Deep learning for protein structure prediction and design-progress and applications

Proteins are the key molecular machines that orchestrate all biological processes of the cell. Most proteins fold into three-dimensional shapes that are critical for their function. Studying the 3D shape of proteins can inform us of the mechanisms that underlie biological processes in living cells and can have practical applications in the study of disease mutations or the discovery of novel drug treatments. Here, we review the progress made in sequence-based prediction of protein structures with a focus on applications that go beyond the prediction of single monomer structures. This includes the application of deep learning methods for the prediction of structures of protein complexes, different conformations, the evolution of protein structures and the application of these methods to protein design. These developments create new opportunities for research that will have impact across many areas of biomedical research.

In vitro Production of Hemin-Based Artificial Metalloenzymes

Developing enzyme alternatives is pivotal to improving and enabling new processes in biotechnology and industry. Artificial metalloenzymes (ArMs) are combinations of protein scaffolds with metal elements, such as metal nanoclusters or metal-containing molecules with specific catalytic properties, which can be customized. Here, we engineered an ArM based on the consensus tetratricopeptide repeat (CTPR) scaffold by introducing a unique histidine residue to coordinate the hemin cofactor. Our results show that this engineered system exhibits robust peroxidase-like catalytic activity driven by the hemin. The expression of the scaffold and subsequent coordination of hemin was achieved by recombinant expression in bulk and through in vitro transcription and translation systems in water-in-oil drops. The ability to synthesize this system in emulsio paves the way to improve its properties by means of droplet microfluidic screenings, facilitating the exploration of the protein combinatorial space to discover improved or novel catalytic activities.

Constructing Photoactivatable Protein with Genetically Encoded Photocaged Glutamic Acid

Genetically replacing an essential residue with the corresponding photocaged analogues via genetic code expansion (GCE) constitutes a useful and unique strategy to directly and effectively generate photoactivatable proteins. However, the application of this strategy is severely hampered by the limited number of encoded photocaged proteinogenic amino acids. Herein, we report the genetic incorporation of photocaged glutamic acid analogues in E. coli and mammalian cells and demonstrate their use in constructing photoactivatable variants of various fluorescent proteins and SpyCatcher. We believe genetically encoded photocaged Glu would significantly promote the design and application of photoactivatable proteins in many areas.

A De Novo-Designed Type 3 Copper Protein Tunes Catechol Substrate Recognition and Reactivity

De novo metalloprotein design is a remarkable approach to shape protein scaffolds toward specific functions. Here, we report the design and characterization of Due Rame 1 (DR1), a de novo designed protein housing a di-copper site and mimicking the Type 3 (T3) copper-containing polyphenol oxidases (PPOs). To achieve this goal, we hierarchically designed the first and the second di-metal coordination spheres to engineer the di-copper site into a simple four-helix bundle scaffold. Spectroscopic, thermodynamic, and functional characterization revealed that DR1 recapitulates the T3 copper site, supporting different copper redox states, and being active in the O₂ -dependent oxidation of catechols to o-quinones. Careful design of the residues lining the substrate access site endows DR1 with substrate recognition, as revealed by Hammet analysis and computational studies on substituted catechols. This study represents a premier example in the construction of a functional T3 copper site into a designed four-helix bundle protein.

A Method for User-defined Mutagenesis by Integrating Oligo Pool Synthesis Technology with Nicking Mutagenesis

Saturation mutagenesis is a fundamental enabling technology for protein engineering and epitope mapping. Nicking mutagenesis (NM) allows the user to rapidly construct libraries of all possible single mutations in a target protein sequence from plasmid DNA in a one-pot procedure. Briefly, one strand of the plasmid DNA is degraded using a nicking restriction endonuclease and exonuclease treatment. Mutagenic primers encoding the desired mutations are annealed to the resulting circular single-stranded DNA, extended with high-fidelity polymerase, and ligated into covalently closed circular DNA by Taq DNA ligase. The heteroduplex DNA is resolved by selective degradation of the template strand. The complementary strand is synthesized and ligated, resulting in a library of mutated covalently closed circular plasmids. It was later shown that because very little primer is used in the procedure, resuspended oligo pools, which normally require amplification before use, can be used directly in the mutagenesis procedure. Because oligo pools can contain tens of thousands of unique oligos, this enables the construction of libraries of tens of thousands of user-defined mutations in a single-pot mutagenesis reaction, which significantly improves the utility of NM as described below. Use of oligo pools afford an economically advantageous approach to mutagenic experiments. First, oligo pool synthesis is much less expensive per nucleotide synthesized than conventional synthesis. Second, a mixed pool may be generated and used for mutagenesis of multiple different genes. To use the same oligo-pool for mutagenesis of a variety of genes, the user must only quantify the fraction of the oligo-pool specific to her mutagenic experiment and adjust the volume and effective concentration of the oligo-pool for use in nicking mutagenesis.

