CrysFormer: Protein structure determination via Patterson maps, deep learning, and partial structure attention.

Determining the atomic-level structure of a protein has been a decades-long challenge. However, recent advances in transformers and related neural network architectures have enabled researchers to significantly improve solutions to this problem. These methods use large datasets of sequence information and corresponding known protein template structures, if available. Yet, such methods only focus on sequence information. Other available prior knowledge could also be utilized, such as constructs derived from x-ray crystallography experiments and the known structures of the most common conformations of amino acid residues, which we refer to as partial structures. To the best of our knowledge, we propose the first transformer-based model that directly utilizes experimental protein crystallographic data and partial structure information to calculate electron density maps of proteins. In particular, we use Patterson maps, which can be directly obtained from x-ray crystallography experimental data, thus bypassing the well-known crystallographic phase problem. We demonstrate that our method, CrysFormer, achieves precise predictions on two synthetic datasets of peptide fragments in crystalline forms, one with two residues per unit cell and the other with fifteen. These predictions can then be used to generate accurate atomic models using established crystallographic refinement programs.

Heterotrimeric collagen helix with high specificity of assembly results in a rapid rate of folding.

The most abundant natural collagens form heterotrimeric triple helices. Synthetic mimics of collagen heterotrimers have been found to fold slowly, even compared to the already slow rates of homotrimeric helices. These prolonged folding rates are not understood. Here we compare the stabilities, specificities and folding rates of three heterotrimeric collagen mimics designed through a computationally assisted approach. The crystal structure of one ABC-type heterotrimer verified a well-controlled composition and register and elucidated the geometry of pairwise cation-π and axial and lateral salt bridges in the assembly. This collagen heterotrimer folds much faster (hours versus days) than comparable, well-designed systems. Circular dichroism and NMR data suggest the folding is frustrated by unproductive, competing heterotrimer species and these species must unwind before refolding into the thermodynamically favoured assembly. The heterotrimeric collagen folding rate is inhibited by the introduction of preformed competing triple-helical assemblies, which suggests that slow heterotrimer folding kinetics are dominated by the frustration of the energy landscape caused by competing triple helices.

Crystal structure of the photosensory module from a PAS-less cyanobacterial phytochrome as Pr shows a mix of dark-adapted and photoactivated features.

Phytochromes (Phys) are a diverse collection of photoreceptors that regulate numerous physiological and developmental processes in microorganisms and plants through photointerconversion between red-light-absorbing Pr and far-red light-absorbing Pfr states. Light is detected by an N-terminal photo-sensing module (PSM) sequentially comprised of Period/ARNT/Sim (PAS), cGMP-phosphodiesterase/adenylyl cyclase/FhlA (GAF), and Phy-specific (PHY) domains, with the bilin chromophore covalently-bound within the GAF domain. Phys sense light via the Pr/Pfr ratio measured by the light-induced rotation of the bilin D-pyrrole ring that triggers conformational changes within the PSM, which for microbial Phys reaches into an output region. A key step is a ß-stranded to α-helical reconfiguration of a hairpin loop extending from the PHY domain to contact the GAF domain. Besides canonical Phys, cyanobacteria express several variants, including a PAS-less subfamily that harbors just the GAF and PHY domains for light detection. Prior 2D-NMR studies of a model PAS-less Phy from Synechococcus_sp._JA-2-3B'a(2-13) (SyB-Cph1) proposed a unique photoconversion mechanism involving an A-pyrrole ring rotation while magic-angle-spinning NMR probing the chromophore proposed the prototypic D-ring flip. To help solve this conundrum, we determined the crystallographic structure of the GAF-PHY region from SyB-Cph1 as Pr. Surprisingly, this structure differs from canonical Phys by having a Pr ZZZsyn,syn,anti bilin configuration but shifted to the activated position in the binding pocket with consequent folding of the hairpin loop to α-helical, an architecture common for Pfr. Collectively, the PSM of SyB-Cph1 as Pr displayed a mix of dark-adapted and photoactivated features whose co-planar A-C pyrrole rings support a D-ring flip mechanism.

Biochemical and Structural Studies of the Carminomycin 4-O-Methyltransferase DnrK.

