Design of novel granulopoietic proteins by topological rescaffolding.

Computational protein design is rapidly becoming more powerful, and improving the accuracy of computational methods would greatly streamline protein engineering by eliminating the need for empirical optimization in the laboratory. In this work, we set out to design novel granulopoietic agents using a rescaffolding strategy with the goal of achieving simpler and more stable proteins. All of the 4 experimentally tested designs were folded, monomeric, and stable, while the 2 determined structures agreed with the design models within less than 2.5 Å. Despite the lack of significant topological or sequence similarity to their natural granulopoietic counterpart, 2 designs bound to the granulocyte colony-stimulating factor (G-CSF) receptor and exhibited potent, but delayed, in vitro proliferative activity in a G-CSF-dependent cell line. Interestingly, the designs also induced proliferation and differentiation of primary human hematopoietic stem cells into mature granulocytes, highlighting the utility of our approach to develop highly active therapeutic leads purely based on computational design.

The breakthrough in protein structure prediction.

Proteins are the essential agents of all living systems. Even though they are synthesized as linear chains of amino acids, they must assume specific three-dimensional structures in order to manifest their biological activity. These structures are fully specified in their amino acid sequences - and therefore in the nucleotide sequences of their genes. However, the relationship between sequence and structure, known as the protein folding problem, has remained elusive for half a century, despite sustained efforts. To measure progress on this problem, a series of doubly blind, biennial experiments called CASP (critical assessment of structure prediction) were established in 1994. We were part of the assessment team for the most recent CASP experiment, CASP14, where we witnessed an astonishing breakthrough by DeepMind, the leading artificial intelligence laboratory of Alphabet Inc. The models filed by DeepMind's structure prediction team using the program AlphaFold2 were often essentially indistinguishable from experimental structures, leading to a consensus in the community that the structure prediction problem for single protein chains has been solved. Here, we will review the path to CASP14, outline the method employed by AlphaFold2 to the extent revealed, and discuss the implications of this breakthrough for the life sciences.

Reconstructing the Remote Origins of a Fold Singleton from a Flavodoxin-Like Ancestor.

Evolutionary processes that led to the emergence of structured protein domains left footprints in the sequences of modern proteins. We searched for such hints employing state-of-the-art sequence analysis and found evidence that the HemD-like fold emerged from the flavodoxin-like fold through segment swap and gene duplication. To verify this hypothesis, we reverted these evolutionary steps experimentally, constructing a HemD-half that resulted in a protein with the canonical flavodoxin-like architecture. These results of fold reconstruction from the sequence of a different fold strongly support our hypothesis of common ancestry. It further illustrates the plasticity of modern proteins to form new folded proteins.

Asymmetric protein design from conserved supersecondary structures.

Computational design with supersecondary structures as building blocks has proven effective in the construction of new proteins with controlled geometries. So far, this approach has primarily exploited amplification, effectively harnessing the internal folding propensity of self-compatible fragments to achieve sufficient enthalpy for folding. Here we exploit an interface-driven strategy to depart from the repeat design realm, constructing an asymmetric, globular domain from heterologous supersecondary structures. We report the successful design of a dRP lyase domain fold, which agrees with the experimental NMR structure at atomic accuracy (backbone RMSD of 0.94 Å). Our results show that the residual folding information within conserved fragments, combined with efficient interface-directed sampling, can effectively yield globular proteins with novel sequences and biophysical properties.

The Intimin periplasmic domain mediates dimerisation and binding to peptidoglycan.

Intimin and Invasin are prototypical inverse (Type Ve) autotransporters and important virulence factors of enteropathogenic Escherichia coli and Yersinia spp. respectively. In addition to a C-terminal extracellular domain and a ß-barrel transmembrane domain, both proteins also contain a short N-terminal periplasmic domain that, in Intimin, includes a lysin motif (LysM), which is thought to mediate binding to peptidoglycan. We show that the periplasmic domain of Intimin does bind to peptidoglycan both in vitro and in vivo, but only under acidic conditions. We were able to determine a dissociation constant of 0.8 µM for this interaction, whereas the Invasin periplasmic domain, which lacks a LysM, bound only weakly in vitro and failed to bind peptidoglycan in vivo. We present the solution structure of the Intimin LysM, which has an additional α-helix conserved within inverse autotransporter LysMs but lacking in others. In contrast to previous reports, we demonstrate that the periplasmic domain of Intimin mediates dimerisation. We further show that dimerisation and peptidoglycan binding are general features of LysM-containing inverse autotransporters. Peptidoglycan binding by the periplasmic domain in the infection process may aid in resisting mechanical and chemical stress during transit through the gastrointestinal tract.

Distinct mechanisms contribute to immunity in the lantibiotic NAI-107 producer strain Microbispora ATCC PTA-5024.