Designing and defining dynamic protein cage nanoassemblies in solution

Central challenges in the design of large and dynamic macromolecular assemblies for synthetic biology lie in developing effective methods for testing design strategies and their outcomes, including comprehensive assessments of solution behavior. We created and validated an advanced design of a 600-kDa protein homododecamer that self-assembles into a symmetric tetrahedral cage. The monomeric unit is composed of a trimerizing apex-forming domain genetically linked to an edge-forming dimerizing domain. Enhancing the crystallographic results, high-throughput small-angle x-ray scattering (SAXS) comprehensively contrasted our modifications under diverse solution conditions. To generate a phase diagram associating structure and assembly, we developed force plots that measure dissimilarity among multiple SAXS data sets. These new tools, which provided effective feedback on experimental constructs relative to design, have general applicability in analyzing the solution behavior of heterogeneous nanosystems and have been made available as a web-based application. Specifically, our results probed the influence of solution conditions and symmetry on stability and structural adaptability, identifying the dimeric interface as the weak point in the assembly. Force plots comparing SAXS data sets further reveal more complex and controllable behavior in solution than captured by our crystal structures. These methods for objectively and comprehensively comparing SAXS profiles for systems critically affected by solvent conditions and structural heterogeneity provide an enabling technology for advancing the design and bioengineering of nanoscale biological materials.

The self-assembly of a mini-fibril with axial periodicity from a designed collagen-mimetic triple helix

In this work we describe the self-assembly of a collagen-like periodic mini-fibril from a recombinant triple helix. The triple helix, designated Col108, is expressed in Escherichia coli using an artificial gene and consists of a 378-residue triple helix domain organized into three pseudo-repeating sequence units. The peptide forms a stable triple helix with a melting temperature of 41 °C. Upon increases of pH and temperature, Col108 self-assembles in solution into smooth mini-fibrils with the cross-striated banding pattern typical of fibrillar collagens. The banding pattern is characterized by an axially repeating feature of ∼35 nm as observed by transmission electron microscopy and atomic force microscopy. Both the negatively stained and the positively stained transmission electron microscopy patterns of the Col108 mini-fibrils are consistent with a staggered arrangement of triple helices having a staggering value of 123 residues, a value closely connected to the size of one repeat sequence unit. A mechanism is proposed for the mini-fibril formation of Col108 in which the axial periodicity is instigated by the built-in sequence periodicity and stabilized by the optimized interactions between the triple helices in a 1-unit staggered arrangement. Lacking hydroxyproline residues and telopeptides, two factors implicated in the fibrillogenesis of native collagen, the Col108 mini-fibrils demonstrate that sequence features of the triple helical domain alone are sufficient to "code" for axially repeating periodicity of fibrils. To our knowledge, Col108 is the first designed triple helix to self-assemble into periodic fibrils and offers a unique opportunity to unravel the specific molecular interactions of collagen fibrillogenesis.

