Prediction and design of protease enzyme specificity using a structure-aware graph convolutional network.

Site-specific proteolysis by the enzymatic cleavage of small linear sequence motifs is a key posttranslational modification involved in physiology and disease. The ability to robustly and rapidly predict protease-substrate specificity would also enable targeted proteolytic cleavage by designed proteases. Current methods for predicting protease specificity are limited to sequence pattern recognition in experimentally derived cleavage data obtained for libraries of potential substrates and generated separately for each protease variant. We reasoned that a more semantically rich and robust model of protease specificity could be developed by incorporating the energetics of molecular interactions between protease and substrates into machine learning workflows. We present Protein Graph Convolutional Network (PGCN), which develops a physically grounded, structure-based molecular interaction graph representation that describes molecular topology and interaction energetics to predict enzyme specificity. We show that PGCN accurately predicts the specificity landscapes of several variants of two model proteases. Node and edge ablation tests identified key graph elements for specificity prediction, some of which are consistent with known biochemical constraints for protease:substrate recognition. We used a pretrained PGCN model to guide the design of protease libraries for cleaving two noncanonical substrates, and found good agreement with experimental cleavage results. Importantly, the model can accurately assess designs featuring diversity at positions not present in the training data. The described methodology should enable the structure-based prediction of specificity landscapes of a wide variety of proteases and the construction of tailor-made protease editors for site-selectively and irreversibly modifying chosen target proteins.

Massively parallel, computationally guided design of a proenzyme.

Confining the activity of a designed protein to a specific microenvironment would have broad-ranging applications, such as enabling cell type-specific therapeutic action by enzymes while avoiding off-target effects. While many natural enzymes are synthesized as inactive zymogens that can be activated by proteolysis, it has been challenging to redesign any chosen enzyme to be similarly stimulus responsive. Here, we develop a massively parallel computational design, screening, and next-generation sequencing-based approach for proenzyme design. For a model system, we employ carboxypeptidase G2 (CPG2), a clinically approved enzyme that has applications in both the treatment of cancer and controlling drug toxicity. Detailed kinetic characterization of the most effectively designed variants shows that they are inhibited by ∼80% compared to the unmodified protein, and their activity is fully restored following incubation with site-specific proteases. Introducing disulfide bonds between the pro- and catalytic domains based on the design models increases the degree of inhibition to 98% but decreases the degree of restoration of activity by proteolysis. A selected disulfide-containing proenzyme exhibits significantly lower activity relative to the fully activated enzyme when evaluated in cell culture. Structural and thermodynamic characterization provides detailed insights into the prodomain binding and inhibition mechanisms. The described methodology is general and could enable the design of a variety of proproteins with precise spatial regulation.

Genetic and Structural Analysis of SARS-CoV-2 Spike Protein for Universal Epitope Selection.

Evaluation of immunogenic epitopes for universal vaccine development in the face of ongoing SARS-CoV-2 evolution remains a challenge. Herein, we investigate the genetic and structural conservation of an immunogenically relevant epitope (C662-C671) of spike (S) protein across SARS-CoV-2 variants to determine its potential utility as a broad-spectrum vaccine candidate against coronavirus diseases. Comparative sequence analysis, structural assessment, and molecular dynamics simulations of C662-C671 epitope were performed. Mathematical tools were employed to determine its mutational cost. We found that the amino acid sequence of C662-C671 epitope is entirely conserved across the observed major variants of SARS-CoV-2 in addition to SARS-CoV. Its conformation and accessibility are predicted to be conserved, even in the highly mutated Omicron variant. Costly mutational rate in the context of energy expenditure in genome replication and translation can explain this strict conservation. These observations may herald an approach to developing vaccine candidates for universal protection against emergent variants of coronavirus.

Macromolecular modeling and design in Rosetta: recent methods and frameworks.

The Rosetta software for macromolecular modeling, docking and design is extensively used in laboratories worldwide. During two decades of development by a community of laboratories at more than 60 institutions, Rosetta has been continuously refactored and extended. Its advantages are its performance and interoperability between broad modeling capabilities. Here we review tools developed in the last 5 years, including over 80 methods. We discuss improvements to the score function, user interfaces and usability. Rosetta is available at http://www.rosettacommons.org.

