Nozomi Nagano

Characterization of the biosynthetic gene cluster for the ribosomally synthesized cyclic peptide ustiloxin B in Aspergillus flavus

Ustiloxin B is a secondary metabolite known to be produced by Ustilaginoidea virens. In our previous paper, we observed the production of this compound by Aspergillus flavus, and identified two A. flavus genes responsible for ustiloxin B biosynthesis (Umemura et al., 2013). The compound is a cyclic tetrapeptide of Tyr-Ala-Ile-Gly, whose tyrosine is modified with a non-protein coding amino acid, norvaline. Although its chemical structure strongly suggested that ustiloxin B is biosynthesized by a non-ribosomal peptide synthetase, in the present study, we observed its synthesis through a ribosomal peptide synthetic (RiPS) pathway by precise sequence analyses after experimental validation of the cluster. The cluster possessed a gene (AFLA_094980), termed ustA, whose translated product, UstA, contains a 16-fold repeated peptide embedding a tetrapeptide, Tyr-Ala-Ile-Gly, that is converted into the cyclic moiety of ustiloxin B. This result strongly suggests that ustiloxin B is biosynthesized through a RiPS pathway and that UstA provides the precursor peptide of the compound. The present work is the first characterization of RiPS in Ascomycetes and the entire RiPS gene cluster in fungi. Based on the sequence analyses, we also proposed a biosynthetic mechanism involving the entire gene cluster. Our finding indicates the possibility that a number of unidentified RiPSs exist in Ascomycetes as the biosynthetic genes of secondary metabolites, and that the feature of a highly repeated peptide sequence in UstA will greatly contribute to the discovery of additional RiPS.

EzCatDB: the Enzyme Catalytic-mechanism Database

The EzCatDB (Enzyme Catalytic-mechanism Database) specifically includes catalytic mechanisms of enzymes in terms of sequences and tertiary structures of enzymes, and proposed catalytic mechanisms, along with ligand structures. The EzCatDB groups enzyme data in the Protein Data Bank (PDB) and the SWISS-PROT database with identical domain compositions, Enzyme Commission (EC) numbers and catalytic mechanisms. The EzCatDB can be queried by the type of catalytic residue, name and type of ligand molecule that interacts with an enzyme as a cofactor, substrate or product. It can provide literature information, other database codes and EC numbers. The EzCatDB provides ligand annotation for enzymes in the PDB as well as literature information on structure and catalytic mechanisms. Furthermore, the EzCatDB also provides a hierarchic classification of catalytic mechanisms. This classification incorporates catalytic mechanisms and active-site structures of enzymes as well as basic reactions and reactive parts of ligand molecules. The EzCatDB is available at http://mbs.cbrc.jp/EzCatDB/.

MIDDAS-M: Motif-Independent De Novo Detection of Secondary Metabolite Gene Clusters through the Integration of Genome Sequencing and Transcriptome Data

Many bioactive natural products are produced as “secondary metabolites” by plants, bacteria, and fungi. During the middle of the 20th century, several secondary metabolites from fungi revolutionized the pharmaceutical industry, for example, penicillin, lovastatin, and cyclosporine. They are generally biosynthesized by enzymes encoded by clusters of coordinately regulated genes, and several motif-based methods have been developed to detect secondary metabolite biosynthetic (SMB) gene clusters using the sequence information of typical SMB core genes such as polyketide synthases (PKS) and non-ribosomal peptide synthetases (NRPS). However, no detection method exists for SMB gene clusters that are functional and do not include core SMB genes at present. To advance the exploration of SMB gene clusters, especially those without known core genes, we developed MIDDAS-M, a motif-independent de novo detection algorithm for SMB gene clusters. We integrated virtual gene cluster generation in an annotated genome sequence with highly sensitive scoring of the cooperative transcriptional regulation of cluster member genes. MIDDAS-M accurately predicted 38 SMB gene clusters that have been experimentally confirmed and/or predicted by other motif-based methods in 3 fungal strains. MIDDAS-M further identified a new SMB gene cluster for ustiloxin B, which was experimentally validated. Sequence analysis of the cluster genes indicated a novel mechanism for peptide biosynthesis independent of NRPS. Because it is fully computational and independent of empirical knowledge about SMB core genes, MIDDAS-M allows a large-scale, comprehensive analysis of SMB gene clusters, including those with novel biosynthetic mechanisms that do not contain any functionally characterized genes.