Aromatic anchor at an invariant hormone-receptor interface: function of insulin residue B24 with application to protein design

Crystallographic studies of insulin bound to fragments of the insulin receptor have recently defined the topography of the primary hormone-receptor interface. Here, we have investigated the role of Phe(B24), an invariant aromatic anchor at this interface and site of a human mutation causing diabetes mellitus. An extensive set of B24 substitutions has been constructed and tested for effects on receptor binding. Although aromaticity has long been considered a key requirement at this position, Met(B24) was found to confer essentially native affinity and bioactivity. Molecular modeling suggests that this linear side chain can serve as an alternative hydrophobic anchor at the hormone-receptor interface. These findings motivated further substitution of Phe(B24) by cyclohexanylalanine (Cha), which contains a nonplanar aliphatic ring. Contrary to expectations, [Cha(B24)]insulin likewise exhibited high activity. Furthermore, its resistance to fibrillation and the rapid rate of hexamer disassembly, properties of potential therapeutic advantage, were enhanced. The crystal structure of the Cha(B24) analog, determined as an R6 zinc-stabilized hexamer at a resolution of 1.5 Å, closely resembles that of wild-type insulin. The nonplanar aliphatic ring exhibits two chair conformations with partial occupancies, each recapitulating the role of Phe(B24) at the dimer interface. Together, these studies have defined structural requirements of an anchor residue within the B24-binding pocket of the insulin receptor; similar molecular principles are likely to pertain to insulin-related growth factors. Our results highlight in particular the utility of nonaromatic side chains as probes of the B24 pocket and suggest that the nonstandard Cha side chain may have therapeutic utility.

Amyloid-β peptide-specific DARPins as a novel class of potential therapeutics for Alzheimer disease

Passive immunization with anti-amyloid-β peptide (Aβ) antibodies is effective in animal models of Alzheimer disease. With the advent of efficient in vitro selection technologies, the novel class of designed ankyrin repeat proteins (DARPins) presents an attractive alternative to the immunoglobulin scaffold. DARPins are small and highly stable proteins with a compact modular architecture ideal for high affinity protein-protein interactions. In this report, we describe the selection, binding profile, and epitope analysis of Aβ-specific DARPins. We further showed their ability to delay Aβ aggregation and prevent Aβ-mediated neurotoxicity in vitro. To demonstrate their therapeutic potential in vivo, mono- and trivalent Aβ-specific DARPins (D23 and 3×D23) were infused intracerebroventricularly into the brains of 11-month-old Tg2576 mice over 4 weeks. Both D23 and 3×D23 treatments were shown to result in improved cognitive performance and reduced soluble Aβ levels. These findings demonstrate the therapeutic potential of Aβ-specific DARPins for the treatment of Alzheimer disease.

Biophysical optimization of a therapeutic protein by nonstandard mutagenesis: studies of an iodo-insulin derivative

Insulin provides a model for the therapeutic application of protein engineering. A paradigm in molecular pharmacology was defined by design of rapid-acting insulin analogs for the prandial control of glycemia. Such analogs, a cornerstone of current diabetes regimens, exhibit accelerated subcutaneous absorption due to more rapid disassembly of oligomeric species relative to wild-type insulin. This strategy is limited by a molecular trade-off between accelerated disassembly and enhanced susceptibility to degradation. Here, we demonstrate that this trade-off may be circumvented by nonstandard mutagenesis. Our studies employed Lys(B28), Pro(B29)-insulin ("lispro") as a model prandial analog that is less thermodynamically stable and more susceptible to fibrillation than is wild-type insulin. We have discovered that substitution of an invariant tyrosine adjoining the engineered sites in lispro (Tyr(B26)) by 3-iodo-Tyr (i) augments its thermodynamic stability (ΔΔGu 0.5 ± 0.2 kcal/mol), (ii) delays onset of fibrillation (lag time on gentle agitation at 37 °C was prolonged by 4-fold), (iii) enhances affinity for the insulin receptor (1.5 ± 0.1-fold), and (iv) preserves biological activity in a rat model of diabetes mellitus. (1)H NMR studies suggest that the bulky iodo-substituent packs within a nonpolar interchain crevice. Remarkably, the 3-iodo-Tyr(B26) modification stabilizes an oligomeric form of insulin pertinent to pharmaceutical formulation (the R6 zinc hexamer) but preserves rapid disassembly of the oligomeric form pertinent to subcutaneous absorption (T6 hexamer). By exploiting this allosteric switch, 3-iodo-Tyr(B26)-lispro thus illustrates how a nonstandard amino acid substitution can mitigate the unfavorable biophysical properties of an engineered protein while retaining its advantages.