The investigation of self-resistance in antibiotic producers is important to understand the emergence of antibiotic resistance in pathogens and to improve antibiotic production. Lantibiotics are ribosomally synthesized antibiotics that mostly target peptidoglycan biosynthesis. The actinomycete Microbispora ATCC PTA-5024 produces the lantibiotic NAI-107, which interferes with peptidoglycan biosynthesis by binding bactoprenol-pyrophosphate-coupled peptidoglycan precursors. In order to understand how Microbispora counteracts the action of its own antibiotic, its peptidoglycan composition was analysed in detail. Microbispora peptidoglycan consists of muropeptides with D-Ala and Gly in similar proportion at the fourth position of the peptide stems and alternative 3-3 cross-links besides the classical 4-3 cross-links. In addition, the NAI-107 biosynthetic gene cluster (mlb) was analysed for the expression of immunity proteins. We show that distinct immunity determinants are encoded in the mlb cluster: the ABC transporter MlbYZ acting cooperatively with the transmembrane protein MlbJ and the lipoprotein MlbQ. NMR structural analysis of MlbQ revealed a hydrophobic surface patch, which is proposed to bind the cognate lantibiotic. This study demonstrates that immunity in Microbispora is not only based on one determinant but on the action of the distinct immunity proteins MlbQ, MlbYZ and MlbJ.

A conserved spider silk domain acts as a molecular switch that controls fibre assembly.

A huge variety of proteins are able to form fibrillar structures, especially at high protein concentrations. Hence, it is surprising that spider silk proteins can be stored in a soluble form at high concentrations and transformed into extremely stable fibres on demand. Silk proteins are reminiscent of amphiphilic block copolymers containing stretches of polyalanine and glycine-rich polar elements forming a repetitive core flanked by highly conserved non-repetitive amino-terminal and carboxy-terminal domains. The N-terminal domain comprises a secretion signal, but further functions remain unassigned. The C-terminal domain was implicated in the control of solubility and fibre formation initiated by changes in ionic composition and mechanical stimuli known to align the repetitive sequence elements and promote beta-sheet formation. However, despite recent structural data, little is known about this remarkable behaviour in molecular detail. Here we present the solution structure of the C-terminal domain of a spider dragline silk protein and provide evidence that the structural state of this domain is essential for controlled switching between the storage and assembly forms of silk proteins. In addition, the C-terminal domain also has a role in the alignment of secondary structural features formed by the repetitive elements in the backbone of spider silk proteins, which is known to be important for the mechanical properties of the fibre.

Crystallographic snapshot of the Escherichia coli EnvZ histidine kinase in an active conformation.

Sensor histidine kinases are important sensors of the extracellular environment and relay signals via conformational changes that trigger autophosphorylation of the kinase and subsequent phosphorylation of a response regulator. The exact mechanism and the regulation of this protein family are a matter of ongoing investigation. Here we present a crystal structure of a functional chimeric protein encompassing the entire catalytic part of the Escherichia coli EnvZ histidine kinase, fused to the HAMP domain of the Archaeoglobus fulgidus Af1503 receptor. The construct is thus equivalent to the full cytosolic part of EnvZ. The structure shows a putatively active conformation of the catalytic domain and gives insight into how this conformation could be brought about in response to sensory input. Our analysis suggests a sequential flip-flop autokinase mechanism.

A soluble mutant of the transmembrane receptor Af1503 features strong changes in coiled-coil periodicity.

Structures of full-length, membrane-bound proteins are essential for understanding transmembrane signaling mechanisms. However, in prokaryotic receptors no such structure has been reported, despite active research for many years. Here we present results of an alternative strategy, whereby a transmembrane receptor is made soluble by selective mutations to the membrane-spanning region, chosen by analysis of helix geometry in the transmembrane regions of chemotaxis receptors. We thus converted the receptor Af1503 from Archaeoglobus fulgidus to a soluble form by deleting transmembrane helix 1 and mutating the surface residues of transmembrane helix 2 to hydrophilic amino acids. Crystallization of this protein resulted in the structure of a tetrameric proteolytic fragment representing the modified transmembrane helices plus the cytoplasmic HAMP domain, a ubiquitous domain of prokaryotic signal transducers. The protein forms a tetramer via native parallel dimerization of the HAMP domain and non-native antiparallel dimerization of the modified transmembrane helices. The latter results in a four-helical coiled coil, characterized by unusually large changes in helix periodicity. The structure offers the first view of the junction between the transmembrane region and HAMP and explains the conservation of a key sequence motif in HAMP domains.

Thalidomide mimics uridine binding to an aromatic cage in cereblon.

