Structures Of Protein Complexes Research Articles

BTB domain mutations perturbing KCTD15 oligomerisation cause a distinctive frontonasal dysplasia syndrome.

KCTD15 encodes an oligomeric BTB domain protein reported to inhibit neural crest formation through repression of Wnt/beta-catenin signalling, as well as transactivation by TFAP2. Heterozygous missense variants in the closely related paralogue KCTD1 cause scalp-ear-nipple syndrome. Exome sequencing was performed on a two-generation family affected by a distinctive phenotype comprising a lipomatous frontonasal malformation, anosmia, cutis aplasia of the scalp and/or sparse hair, and congenital heart disease. Identification of a de novo missense substitution within KCTD15 led to targeted sequencing of DNA from a similarly affected sporadic patient, revealing a different missense mutation. Structural and biophysical analyses were performed to assess the effects of both amino acid substitutions on the KCTD15 protein. A heterozygous c.310G>C variant encoding p.(Asp104His) within the BTB domain of KCTD15 was identified in an affected father and daughter and segregated with the phenotype. In the sporadically affected patient, a de novo heterozygous c.263G>A variant encoding p.(Gly88Asp) was present in KCTD15. Both substitutions were found to perturb the pentameric assembly of the BTB domain. A crystal structure of the BTB domain variant p.(Gly88Asp) revealed a closed hexameric assembly, whereas biophysical analyses showed that the p.(Asp104His) substitution resulted in a monomeric BTB domain likely to be partially unfolded at physiological temperatures. BTB domain substitutions in KCTD1 and KCTD15 cause clinically overlapping phenotypes involving craniofacial abnormalities and cutis aplasia. The structural analyses demonstrate that missense substitutions act through a dominant negative mechanism by disrupting the higher order structure of the KCTD15 protein complex.

DEMO-EM2: assembling protein complex structures from cryo-EM maps through intertwined chain and domain fitting.

The breakthrough in cryo-electron microscopy (cryo-EM) technology has led to an increasing number of density maps of biological macromolecules. However, constructing accurate protein complex atomic structures from cryo-EM maps remains a challenge. In this study, we extend our previously developed DEMO-EM to present DEMO-EM2, an automated method for constructing protein complex models from cryo-EM maps through an iterative assembly procedure intertwining chain- and domain-level matching and fitting for predicted chain models. The method was carefully evaluated on 27 cryo-electron tomography (cryo-ET) maps and 16 single-particle EM maps, where DEMO-EM2 models achieved an average TM-score of 0.92, outperforming those of state-of-the-art methods. The results demonstrate an efficient method that enables the rapid and reliable solution of challenging cryo-EM structure modeling problems.

Structure of a tripartite protein complex that targets toxins to the type VII secretion system.

Type VII secretion systems are membrane-embedded nanomachines used by Gram-positive bacteria to export effector proteins from the cytoplasm to the extracellular environment. Many of these effectors are polymorphic toxins comprised of an N-terminal Leu-x-Gly (LXG) domain of unknown function and a C-terminal toxin domain that inhibits the growth of bacterial competitors. In recent work, it was shown that LXG effectors require two cognate Lap proteins for T7SS-dependent export. Here, we present the 2.6 Å structure of the LXG domain of the TelA toxin from the opportunistic pathogen Streptococcus intermedius in complex with both of its cognate Lap targeting factors. The structure reveals an elongated α-helical bundle within which each Lap protein makes extensive hydrophobic contacts with either end of the LXG domain. Remarkably, despite low overall sequence identity, we identify striking structural similarity between our LXG complex and PE-PPE heterodimers exported by the distantly related ESX type VII secretion systems of Mycobacteria implying a conserved mechanism of effector export among diverse Gram-positive bacteria. Overall, our findings demonstrate that LXG domains, in conjunction with their cognate Lap targeting factors, represent a tripartite secretion signal for a widespread family of T7SS toxins.

Cisplatin and Procaterol Combination in Gastric Cancer? Targeting Checkpoint Kinase 1 for Cancer Drug Discovery and Repurposing by an Integrated Computational and Experimental Approach.