Exploring the structure of the voltage-gated Na+ channel by an engineered drug access pathway to the receptor site for local anesthetics

Despite the availability of several crystal structures of bacterial voltage-gated Na(+) channels, the structure of eukaryotic Na(+) channels is still undefined. We used predictions from available homology models and crystal structures to modulate an external access pathway for the membrane-impermeant local anesthetic derivative QX-222 into the internal vestibule of the mammalian rNaV1.4 channel. Potassium channel-based homology models predict amino acid Ile-1575 in domain IV segment 6 to be in close proximity to Lys-1237 of the domain III pore-loop selectivity filter. The mutation K1237E has been shown previously to increase the diameter of the selectivity filter. We found that an access pathway for external QX-222 created by mutations of Ile-1575 was abolished by the additional mutation K1237E, supporting the notion of a close spatial relationship between sites 1237 and 1575. Crystal structures of bacterial voltage-gated Na(+) channels predict that the side chain of rNaV1.4 Trp-1531 of the domain IV pore-loop projects into the space between domain IV segment 6 and domain III pore-loop and, therefore, should obstruct the putative external access pathway. Indeed, mutations W1531A and W1531G allowed for exceptionally rapid access of QX-222. In addition, W1531G created a second non-selective ion-conducting pore, bypassing the outer vestibule but probably merging into the internal vestibule, allowing for control by the activation gate. These data suggest a strong structural similarity between bacterial and eukaryotic voltage-gated Na(+) channels.

Insight into the role of substrate-binding residues in conferring substrate specificity for the multifunctional polysaccharide lyase Smlt1473

Anionic polysaccharides are of growing interest in the biotechnology industry due to their potential pharmaceutical applications in drug delivery and wound treatment. Chemical composition and polymer length strongly influence the physical and biological properties of the polysaccharide and thus its potential industrial and medical applications. One promising approach to determining monomer composition and controlling the degree of polymerization involves the use of polysaccharide lyases, which catalyze the depolymerization of anionic polysaccharides via a β-elimination mechanism. Utilization of these enzymes for the production of custom-made oligosaccharides requires a high degree of control over substrate specificity. Previously, we characterized a polysaccharide lyase (Smlt1473) from Stenotrophomonas maltophilia k279a, which exhibited significant activity against hyaluronan (HA), poly-β-d-glucuronic acid (poly-GlcUA), and poly-β-d-mannuronic acid (poly-ManA) in a pH-regulated manner. Here, we utilize a sequence structure guided approach based on a homology model of Smlt1473 to identify nine putative substrate-binding residues and examine their effect on substrate specificity via site-directed mutagenesis. Interestingly, single point mutations H221F and R312L resulted in increased activity and specificity toward poly-ManA and poly-GlcUA, respectively. Furthermore, a W171A mutant nearly eliminated HA activity, while increasing poly-ManA and poly-GlcUA activity by at least 35%. The effect of these mutations was analyzed by comparison with the high resolution structure of Sphingomonas sp. A1-III alginate lyase in complex with poly-ManA tetrasaccharide and by taking into account the structural differences between HA, poly-GlcUA, and poly-ManA. Overall, our results demonstrate that even minor changes in active site architecture have a significant effect on the substrate specificity of Smlt1473, whose structural plasticity could be applied to the design of highly active and specific polysaccharide lyases.

Structural similarity between defense peptide from wheat and scorpion neurotoxin permits rational functional design

In this study, we present the spatial structure of the wheat antimicrobial peptide (AMP) Tk-AMP-X2 studied using NMR spectroscopy. This peptide was found to adopt a disulfide-stabilized α-helical hairpin fold and therefore belongs to the α-hairpinin family of plant defense peptides. Based on Tk-AMP-X2 structural similarity to cone snail and scorpion potassium channel blockers, a mutant molecule, Tk-hefu, was engineered by incorporating the functionally important residues from κ-hefutoxin 1 onto the Tk-AMP-X2 scaffold. The designed peptide contained the so-called essential dyad of amino acid residues significant for channel-blocking activity. Electrophysiological studies showed that although the parent peptide Tk-AMP-X2 did not present any activity against potassium channels, Tk-hefu blocked Kv1.3 channels with similar potency (IC50 ∼ 35 μm) to κ-hefutoxin 1 (IC50 ∼ 40 μm). We conclude that α-hairpinins are attractive in their simplicity as structural templates, which may be used for functional engineering and drug design.

