Structure-aware machine learning strategies for antimicrobial peptide discovery.

Machine learning models are revolutionizing our approaches to discovering and designing bioactive peptides. These models often need protein structure awareness, as they heavily rely on sequential data. The models excel at identifying sequences of a particular biological nature or activity, but they frequently fail to comprehend their intricate mechanism(s) of action. To solve two problems at once, we studied the mechanisms of action and structural landscape of antimicrobial peptides as (i) membrane-disrupting peptides, (ii) membrane-penetrating peptides, and (iii) protein-binding peptides. By analyzing critical features such as dipeptides and physicochemical descriptors, we developed models with high accuracy (86-88%) in predicting these categories. However, our initial models (1.0 and 2.0) exhibited a bias towards α-helical and coiled structures, influencing predictions. To address this structural bias, we implemented subset selection and data reduction strategies. The former gave three structure-specific models for peptides likely to fold into α-helices (models 1.1 and 2.1), coils (1.3 and 2.3), or mixed structures (1.4 and 2.4). The latter depleted over-represented structures, leading to structure-agnostic predictors 1.5 and 2.5. Additionally, our research highlights the sensitivity of important features to different structure classes across models.

Before and after AlphaFold2: An overview of protein structure prediction.

Three-dimensional protein structure is directly correlated with its function and its determination is critical to understanding biological processes and addressing human health and life science problems in general. Although new protein structures are experimentally obtained over time, there is still a large difference between the number of protein sequences placed in Uniprot and those with resolved tertiary structure. In this context, studies have emerged to predict protein structures by methods based on a template or free modeling. In the last years, different methods have been combined to overcome their individual limitations, until the emergence of AlphaFold2, which demonstrated that predicting protein structure with high accuracy at unprecedented scale is possible. Despite its current impact in the field, AlphaFold2 has limitations. Recently, new methods based on protein language models have promised to revolutionize the protein structural biology allowing the discovery of protein structure and function only from evolutionary patterns present on protein sequence. Even though these methods do not reach AlphaFold2 accuracy, they already covered some of its limitations, being able to predict with high accuracy more than 200 million proteins from metagenomic databases. In this mini-review, we provide an overview of the breakthroughs in protein structure prediction before and after AlphaFold2 emergence.

Characterization of a virulence-modifying protein of Leptospira interrogans identified by shotgun phage display.

Pathogenic species of Leptospira are etiologic agents of leptospirosis, an emerging zoonotic disease of worldwide extent and endemic in tropical regions. The growing number of identified leptospiral species sheds light to their genetic diversity and unique virulence mechanisms, many of them still remain unknown. Toxins and adhesins are important virulence factors in several pathogens, constituting promising antigens for the development of vaccines with cross-protection and long-lasting effect against leptospirosis. For this aim, we used the shotgun phage display technique to unravel new proteins with adhesive properties. A shotgun library was constructed using fragmented genomic DNA from Leptospira interrogans serovar Copenhageni strain Fiocruz L1-130 and pG8SAET phagemid vector. Selection of phages bearing new possible cell-binding antigens was performed against VERO cells, using BRASIL biopanning methodology. Analysis of selected clones revealed the hypothetical protein LIC10778, a potentially exposed virulence factor that belongs to the virulence-modifying (VM) protein family (PF07598), composed of 13 members in the leptospiral strain Fiocruz L1-130. Prediction of LIC10778 tertiary structure indicates that the protein contains a cellular-binding domain (N-terminal portion) and an unknown domain of no assigned activity (C-terminal portion). The predicted N-terminal domain shared structural similarities with the cell-binding and internalization domain of toxins like Ricin and Abrin, as well as to the Community-Acquired Respiratory Distress Syndrome (CARDS) toxin in Mycoplasma pneumoniae. Interestingly, recombinant portions of the N-terminal region of LIC10778 protein showed binding to laminin, collagens I and IV, vitronectin, and plasma and cell fibronectins using overlay blotting technique, especially regarding the binding site identified by phage display. These data validate our preliminary phage display biopanning and support the predicted three-dimensional models of LIC10778 protein and other members of PF07598 protein family, confirming the identification of the N-terminal cell-binding domains that are similar to ricin-like toxins. Moreover, fluorescent fused proteins also confirmed that N-terminal region of LIC10778 is capable of binding to VERO and A549 cell lines, further highlighting its virulence role during host-pathogen interaction in leptospirosis probably mediated by its C-terminal domain. Indeed, recent results in the literature confirmed this assumption by demonstrating the cytotoxicity of a closely related PF07598 member.