Structural and functional studies of the carminomycin 4-O-methyltransferase DnrK are described, with an emphasis on interrogating the acceptor substrate scope of DnrK. Specifically, the evaluation of 100 structurally and functionally diverse natural products and natural product mimetics revealed an array of pharmacophores as productive DnrK substrates. Representative newly identified DnrK substrates from this study included anthracyclines, angucyclines, anthraquinone-fused enediynes, flavonoids, pyranonaphthoquinones, and polyketides. The ligand-bound structure of DnrK bound to a non-native fluorescent hydroxycoumarin acceptor, 4-methylumbelliferone, along with corresponding DnrK kinetic parameters for 4-methylumbelliferone and native acceptor carminomycin are also reported for the first time. The demonstrated unique permissivity of DnrK highlights the potential for DnrK as a new tool in future biocatalytic and/or strain engineering applications. In addition, the comparative bioactivity assessment (cancer cell line cytotoxicity, 4E-BP1 phosphorylation, and axolotl embryo tail regeneration) of a select set of DnrK substrates/products highlights the ability of anthracycline 4-O-methylation to dictate diverse functional outcomes.

A deep learning solution for crystallographic structure determination.

The general de novo solution of the crystallographic phase problem is difficult and only possible under certain conditions. This paper develops an initial pathway to a deep learning neural network approach for the phase problem in protein crystallography, based on a synthetic dataset of small fragments derived from a large well curated subset of solved structures in the Protein Data Bank (PDB). In particular, electron-density estimates of simple artificial systems are produced directly from corresponding Patterson maps using a convolutional neural network architecture as a proof of concept.

Protein target highlights in CASP15: Analysis of models by structure providers.

We present an in-depth analysis of selected CASP15 targets, focusing on their biological and functional significance. The authors of the structures identify and discuss key protein features and evaluate how effectively these aspects were captured in the submitted predictions. While the overall ability to predict three-dimensional protein structures continues to impress, reproducing uncommon features not previously observed in experimental structures is still a challenge. Furthermore, instances with conformational flexibility and large multimeric complexes highlight the need for novel scoring strategies to better emphasize biologically relevant structural regions. Looking ahead, closer integration of computational and experimental techniques will play a key role in determining the next challenges to be unraveled in the field of structural molecular biology.

An NmrA-like enzyme-catalysed redox-mediated Diels-Alder cycloaddition with anti-selectivity.

The Diels-Alder cycloaddition is one of the most powerful approaches in organic synthesis and is often used in the synthesis of important pharmaceuticals. Yet, strictly controlling the stereoselectivity of the Diels-Alder reactions is challenging, and great efforts are needed to construct complex molecules with desired chirality via organocatalysis or transition-metal strategies. Nature has evolved different types of enzymes to exquisitely control cyclization stereochemistry; however, most of the reported Diels-Alderases have been shown to only facilitate the energetically favourable diastereoselective cycloadditions. Here we report the discovery and characterization of CtdP, a member of a new class of bifunctional oxidoreductase/Diels-Alderase, which was previously annotated as an NmrA-like transcriptional regulator. We demonstrate that CtdP catalyses the inherently disfavoured cycloaddition to form the bicyclo[2.2.2]diazaoctane scaffold with a strict α-anti-selectivity. Guided by computational studies, we reveal a NADP+/NADPH-dependent redox mechanism for the CtdP-catalysed inverse electron demand Diels-Alder cycloaddition, which serves as the first example of a bifunctional Diels-Alderase that utilizes this mechanism.

Expanding the eukaryotic genetic code with a biosynthesized 21st amino acid.

Genetic code expansion technology allows for the use of noncanonical amino acids (ncAAs) to create semisynthetic organisms for both biochemical and biomedical applications. However, exogenous feeding of chemically synthesized ncAAs at high concentrations is required to compensate for the inefficient cellular uptake and incorporation of these components into proteins, especially in the case of eukaryotic cells and multicellular organisms. To generate organisms capable of autonomously biosynthesizing an ncAA and incorporating it into proteins, we have engineered a metabolic pathway for the synthesis of O-methyltyrosine (OMeY). Specifically, we endowed organisms with a marformycins biosynthetic pathway-derived methyltransferase that efficiently converts tyrosine to OMeY in the presence of the co-factor S-adenosylmethionine. The resulting cells can produce and site-specifically incorporate OMeY into proteins at much higher levels than cells exogenously fed OMeY. To understand the structural basis for the substrate selectivity of the transferase, we solved the X-ray crystal structures of the ligand-free and tyrosine-bound enzymes. Most importantly, we have extended this OMeY biosynthetic system to both mammalian cells and the zebrafish model to enhance the utility of genetic code expansion. The creation of autonomous eukaryotes using a 21st amino acid will make genetic code expansion technology more applicable to multicellular organisms, providing valuable vertebrate models for biological and biomedical research.

The Structure of the Arabidopsis PEX4-PEX22 Peroxin Complex-Insights Into Ubiquitination at the Peroxisomal Membrane.