Enhanced enzyme kinetic stability by increasing rigidity within the active site

Enzyme stability is an important issue for protein engineers. Understanding how rigidity in the active site affects protein kinetic stability will provide new insight into enzyme stabilization. In this study, we demonstrated enhanced kinetic stability of Candida antarctica lipase B (CalB) by mutating the structurally flexible residues within the active site. Six residues within 10 Å of the catalytic Ser(105) residue with a high B factor were selected for iterative saturation mutagenesis. After screening 2200 colonies, we obtained the D223G/L278M mutant, which exhibited a 13-fold increase in half-life at 48 °C and a 12 °C higher T50(15), the temperature at which enzyme activity is reduced to 50% after a 15-min heat treatment. Further characterization showed that global unfolding resistance against both thermal and chemical denaturation also improved. Analysis of the crystal structures of wild-type CalB and the D223G/L278M mutant revealed that the latter formed an extra main chain hydrogen bond network with seven structurally coupled residues within the flexible α10 helix that are primarily involved in forming the active site. Further investigation of the relative B factor profile and molecular dynamics simulation confirmed that the enhanced rigidity decreased fluctuation of the active site residues at high temperature. These results indicate that enhancing the rigidity of the flexible segment within the active site may provide an efficient method for improving enzyme kinetic stability.

Tracing primordial protein evolution through structurally guided stepwise segment elongation

The understanding of how primordial proteins emerged has been a fundamental and longstanding issue in biology and biochemistry. For a better understanding of primordial protein evolution, we synthesized an artificial protein on the basis of an evolutionary hypothesis, segment-based elongation starting from an autonomously foldable short peptide. A 10-residue protein, chignolin, the smallest foldable polypeptide ever reported, was used as a structural support to facilitate higher structural organization and gain-of-function in the development of an artificial protein. Repetitive cycles of segment elongation and subsequent phage display selection successfully produced a 25-residue protein, termed AF.2A1, with nanomolar affinity against the Fc region of immunoglobulin G. AF.2A1 shows exquisite molecular recognition ability such that it can distinguish conformational differences of the same molecule. The structure determined by NMR measurements demonstrated that AF.2A1 forms a globular protein-like conformation with the chignolin-derived β-hairpin and a tryptophan-mediated hydrophobic core. Using sequence analysis and a mutation study, we discovered that the structural organization and gain-of-function emerged from the vicinity of the chignolin segment, revealing that the structural support served as the core in both structural and functional development. Here, we propose an evolutionary model for primordial proteins in which a foldable segment serves as the evolving core to facilitate structural and functional evolution. This study provides insights into primordial protein evolution and also presents a novel methodology for designing small sized proteins useful for industrial and pharmaceutical applications.

Circular permutation directs orthogonal assembly in complex collagen peptide mixtures

Multiple types of natural collagens specifically assemble and co-exist in the extracellular matrix. Although noncollagenous trimerization domains facilitate the folding of triple-helical regions, it is intriguing to ask whether collagen sequences are also capable of controlling heterospecific association. In this study, we designed a model system mimicking simultaneous specific assembly of two collagen heterotrimers using a genetically inspired operation, circular permutation. Previously, surface charge-pair interactions were optimized on three collagen peptides to promote the formation of an abc-type heterotrimer. Circular permutation of these sequences retained networks of stabilizing interactions, preserving both triple-helical structure and heterospecificity of assembly. Combining original peptides A, B, and C and permuted peptides D, E, and F resulted primarily in formation of A:B:C and D:E:F, a heterospecificity of 2 of 56 possible stoichiometries. This degree of specificity in collagen molecular recognition is unprecedented in natural or synthetic collagens. Analysis of natural collagen sequences indicates low similarity between the neighboring exons. Combining the synthetic collagen model and bioinformatic analysis provides insight on how fibrillar collagens might have arisen from the duplication of smaller domains.

