Deep learning-driven insights into super protein complexes for outer membrane protein biogenesis in bacteria

Predicting protein interactions of the kinase Lck critical to T cell modulation

Protein-protein interactions (PPIs) play pivotal roles in directing T cell fate. One key player is the non-receptor tyrosine protein kinase Lck that helps to transduce T cell activation signals. Lck is mediated by other proteins via interactions that are inadequately understood. Here, we use the deep learning method AF2Complex to predict PPIs involving Lck, by screening it against ∼1,000 proteins implicated in immune responses, followed by extensive structural modeling for selected interactions. Remarkably, we describe how Lck may be specifically targeted by a palmitoyltransferase using a phosphotyrosine motif. We uncover "hotspot" interactions between Lck and the tyrosine phosphatase CD45, leading to a significant conformational shift of Lck for activation. Lastly, we present intriguing interactions between the phosphotyrosine-binding domain of Lck and the cytoplasmic tail of the immune checkpoint LAG3 and propose a molecular mechanism for its inhibitory role. Together, this multifaceted study provides valuable insights into T cell regulation and signaling.

Improved deep learning prediction of antigen-antibody interactions

Identifying antibodies that neutralize specific antigens is crucial for developing effective immunotherapies, but this task remains challenging for many target antigens. The rise of deep learning-based computational approaches presents a promising avenue to address this challenge. Here, we assess the performance of a deep learning approach through two benchmark tests aimed at predicting antibodies for the receptor-binding domain of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike protein. Three different strategies for constructing input sequence alignments are employed for predicting structural models of antigen-antibody complexes. In our initial testing set, which comprises known experimental structures, these strategies collectively yield a significant top-ranked prediction for 61% of cases and a success rate of 47%. Notably, one strategy that utilizes the sequences of known antigen binders outperforms the other two, achieving a precision of 90% in a subsequent test set of ~1,000 antibodies, balanced between true and control antibodies for the antigen, albeit with a lower recall of 25%. Our results underscore the potential of integrating deep learning methods with single B cell sequencing techniques to enhance the prediction accuracy of antigen-antibody interactions.

Dual client binding sites in the ATP-independent chaperone SurA

The ATP-independent chaperone SurA protects unfolded outer membrane proteins (OMPs) from aggregation in the periplasm of Gram-negative bacteria, and delivers them to the β-barrel assembly machinery (BAM) for folding into the outer membrane (OM). Precisely how SurA recognises and binds its different OMP clients remains unclear. Escherichia coli SurA comprises three domains: a core and two PPIase domains (P1 and P2). Here, by combining methyl-TROSY NMR, single-molecule Förster resonance energy transfer (smFRET), and bioinformatics analyses we show that SurA client binding is mediated by two binding hotspots in the core and P1 domains. These interactions are driven by aromatic-rich motifs in the client proteins, leading to SurA core/P1 domain rearrangements and expansion of clients from collapsed, non-native states. We demonstrate that the core domain is key to OMP expansion by SurA, and uncover a role for SurA PPIase domains in limiting the extent of expansion. The results reveal insights into SurA-OMP recognition and the mechanism of activation for an ATP-independent chaperone, and suggest a route to targeting the functions of a chaperone key to bacterial virulence and OM integrity.

Outer membrane protein assembly mediated by BAM-SurA complexes

The outer membrane is a formidable barrier that protects Gram-negative bacteria against environmental threats. Its integrity requires the correct folding and insertion of outer membrane proteins (OMPs) by the membrane-embedded β-barrel assembly machinery (BAM). Unfolded OMPs are delivered to BAM by the periplasmic chaperone SurA, but how SurA and BAM work together to ensure successful OMP delivery and folding remains unclear. Here, guided by AlphaFold2 models, we use disulphide bond engineering in an attempt to trap SurA in the act of OMP delivery to BAM, and solve cryoEM structures of a series of complexes. The results suggest that SurA binds BAM at its soluble POTRA-1 domain, which may trigger conformational changes in both BAM and SurA that enable transfer of the unfolded OMP to the BAM lateral gate for insertion into the outer membrane. Mutations that disrupt the interaction between BAM and SurA result in outer membrane assembly defects, supporting the key role of SurA in outer membrane biogenesis.