Peroxisomes are eukaryotic organelles that sequester critical oxidative reactions and process the resulting reactive oxygen species into less toxic byproducts. Peroxisome function and formation are coordinated by peroxins (PEX proteins) that guide peroxisome biogenesis and division and shuttle proteins into the lumen and membrane of the organelle. Despite the importance of peroxins in plant metabolism and development, no plant peroxin structures have been reported. Here we report the X-ray crystal structure of the PEX4-PEX22 peroxin complex from the reference plant Arabidopsis thaliana. PEX4 is a ubiquitin-conjugating enzyme (UBC) that ubiquitinates proteins associated with the peroxisomal membrane, and PEX22 is a peroxisomal membrane protein that anchors PEX4 to the peroxisome and facilitates PEX4 activity. We co-expressed Arabidopsis PEX4 as a translational fusion with the soluble PEX4-interacting domain of PEX22 in E. coli. The fusion was linked via a protease recognition site, allowing us to separate PEX4 and PEX22 following purification and solve the structure of the complex. We compared the structure of the PEX4-PEX22 complex to the previously published structures of yeast orthologs. Arabidopsis PEX4 displays the typical UBC structure expected from its sequence. Although Arabidopsis PEX22 lacks notable sequence identity to yeast PEX22, it maintains a similar Rossmann fold-like structure. Several salt bridges are positioned to contribute to the specificity of PEX22 for PEX4 versus other Arabidopsis UBCs, and the long unstructured PEX22 tether would allow PEX4-mediated ubiquitination of distant peroxisomal membrane targets without dissociation from PEX22. The Arabidopsis PEX4-PEX22 structure also revealed that the residue altered in pex4-1 (P123L), a mutant previously isolated via a forward-genetic screen for peroxisomal dysfunction, is near the active site cysteine of PEX4. We demonstrated in vitro UBC activity for the PEX4-PEX22 complex and found that the pex4-1 enzyme has reduced in vitro ubiquitin-conjugating activity and altered specificity compared to PEX4. Our findings illuminate the role of PEX4 and PEX22 in peroxisome structure and function and provide tools for future exploration of ubiquitination at the peroxisome surface.

The crystal structure of DynF from the dynemicin-biosynthesis pathway of Micromonospora chersina.

Dynemicin is an enediyne natural product from Micromonospora chersina ATCC53710. Access to the biosynthetic gene cluster of dynemicin has enabled the in vitro study of gene products within the cluster to decipher their roles in assembling this unique molecule. This paper reports the crystal structure of DynF, the gene product of one of the genes within the biosynthetic gene cluster of dynemicin. DynF is revealed to be a dimeric eight-stranded ß-barrel structure with palmitic acid bound within a cavity. The presence of palmitic acid suggests that DynF may be involved in binding the precursor polyene heptaene, which is central to the synthesis of the ten-membered ring of the enediyne core.

Observation of substrate diffusion and ligand binding in enzyme crystals using high-repetition-rate mix-and-inject serial crystallography.

Here, we illustrate what happens inside the catalytic cleft of an enzyme when substrate or ligand binds on single-millisecond timescales. The initial phase of the enzymatic cycle is observed with near-atomic resolution using the most advanced X-ray source currently available: the European XFEL (EuXFEL). The high repetition rate of the EuXFEL combined with our mix-and-inject technology enables the initial phase of ceftriaxone binding to the Mycobacterium tuberculosis ß-lactamase to be followed using time-resolved crystallography in real time. It is shown how a diffusion coefficient in enzyme crystals can be derived directly from the X-ray data, enabling the determination of ligand and enzyme-ligand concentrations at any position in the crystal volume as a function of time. In addition, the structure of the irreversible inhibitor sulbactam bound to the enzyme at a 66 ms time delay after mixing is described. This demonstrates that the EuXFEL can be used as an important tool for biomedically relevant research.

Structure and Function of a Dual Reductase-Dehydratase Enzyme System Involved in p-Terphenyl Biosynthesis.

We report the identification of the ter gene cluster responsible for the formation of the p-terphenyl derivatives terfestatins B and C and echoside B from the Appalachian Streptomyces strain RM-5-8. We characterize the function of TerB/C, catalysts that work together as a dual enzyme system in the biosynthesis of natural terphenyls. TerB acts as a reductase and TerC as a dehydratase to enable the conversion of polyporic acid to a terphenyl triol intermediate. X-ray crystallography of the apo and substrate-bound forms for both enzymes provides additional mechanistic insights. Validation of the TerC structural model via mutagenesis highlights a critical role of arginine 143 and aspartate 173 in catalysis. Cumulatively, this work highlights a set of enzymes acting in harmony to control and direct reactive intermediates and advances fundamental understanding of the previously unresolved early steps in terphenyl biosynthesis.

