MEnTaT: A machine-learning approach for the identification of mutations to increase protein stability.

Enhancing protein thermal stability is important for biomedical and industrial applications as well as in the research laboratory. Here, we describe a simple machine-learning method which identifies amino acid substitutions that contribute to thermal stability based on comparison of the amino acid sequences of homologous proteins derived from bacteria that grow at different temperatures. A key feature of the method is that it compares the sequences based not simply on the amino acid identity, but rather on the structural and physicochemical properties of the side chain. The method accurately identified stabilizing substitutions in three well-studied systems and was validated prospectively by experimentally testing predicted stabilizing substitutions in a polyamine oxidase. In each case, the method outperformed the widely used bioinformatic consensus approach. The method can also provide insight into fundamental aspects of protein structure, for example, by identifying how many sequence positions in a given protein are relevant to temperature adaptation.

Structural Insights into the Substrate Range of a Bacterial Monoamine Oxidase.

Monoamine oxidases (MAOs) play a key role in the breakdown of primary and secondary amines. In eukaryotic organisms, these enzymes are vital to the regulation of monoamine neurotransmitters and the degradation of dietary monoamines. MAOs have also been identified in prokaryotic species, although their role in these organisms is not well understood. Here, we report the biophysical and structural properties of a promiscuous, bacterial MAO from Corynebacterium ammoniagenes (caMAO). caMAO catalyzes the oxidation of a number of monoamine substrates including dopamine and norepinephrine, as well as exhibiting some activity with polyamine substrates such as cadaverine. The X-ray crystal structures of Michaelis complexes with seven substrates show that conserved hydrophobic interactions and hydrogen-bonding pattern (for polar substrates) allow the broad specificity range. The structure of caMAO identifies an unusual cysteine (Cys424) residue in the so-called "aromatic cage", which flanks the flavin isoalloxazine ring in the active site. Site-directed mutagenesis, steady-state kinetics in air-saturated buffer, and UV-vis spectroscopy revealed that Cys424 plays a role in the pH dependence and modulation of electrostatics within the caMAO active site. Notably, bioinformatic analysis shows a propensity for variation at this site within the "aromatic cage" of the flavin amine oxidase (FAO) superfamily. Structural analysis also identified the conservation of a secondary substrate inhibition site, present in a homologous member of the superfamily. Finally, genome neighborhood diagram analysis of caMAO in the context of the FAO superfamily allows us to propose potential roles for these bacterial MAOs in monoamine and polyamine degradation and catabolic pathways related to scavenging of nitrogen.

Structure-Guided Insight into the Specificity and Mechanism of a Parasitic Nucleoside Hydrolase.

Trichomonas vaginalis is the causative parasitic protozoan of the disease trichomoniasis, the most prevalent, nonviral sexually transmitted disease in the world. T. vaginalis is a parasite that scavenges nucleosides from the host organism via catalysis by nucleoside hydrolase (NH) enzymes to yield purine and pyrimidine bases. One of the four NH enzymes identified within the genome of T. vaginalis displays unique specificity toward purine nucleosides, adenosine and guanosine, but not inosine, and atypically shares greater sequence similarity to the pyrimidine hydrolases. Bioinformatic analysis of this enzyme, adenosine/guanosine-preferring nucleoside ribohydrolase (AGNH), was incapable of identifying the residues responsible for this uncommon specificity, highlighting the need for structural information. Here, we report the X-ray crystal structures of holo, unliganded AGNH and three additional structures of the enzyme bound to fragment and small-molecule inhibitors. Taken together, these structures facilitated the identification of residue Asp231, which engages in substrate interactions in the absence of those residues that typically support the canonical purine-specific tryptophan-stacking specificity motif. An altered substrate-binding pose is mirrored by repositioning within the protein scaffold of the His80 general acid/base catalyst. The newly defined structure-determined sequence markers allowed the assignment of additional NH orthologs, which are proposed to exhibit the same specificity for adenosine and guanosine alone and further delineate specificity classes for these enzymes.

Recapitulating the Binding Affinity of Nrf2 for KEAP1 in a Cyclic Heptapeptide, Guided by NMR, X-ray Crystallography, and Machine Learning.