Checkpoint kinase 1 (CHK1), a serine/threonine kinase, plays a crucial role in cell cycle arrest and is a promising therapeutic target for drug development against cancers. CHK1 coordinates cell cycle checkpoints in response to DNA damage, facilitating repair of single-strand breaks, and maintains the genome integrity in response to replication stress. In this study, we employed an integrated computational and experimental approach to drug discovery and repurposing, aiming to identify a potent CHK1 inhibitor among existing drugs. An e-pharmacophore model was developed based on the three-dimensional crystal structure of the CHK1 protein in complex with CCT245737. This model, characterized by seven key molecular features, guided the screening of a library of drugs through molecular docking. The top 10% of scored ligands were further examined, with procaterol emerging as the leading candidate. Procaterol demonstrated interaction patterns with the CHK1 active site similar to CHK1 inhibitor (CCT245737), as shown by molecular dynamics analysis. Subsequent in vitro assays, including cell proliferation, colony formation, and cell cycle analysis, were conducted on gastric adenocarcinoma cells treated with procaterol, both as a monotherapy and in combination with cisplatin. Procaterol, in synergy with cisplatin, significantly inhibited cell growth, suggesting a potentiated therapeutic effect. Thus, we propose the combined application of cisplatin and procaterol as a novel potential therapeutic strategy against human gastric cancer. The findings also highlight the relevance of CHK1 kinase as a drug target for enhancing the sensitivity of cytotoxic agents in cancer.

Molecular forces driving protein complexation of lentil and whey proteins: Structure-function relationships of trehalose-conjugated protein complexes on protein digestibility and solubility

Plant-based proteins are often associated with a range of health benefits. Most research primarily investigates pea and soy proteins, while lentil proteins received minimal attention. This study evaluates the effect of protein complexation (using the pH-shifting technique) coupled with trehalose conjugation on lentil and whey proteins. The protein structures after the modification were analysed using spectroscopic methods: Fourier-transform infrared, ultraviolet spectra, and fluorescence spectra. The amide group I, conformation protein, and tertiary structure of the trehalose-conjugated lentil-whey protein complexes (T-LWPs) showed significant changes (P < 0.05). Moreover, the surface properties (surface hydrophobicity and charges) of T-LWPs were significantly modified (P < 0.05), from 457 to 324 a.u and from 36 to −40 mV, respectively. Due to these modifications on the protein structures, the protein digestibility (80–86%) and water solubility (90–94.5%) of T-LWPs increased significantly (P < 0.05) with the increase in the trehalose concentration, from 0 (control) to 5% (w/w), respectively. This study suggested that coupling protein complexation and trehalose conjugation can enhance the overall properties of lentil-based protein complexes. With this enhancement, more opportunities in the utilisation of lentils are to be expected.

Toward quantitative super-resolution methods for cryo-CLEM.

Cryogenic ultrastructural imaging techniques such as cryo-electron tomography have produced a revolution in how the structure of biological systems is investigated by enabling the determination of structures of protein complexes immersed in a complex biological matrix within vitrified cell and model organisms. However, so far, the portfolio of successes has been mostly limited to highly abundant complexes or to structures that are relatively unambiguous and easy to identify through electron microscopy. In order to realize the full potential of this revolution, researchers would have to be able to pinpoint lower abundance species and obtain functional annotations on the state of objects of interest which would then be correlated to ultrastructural information to build a complete picture of the structure-function relationships underpinning biological processes. Fluorescence imaging at cryogenic conditions has the potential to be able to meet these demands. However, wide-field images acquired at low numeric aperture (NA) using air immersion objective have a low resolving power and cannot provide accurate enough three-dimensional (3D) localization to enable the assignment of functional annotations to individual objects of interest or target sample debulking to ensure the preservation of the structures of interest. It is therefore necessary to develop super-resolved cryo-fluorescence workflows capable of fulfilling this role and enabling new biological discoveries. In this chapter, we present the current state of development of two super-resolution cryogenic fluorescence techniques, superSIL-STORM and astigmatism-based 3D STORM, show their application to a variety of biological systems and discuss their advantages and limitations. We further discuss the future applicability to cryo-CLEM workflows though examples of practical application to the study of membrane protein complexes both in mammalian cells and in Escherichia coli.

Contribution of Artificial Intelligence to the Identification of Protein-Protein Interactions: A Case Study on PAR-3 and Its Partner Adapter Molecule Crk.