Integrative modeling meets deep learning: Recent advances in modeling protein assemblies

Recent progress in protein structure prediction based on deep learning revolutionized the field of Structural Biology. Beyond single proteins, it also enabled high-throughput prediction of structures of protein-protein interactions. Despite the success in predicting complex structures, large macromolecular assemblies still require specialized approaches. Here we describe recent advances in modeling macromolecular assemblies using integrative and hierarchical approaches. We highlight applications that predict protein-protein interactions and challenges in modeling complexes based on the interaction networks, including the prediction of complex stoichiometry and heterogeneity.

A team of chaperones play to win in the bacterial periplasm

The survival and virulence of Gram-negative bacteria require proper biogenesis and maintenance of the outer membrane (OM), which is densely packed with β-barrel OM proteins (OMPs). Before reaching the OM, precursor unfolded OMPs (uOMPs) must cross the whole cell envelope. A network of periplasmic chaperones and proteases maintains unfolded but folding-competent conformations of these membrane proteins in the aqueous periplasm while simultaneously preventing off-pathway aggregation. These periplasmic proteins utilize different strategies, including conformational heterogeneity, oligomerization, multivalency, and kinetic partitioning, to perform and regulate their functions. Redundant and unique characteristics of the individual periplasmic players synergize to create a protein quality control team capable responding to changing environmental stresses.

Overview of AlphaFold2 and breakthroughs in overcoming its limitations

Predicting three-dimensional (3D) protein structures has been challenging for decades. The emergence of AlphaFold2 (AF2), a deep learning-based machine learning method developed by DeepMind, became a game changer in the protein folding community. AF2 can predict a protein's three-dimensional structure with high confidence based on its amino acid sequence. Accurate prediction of protein structures can dramatically accelerate our understanding of biological mechanisms and provide a solid foundation for reliable drug design. Although AF2 breaks through the barriers in predicting protein structures, many rooms remain to be further studied. This review provides a brief historical overview of the development of protein structure prediction, covering template-based, template-free, and machine learning-based methods. In addition to reviewing the potential benefits (Pros) and considerations (Cons) of using AF2, this review summarizes the diverse applications, including protein structure predictions, dynamic changes, point mutation, integration of language model and experimental data, protein complex, and protein-peptide interaction. It underscores recent advancements in efficiency, reliability, and broad application of AF2. This comprehensive review offers valuable insights into the applications of AF2 and AF2-inspired AI methods in structural biology and its potential for clinically significant drug target discovery.

Enhancing Cytochrome C Recognition and Adsorption through Epitope-Imprinted Mesoporous Silica with a Tailored Pore Size

We have reported the synthesis of epitope-imprinted mesoporous silica (EIMS) with an average pore size of 6.2 nm, which is similar to the geometrical size of the target protein, cytochrome C (Cyt c, 2.6 × 3.2 × 3.3 nm3), showing great recognition and large-scale adsorption performance. The characteristic fragment of Cyt c was used as a template and docked onto the surface of C16MIMCl micelles via multiple interactions. Nitrogen adsorption-desorption and transmission electron microscopy confirmed the successful preparation of EIMS. Due to the ordered pore structure, larger pore size, and high specific surface area, the prepared EIMS show superior specificity (IF = 3.8), excellent selectivity toward Cyt c, high adsorption capacity (249.6 mg g-1), and fast adsorption equilibrium (10 min). This study demonstrates the potential application of EIMS with a controllable pore size for high-effective and large-scale separation of Cyt c, providing a new approach for effective biomacromolecular recognition.