Monomer Choice Influences N-Acryloyl Amino Acid Grafter Conversion via Protease Catalysis.

End-capped peptides modified with reactive functional groups on the N-terminus provide a route to prepare peptide-polymer conjugates for a broad range of applications. Unfortunately, current chemical methods to construct modified peptides rely largely on solid-phase peptide synthesis (SPPS), which lacks green preparative characteristics and is costly, thus limiting its applicability to specialty applications such as regenerative medicine. This work evaluates N-terminally modified N-acryloyl-glutamic acid diethyl ester, N-acryloyl-leucine ethyl ester, and N-acryloyl-alanine ethyl ester as grafters and papain as the protease for the direct addition of amino acid ethyl ester (AA-OEt) monomers via protease-catalyzed peptide synthesis (PCPS) and the corresponding formation of N-acryloyl-functionalized oligopeptides in a one-pot aqueous reaction. It was hypothesized that by building N-acryloyl grafters from AA-OEt monomers that are known to be good substrates for papain in PCPS, the corresponding grafters would yield high grafter conversions, high ratio of grafter-oligopeptide to free NH2-oligopeptide, and high overall yield. However, this work demonstrates based on the grafter/monomers studied herein that the dominant factor in N-acryloyl-AA-OEt grafter conversion is the co-monomer used in co-oligomerizations. Computational modeling using Rosetta qualitatively recapitulates the results and provides insight into the structural and energetic bases underlying substrate selectivity. The findings herein expand our knowledge of factors that determine the efficiency of preparing N-acryloyl-terminated oligopeptides by PCPS that could provide practical routes to peptide macromers for conjugation to polymers and surfaces for a broad range of applications.

Tuning Enzyme Thermostability via Computationally Guided Covalent Stapling and Structural Basis of Enhanced Stabilization.

Enhancing the thermostability of enzymes without impacting their catalytic function represents an important yet challenging goal in protein engineering and biocatalysis. We recently introduced a novel method for enzyme thermostabilization that relies on the computationally guided installation of genetically encoded thioether "staples" into a protein via cysteine alkylation with the noncanonical amino acid O-2-bromoethyl tyrosine (O2beY). Here, we demonstrate the functionality of an expanded set of electrophilic amino acids featuring chloroacetamido, acrylamido, and vinylsulfonamido side-chain groups for protein stapling using this strategy. Using a myoglobin-based cyclopropanase as a model enzyme, our studies show that covalent stapling with p-chloroacetamido-phenylalanine (pCaaF) provides higher stapling efficiency and enhanced stability (thermodynamic and kinetic) compared to the other stapled variants and the parent protein. Interestingly, molecular simulations of conformational flexibility of the cross-links show that the pCaaF staple allows fewer energetically feasible conformers than the other staples, and this property may be a broader indicator of stability enhancement. Using this strategy, pCaaF-stapled variants with significantly enhanced stability against thermal denaturation (ΔTm' = +27 °C) and temperature-induced heme loss (ΔT50 = +30 °C) were obtained while maintaining high levels of catalytic activity and stereoselectivity. Crystallographic analyses of singly and doubly stapled variants provide key insights into the structural basis for stabilization, which includes both direct interactions of the staples with protein residues and indirect interactions through adjacent residues involved in heme binding. This work expands the toolbox of protein stapling strategies available for protein stabilization.

Enantioselective Synthesis of α-Trifluoromethyl Amines via Biocatalytic N-H Bond Insertion with Acceptor-Acceptor Carbene Donors.