Protein-protein interactions (PPIs) are known to be involved in most cellular functions, and a detailed knowledge of such interactions is essential for studying their role in normal and pathological conditions. Significant progress is being made in the identification of PPIs through advances in computational methods. In particular, the AlphaFold2 machine learning-based model has been shown to accelerate drug discovery process by predicting the 3D structure of protein complexes. In this chapter, a straightforward protocol for predicting interprotein interactions between PAR-3 and its protein partner adapter molecule crk is provided. Such artificial intelligence-based and publicly available approaches can provide a resource for further investigation of therapeutic drug targets.

Characterization of Posttranslationally Modified PHF-1 Tau Peptides Using Gaussian Accelerated Molecular Dynamics Simulation.

The microtubule-associated protein, Tau, is an intrinsically disordered protein that plays a crucial role in neurodegenerative diseases like Alzheimer's disease. The posttranslational modifications across the Tau protein domains are involved in regulating Tau protein's function and disease onset. Of the various posttranslational modifications at Ser, Thr, and Tyr sites, O-GlcNAcylation and phosphorylation are the most critical ones, playing a vital role in Tau aggregation and tauopathies. To understand the function, it is essential to characterize the structural changes associated withTaumodification. Previous experimental studies have focused on high-resolution nuclear magnetic resonance techniques to structurally characterize the effect of phosphorylation, O-GlcNAcylation, and combination of both PTMs on Tau conformation in small peptides centered on the PHF-1 epitope from amino acid 392 to 411. The structural characterization using atomistic molecular dynamics simulation of such disordered peptides requires long simulation time, proper sampling method, and utilization of appropriate force fields for accurate determination of conformational ensembles, resembling the experimental data. This chapter details the protocol for the structural characterization of modified Tau peptides using the CHARMM36m force field and enhanced sampling methods like Gaussian accelerated molecular dynamics (GaMD) simulation. We have focused on a detailed explanation of the GaMD method and analyses of molecular dynamics trajectories to explain the relationship between two modifications, phospho- and glyco-, at C-terminus of Tau protein and its stable conformation over the longer simulation timeframes. The analyses involve energetics reweighting, clustering of simulation trajectories, and characterization of secondary structure using circular dichroism data from the simulation. The reader can utilize this protocol to investigate the structures of complex proteins, especially the disordered ones.

What are the minimal folding seeds in proteins? Experimental and theoretical assessment of secondary structure propensities of small peptide fragments.

Certain peptide sequences, some of them as short as amino acid triplets, are significantly overpopulated in specific secondary structure motifs in folded protein structures. For example, 74% of the EAM triplet is found in α-helices, and only 3% occurs in the extended parts of proteins (typically β-sheets). In contrast, other triplets (such as VIV and IYI) appear almost exclusively in extended parts (79% and 69%, respectively). In order to determine whether such preferences are structurally encoded in a particular peptide fragment or appear only at the level of a complex protein structure, NMR, VCD, and ECD experiments were carried out on selected tripeptides: EAM (denoted as pro-'α-helical' in proteins), KAM(α), ALA(α), DIC(α), EKF(α), IYI(pro-β-sheet or more generally, pro-extended), and VIV(β), and the reference α-helical CATWEAMEKCK undecapeptide. The experimental data were in very good agreement with extensive quantum mechanical conformational sampling. Altogether, we clearly showed that the pro-helical vs. pro-extended propensities start to emerge already at the level of tripeptides and can be fully developed at longer sequences. We postulate that certain short peptide sequences can be considered minimal "folding seeds". Admittedly, the inherent secondary structure propensity can be overruled by the large intramolecular interaction energies within the folded and compact protein structures. Still, the correlation of experimental and computational data presented herein suggests that the secondary structure propensity should be considered as one of the key factors that may lead to understanding the underlying physico-chemical principles of protein structure and folding from the first principles.

Recent Advances and Challenges in Protein Structure Prediction.

Artificial intelligence has made significant advances in the field of protein structure prediction in recent years. In particular, DeepMind's end-to-end model, AlphaFold2, has demonstrated the capability to predict three-dimensional structures of numerous unknown proteins with accuracy levels comparable to those of experimental methods. This breakthrough has opened up new possibilities for understanding protein structure and function as well as accelerating drug discovery and other applications in the field of biology and medicine. Despite the remarkable achievements of artificial intelligence in the field, there are still some challenges and limitations. In this Review, we discuss the recent progress and some of the challenges in protein structure prediction. These challenges include predicting multidomain protein structures, protein complex structures, multiple conformational states of proteins, and protein folding pathways. Furthermore, we highlight directions in which further improvements can be conducted.