Structural changes and adaptative evolutionary constraints in FLOWERING LOCUS T and TERMINAL FLOWER1-like genes of flowering plants.

Regulation of flowering is a crucial event in the evolutionary history of angiosperms. The production of flowers is regulated through the integration of different environmental and endogenous stimuli, many of which involve the activation of different genes in a hierarchical and complex signaling network. The FLOWERING LOCUS T/TERMINAL FLOWER 1 (FT/TFL1) gene family is known to regulate important aspects of flowering in plants. To better understand the pivotal events that changed FT and TFL1 functions during the evolution of angiosperms, we reconstructed the ancestral sequences of FT/TFL1-like genes and predicted protein structures through in silico modeling to identify determinant sites that evolved in both proteins and allowed the adaptative diversification in the flowering phenology and developmental processes. In addition, we demonstrate that the occurrence of destabilizing mutations in residues located at the phosphatidylcholine binding sites of FT structure are under positive selection, and some residues of 4th exon are under negative selection, which is compensated by the occurrence of stabilizing mutations in key regions and the P-loop to maintain the overall protein stability. Our results shed light on the evolutionary history of key genes involved in the diversification of angiosperms.

Characterization of a virulence-modifying protein of Leptospira interrogans identified by shotgun phage display

Pathogenic species of Leptospira are etiologic agents of leptospirosis, an emerging zoonotic disease of worldwide extent and endemic in tropical regions. The growing number of identified leptospiral species sheds light to their genetic diversity and unique virulence mechanisms, many of them still remain unknown. Toxins and adhesins are important virulence factors in several pathogens, constituting promising antigens for the development of vaccines with cross-protection and long-lasting effect against leptospirosis. For this aim, we used the shotgun phage display technique to unravel new proteins with adhesive properties. A shotgun library was constructed using fragmented genomic DNA from Leptospira interrogans serovar Copenhageni strain Fiocruz L1-130 and pG8SAET phagemid vector. Selection of phages bearing new possible cell-binding antigens was performed against VERO cells, using BRASIL biopanning methodology. Analysis of selected clones revealed the hypothetical protein LIC10778, a potentially exposed virulence factor that belongs to the virulence-modifying (VM) protein family (PF07598), composed of 13 members in the leptospiral strain Fiocruz L1-130. Prediction of LIC10778 tertiary structure indicates that the protein contains a cellular-binding domain (N-terminal portion) and an unknown domain of no assigned activity (C-terminal portion). The predicted N-terminal domain shared structural similarities with the cell-binding and internalization domain of toxins like Ricin and Abrin, as well as to the Community-Acquired Respiratory Distress Syndrome (CARDS) toxin in Mycoplasma pneumoniae. Interestingly, recombinant portions of the N-terminal region of LIC10778 protein showed binding to laminin, collagens I and IV, vitronectin, and plasma and cell fibronectins using overlay blotting technique, especially regarding the binding site identified by phage display. These data validate our preliminary phage display biopanning and support the predicted three-dimensional models of LIC10778 protein and other members of PF07598 protein family, confirming the identification of the N-terminal cell-binding domains that are similar to ricin-like toxins. Moreover, fluorescent fused proteins also confirmed that N-terminal region of LIC10778 is capable of binding to VERO and A549 cell lines, further highlighting its virulence role during host-pathogen interaction in leptospirosis probably mediated by its C-terminal domain. Indeed, recent results in the literature confirmed this assumption by demonstrating the cytotoxicity of a closely related PF07598 member.

Analysis of the Zonula occludens Toxin Found in the Genome of the Chilean Non-toxigenic Vibrio parahaemolyticus Strain PMC53.7.