Structural characterization of DynU16, a START/Bet v1-like protein involved in dynemicin biosynthesis.

The 1.5 Šresolution crystal structure of DynU16, a protein identified in the dynemicin-biosynthetic gene cluster, is reported. The structure adopts a di-domain helix-grip fold with a uniquely positioned open cavity connecting the domains. The elongated dimensions of the cavity appear to be compatible with the geometry of a linear polyene, suggesting the involvement of DynU16 in the upstream steps of dynemicin biosynthesis.

Moving beyond static snapshots: Protein dynamics and the Protein Data Bank.

Proteins are the molecular machines of living systems. Their dynamics are an intrinsic part of their evolutionary selection in carrying out their biological functions. Although the dynamics are more difficult to observe than a static, average structure, we are beginning to observe these dynamics and form sound mechanistic connections between structure, dynamics, and function. This progress is highlighted in case studies from myoglobin and adenylate kinase to the ribosome and molecular motors where these molecules are being probed with a multitude of techniques across many timescales. New approaches to time-resolved crystallography are allowing simple "movies" to be taken of proteins in action, and new methods of mapping the variations in cryo-electron microscopy are emerging to reveal a more complete description of life's machines. The results of these new methods are aided in their dissemination by continual improvements in curation and distribution by the Protein Data Bank and their partners around the world.

A collagen glucosyltransferase drives lung adenocarcinoma progression in mice.

Cancer cells are a major source of enzymes that modify collagen to create a stiff, fibrotic tumor stroma. High collagen lysyl hydroxylase 2 (LH2) expression promotes metastasis and is correlated with shorter survival in lung adenocarcinoma (LUAD) and other tumor types. LH2 hydroxylates lysine (Lys) residues on fibrillar collagen's amino- and carboxy-terminal telopeptides to create stable collagen cross-links. Here, we show that electrostatic interactions between the LH domain active site and collagen determine the unique telopeptidyl lysyl hydroxylase (tLH) activity of LH2. However, CRISPR/Cas-9-mediated inactivation of tLH activity does not fully recapitulate the inhibitory effect of LH2 knock out on LUAD growth and metastasis in mice, suggesting that LH2 drives LUAD progression, in part, through a tLH-independent mechanism. Protein homology modeling and biochemical studies identify an LH2 isoform (LH2b) that has previously undetected collagen galactosylhydroxylysyl glucosyltransferase (GGT) activity determined by a loop that enhances UDP-glucose-binding in the GLT active site and is encoded by alternatively spliced exon 13 A. CRISPR/Cas-9-mediated deletion of exon 13 A sharply reduces the growth and metastasis of LH2b-expressing LUADs in mice. These findings identify a previously unrecognized collagen GGT activity that drives LUAD progression.

The crystal structure of AbsH3: A putative flavin adenine dinucleotide-dependent reductase in the abyssomicin biosynthesis pathway.

Natural products and natural product-derived compounds have been widely used for pharmaceuticals for many years, and the search for new natural products that may have interesting activity is ongoing. Abyssomicins are natural product molecules that have antibiotic activity via inhibition of the folate synthesis pathway in microbiota. These compounds also appear to undergo a required [4 + 2] cycloaddition in their biosynthetic pathway. Here we report the structure of an flavin adenine dinucleotide-dependent reductase, AbsH3, from the biosynthetic gene cluster of novel abyssomicins found in Streptomyces sp. LC-6-2.

Molecular-replacement phasing using predicted protein structures from AWSEM-Suite.

The phase problem in X-ray crystallography arises from the fact that only the intensities, and not the phases, of the diffracting electromagnetic waves are measured directly. Molecular replacement can often estimate the relative phases of reflections starting with those derived from a template structure, which is usually a previously solved structure of a similar protein. The key factor in the success of molecular replacement is finding a good template structure. When no good solved template exists, predicted structures based partially on templates can sometimes be used to generate models for molecular replacement, thereby extending the lower bound of structural and sequence similarity required for successful structure determination. Here, the effectiveness is examined of structures predicted by a state-of-the-art prediction algorithm, the Associative memory, Water-mediated, Structure and Energy Model Suite (AWSEM-Suite), which has been shown to perform well in predicting protein structures in CASP13 when there is no significant sequence similarity to a solved protein or only very low sequence similarity to known templates. The performance of AWSEM-Suite structures in molecular replacement is discussed and the results show that AWSEM-Suite performs well in providing useful phase information, often performing better than I-TASSER-MR and the previous algorithm AWSEM-Template.

Covalent Capture of Collagen Triple Helices Using Lysine-Aspartate and Lysine-Glutamate Pairs.