When MELD Meets GaMD: Accelerating Biomolecular Landscape Exploration.

We introduce Gaussian accelerated MELD (GaMELD) as a new method for exploring the energy landscape of biomolecules. GaMELD combines the strengths of Gaussian accelerated molecular dynamics (GaMD) and modeling employing limited data (MELD) to navigate complex energy landscapes. MELD uses replica-exchange molecular simulations to integrate limited and uncertain data into simulations via Bayesian inference. MELD has been successfully applied to problems of structure prediction like protein folding and complex structure prediction. However, the computational cost for MELD simulations has limited its broader applicability. The synergy of GaMD and MELD surmounts this limitation efficiently sampling the energy landscape at a lower computational cost (reducing the computational cost by a factor of 2 to six). Effectively, GaMD is used to shift energy distributions along replicas to increase the overlap in energy distributions across replicas, facilitating a random walk in replica space. We tested GaMELD on a benchmark set of 12 small proteins that have been previously studied through MELD and conventional MD. GaMELD consistently achieves accurate predictions with fewer replicas. By increasing the efficacy of replica exchange, GaMELD effectively accelerates convergence in the conformational space, enabling improved sampling across a diverse set of systems.

Multi-domain and complex protein structure prediction using inter-domain interactions from deep learning

Accurately capturing domain-domain interactions is key to understanding protein function and designing structure-based drugs. Although AlphaFold2 has made a breakthrough on single domain, it should be noted that the structure modeling for multi-domain protein and complex remains a challenge. In this study, we developed a multi-domain and complex structure assembly protocol, named DeepAssembly, based on domain segmentation and single domain modeling algorithms. Firstly, DeepAssembly uses a population-based evolutionary algorithm to assemble multi-domain proteins by inter-domain interactions inferred from a developed deep learning network. Secondly, protein complexes are assembled by means of domains rather than chains using DeepAssembly. Experimental results show that on 219 multi-domain proteins, the average inter-domain distance precision by DeepAssembly is 22.7% higher than that of AlphaFold2. Moreover, DeepAssembly improves accuracy by 13.1% for 164 multi-domain structures with low confidence deposited in AlphaFold database. We apply DeepAssembly for the prediction of 247 heterodimers. We find that DeepAssembly successfully predicts the interface (DockQ ≥ 0.23) for 32.4% of the dimers, suggesting a lighter way to assemble complex structures by treating domains as assembly units and using inter-domain interactions learned from monomer structures.

Fully activated structure of the sterol-bound Smoothened GPCR-Gi protein complex.

Smoothened (SMO) is an oncoprotein and signal transducer in the Hedgehog signaling pathway that regulates cellular differentiation and embryogenesis. As a member of the Frizzled (Class F) family of G protein-coupled receptors (GPCRs), SMO biochemically and functionally interacts with Gi family proteins. However, key molecular features of fully activated, G protein-coupled SMO remain elusive. We present the atomistic structure of activated human SMO complexed with the heterotrimeric Gi protein and two sterol ligands, equilibrated at 310 K in a full lipid bilayer at physiological salt concentration and pH. In contrast to previous experimental structures, our equilibrated SMO complex exhibits complete breaking of the pi-cation interaction between R4516.32 and W5357.55, a hallmark of Class F receptor activation. The Gi protein couples to SMO at seven strong anchor points similar to those in Class A GPCRs: intracellular loop 1, intracellular loop 2, transmembrane helix 6, and helix 8. On the path to full activation, we find that the extracellular cysteine-rich domain (CRD) undergoes a dramatic tilt, following a trajectory suggested by positions of the CRD in active and inactive experimental SMO structures. Strikingly, a sterol ligand bound to a shallow transmembrane domain (TMD) site in the initial structure migrates to a deep TMD pocket found exclusively in activator-bound SMO complexes. Thus, our results indicate that SMO interacts with Gi prior to full activation to break the molecular lock, form anchors with Gi subunits, tilt the CRD, and facilitate migration of a sterol ligand in the TMD to an activated position.