Interrogation and validation of the interactome of neuronal Munc18-interacting Mint proteins with AlphaFold2

Munc18-interacting proteins (Mints) are multidomain adaptors that regulate neuronal membrane trafficking, signaling, and neurotransmission. Mint1 and Mint2 are highly expressed in the brain with overlapping roles in the regulation of synaptic vesicle fusion required for neurotransmitter release by interacting with the essential synaptic protein Munc18-1. Here, we have used AlphaFold2 to identify and then validate the mechanisms that underpin both the specific interactions of neuronal Mint proteins with Munc18-1 as well as their wider interactome. We found that a short acidic α-helical motif within Mint1 and Mint2 is necessary and sufficient for specific binding to Munc18-1 and binds a conserved surface on Munc18-1 domain3b. In Munc18-1/2 double knockout neurosecretory cells, mutation of the Mint-binding site reduces the ability of Munc18-1 to rescue exocytosis, and although Munc18-1 can interact with Mint and Sx1a (Syntaxin1a) proteins simultaneously in vitro, we find that they have mutually reduced affinities, suggesting an allosteric coupling between the proteins. Using AlphaFold2 to then examine the entire cellular network of putative Mint interactors provides a structural model for their assembly with a variety of known and novel regulatory and cargo proteins including ADP-ribosylation factor (ARF3/ARF4) small GTPases and the AP3 clathrin adaptor complex. Validation of Mint1 interaction with a new predicted binder TJAP1 (tight junction-associated protein 1) provides experimental support that AlphaFold2 can correctly predict interactions across such large-scale datasets. Overall, our data provide insights into the diversity of interactions mediated by the Mint family and show that Mints may help facilitate a key trigger point in SNARE (soluble N-ethylmaleimide-sensitive factor attachment receptor) complex assembly and vesicle fusion.

Unrealized targets in the discovery of antibiotics for Gram-negative bacterial infections

Advances in areas that include genomics, systems biology, protein structure determination and artificial intelligence provide new opportunities for target-based antibacterial drug discovery. The selection of a 'good' new target for direct-acting antibacterial compounds is the first decision, for which multiple criteria must be explored, integrated and re-evaluated as drug discovery programmes progress. Criteria include essentiality of the target for bacterial survival, its conservation across different strains of the same species, bacterial species and growth conditions (which determines the spectrum of activity of a potential antibiotic) and the level of homology with human genes (which influences the potential for selective inhibition). Additionally, a bacterial target should have the potential to bind to drug-like molecules, and its subcellular location will govern the need for inhibitors to penetrate one or two bacterial membranes, which is a key challenge in targeting Gram-negative bacteria. The risk of the emergence of target-based drug resistance for drugs with single targets also requires consideration. This Review describes promising but as-yet-unrealized targets for antibacterial drugs against Gram-negative bacteria and examples of cognate inhibitors, and highlights lessons learned from past drug discovery programmes.

Atomic Models of All Major Trans-Envelope Complexes Involved in Lipid Trafficking in Escherichia Coli Constructed Using a Combination of AlphaFold2, AF2Complex, and Membrane Morphing Simulations

In Gram-negative bacteria, several trans-envelope complexes (TECs) have been identified that span the periplasmic space in order to facilitate lipid transport between the inner- and outer- membranes. While partial or near-complete structures of some of these TECs have been solved by conventional experimental techniques, most remain incomplete. Here we describe how a combination of computational approaches, constrained by experimental data, can be used to build complete atomic models for four TECs implicated in lipid transport in Escherichia coli . We use DeepMind's protein structure prediction algorithm, AlphaFold2, and a variant of it designed to predict protein complexes, AF2Complex, to predict the oligomeric states of key components of TECs and their likely interfaces with other components. After obtaining initial models of the complete TECs by superimposing predicted structures of subcomplexes, we use the membrane orientation prediction algorithm OPM to predict the likely orientations of the inner- and outer- membrane components in each TEC. Since, in all cases, the predicted membrane orientations in these initial models are tilted relative to each other, we devise a novel molecular mechanics-based strategy that we call "membrane morphing" that adjusts each TEC model until the two membranes are properly aligned with each other and separated by a distance consistent with estimates of the periplasmic width in E. coli . The study highlights the potential power of combining computational methods, operating within limits set by both experimental data and by cell physiology, for producing useable atomic structures of very large protein complexes.