БИОИНФОРМАТИЧЕСКИЙ АНАЛИЗ ГЕНОМНЫХ ОСТРОВОВ В ЦЕЛЫХ ГЕНОМАХ ШТАММОВ PISCIRICKETTSIA SALMONIS

The control of piscirickettsiosis in Chile has been addressed using antibiotics, which has decreased the susceptibility of the pathogen to these treatments

P. salmonis generates outbreaks of piscirickettsiosis in host fish population, producing systemic injuries with a variety of clinical signs. It is also characterized by functional mechanisms that increase its survival like biofilms formation on different parts of the host and form viable structures on inorganic materials, highlighting its role as a facultative pathogen

Bioinformatic tools and approaches can provide deeper insights into the genomic foundations of microorganism’s pathological activities. Genomic islands, ranging in size from 10 to 200 kb, play a key role in this context as they contain genes with functional interest that are grouped by mobile genetic elements

Genomic islands have been categorized into 'fitness', 'ecological' and 'pathogenicity islands'

A previous report analyzed a pathogenicity island with 38.57% GC content, where 23 genes were reported, registering both hypothetical proteins and genes associated with virulence, using PIPS (Pathogenicity Island Prediction Software). The virulence mechanisms of this pathogen and many others can be understood by analyzing pathogenicity islands, mobile genetic elements that rotate between chromosomes and plasmids, possessing genes involved in virulent processes and/or antibiotic resistance. Three genes associated with virulence were evaluated: transcription-repair coupling protein (tcf), Mg-dependent DNase (dnase) gene and lytic transglycosylase (liso). They found a similarity between the above genes and those found in Neptunomonas antarctica, Endozoicomonas sp. and Legionella spiritensis, respectively, and hypothetical proteins that have the potential of being compared against a database for future identification and classification

The National Center for Biotechnology Information (NCBI) has a database containing annotated complete genomes of various strains of P. salmonis, making it possible to compare complete, assembled, and annotated strains and genomes. This includes the complete genome of the original strain isolated in Chile in 1989 (ATCC VR-1361)

This work focused on analyzing the genomic islands in P. salmonis, evaluating the relationship between the number of pathogenicity islands and their virulence genes in complete genomes of different strains of the EM and LF genogroups, as well as determining the presence of fitness islands.

2.1. Bioinformatics tools and P. salmonis strains

P. salmonis strains Psal-006b, Psal-009 and Psal-011, Psal-001, Psal-005, Psal-069 and Psal-072 were selected because they were isolate from southern Chile after LF-89 isolation date, besides LF-89 as original strain. Genome information was downloaded from NCBI in gbk format (Table 1).

Two bioinformatics tools were selected for double prediction confirmation of candidate genomic islands. Genomic Island Prediction Software (GIPSy) v.1.1.3 (www.bioinformatics.org/groups/?group_id=1180)

2.2. Locus conservation between genomes and islands

The selected islands were ordered and broken down by each locus present, and the resulting matrices from BLASTp in GIPSy were used as a database for comparison against the gbk file format database.

The identity percentage of each intersected island was averaged to determine the degree of conservation between strains and genogroups. However, strains LF-89 and Psal-006b were determined differently from the rest, so the average included them. The same process was repeated but contrasting the ordered loci with respect to the same islands previously determined. The identity values obtained for each locus were averaged and grouped by island, both to compare with respect to genogroup and island, and to culminate in a comparison between genomes, averaging the values of the islands in each genome.

2.3. Search for non-pathogenic elements in genomic islands

The non-virulent factors determined by BLASTp against databases COG

The islands were broken down by loci, assigning an individual percentage of identity against the database and a percentage of identity with respect to LF-89 as the original strain, to distribute the information regarding the genes encountered in contrast with each strain.

2.4. Description of specific pathogenicity islands

The selection criteria used to analyze islands were most conserved in general, most conserved per genogroup, and less conserved, focusing on total values above 0%. These criteria were used to prioritize the value of the mean difference between genogroups.

Three islands for each category were selected for the subsequent description of their gene identity. GIPSy was used to confirm the islands, and Artemis (https://www.sanger.ac.uk/tool/artemis/), for genome visualization

2.5. Visualization of genomic rearrangement in bacterial chromosomes

The eight bacterial strains under study were analyzed for synteny with the pyGenomeViz library

This study is an update of the previous report

BLASTp was performed against GIPSy for all sequences, yielding a total of 56 databases. The results revealed a total of 100 islands among all strains, with an average of 12 loci per island. The largest islands were found in Psal-011 and Psal-069 with 47 loci. The averaged similarity between strains was established, evidencing that the islands from LF-89 and Psal-006 presented values lower than 75% at the genome level. The strains of the EM genogroup exhibited a higher percentage of identity, indicating that, with respect to all the genomes, they do present higher percent identity values. Another remarkable aspect was the decrease in the identity percentage of the islands with regard to the genome of LF-89, unlike its counterpart, the genome of Psal-072 shown in the eighth column, that was constantly increasing (Table 3).