Engineering the pattern of protein glycosylation modulates the thermostability of a GH11 xylanase

Protein glycosylation is a common post-translational modification, the effect of which on protein conformational and stability is incompletely understood. Here we have investigated the effects of glycosylation on the thermostability of Bacillus subtilis xylanase A (XynA) expressed in Pichia pastoris. Intact mass analysis of the heterologous wild-type XynA revealed two, three, or four Hex(8-16)GlcNAc2 modifications involving asparagine residues at positions 20, 25, 141, and 181. Molecular dynamics (MD) simulations of the XynA modified with various combinations of branched Hex9GlcNAc2 at these positions indicated a significant contribution from protein-glycan interactions to the overall energy of the glycoproteins. The effect of glycan content and glycosylation position on protein stability was evaluated by combinatorial mutagenesis of all six potential N-glycosylation sites. The majority of glycosylated enzymes expressed in P. pastoris presented increased thermostability in comparison with their unglycosylated counterparts expressed in Escherichia coli. Steric effects of multiple glycosylation events were apparent, and glycosylation position rather than the number of glycosylation events determined increases in thermostability. The MD simulations also indicated that clustered glycan chains tended to favor less stabilizing glycan-glycan interactions, whereas more dispersed glycosylation patterns favored stabilizing protein-glycan interactions.

Allosteric coupling in the bacterial adhesive protein FimH

The protein FimH is expressed by the majority of commensal and uropathogenic strains of Escherichia coli on the tips of type 1 fimbriae and mediates adhesion via a catch bond to its ligand mannose. Crystal structures of FimH show an allosteric conformational change, but it remains unclear whether all of the observed structural differences are part of the allosteric mechanism. Here we use the protein structural analysis tool RosettaDesign combined with human insight to identify and synthesize 10 mutations in four regions that we predicted would stabilize one of the conformations of that region. The function of each variant was characterized by measuring binding to the ligand mannose, whereas the allosteric state was determined using a conformation-specific monoclonal antibody. These studies demonstrated that each region investigated was indeed part of the FimH allosteric mechanism. However, the studies strongly suggested that some regions were more tightly coupled to mannose binding and others to antibody binding. In addition, we identified many FimH variants that appear locked in the low affinity state. Knowledge of regulatory sites outside the active and effector sites as well as the ability to make FimH variants locked in the low affinity state may be crucial to the future development of novel antiadhesive and antimicrobial therapies using allosteric regulation to inhibit FimH.

Anti-antimicrobial peptides: folding-mediated host defense antagonists

Antimicrobial or host defense peptides are innate immune regulators found in all multicellular organisms. Many of them fold into membrane-bound α-helices and function by causing cell wall disruption in microorganisms. Herein we probe the possibility and functional implications of antimicrobial antagonism mediated by complementary coiled-coil interactions between antimicrobial peptides and de novo designed antagonists: anti-antimicrobial peptides. Using sequences from native helical families such as cathelicidins, cecropins, and magainins we demonstrate that designed antagonists can co-fold with antimicrobial peptides into functionally inert helical oligomers. The properties and function of the resulting assemblies were studied in solution, membrane environments, and in bacterial culture by a combination of chiroptical and solid-state NMR spectroscopies, microscopy, bioassays, and molecular dynamics simulations. The findings offer a molecular rationale for anti-antimicrobial responses with potential implications for antimicrobial resistance.

Identification of a highly conserved surface on Tat variants

Extracellular Tat is suspected to protect HIV-1-infected cells from cellular immunity. Seropositive patients are unable to produce neutralizing antibodies against Tat, and Tat is still secreted under antiviral treatment. In mice, the Tat OYI vaccine candidate generates neutralizing antibodies such as the mAb 7G12. A peptide called MIMOOX was designed from fragments of Tat OYI identified as the possible binding site for mAb 7G12. MIMOOX was chemically synthesized, and its structure was stabilized with a disulfide bridge. Circular dichroism spectra showed that MIMOOX had mainly β turns but no α helix as Tat OYI. MIMOOX was recognized by mAb 7G12 in ELISA only in reduced conditions. Moreover, a competitive recognition assay with mAb 7G12 between MIMOOX and Tat variants showed that MIMOOX mimics a highly conserved surface in Tat variants. Rat immunizations with MIMOOX induce antibodies recognizing Tat variants from the main HIV-1 subtypes and confirm the Tat OYI vaccine approach.