Vibrio parahaemolyticus non-toxigenic strains are responsible for about 10% of acute gastroenteritis associated with this species, suggesting they harbor unique virulence factors. Zonula occludens toxin (Zot), firstly described in Vibrio cholerae, is a secreted toxin that increases intestinal permeability. Recently, we identified Zot-encoding genes in the genomes of highly cytotoxic Chilean V. parahaemolyticus strains, including the non-toxigenic clinical strain PMC53.7. To gain insights into a possible role of Zot in V. parahaemolyticus, we analyzed whether it could be responsible for cytotoxicity. However, we observed a barely positive correlation between Caco-2 cell membrane damage and Zot mRNA expression during PMC53.7 infection and non-cytotoxicity induction in response to purified PMC53.7-Zot. Unusually, we observed a particular actin disturbance on cells infected with PMC53.7. Based on this observation, we decided to compare the sequence of PMC53.7-Zot with Zot of human pathogenic species such as V. cholerae, Campylobacter concisus, Neisseria meningitidis, and other V. parahaemolyticus strains, using computational tools. The PMC53.7-Zot was compared with other toxins and identified as an endotoxin with conserved motifs in the N-terminus and a variable C-terminal region and without FCIGRL peptide. Notably, the C-terminal diversity among Zots meant that not all of them could be identified as toxins. Structurally, PMC53.7-Zot was modeled as a transmembrane protein. Our results suggested that it has partial 3D structure similarity with V. cholerae-Zot. Probably, the PMC53.7-Zot would affect the actin cytoskeletal, but, in the absence of FCIGRL, the mechanisms of actions must be elucidated.

Prediction of Structure and Molecular Interaction with DNA of BvrR, a Virulence-Associated Regulatory Protein of Brucella.

Brucellosis, also known as "undulant fever" is a zoonotic disease caused by Brucella, which is a facultative intracellular bacterium. Despite efforts to eradicate this disease, infection in uncontrolled domestic animals persists in several countries and therefore transmission to humans is common. Brucella evasion of the innate immune system depends on its ability to evade the mechanisms of intracellular death in phagocytic cells. The BvrR-BvrS two-component system allows the bacterium to detect adverse conditions in the environment. The BvrS protein has been associated with genes of virulence factors, metabolism, and membrane transport. In this study, we predicted the DNA sequence recognized by BvrR with Gibbs Recursive Sampling and identified the three-dimensional structure of BvrR using I-TASSER suite, and the interaction mechanism between BvrR and DNA with Protein-DNA docking and molecular dynamics (MD) simulation. Based on the Gibbs recursive Sampling analysis, we found the motif AAHTGC (H represents A, C, and T nucleotides) as a possible sequence recognized by BvrR. The docking and EMD simulation results showed that C-terminal effector domain of BvrR protein is likely to interact with AAHTGC sequence. In conclusion, we predicted the structure, recognition motif, and interaction of BvrR with DNA.

Supersecondary Structures and Fragment Libraries.

The use of smotifs and fragment libraries has proven useful to both simplify and increase the quality of protein models. Here, we present Profrager, a tool that automatically generates putative structural fragments to reproduce local motifs of proteins given a target sequence. Profrager is highly customizable, allowing the user to select the number of fragments per library, the ranking method is able to generate fragments of all sizes, and it was recently modified to include the possibility of output exclusively smotifs.

Myosin B of Plasmodium falciparum (PfMyoB): in silico prediction of its three-dimensional structure and its possible interaction with MTIP.

The mobility and invasion strategy of Plasmodium falciparum is governed by a protein complex known as the glideosome, which contains an actin-myosin motor. It has been shown that myosin A of the parasite (PfMyoA) is the myosin of the glideosome, and the interaction of PfMyoA with myosin tail domain interacting protein (MTIP) determines its correct location and its ability to function in the complex. Because PfMyoA and myosin B of P. falciparum (PfMyoB) share high sequence identity, are both small proteins without a tail domain, belong to the class XIV myosins, and are expressed in late schizonts and merozoites, we suspect that these myosins may have similar or redundant functions. Therefore, this work examined the structural similarity between PfMyoA and PfMyoB and performed a molecular docking between PfMyoB and MTIP. Three-dimensional (3D) models obtained for PfMyoA and PfMyoB achieved high scores in the structural validation programs used, and their superimposition revealed high structural similarity, supporting the hypothesis of possible similar functions for these two proteins. The 3D interaction models obtained and energy values found suggested that interaction between PfMyoB and MTIP is possible. Given the apparent abundance of PfMyoA relative to PfMyoB in the parasite, we believe that the interaction between PfMyoB and MTIP would only be detectable in specific cellular environments because under normal circumstances, it would be masked by the interaction between PfMyoA and MTIP.