The biocatalytic toolbox has recently been expanded to include enzyme-catalyzed carbene transfer reactions not occurring in Nature. Herein, we report the development of a biocatalytic strategy for the synthesis of enantioenriched α-trifluoromethyl amines through an asymmetric N-H carbene insertion reaction catalyzed by engineered variants of cytochrome c552 from Hydrogenobacter thermophilus. Using a combination of protein and substrate engineering, this metalloprotein scaffold was redesigned to enable the synthesis of chiral α-trifluoromethyl amino esters with up to >99% yield and 95:5 er using benzyl 2-diazotrifluoropropanoate as the carbene donor. When the diazo reagent was varied, the enantioselectivity of the enzyme could be inverted to produce the opposite enantiomers of these products with up to 99.5:0.5 er. This methodology is applicable to a broad range of aryl amine substrates, and it can be leveraged to obtain chemoenzymatic access to enantioenriched ß-trifluoromethyl-ß-amino alcohols and halides. Computational analyses provide insights into the interplay of protein- and reagent-mediated control on the enantioselectivity of this reaction. This work introduces the first example of a biocatalytic N-H carbenoid insertion with an acceptor-acceptor carbene donor, and it offers a biocatalytic solution for the enantioselective synthesis of α-trifluoromethylated amines as valuable synthons for medicinal chemistry and the synthesis of bioactive molecules.

Data-driven supervised learning of a viral protease specificity landscape from deep sequencing and molecular simulations.

Biophysical interactions between proteins and peptides are key determinants of molecular recognition specificity landscapes. However, an understanding of how molecular structure and residue-level energetics at protein-peptide interfaces shape these landscapes remains elusive. We combine information from yeast-based library screening, next-generation sequencing, and structure-based modeling in a supervised machine learning approach to report the comprehensive sequence-energetics-function mapping of the specificity landscape of the hepatitis C virus (HCV) NS3/4A protease, whose function-site-specific cleavages of the viral polyprotein-is a key determinant of viral fitness. We screened a library of substrates in which five residue positions were randomized and measured cleavability of ∼30,000 substrates (∼1% of the library) using yeast display and fluorescence-activated cell sorting followed by deep sequencing. Structure-based models of a subset of experimentally derived sequences were used in a supervised learning procedure to train a support vector machine to predict the cleavability of 3.2 million substrate variants by the HCV protease. The resulting landscape allows identification of previously unidentified HCV protease substrates, and graph-theoretic analyses reveal extensive clustering of cleavable and uncleavable motifs in sequence space. Specificity landscapes of known drug-resistant variants are similarly clustered. The described approach should enable the elucidation and redesign of specificity landscapes of a wide variety of proteases, including human-origin enzymes. Our results also suggest a possible role for residue-level energetics in shaping plateau-like functional landscapes predicted from viral quasispecies theory.

MHCEpitopeEnergy, a Flexible Rosetta-Based Biotherapeutic Deimmunization Platform.

As non-"self" macromolecules, biotherapeutics can trigger an immune response that can reduce drug efficacy, require patients to be taken off therapy, or even cause life-threatening reactions. To enable the flexible and facile design of protein biotherapeutics while reducing the prevalence of T-cell epitopes that drive immune recognition, we have integrated into the Rosetta protein design suite a new scoring term that allows design protocols to account for predicted or experimentally identified epitopes in the optimized objective function. This flexible scoring term can be used in any Rosetta design trajectory, can be targeted to specific regions of a protein, and can be readily extended to work with a variety of epitope predictors. By performing extensive design runs with varied design parameter choices for three case study proteins as well as a larger diverse benchmark, we show that the incorporation of this scoring term enables the effective exploration of an alternative, deimmunized sequence space to discover diverse proteins that are potentially highly deimmunized while retaining physical and chemical qualities similar to those yielded by equivalent nondeimmunizing sequence design protocols.

Enzyme stabilization via computationally guided protein stapling.

Thermostabilization represents a critical and often obligatory step toward enhancing the robustness of enzymes for organic synthesis and other applications. While directed evolution methods have provided valuable tools for this purpose, these protocols are laborious and time-consuming and typically require the accumulation of several mutations, potentially at the expense of catalytic function. Here, we report a minimally invasive strategy for enzyme stabilization that relies on the installation of genetically encoded, nonreducible covalent staples in a target protein scaffold using computational design. This methodology enables the rapid development of myoglobin-based cyclopropanation biocatalysts featuring dramatically enhanced thermostability (ΔTm = +18.0 °C and ΔT50 = +16.0 °C) as well as increased stability against chemical denaturation [ΔCm (GndHCl) = 0.53 M], without altering their catalytic efficiency and stereoselectivity properties. In addition, the stabilized variants offer superior performance and selectivity compared with the parent enzyme in the presence of a high concentration of organic cosolvents, enabling the more efficient cyclopropanation of a water-insoluble substrate. This work introduces and validates an approach for protein stabilization which should be applicable to a variety of other proteins and enzymes.