Thalidomide and its derivatives lenalidomide and pomalidomide are important anticancer agents but can cause severe birth defects via an interaction with the protein cereblon. The ligand-binding domain of cereblon is found, with a high degree of conservation, in both bacteria and eukaryotes. Using a bacterial model system, we reveal the structural determinants of cereblon substrate recognition, based on a series of high-resolution crystal structures. For the first time, we identify a cellular ligand that is universally present: we show that thalidomide and its derivatives mimic and compete for the binding of uridine, and validate these findings in vivo. The nature of the binding pocket, an aromatic cage of three tryptophan residues, further suggests a role in the recognition of cationic ligands. Our results allow for general evaluation of pharmaceuticals for potential cereblon-dependent teratogenicity.

Your personalized protein structure: Andrei N. Lupas fused to GCN4 adaptors.

This work presents a protein structure that has been designed purely for aesthetic reasons, symbolizing decades of coiled-coil research and praising its most fundamental model system, the GCN4 leucine zipper. The GCN4 leucine zipper is a highly stable coiled coil which can be tuned to adopt different oligomeric states via mutation of its core residues. For these reasons it is used in structural studies as a stabilizing fusion adaptor. On the occasion of the 50th birthday of Andrei N. Lupas, we used it to create the first personalized protein structure: we fused the sequence ANDREI-N-LVPAS in heptad register to trimeric GCN4 adaptors and determined its structure by X-ray crystallography. The structure demonstrates the robustness and versatility of GCN4 as a fusion adaptor. We learn how proline can be accommodated in trimeric coiled coils, and put the structure into the context of the other GCN4-fusion structures known to date.

Potential of fragment recombination for rational design of proteins.

It is hypothesized that protein domains evolved from smaller intrinsically stable subunits via combinatorial assembly. Illegitimate recombination of fragments that encode protein subunits could have quickly led to diversification of protein folds and their functionality. This evolutionary concept presents an attractive strategy to protein engineering, e.g., to create new scaffolds for enzyme design. We previously combined structurally similar parts from two ancient protein folds, the (ßα)(8)-barrel and the flavodoxin-like fold. The resulting "hopeful monster" differed significantly from the intended (ßα)(8)-barrel fold by an extra ß-strand in the core. In this study, we ask what modifications are necessary to form the intended structure and what potential this approach has for the rational design of functional proteins. Guided by computational design, we optimized the interface between the fragments with five targeted mutations yielding a stable, monomeric protein whose predicted structure was verified experimentally. We further tested binding of a phosphorylated compound and detected that some affinity was already present due to an intact phosphate-binding site provided by one fragment. The affinity could be improved quickly to the level of natural proteins by introducing two additional mutations. The study illustrates the potential of recombining protein fragments with unique properties to design new and functional proteins, offering both a possible pathway of protein evolution and a protocol to rapidly engineer proteins for new applications.

Screening of a Halogen-Enriched Fragment Library Leads to Unconventional Binding Modes.

We conceived the Halogen-Enriched Fragment Library (HEFLib) to investigate the potential of halogen bonds in the early stages of drug discovery. As the number of competitive interactions increases with ligand size, we reasoned that a binding mode relying on halogen bonding is more likely for fragments than highly decorated molecules. Thus, fragments could feature unexplored binding modes. We screened the HEFLib against the human kinase DYRK1a and verified micromolar binding fragments via isothermal titration calorimetry (ITC). The crystal structure of one fragment revealed a noncanonical binding mode, despite the fragment's classical hinge binding motif. In addition, the fragment occupies a secondary binding site. Both binding modes feature a halogen bond, which we evaluated by ab initio calculations. Structure-affinity relationship (SAR) from a set of analogues improves the affinity, provides a promising fragment-growth vector, and highlights the benefits and applicability of halogen bonds in early lead development.

Revisiting a challenging p53 binding site: a diversity-optimized HEFLib reveals diverse binding modes in T-p53C-Y220C.

The cellular tumor antigen p53 is a key component in cell cycle control. The mutation Y220C heavily destabilizes the protein thermally but yields a druggable crevice. We have screened the diversity-optimized halogen-enriched fragment library against T-p53C-Y220C with STD-NMR and DSF to identify hits, which we validated by 1H,15N-HSQC NMR. We could identify four hits binding in the Y220C cleft, one hit binding covalently and four hits binding to an uncharacterized binding site. Compound 1151 could be crystallized showing a flip of C220 and thus opening subsite 3. Additionally, 4482 was identified to alkylate cysteines. Data shows that the diversity-optimized HEFLib leads to multiple diverse hits. The identified scaffolds can be used to further optimize interactions with T-p53C-Y220C and increase thermal stability.

A topological refactoring design strategy yields highly stable granulopoietic proteins.