Human cystatin C induces the disaggregation process of selected amyloid beta peptides: a structural and kinetic view

Neurodegenerative diseases, such as Alzheimer’s disease (AD) and various types of amyloidosis, are incurable; therefore, understanding the mechanisms of amyloid decomposition is crucial to develop an effective drug against them for future therapies. It has been reported that one out of three people over the age of 85 are suffering from dementia as a comorbidity to AD. Amyloid beta (Aβ), the hallmark of AD, transforms structurally from monomers into β-stranded aggregates (fibrils) via multiple oligomeric states. Astrocytes in the central nervous system secrete the human cystatin C protein (HCC) in response to various proteases and cytokines. The codeposition of Aβ and HCC in the brains of patients with AD led to the hypothesis that cystatin C is implicated in the disease process. In this study, we investigate the intermolecular interactions between different atomic structures of fibrils formed by Aβ peptides and HCC to understand the pathological aggregation of these polypeptides into neurotoxic oligomers and then amyloid plaques. To characterize the interactions between Aβ and HCC, we used a complementary approach based on the combination of small-angle neutron scattering analysis, atomic force microscopy and computational modelling, allowing the exploration of the structures of multicomponent protein complexes. We report here an optimized protocol to study that interaction. The results show a dependency of the sequence length of the Aβ peptide on the ability of the associated HCC to disaggregate it.

Structural coverage of the human interactome.

Complex biological processes in cells are embedded in the interactome, representing the complete set of protein-protein interactions. Mapping and analyzing the protein structures are essential to fully comprehending these processes' molecular details. Therefore, knowing the structural coverage of the interactome is important to show the current limitations. Structural modeling of protein-protein interactions requires accurate protein structures. In this study, we mapped all experimental structures to the reference human proteome. Later, we found the enrichment in structural coverage when complementary methods such as homology modeling and deep learning (AlphaFold) were included. We then collected the interactions from the literature and databases to form the reference human interactome, resulting in 117897 non-redundant interactions. When we analyzed the structural coverage of the interactome, we found that the number of experimentally determined protein complex structures is scarce, corresponding to 3.95% of all binary interactions. We also analyzed known and modeled structures to potentially construct the structural interactome with a docking method. Our analysis showed that 12.97% of the interactions from HuRI and 73.62% and 32.94% from the filtered versions of STRING and HIPPIE could potentially be modeled with high structural coverage or accuracy, respectively. Overall, this paper provides an overview of the current state of structural coverage of the human proteome and interactome.

Enhancing alphafold-multimer-based protein complex structure prediction with MULTICOM in CASP15

To enhance the AlphaFold-Multimer-based protein complex structure prediction, we developed a quaternary structure prediction system (MULTICOM) to improve the input fed to AlphaFold-Multimer and evaluate and refine its outputs. MULTICOM samples diverse multiple sequence alignments (MSAs) and templates for AlphaFold-Multimer to generate structural predictions by using both traditional sequence alignments and Foldseek-based structure alignments, ranks structural predictions through multiple complementary metrics, and refines the structural predictions via a Foldseek structure alignment-based refinement method. The MULTICOM system with different implementations was blindly tested in the assembly structure prediction in the 15th Critical Assessment of Techniques for Protein Structure Prediction (CASP15) in 2022 as both server and human predictors. MULTICOM_qa ranked 3rd among 26 CASP15 server predictors and MULTICOM_human ranked 7th among 87 CASP15 server and human predictors. The average TM-score of the first predictions submitted by MULTICOM_qa for CASP15 assembly targets is ~0.76, 5.3% higher than ~0.72 of the standard AlphaFold-Multimer. The average TM-score of the best of top 5 predictions submitted by MULTICOM_qa is ~0.80, about 8% higher than ~0.74 of the standard AlphaFold-Multimer. Moreover, the Foldseek Structure Alignment-based Multimer structure Generation (FSAMG) method outperforms the widely used sequence alignment-based multimer structure generation.

Recognition of methamphetamine and other amines by trace amine receptor TAAR1.