Ustiloxins, fungal cyclic peptides, are ribosomally synthesized in Ustilaginoidea virens

MOTIVATION Ustiloxins A and B are toxic cyclic tetrapeptides, Tyr-Val/Ala-Ile-Gly (Y-V/A-I-G), that were originally identified from Ustilaginoidea virens, a pathogenic fungus affecting rice plants. Contrary to our report that ustiloxin B is ribosomally synthesized in Aspergillus flavus, a recent report suggested that ustiloxins are synthesized by a non-ribosomal peptide synthetase in U.virens. Thus, we analyzed the U.virens genome, to identify the responsible gene cluster. RESULTS The biosynthetic gene cluster was identified from the genome of U.virens based on homologies to the ribosomal peptide biosynthetic gene cluster for ustiloxin B identified from A.flavus. It contains a gene encoding precursor protein having five Tyr-Val-Ile-Gly and three Tyr-Ala-Ile-Gly motifs for ustiloxins A and B, respectively, strongly indicating that ustiloxins A and B from U.virens are ribosomally synthesized. AVAILABILITY AND IMPLEMENTATION Accession codes of the U.virens and A.flavus gene clusters in NCBI are BR001221 and BR001206, respectively. Supplementary data are available at Bioinformatics online.

Prediction of detailed enzyme functions and identification of specificity determining residues by random forests.

Determining enzyme functions is essential for a thorough understanding of cellular processes. Although many prediction methods have been developed, it remains a significant challenge to predict enzyme functions at the fourth-digit level of the Enzyme Commission numbers. Functional specificity of enzymes often changes drastically by mutations of a small number of residues and therefore, information about these critical residues can potentially help discriminate detailed functions. However, because these residues must be identified by mutagenesis experiments, the available information is limited, and the lack of experimentally verified specificity determining residues (SDRs) has hindered the development of detailed function prediction methods and computational identification of SDRs. Here we present a novel method for predicting enzyme functions by random forests, EFPrf, along with a set of putative SDRs, the random forests derived SDRs (rf-SDRs). EFPrf consists of a set of binary predictors for enzymes in each CATH superfamily and the rf-SDRs are the residue positions corresponding to the most highly contributing attributes obtained from each predictor. EFPrf showed a precision of 0.98 and a recall of 0.89 in a cross-validated benchmark assessment. The rf-SDRs included many residues, whose importance for specificity had been validated experimentally. The analysis of the rf-SDRs revealed both a general tendency that functionally diverged superfamilies tend to include more active site residues in their rf-SDRs than in less diverged superfamilies, and superfamily-specific conservation patterns of each functional residue. EFPrf and the rf-SDRs will be an effective tool for annotating enzyme functions and for understanding how enzyme functions have diverged within each superfamily.

Metric learning for enzyme active-site search

Motivation: Finding functionally analogous enzymes based on the local structures of active sites is an important problem. Conventional methods use templates of local structures to search for analogous sites, but their performance depends on the selection of atoms for inclusion in the templates. Results: The automatic selection of atoms so that site matches can be discriminated from mismatches. The algorithm provides not only good predictions, but also some insights into which atoms are important for the prediction. Our experimental results suggest that the metric learning automatically provides more effective templates than those whose atoms are selected manually. Availability: Online software is available at http://www.net-machine.net/∼kato/lpmetric1/ Contact: kato-tsuyoshi@k.u-tokyo.ac.jp Supplementary information: Supplementary data are available at Bioinformatics online.

Key challenges for the creation and maintenance of specialist protein resources

As the volume of data relating to proteins increases, researchers rely more and more on the analysis of published data, thus increasing the importance of good access to these data that vary from the supplemental material of individual articles, all the way to major reference databases with professional staff and long‐term funding. Specialist protein resources fill an important middle ground, providing interactive web interfaces to their databases for a focused topic or family of proteins, using specialized approaches that are not feasible in the major reference databases. Many are labors of love, run by a single lab with little or no dedicated funding and there are many challenges to building and maintaining them. This perspective arose from a meeting of several specialist protein resources and major reference databases held at the Wellcome Trust Genome Campus (Cambridge, UK) on August 11 and 12, 2014. During this meeting some common key challenges involved in creating and maintaining such resources were discussed, along with various approaches to address them. In laying out these challenges, we aim to inform users about how these issues impact our resources and illustrate ways in which our working together could enhance their accuracy, currency, and overall value. Proteins 2015; 83:1005–1013.