Macrocycles, including macrocyclic peptides, have shown promise for targeting challenging protein-protein interactions (PPIs). One PPI of high interest is between Kelch-like ECH-Associated Protein-1 (KEAP1) and Nuclear Factor (Erythroid-derived 2)-like 2 (Nrf2). Guided by X-ray crystallography, NMR, modeling, and machine learning, we show that the full 20 nM binding affinity of Nrf2 for KEAP1 can be recapitulated in a cyclic 7-mer peptide, c[(D)-ß-homoAla-DPETGE]. This compound was identified from the Nrf2-derived linear peptide GDEETGE (KD = 4.3 µM) solely by optimizing the conformation of the cyclic compound, without changing any KEAP1 interacting residue. X-ray crystal structures were determined for each linear and cyclic peptide variant bound to KEAP1. Despite large variations in affinity, no obvious differences in the conformation of the peptide binding residues or in the interactions they made with KEAP1 were observed. However, analysis of the X-ray structures by machine learning showed that locations of strain in the bound ligand could be identified through patterns of subangstrom distortions from the geometry observed for unstrained linear peptides. We show that optimizing the cyclic peptide affinity was driven partly through conformational preorganization associated with a proline substitution at position 78 and with the geometry of the noninteracting residue Asp77 and partly by decreasing strain in the ETGE motif itself. This approach may have utility in dissecting the trade-off between conformational preorganization and strain in other ligand-receptor systems. We also identify a pair of conserved hydrophobic residues flanking the core DxETGE motif which play a conformational role in facilitating the high-affinity binding of Nrf2 to KEAP1.

Interaction Energetics and Druggability of the Protein-Protein Interaction between Kelch-like ECH-Associated Protein 1 (KEAP1) and Nuclear Factor Erythroid 2 Like 2 (Nrf2).

Development of small molecule inhibitors of protein-protein interactions (PPIs) is hampered by our poor understanding of the druggability of PPI target sites. Here, we describe the combined application of alanine-scanning mutagenesis, fragment screening, and FTMap computational hot spot mapping to evaluate the energetics and druggability of the highly charged PPI interface between Kelch-like ECH-associated protein 1 (KEAP1) and nuclear factor erythroid 2 like 2 (Nrf2), an important drug target. FTMap identifies four binding energy hot spots at the active site. Only two of these are exploited by Nrf2, which alanine scanning of both proteins shows to bind primarily through E79 and E82 interacting with KEAP1 residues S363, R380, R415, R483, and S508. We identify fragment hits and obtain X-ray complex structures for three fragments via crystal soaking using a new crystal form of KEAP1. Combining these results provides a comprehensive and quantitative picture of the origins of binding energy at the interface. Our findings additionally reveal non-native interactions that might be exploited in the design of uncharged synthetic ligands to occupy the same site on KEAP1 that has evolved to bind the highly charged DEETGE binding loop of Nrf2. These include π-stacking with KEAP1 Y525 and interactions at an FTMap-identified hot spot deep in the binding site. Finally, we discuss how the complementary information provided by alanine-scanning mutagenesis, fragment screening, and computational hot spot mapping can be integrated to more comprehensively evaluate PPI druggability.

Crystallography Coupled with Kinetic Analysis Provides Mechanistic Underpinnings of a Nicotine-Degrading Enzyme.

Nicotine oxidoreductase (NicA2) is a bacterial flavoenzyme, which catalyzes the first step of nicotine catabolism by oxidizing S-nicotine into N-methyl-myosmine. It has been proposed as a biotherapeutic for nicotine addiction because of its nanomolar substrate binding affinity. The first crystal structure of NicA2 has been reported, establishing NicA2 as a member of the monoamine oxidase (MAO) family. However, substrate specificity and structural determinants of substrate binding and/or catalysis have not been explored. Herein, analysis of the pH-rate profile, single-turnover kinetics, and binding data establish that pH does not significantly affect the catalytic rate and product release is not rate-limiting. The X-ray crystal structure of NicA2 with S-nicotine refined to 2.65 Å resolution reveals a hydrophobic binding site with a solvent exclusive cavity. Hydrophobic interactions predominantly orient the substrate, promoting the binding of a deprotonated species and supporting a hydride-transfer mechanism. Notably, NicA2 showed no activity against neurotransmitters oxidized by the two isoforms of human MAO. To further probe the substrate range of NicA2, enzyme activity was evaluated using a series of substrate analogues, indicating that S-nicotine is the optimal substrate and substitutions within the pyridyl ring abolish NicA2 activity. Moreover, mutagenesis and kinetic analysis of active-site residues reveal that removal of a hydrogen bond between the pyridyl ring of S-nicotine and the hydroxyl group of T381 has a 10-fold effect on KM, supporting the role of this bond in positioning the catalytically competent form of the substrate. Together, crystallography combined with kinetic analysis provides a deeper understanding of this enzyme's remarkable specificity.