Structures of the peptide-modifying radical SAM enzyme SuiB elucidate the basis of substrate recognition.

Posttranslational modification of ribosomally synthesized peptides provides an elegant means for the production of biologically active molecules known as RiPPs (ribosomally synthesized and posttranslationally modified peptides). Although the leader sequence of the precursor peptide is often required for turnover, the exact mode of recognition by the modifying enzymes remains unclear for many members of this class of natural products. Here, we have used X-ray crystallography and computational modeling to examine the role of the leader peptide in the biosynthesis of a homolog of streptide, a recently identified peptide natural product with an intramolecular lysine-tryptophan cross-link, which is installed by the radical S-adenosylmethionine (SAM) enzyme, StrB. We present crystal structures of SuiB, a close ortholog of StrB, in various forms, including apo SuiB, SAM-bound SuiB, and a complex of SuiB with SAM and its peptide substrate, SuiA. Although the N-terminal domain of SuiB adopts a typical RRE (RiPP recognition element) motif, which has been implicated in precursor peptide recognition, we observe binding of the leader peptide in the catalytic barrel rather than the N-terminal domain. Computational simulations support a mechanism in which the leader peptide guides posttranslational modification by positioning the cross-linking residues of the precursor peptide within the active site. Together the results shed light onto binding of the precursor peptide and the associated conformational changes needed for the formation of the unique carbon-carbon cross-link in the streptide family of natural products.

Computational design of ligand-binding proteins with high affinity and selectivity.

The ability to design proteins with high affinity and selectivity for any given small molecule is a rigorous test of our understanding of the physiochemical principles that govern molecular recognition. Attempts to rationally design ligand-binding proteins have met with little success, however, and the computational design of protein-small-molecule interfaces remains an unsolved problem. Current approaches for designing ligand-binding proteins for medical and biotechnological uses rely on raising antibodies against a target antigen in immunized animals and/or performing laboratory-directed evolution of proteins with an existing low affinity for the desired ligand, neither of which allows complete control over the interactions involved in binding. Here we describe a general computational method for designing pre-organized and shape complementary small-molecule-binding sites, and use it to generate protein binders to the steroid digoxigenin (DIG). Of seventeen experimentally characterized designs, two bind DIG; the model of the higher affinity binder has the most energetically favourable and pre-organized interface in the design set. A comprehensive binding-fitness landscape of this design, generated by library selections and deep sequencing, was used to optimize its binding affinity to a picomolar level, and X-ray co-crystal structures of two variants show atomic-level agreement with the corresponding computational models. The optimized binder is selective for DIG over the related steroids digitoxigenin, progesterone and ß-oestradiol, and this steroid binding preference can be reprogrammed by manipulation of explicitly designed hydrogen-bonding interactions. The computational design method presented here should enable the development of a new generation of biosensors, therapeutics and diagnostics.

A pH-dependent switch promotes ß-synuclein fibril formation via glutamate residues.

α-Synuclein (αS) is the primary protein associated with Parkinson's disease, and it undergoes aggregation from its intrinsically disordered monomeric form to a cross-ß fibrillar form. The closely related homolog ß-synuclein (ßS) is essentially fibril-resistant under cytoplasmic physiological conditions. Toxic gain-of-function by ßS has been linked to dysfunction, but the aggregation behavior of ßS under altered pH is not well-understood. In this work, we compare fibril formation of αS and ßS at pH 7.3 and mildly acidic pH 5.8, and we demonstrate that pH serves as an on/off switch for ßS fibrillation. Using αS/ßS domain-swapped chimera constructs and single residue substitutions in ßS, we localized the switch to acidic residues in the N-terminal and non-amyloid component domains of ßS. Computational models of ßS fibril structures indicate that key glutamate residues (Glu-31 and Glu-61) in these domains may be sites of pH-sensitive interactions, and variants E31A and E61A show dramatically altered pH sensitivity for fibril formation supporting the importance of these charged side chains in fibril formation of ßS. Our results demonstrate that relatively small changes in pH, which occur frequently in the cytoplasm and in secretory pathways, may induce the formation of ßS fibrils and suggest a complex role for ßS in synuclein cellular homeostasis and Parkinson's disease.