The two strains with the highest percentage of conserved islands at the level of total genomes correspond to Psal-069 and Psal-005, respectively. Considering genome, the strain with the most conserved islands corresponds to Psal-001, with an average identity of 88.77% between genomes. The most conserved in the LF group was LF-89 with 85.81%, and with respect to EM, it was Psal-001 with 99.01%. The strains with the least general conservation at the chromosome level were LF-89, Psal-006b and Psal-009, but simultaneously showed the lowest variation between genogroups. Psal-011, despite being from the LF genogroup, both values were more alike to those of the EM genogroup, reaching values of 98.82% and 76.13% respectively.

The LF-89 and Psal-006b strains presented the least conserved islands, with values close to 3% for LF-89 and less than 1% for Psal-006b, except for their relationship with Psal-009, with 4.09% conservation between islands. Psal-009 was the one that showed the least difference between genogroups (Table 4).

GIPSy analysis revealed that the islands showing pathogenicity features were the most abundant in all genomes, while those classified as metabolic represented 18%. Antibiotic resistance were only 5% of those evaluated, and symbiotic reached 7%, with identical percentages observed in LF- and EM-type genomes (Table 5).

The most conserved islands between strains were LF-89 island-2, Psal-006b island-2 and Psal-001 island-6, with a mean total of 96.63% with a difference of 4.1% between genogroups. From the LF genogroup, LF-89 island-2, Psal-006b island-2 and LF-89 island-8 had higher values of 98.35%, 97.90% and 96.80%, respectively. From the EM genogroup, Psal-009 island-9, Psal-009 island-16 and Psal-011 island-11 had values of 100% for the EM genogroup and a difference between genogroups of 50.00%, 14.90% and 10.00%, respectively. Finally, Psal-009 island-5, Psal-006b island-1, and Psal-006b island-4 stood out.

Similar genes were observed among strains, with over 90% identity on a proportional basis for all genomes. Strain Psal-009 exhibited the largest chromosome and the greatest difference regarding Psal-006b, but with similarities between them, but in different locations. Strain Psal-011 and Psal-001 showed regions smaller than 0.4 Mb and with a similarity denoted in green, the same as between Psal-001 and Psal-005. Additionally, regions above 1 Mb are also green-colored and present both in strains Psal-009 and Psal-011, and Psal-069 and Psal-072. An abundance of red color can be noticed throughout the image, indicating preservation of genes in the same orientation and with identity values over 90%. It must be noted that the genome of Psal-006b had to be turned around because it was not clear in the image (Fig 1).

Figure 1 - Complete synteny of bacterial chromosomes of strains of genogroup LF (four above) and genogroup EM (four below), in proportional scale of 3M base pairs with barcolor according to gene orientation comparison, being red for same strand and green in opposite strand

Note: links between genes were filtered, registering only links over 90% identity percentage

DOI:10.18454/jbg.2023.22.4.6

The genes from the antibiotic resistance database identified in P. salmonis genomes corresponded to sensor proteins for BaeR, TnpR, alkaline phosphatase synthesis, and Populus trichocarpa predictor protein. For metabolic type, analyses found a relationship with NAD-dependent aldehyde dehydrogenase, arabinose efflux permease, amino acid transporters, acetyl CoA dehydrogenase, amidase related to nicotinamidase, endonuclease, PEP phosphonomutase and dipeptide/tripeptide permease, highlighting PEP for sharing 60% identity with Vibrio cholerae methylisocitrate lyase. For symbiotic factors, a relationship was found with SDR short-chain dehydrogenase/reductase, Bradyrhizobium diazoefficiens gene, and two-component phosphate regulon (Table S3).

The most conserved genes observed in each category were mobile elements, such as transposases and integrases, and genes described in Swiss-Prot, such as fur, bamE, recN, ppnK, and hslO. The second category was the most diverse for presenting bamE, fur, recN, and ppnK, mainly due to the reappearance of islands in both Psal-006b and LF-89. The most conserved islands in the EM genogroup analyzed showed a total of 12 genes, finding genes for mobile elements, transporters, cobalt chelatase, and hypothetical proteins. Finally, 28 genes were found in the least conserved chromosomal islands, with a predominance of hypothetical proteins, genes related to phages, and components of the type IV secretion system (Table S4).