Trace amine-associated receptor 1 (TAAR1), the founding member of a nine-member family of trace amine receptors, is responsible for recognizing a range of biogenic amines in the brain, including the endogenous β-phenylethylamine (β-PEA)1 as well as methamphetamine2, an abused substance that has posed a severe threat to human health and society3. Given its unique physiological role in the brain, TAAR1 is also an emerging target for a range of neurological disorders including schizophrenia, depression and drug addiction2,4,5. Here we report structures of human TAAR1-G-protein complexes bound to methamphetamine and β-PEA as well as complexes bound to RO5256390, a TAAR1-selective agonist, and SEP-363856, a clinical-stage dual agonist for TAAR1 and serotonin receptor 5-HT1AR (refs. 6,7). Together with systematic mutagenesis and functional studies, the structures reveal the molecular basis of methamphetamine recognition and underlying mechanisms of ligand selectivity and polypharmacology between TAAR1 and other monoamine receptors. We identify a lid-like extracellular loop 2 helix/loop structure and a hydrogen-bonding network in the ligand-binding pockets, which may contribute to the ligand recognition in TAAR1. These findings shed light on the ligand recognition mode and activation mechanism for TAAR1 and should guide the development of next-generation therapeutics for drug addiction and various neurological disorders.

Structure and energetics of serum protein complex of tea adulterant dye Bismarck brown Y using experimental and computational methods

Tea, a widely consumed aromatic beverage, is often adulterated with dyes such as Bismarck brown Y (C.I. 21000) (BBY), Prussian blue, and Plumbago, which pose potential health risks. The objective of this study is to analyze how the food dye BBY interacts with serum protein, bovine serum albumin (BSA). This study investigated the BBY-BSA interaction at the molecular level. Fluorescence spectroscopy results showed that the quenching of BSA by BBY is carried out by dynamic quenching mechanism. The displacement assay and molecular docking studies revealed that BBY binds at the flavanone binding site of BSA with hydrophobic interactions. Circular Dichroism results indicate the structural stability of the protein upon BBY binding. Molecular dynamics simulations demonstrated the stability of the complex in a dynamic solvent system, and quantum mechanics calculations showed slight conformational changes of the diaminophenyl ring due to increased hydrophobic interaction. The energetics of gas phase optimized and stable MD structures of BBY indicated similar values which further confirmed that the conformational changes were minor, and it also exhibited a moderate binding with BSA as shown by the MM/PBSA results. This study enhances our understanding of the molecular-level interactions between BBY and BSA, emphasizing the critical role of hydrophobic interactions.

Topological links in predicted protein complex structures reveal limitations of AlphaFold

AlphaFold is making great progress in protein structure prediction, not only for single-chain proteins but also for multi-chain protein complexes. When using AlphaFold-Multimer to predict protein‒protein complexes, we observed some unusual structures in which chains are looped around each other to form topologically intertwining links at the interface. Based on physical principles, such topological links should generally not exist in native protein complex structures unless covalent modifications of residues are involved. Although it is well known and has been well studied that protein structures may have topologically complex shapes such as knots and links, existing methods are hampered by the chain closure problem and show poor performance in identifying topologically linked structures in protein‒protein complexes. Therefore, we address the chain closure problem by using sliding windows from a local perspective and propose an algorithm to measure the topological–geometric features that can be used to identify topologically linked structures. An application of the method to AlphaFold-Multimer-predicted protein complex structures finds that approximately 1.72% of the predicted structures contain topological links. The method presented in this work will facilitate the computational study of protein‒protein interactions and help further improve the structural prediction of multi-chain protein complexes.

Co-evolution at protein-protein interfaces guides inference of stoichiometry of oligomeric protein complexes by de novo structure prediction.

The quaternary structure with specific stoichiometry is pivotal to the specific function of protein complexes. However, determining the structure of many protein complexes experimentally remains a major bottleneck. Structural bioinformatics approaches, such as the deep learning algorithm Alphafold2-multimer (AF2-multimer), leverage the co-evolution of amino acids and sequence-structure relationships for accurate de novo structure and contact prediction. Pseudo-likelihood maximization direct coupling analysis (plmDCA) has been used to detect co-evolving residue pairs by statistical modeling. Here, we provide evidence that combining both methods can be used for de novo prediction of the quaternary structure and stoichiometry of a protein complex. We achieve this by augmenting the existing AF2-multimer confidence metrics with an interpretable score to identify the complex with an optimal fraction of native contacts of co-evolving residue pairs at intermolecular interfaces. We use this strategy to predict the quaternary structure and non-trivial stoichiometries of Bacillus subtilis spore germination protein complexes with unknown structures. Co-evolution at intermolecular interfaces may therefore synergize with AI-based de novo quaternary structure prediction of structurally uncharacterized bacterial protein complexes.