Adenosine/guanosine preferring nucleoside ribohydrolase is a distinct, druggable antitrichomonal target.

Nucleoside salvage pathway enzymes used by Trichomonas vaginalis are distinct from the pathway involved in activation of existing 5-nitroimidazole drugs. They thus represent excellent targets for developing novel, mechanism-based antitrichomonal agents. The purine-specific adenosine/guanosine preferring ribohydrolase (AGNH) was screened against the NIH Clinical Collection to assess its druggability. Eight compounds, including five flavonoids, were identified with IC50 values ⩽10 µM and confirmed in counter screens run in the presence of detergent. The inhibitors are structurally distinct from inhibitors of the pyrimidine-specific uridine ribohydrolase (UNH) thus indicating that AGNH is a distinct, druggable target from UNH.

Ligand-Efficient Inhibitors of Trichomonas vaginalis Adenosine/Guanosine Preferring Nucleoside Ribohydrolase.

Trichomoniasis is caused by the parasitic protozoan Trichomonas vaginalis and is the most prevalent, nonviral sexually transmitted disease. The parasite has shown increasing resistance to the current 5-nitroimidazole therapies indicating the need for new therapies with different mechanisms. T. vaginalis is an obligate parasite that scavenges nucleosides from host cells and then uses salvage pathway enzymes to obtain the nucleobases. The adenosine/guanosine preferring nucleoside ribohydrolase was screened against a 2000-compound diversity fragment library using a 1H NMR-based activity assay. Three classes of inhibitors with more than five representatives were identified: bis-aryl phenols, amino bicyclic pyrimidines, and aryl acetamides. Among the active fragments were 10 compounds with ligand efficiency values greater than 0.5, including five with IC50 values <10 µM. Jump-dilution and detergent counter screens validated reversible, target-specific activity. The data reveals an emerging SAR that is guiding our medicinal chemistry efforts aimed at discovering compounds with nanomolar potency.

NMR-Based Activity Assays for Determining Compound Inhibition, IC50 Values, Artifactual Activity, and Whole-Cell Activity of Nucleoside Ribohydrolases.

NMR spectroscopy is often used for the identification and characterization of enzyme inhibitors in drug discovery, particularly in the context of fragment screening. NMR-based activity assays are ideally suited to work at the higher concentrations of test compounds required to detect these weaker inhibitors. The dynamic range and chemical shift dispersion in an NMR experiment can easily resolve resonances from substrate, product, and test compounds. This contrasts with spectrophotometric assays, in which read-out interference problems often arise from compounds with overlapping UV-vis absorption profiles. In addition, since they lack reporter enzymes, the single-enzyme NMR assays are not prone to coupled-assay false positives. This attribute makes them useful as orthogonal assays, complementing traditional high throughput screening assays and benchtop triage assays. Detailed protocols are provided for initial compound assays at 500 µM and 250 µM, dose-response assays for determining IC50 values, detergent counter screen assays, jump-dilution counter screen assays, and assays in E. coli whole cells. The methods are demonstrated using two nucleoside ribohydrolase enzymes. The use of 1H NMR is shown for the purine-specific enzyme, while 19F NMR is shown for the pyrimidine-specific enzyme. The protocols are generally applicable to any enzyme where substrate and product resonances can be observed and distinguished by NMR spectroscopy. To be the most useful in the context of drug discovery, the final concentration of substrate should be no more than 2-3x its Km value. The choice of NMR experiment depends on the enzyme reaction and substrates available as well as available NMR instrumentation.