Collagen mimetic peptides (CMPs) self-assemble into a triple helix reproducing the most fundamental aspect of the collagen structural hierarchy. They are therefore important for both further understanding this complex family of proteins and use in a wide range of biomaterials and biomedical applications. CMP self-assembly is complicated by a number of factors which limit the use of CMPs including their slow rate of folding, relatively poor monomer-trimer equilibrium, and the large number of competing species possible in heterotrimeric helices. All of these problems can be solved through the formation of isopeptide bonds between lysine and either aspartate or glutamate. These amino acids serve two purposes: they first direct self-assemble, allowing for composition and register control within the triple helix, and subsequently can be covalently linked, fixing the composition and register of the assembled structure without perturbing the triple helical conformation. This self-assembly and covalent capture are demonstrated here with four different triple helices. The formation of an isopeptide bond between lysine and glutamate (K-E) is shown to be a faster and higher yielding reaction than lysine with aspartate (K-D). Additionally, K-E amide bonds increase the thermal stability, improve the refolding capabilities, and enhance the triple helical structure as compared to K-E supramolecular interactions, observed by circular dichroism. In contrast, covalent capture of triple helices with K-D amide bonds occurs slower, and the captured triple helices do not have enhanced helical structure. The crystal structure of a triple helix captured through the formation of three K-E isopeptide bonds unequivocally demonstrates the connectivity of the amide bonds formed while also confirming the preservation of the canonical triple helix. The rate of reaction and yield for covalently captured K-E triple helices along with the excellent preservation of triple helical structure demonstrate that this approach can be used to effectively capture and stabilize this important biological motif for biological and biomedical applications.

Prochlorococcus phage ferredoxin: structural characterization and electron transfer to cyanobacterial sulfite reductases.

Marine cyanobacteria are infected by phages whose genomes encode ferredoxin (Fd) electron carriers. These Fds are thought to redirect the energy harvested from light to phage-encoded oxidoreductases that enhance viral fitness, but it is unclear how the biophysical properties and partner specificities of phage Fds relate to those of photosynthetic organisms. Here, results of a bioinformatics analysis using a sequence similarity network revealed that phage Fds are most closely related to cyanobacterial Fds that transfer electrons from photosystems to oxidoreductases involved in nutrient assimilation. Structural analysis of myovirus P-SSM2 Fd (pssm2-Fd), which infects the cyanobacterium Prochlorococcus marinus, revealed high levels of similarity to cyanobacterial Fds (root mean square deviations of ≤0.5 Å). Additionally, pssm2-Fd exhibited a low midpoint reduction potential (-336 mV versus a standard hydrogen electrode), similar to other photosynthetic Fds, although it had lower thermostability (Tm = 28 °C) than did many other Fds. When expressed in an Escherichia coli strain deficient in sulfite assimilation, pssm2-Fd complemented bacterial growth when coexpressed with a P. marinus sulfite reductase, revealing that pssm2-Fd can transfer electrons to a host protein involved in nutrient assimilation. The high levels of structural similarity with cyanobacterial Fds and reactivity with a host sulfite reductase suggest that phage Fds evolved to transfer electrons to cyanobacterially encoded oxidoreductases.

Methionine Adenosyltransferase Engineering to Enable Bioorthogonal Platforms for AdoMet-Utilizing Enzymes.

The structural conservation among methyltransferases (MTs) and MT functional redundancy is a major challenge to the cellular study of individual MTs. As a first step toward the development of an alternative biorthogonal platform for MTs and other AdoMet-utilizing enzymes, we describe the evaluation of 38 human methionine adenosyltransferase II-α (hMAT2A) mutants in combination with 14 non-native methionine analogues to identify suitable bioorthogonal mutant/analogue pairings. Enabled by the development and implementation of a hMAT2A high-throughput (HT) assay, this study revealed hMAT2A K289L to afford a 160-fold inversion of the hMAT2A selectivity index for a non-native methionine analogue over the native substrate l-Met. Structure elucidation of K289L revealed the mutant to be folded normally with minor observed repacking within the modified substrate pocket. This study highlights the first example of exchanging l-Met terminal carboxylate/amine recognition elements within the hMAT2A active-site to enable non-native bioorthgonal substrate utilization. Additionally, several hMAT2A mutants and l-Met substrate analogues produced AdoMet analogue products with increased stability. As many AdoMet-producing (e.g., hMAT2A) and AdoMet-utlizing (e.g., MTs) enzymes adopt similar active-site strategies for substrate recognition, the proof of concept first generation hMAT2A engineering highlighted herein is expected to translate to a range of AdoMet-utilizing target enzymes.