Protein therapeutics frequently face major challenges, including complicated production, instability, poor solubility, and aggregation. De novo protein design can readily address these challenges. Here, we demonstrate the utility of a topological refactoring strategy to design novel granulopoietic proteins starting from the granulocyte-colony stimulating factor (G-CSF) structure. We change a protein fold by rearranging the sequence and optimising it towards the new fold. Testing four designs, we obtain two that possess nanomolar activity, the most active of which is highly thermostable and protease-resistant, and matches its designed structure to atomic accuracy. While the designs possess starkly different sequence and structure from the native G-CSF, they show specific activity in differentiating primary human haematopoietic stem cells into mature neutrophils. The designs also show significant and specific activity in vivo. Our topological refactoring approach is largely independent of sequence or structural context, and is therefore applicable to a wide range of protein targets.

A glutamine-based single α-helix scaffold to target globular proteins.

The binding of intrinsically disordered proteins to globular ones can require the folding of motifs into α-helices. These interactions offer opportunities for therapeutic intervention but their modulation with small molecules is challenging because they bury large surfaces. Linear peptides that display the residues that are key for binding can be targeted to globular proteins when they form stable helices, which in most cases requires their chemical modification. Here we present rules to design peptides that fold into single α-helices by instead concatenating glutamine side chain to main chain hydrogen bonds recently discovered in polyglutamine helices. The resulting peptides are uncharged, contain only natural amino acids, and their sequences can be optimized to interact with specific targets. Our results provide design rules to obtain single α-helices for a wide range of applications in protein engineering and drug design.

Cradle-loop barrels and the concept of metafolds in protein classification by natural descent.

Current classification systems for protein structure show many inconsistencies both within and between systems. The metafold concept was introduced to identify fold similarities by consensus and thus provide a more unified view of fold space. Using cradle-loop barrels as an example, we propose to use the metafold as the next hierarchical level above the fold, encompassing a group of topologically related folds for which a homologous relationship has been substantiated. We see this as an important step on the way to a classification of proteins by natural descent.

Diagonal-free 3D/4D HN,HN-TROSY-NOESY-TROSY.

Structural biology by NMR spectroscopy relies on measuring interproton distances via NOE cross-signals in nuclear Overhauser effect spectroscopy (NOESY) spectra. In proteins, the subset of H(N)-H'(N) NOE contacts is most important for deriving initial structural models and for spectral assignment by "NOE walking". Here we present a fully optimized NMR experiment for measuring these pivotal contacts: diagonal-free 3D/4D HN,HN-TROSY-NOESY-TROSY. It combines all of the critical requirements for extracting the optimal H(N)-H'(N) distance information: the highest resolution by consistent transverse relaxation-optimized spectroscopy (TROSY) evolution, the largest spectral dispersion in two (15)N dimensions, and maximal coverage and purity through specific suppression of the intense diagonal signals that are the main source of overlap, artifacts, and bias in any NOESY spectrum. Most notably, diagonal suppression here comes without compromising the NOE cross-signal intensities. This optimized experiment appears to be ideal for a broad range of structural studies, particularly on large deuterated, partially unfolded, helical, and membrane proteins.

Optimized measurement temperature gives access to the solution structure of a 49 kDa homohexameric ß-propeller.

Ph1500 is a homohexameric, two-domain protein of unknown function from the hyperthermophilic archaeon Pyrococcus horikoshii. The C-terminal hexamerization domain (Ph1500C) is of particular interest, as it lacks sequence homology to proteins of known structure. However, it resisted crystallization for X-ray analysis, and proteins of this size (49 kDa) present a considerable challenge to NMR structure determination in solution. We solved the high-resolution structure of Ph1500C, exploiting the hyperthermophilic nature of the protein to minimize unfavorable relaxation properties by high-temperature measurement. Thus, the side chain assignment (97%) and structure determination became possible at full proton density. To our knowledge, Ph1500C is the largest protein for which this has been achieved. To minimize detrimental fast water exchange of amide protons at increased temperature, we employed a strategy where the temperature was optimized separately for backbone and side chain experiments.

SimShiftDB; local conformational restraints derived from chemical shift similarity searches on a large synthetic database.

We present SimShiftDB, a new program to extract conformational data from protein chemical shifts using structural alignments. The alignments are obtained in searches of a large database containing 13,000 structures and corresponding back-calculated chemical shifts. SimShiftDB makes use of chemical shift data to provide accurate results even in the case of low sequence similarity, and with even coverage of the conformational search space. We compare SimShiftDB to HHSearch, a state-of-the-art sequence-based search tool, and to TALOS, the current standard tool for the task. We show that for a significant fraction of the predicted similarities, SimShiftDB outperforms the other two methods. Particularly, the high coverage afforded by the larger database often allows predictions to be made for residues not involved in canonical secondary structure, where TALOS predictions are both less frequent and more error prone. Thus SimShiftDB can be seen as a complement to currently available methods.