCryoPicker: fully automated deep neural network for single protein particle picking in cryo-EM.

BACKGROUND: Cryo-electron microscopy (Cryo-EM) is widely used in the determination of the three-dimensional (3D) structures of macromolecules. Particle picking from 2D micrographs remains a challenging early step in the Cryo-EM pipeline due to the diversity of particle shapes and the extremely low signal-to-noise ratio of micrographs. Because of these issues, significant human intervention is often required to generate a high-quality set of particles for input to the downstream structure determination steps. RESULTS: Here we propose a fully automated approach (DeepCryoPicker) for single particle picking based on deep learning. It first uses automated unsupervised learning to generate particle training datasets. Then it trains a deep neural network to classify particles automatically. Results indicate that the DeepCryoPicker compares favorably with semi-automated methods such as DeepEM, DeepPicker, and RELION, with the significant advantage of not requiring human intervention. CONCLUSIONS: Our framework combing supervised deep learning classification with automated un-supervised clustering for generating training data provides an effective approach to pick particles in cryo-EM images automatically and accurately.

Structural Determinants of Flavin Dynamics in a Class B Monooxygenase.

The ornithine hydroxylase known as SidA is a class B flavin monooxygenase that catalyzes the first step in the biosynthesis of hydroxamate-containing siderophores in Aspergillus fumigatus. Crystallographic studies of SidA revealed that the FAD undergoes dramatic conformational changes between out and in states during the catalytic cycle. We sought insight into the origins and purpose of flavin motion in class B monooxygenases by probing the function of Met101, a residue that contacts the pyrimidine ring of the in FAD. Steady-state kinetic measurements showed that the mutant variant M101A has a 25-fold lower turnover number. Pre-steady-state kinetic measurements, pH profiles, and solvent kinetic isotope effect measurements were used to isolate the microscopic step that is responsible for the reduced steady-state activity. The data are consistent with a bottleneck in the final step of the mechanism, which involves flavin dehydration and the release of hydroxy-l-ornithine and NADP+. Crystal structures were determined for M101A in the resting state and complexed with NADP+. The resting enzyme structure is similar to that of wild-type SidA, consistent with M101A exhibiting normal kinetics for flavin reduction by NADPH and wild-type affinity for NADPH. In contrast, the structure of the M101A-NADP+ complex unexpectedly shows the FAD adopting the out conformation and may represent a stalled conformation that is responsible for the slow kinetics. Altogether, our data support a previous proposal that one purpose of the FAD conformational change from in to out in class B flavin monooxygenases is to eject spent NADP+ in preparation for a new catalytic cycle.