A computational method for the design of nested proteins by loop-directed domain insertion.

The computational design of novel nested proteins-in which the primary structure of one protein domain (insert) is flanked by the primary structure segments of another (parent)-would enable the generation of multifunctional proteins. Here we present a new algorithm, called Loop-Directed Domain Insertion (LooDo), implemented within the Rosetta software suite, for the purpose of designing nested protein domain combinations connected by flexible linker regions. Conformational space for the insert domain is sampled using large libraries of linker fragments for linker-to-parent domain superimposition followed by insert-to-linker superimposition. The relative positioning of the two domains (treated as rigid bodies) is sampled efficiently by a grid-based, mutual placement compatibility search. The conformations of the loop residues, and the identities of loop as well as interface residues, are simultaneously optimized using a generalized kinematic loop closure algorithm and Rosetta EnzymeDesign, respectively, to minimize interface energy. The algorithm was found to consistently sample near-native conformations and interface sequences for a benchmark set of structurally similar but functionally divergent domain-inserted enzymes from the α/ß hydrolase superfamily, and discriminates well between native and nonnative conformations and sequences, although loop conformations tended to deviate from the native conformations. Furthermore, in cross-domain placement tests, native insert-parent domain combinations were ranked as the best-scoring structures compared to nonnative domain combinations. This algorithm should be broadly applicable to the design of multi-domain protein complexes with any combination of inserted or tandem domain connections.

Computational Design of a Photocontrolled Cytosine Deaminase.

There is growing interest in designing spatiotemporal control over enzyme activities using noninvasive stimuli such as light. Here, we describe a structure-based, computation-guided predictive method for reversibly controlling enzyme activity using covalently attached photoresponsive azobenzene groups. Applying the method to the therapeutically useful enzyme yeast cytosine deaminase, we obtained a ∼3-fold change in enzyme activity by the photocontrolled modulation of the enzyme's active site lid structure, while fully maintaining thermostability. Multiple cycles of switching, controllable in real time, are possible. The predictiveness of the method is demonstrated by the construction of a variant that does not photoswitch as expected from computational modeling. Our design approach opens new avenues for optically controlling enzyme function. The designed photocontrolled cytosine deaminases may also aid in improving chemotherapy approaches that utilize this enzyme.

MFPred: Rapid and accurate prediction of protein-peptide recognition multispecificity using self-consistent mean field theory.

Multispecificity-the ability of a single receptor protein molecule to interact with multiple substrates-is a hallmark of molecular recognition at protein-protein and protein-peptide interfaces, including enzyme-substrate complexes. The ability to perform structure-based prediction of multispecificity would aid in the identification of novel enzyme substrates, protein interaction partners, and enable design of novel enzymes targeted towards alternative substrates. The relatively slow speed of current biophysical, structure-based methods limits their use for prediction and, especially, design of multispecificity. Here, we develop a rapid, flexible-backbone self-consistent mean field theory-based technique, MFPred, for multispecificity modeling at protein-peptide interfaces. We benchmark our method by predicting experimentally determined peptide specificity profiles for a range of receptors: protease and kinase enzymes, and protein recognition modules including SH2, SH3, MHC Class I and PDZ domains. We observe robust recapitulation of known specificities for all receptor-peptide complexes, and comparison with other methods shows that MFPred results in equivalent or better prediction accuracy with a ~10-1000-fold decrease in computational expense. We find that modeling bound peptide backbone flexibility is key to the observed accuracy of the method. We used MFPred for predicting with high accuracy the impact of receptor-side mutations on experimentally determined multispecificity of a protease enzyme. Our approach should enable the design of a wide range of altered receptor proteins with programmed multispecificities.

Structural principles for computational and de novo design of 4Fe-4S metalloproteins.

