Computational protocol: The Pan-Genome of the Animal Pathogen Corynebacterium pseudotuberculosis Reveals Differences in Genome Plasticity between the Biovar ovis and equi Strains

Similar protocols

Protocol publication

[…] The Gegenees (version 1.1.4) software was used to perform the phylogenomic analyses at the genus level and to retrieve the GenBank sequences of all the complete Corynebacterium genomes from the NCBI ftp site. Briefly, Gegenees was used to divide the genomes into small sequences and to perform an all-versus-all similarity search to determine the minimum content shared by all the genomes. Next, the minimum shared content was subtracted from all the genomes, resulting in the variable content, which was compared with all the other strains to generate the percentages of similarity. Finally, these percentages were plotted in a heatmap chart with a spectrum ranging from red (low similarity) to green (high similarity) . The Gegenees data can also be exported as a distance matrix file in nexus format. Here, we used the distance matrix as an input file for the SplitsTree (version 4.12.6) software to generate a phylogenomic tree using the UPGMA method , . [...] This section describes the analyses that were performed for all of the following three datasets: A) all strains, using C. pseudotuberculosis strain 1002 as a reference; B) the biovar ovis strains, using C. pseudotuberculosis strain 1002 as a reference; and C) the biovar equi strains, using C. pseudotuberculosis strain CIP52.97 as a reference. To calculate the pan-genome, core genome and singletons of the C. pseudotuberculosis species, we used EDGAR (version 1.2), multiple-strain genome comparison software that performs homology analyses based on a specific cutoff that is automatically adjusted to the query dataset . Initially, the genome sequences of C. pseudotuberculosis were retrieved from GenBank, and a new project was created on the annotation platform GenDB (version 2.4) to homogenize the genome annotations . Subsequently, an EDGAR project was created based on the GenDB annotations, and homology calculations based on BLAST Score Ratio Values (SRVs) were performed. According to the SRV method, instead of using raw BLAST scores or E-values, a normalization of each BLAST bit score is calculated by considering the maximum possible bit score (i.e., the bit score of the subject gene against itself). This results in a value ranging from 0 to 1 , which is multiplied by 100 and rounded in a percentage value of homology. Finally, a sliding window on the SRV distribution pattern was used to automatically calculate the SRV cutoff with EDGAR . For this work, a SRV cuttof of 59 was estimated. Pairs of genes exhibiting a Bidirectional Best Hit where both single hits have a SRV higher than the specific cutoff were considered to be orthologous genes.The core genome was calculated as the subset of genes presenting orthologs in all the selected strains. The gene set of subject strain A was compared with the gene set of query strain B, and only genes with orthologs in both strains were members of core AB. The resulting subset was then compared with the gene set of query strain C, and the comparisons continued in a reductive manner. The pan-genome was calculated in the same way, but in an additive manner: the initial pan-genome was composed of strain A, and the non-orthologous genes of strain B were added to pan-genome A to create the pan-genome AB. The resulting set of genes was then compared with strain C, and the comparisons continued in the same manner. Finally, the singletons were calculated as genes that were present in only one strain and thus did not present orthologs in any other C. pseudotuberculosis sequenced strain.The developments of the core genome, pan-genome and singletons of C. pseudotuberculosis were calculated based on permutations of all the sequenced genomes. The developments of the core genome and singletons were calculated using the least-squares fit of the exponential regression decay to the mean values. In contrast, the statistical computing language R was used to calculate the pan-genome extrapolation using Heaps’ Law by estimating the parameters κ and γ using the nonlinear least-squares curve fit to the mean values , .The core genes of all the strains, including the biovar ovis strains and the biovar equi strains, were classified by their Cluster of Orthologous Genes (COG) functional category as the following: 1. information storage and processing; 2. cellular processes and signaling; 3. metabolism; and 4. poorly characterized. To perform this analysis, the query sets of core genes were submitted to BLAST protein (blastp) similarity searches against the COG database, the proteins with E-values higher than 10−6 were discarded, and the best BLAST results for each protein were considered for the COG functional category information retrieval. Finally, the whole-genome comparison maps were visualized using the software CGView Comparison Tool (CCT) . All the strains were plotted against C. pseudotuberculosis strains 1002 and CIP52.97 to generate two genome comparison maps. [...] The plasticity of the 15 genomes was assessed using PIPS: Pathogenicity Island Prediction Software (version 1.1.2). PIPS is a multi-pronged approach that predicts pathogenicity islands (PAIs) based on common features, such as G+C content, codon usage deviation, high concentrations of virulence factors and hypothetical proteins, the presence of transposases and tRNA flanking sequences, and the absence of the query region in non-pathogenic organisms of the same genus or related species . C. glutamicum strain ATCC 13032 was selected as the non-pathogenic organism of the same genus , and separate predictions were performed for each strain. The sizes of the islands were compared with those of all the other strains via ACT: Artemis Comparison Tool (version 10.2.0) and CCT , . Following the curation of the PAIs, the genes of all the islands in each strain were assessed for their presence/absence in all the other strains using the pan-genome data generated by EDGAR. The overall number of genes in the PAIs of the subject strain that were shared by the query strains was expressed as a percentage and plotted in a heatmap. The percentages were also converted into a nexus file, which was used in SplitsTree (version 4.12.6) to create a phylogenomic tree using the UPGMA method , . Finally, zoomed PAI figures were created using a script from CCT (create_zoomed_maps.sh) with the zoom option selected as 30×. […]

Pipeline specifications

Software tools Gegenees, SplitsTree, GenDB, BLASTP, CCT, PIPS, ACT
Databases COGs
Applications Genome annotation, Phylogenetics, Nucleotide sequence alignment, Genome data visualization
Organisms Corynebacterium pseudotuberculosis
Diseases Lymphadenitis, Mastitis, Skin Diseases, Yersinia pseudotuberculosis Infections