Structural and dynamic insights into the C-terminal extension of cysteine proteinase B from Leishmania amazonensis.

Cysteine proteinase B (CPB) is a significant virulence factor for Leishmania infections. Upon processing from its zymogen form, it happens a release of the immunomodulatory CPB C-terminal extension (cyspep) into the cytoplasm of the macrophage. Epitopes derived from this fragment were shown to influence the proportion of lymphocytes CD8+ upon infection, favoring the parasite escaping from the host́s immune system. At present, there is no available structural data of cyspep, which impairs a proper understanding of its biological functions. Here, we attempted to build molecular models for this fragment and subsequently evaluate their stabilities in aqueous solution from molecular dynamics simulations analysis. Characterization of our models obtained with distinct techniques (comparative modeling, threading, and ab initio) indicates a prevalence of ß-sheets in agreement with consensus secondary structure predictions. Simulation data supported this finding since the formation of new strands, along with a rapid disruption of helical content, were observed. Overall, this study provides a rationalization of epitope mapping data and an improved understanding of cyspep antigenicity.

APL: An angle probability list to improve knowledge-based metaheuristics for the three-dimensional protein structure prediction.

Tertiary protein structure prediction is one of the most challenging problems in structural bioinformatics. Despite the advances in algorithm development and computational strategies, predicting the folded structure of a protein only from its amino acid sequence remains as an unsolved problem. We present a new computational approach to predict the native-like three-dimensional structure of proteins. Conformational preferences of amino acid residues and secondary structure information were obtained from protein templates stored in the Protein Data Bank and represented as an Angle Probability List. Two knowledge-based prediction methods based on Genetic Algorithms and Particle Swarm Optimization were developed using this information. The proposed method has been tested with twenty-six case studies selected to validate our approach with different classes of proteins and folding patterns. Stereochemical and structural analysis were performed for each predicted three-dimensional structure. Results achieved suggest that the Angle Probability List can improve the effectiveness of metaheuristics used to predicted the three-dimensional structure of protein molecules by reducing its conformational search space.

Insights on the structure and stability of Licanantase: a trimeric acid-stable coiled-coil lipoprotein from Acidithiobacillus thiooxidans.

Licanantase (Lic) is the major component of the secretome of Acidithiobacillus thiooxidans when grown in elemental sulphur. When used as an additive, Lic improves copper recovery from bioleaching processes. However, this recovery enhancement is not fully understood. In this context, our aim is to predict the 3D structure of Lic, to shed light on its structure-function relationships. Bioinformatics analyses on the amino acid sequence of Lic showed a great similarity with Lpp, an Escherichia coli Lipoprotein that can form stable trimers in solution. Lic and Lpp share the secretion motif, intracellular processing and alpha helix structure, as well as the distribution of hydrophobic residues in heptads forming a hydrophobic core, typical of coiled-coil structures. Cross-linking experiments showed the presence of Lic trimers, supporting our predictions. Taking the in vitro and in silico evidence as a whole, we propose that the most probable structure for Lic is a trimeric coiled-coil. According to this prediction, a suitable model for Lic was produced using the de novo algorithm "Rosetta Fold-and-Dock". To assess the structural stability of our model, Molecular Dynamics (MD) and Replica Exchange MD simulations were performed using the structure of Lpp and a 14-alanine Lpp mutant as controls, at both acidic and neutral pH. Our results suggest that Lic was the most stable structure among the studied proteins in both pH conditions. This increased stability can be explained by a higher number of both intermonomer hydrophobic contacts and hydrogen bonds, key elements for the stability of Lic's secondary and tertiary structure.