EzCatDB: the enzyme reaction database, 2015 update

The EzCatDB database (http://ezcatdb.cbrc.jp/EzCatDB/) has emphasized manual classification of enzyme reactions from the viewpoints of enzyme active-site structures and their catalytic mechanisms based on literature information, amino acid sequences of enzymes (UniProtKB) and the corresponding tertiary structures from the Protein Data Bank (PDB). Reaction types such as hydrolysis, transfer, addition, elimination, isomerization, hydride transfer and electron transfer have been included in the reaction classification, RLCP. This database includes information related to ligand molecules on the enzyme structures in the PDB data, classified in terms of cofactors, substrates, products and intermediates, which are also necessary to elucidate the catalytic mechanisms. Recently, the database system was updated. The 3D structures of active sites for each PDB entry can be viewed using Jmol or Rasmol software. Moreover, sequence search systems of two types were developed for the EzCatDB database: EzCat-BLAST and EzCat-FORTE. EzCat-BLAST is suitable for quick searches, adopting the BLAST algorithm, whereas EzCat-FORTE is more suitable for detecting remote homologues, adopting the algorithm for FORTE protein structure prediction software. Another system, EzMetAct, is also available to searching for major active-site structures in EzCatDB, for which PDB-formatted queries can be searched.

SAHG, a comprehensive database of predicted structures of all human proteins

Most proteins from higher organisms are known to be multi-domain proteins and contain substantial numbers of intrinsically disordered (ID) regions. To analyse such protein sequences, those from human for instance, we developed a special protein-structure-prediction pipeline and accumulated the products in the Structure Atlas of Human Genome (SAHG) database at http://bird.cbrc.jp/sahg. With the pipeline, human proteins were examined by local alignment methods (BLAST, PSI-BLAST and Smith–Waterman profile–profile alignment), global–local alignment methods (FORTE) and prediction tools for ID regions (POODLE-S) and homology modeling (MODELLER). Conformational changes of protein models upon ligand-binding were predicted by simultaneous modeling using templates of apo and holo forms. When there were no suitable templates for holo forms and the apo models were accurate, we prepared holo models using prediction methods for ligand-binding (eF-seek) and conformational change (the elastic network model and the linear response theory). Models are displayed as animated images. As of July 2010, SAHG contains 42 581 protein-domain models in approximately 24 900 unique human protein sequences from the RefSeq database. Annotation of models with functional information and links to other databases such as EzCatDB, InterPro or HPRD are also provided to facilitate understanding the protein structure-function relationships.

Relationships between functional subclasses and information contained in active-site and ligand-binding residues in diverse superfamilies.

To investigate the relationships between functional subclasses and sequence and structural information contained in the active‐site and ligand‐binding residues (LBRs), we performed a detailed analysis of seven diverse enzyme superfamilies: aldolase class I, TIM‐barrel glycosidases, α/β‐hydrolases, P‐loop containing nucleotide triphosphate hydrolases, collagenase, Zn peptidases, and glutamine phosphoribosylpyrophosphate, subunit 1, domain 1. These homologous superfamilies, as defined in CATH, were selected from the enzyme catalytic‐mechanism database. We defined active‐site and LBRs based solely on the literature information and complex structures in the Protein Data Bank. From a structure‐based multiple sequence alignment for each CATH homologous superfamily, we extracted subsequences consisting of the aligned positions that were used as an active‐site or a ligand‐binding site by at least one sequence. Using both the subsequences and full‐length alignments, we performed cluster analysis with three sequence distance measures. We showed that the cluster analysis using the subsequences was able to detect functional subclasses more accurately than the clustering using the full‐length alignments. The subsequences determined by only the literature information and complex structures, thus, had sufficient information to detect the functional subclasses. Detailed examination of the clustering results provided new insights into the mechanism of functional diversification for these superfamilies. Proteins 2010.