Iron-sulfur centers in metalloproteins can access multiple oxidation states over a broad range of potentials, allowing them to participate in a variety of electron transfer reactions and serving as catalysts for high-energy redox processes. The nitrogenase FeMoCO cluster converts di-nitrogen to ammonia in an eight-electron transfer step. The 2(Fe4S4) containing bacterial ferredoxin is an evolutionarily ancient metalloprotein fold and is thought to be a primordial progenitor of extant oxidoreductases. Controlling chemical transformations mediated by iron-sulfur centers such as nitrogen fixation, hydrogen production as well as electron transfer reactions involved in photosynthesis are of tremendous importance for sustainable chemistry and energy production initiatives. As such, there is significant interest in the design of iron-sulfur proteins as minimal models to gain fundamental understanding of complex natural systems and as lead-molecules for industrial and energy applications. Herein, we discuss salient structural characteristics of natural iron-sulfur proteins and how they guide principles for design. Model structures of past designs are analyzed in the context of these principles and potential directions for enhanced designs are presented, and new areas of iron-sulfur protein design are proposed. This article is part of a Special issue entitled Biodesign for Bioenergetics--the design and engineering of electronic transfer cofactors, protein networks, edited by Ronald L. Koder and J.L Ross Anderson.

Design and Evolution of a Macrocyclic Peptide Inhibitor of the Sonic Hedgehog/Patched Interaction.

The hedgehog (Hh) signaling pathway plays a central role during embryonic development, and its aberrant activation has been implicated in the development and progression of several human cancers. Major efforts toward the identification of chemical modulators of the hedgehog pathway have yielded several antagonists of the GPCR-like smoothened receptor. In contrast, potent inhibitors of the sonic hedgehog/patched interaction, the most upstream event in ligand-induced activation of this signaling pathway, have been elusive. To address this gap, a genetically encoded cyclic peptide was designed based on the sonic hedgehog (Shh)-binding loop of hedgehog-interacting protein (HHIP) and subjected to multiple rounds of affinity maturation through the screening of macrocyclic peptide libraries produced in E. coli cells. Using this approach, an optimized macrocyclic peptide inhibitor (HL2-m5) was obtained that binds Shh with a KD of 170 nM, which corresponds to a 120-fold affinity improvement compared to the parent molecule. Importantly, HL2-m5 is able to effectively suppress Shh-mediated hedgehog signaling and Gli-controlled gene transcription in living cells (IC50 = 230 nM), providing the most potent inhibitor of the sonic hedgehog/patched interaction reported to date. This first-in-class macrocyclic peptide modulator of the hedgehog pathway is expected to provide a valuable probe for investigating and targeting ligand-dependent hedgehog pathway activation in cancer and other pathologies. This work also introduces a general strategy for the development of cyclopeptide inhibitors of protein-protein interactions.

Computation-Guided Design of a Stimulus-Responsive Multienzyme Supramolecular Assembly.

The construction of stimulus-responsive supramolecular complexes of metabolic pathway enzymes, inspired by natural multienzyme assemblies (metabolons), provides an attractive avenue for efficient and spatiotemporally controllable one-pot biotransformations. We have constructed a phosphorylation- and optically responsive metabolon for the biodegradation of the environmental pollutant 1,2,3-trichloropropane.

CSAR Benchmark Exercise 2013: Evaluation of Results from a Combined Computational Protein Design, Docking, and Scoring/Ranking Challenge.

Community Structure-Activity Resource (CSAR) conducted a benchmark exercise to evaluate the current computational methods for protein design, ligand docking, and scoring/ranking. The exercise consisted of three phases. The first phase required the participants to identify and rank order which designed sequences were able to bind the small molecule digoxigenin. The second phase challenged the community to select a near-native pose of digoxigenin from a set of decoy poses for two of the designed proteins. The third phase investigated the ability of current methods to rank/score the binding affinity of 10 related steroids to one of the designed proteins (pKd = 4.1 to 6.7). We found that 11 of 13 groups were able to correctly select the sequence that bound digoxigenin, with most groups providing the correct three-dimensional structure for the backbone of the protein as well as all atoms of the active-site residues. Eleven of the 14 groups were able to select the appropriate pose from a set of plausible decoy poses. The ability to predict absolute binding affinities is still a difficult task, as 8 of 14 groups were able to correlate scores to affinity (Pearson-r > 0.7) of the designed protein for congeneric steroids and only 5 of 14 groups were able to correlate the ranks of the 10 related ligands (Spearman-ρ > 0.7).