Two islands from Psal-011 and Psal-001, island-4 in each, were composed of two IS91 family transposase genes, a group II intron reverse transcriptase/maturase, and five hypothetical proteins. Of the three hypothetical proteins present in both islands, analyzed by BLASTp at NCBI, it was found that one shared 34.96% similarity, on the part of the two strains, with -lactamase coming from a different strain of P. salmonis, whereas the other two showed similarity to only hypothetical proteins from different strains of the same bacterium. Despite the code difference, both strains shared 86.99% identity with glycine dehydrogenase subunit I and only a reduced relationship with other hypothetical proteins (Table 6).

Mechanism of acquisition of genomic islands have been studied, such as the high-pathogenicity island in Enterobacteriaceae linked to integrative and conjugative elements

EM genogroup may have rearranged its islands and genes more efficiently, while the LF genogroup showed fewer independent islands when using the GIPSy tool. This difference in rearrangements can be correlated with the difference in pathogenicity between strains from these genogroups. A comparative study evaluating pathogenesis and virulence in isolates from both genogroups highlighted that the EM-90 strain, equivalent in code to Psal-001, had a higher accumulated mortality and a reduced time to death, compared to LF-89

The islands analyzed were characterized by exhibiting only mobile elements and low GC content. However, it has been found that P. salmonis induces an inflammatory response together with a positive regulation to produce interleukin-8

These have been studied and identified in other microorganisms, such as Staphylococcus aureus, contributing to virulence

Databases revealed that 18 non-virulent genes were quantified, highlighting in resistance genes: phosphate regulon sharing 30% identity with the gene encoding BaeR sensor protein in Shigella dysenteriae, and response regulator sharing 43% identity with alkaline phosphatase synthesis in Lysteria monocytogenes. There is evidence of BaeR response regulator in E. coli, related to the drug exporter gene expression in the multidrug resistance system conferred by efflux transporters from the resistance-nodulation-cell division (RND) family

Metabolic genes elbB shared 53% identity for uncharacterized protein involved in early stage of isoprenoid biosynthesis by Pseudomonas aeruginosa , important during intracellular survival of pathogenic bacteria

The third group of genes identified are symbiosis factors, like the regulon phosphate gene of Rhizobium etli and the phoR gene. Reference genes in P. salmonis were the same for resistance factors category.

P. salmonis has many hypothetical proteins in islands, maybe because of the scarcity of gene characterization or curation. This is necessary to further elucidate the mechanisms in P. salmonis or understand the importance of specific domains for antibacterial agents’ development. Based on this and the percentage of factors within bacterial chromosomes (Table 5), it may be suggested that would be use metabolic mechanisms rather than antibiotic resistance.

The analysis of genes presents in selected islands and their follow-up revealed that hypothetical proteins stand out in the first place with a total of 17 genes, followed by mobile elements of IS group with almost 350 among all types, including unclassified transposases. It is speculated that the IS91 family originated by plasmid evolution

Nonetheless, crucial genes related to pathogenic process has been found, such as the fur, recN and hslO gene. Fur is associated with iron uptake

The mazE gene was found in island 1 of Psal-009. It has been demonstrated that mazEF complex overexpression leads to biofilm formation under stressful conditions

Considering the high rearrangement potential, justified by the finding of a large number of mobile elements, a representative figure was elaborated to assess the degree of gene conservation within already established genomic islands (Tables S1 & S2). This allows for determining the degree of conservation in plasmids of the same strain, in the bacterial chromosome, and between different plasmids (Fig. 2). Based on this, it can be stated that the mazEF complex is also found in plasmids of strains LF-89, Psal-001, Psal-005, Psal-011, Psal-069 and Psal-072. Also, Psal-011 shares a rearrangement resembling more that of the EM genogroup than that of the LF genogroup. Also, the EM genogroup presents a pattern of gene conservation in the chromosomes over plasmids, unlike Psal-009.

Figure 2 - Conservation of pathogenicity islands in genomes of P. salmonis strains, measured by BLASTp, distributed by strains, and ordered by chromosome location

Note: a – conservation of pathogenicity islands in plasmids of the same strain; b – conservation of pathogenicity islands in bacterial chromosomes; c – conservation of pathogenicity islands in bacterial plasmids

DOI:10.18454/jbg.2023.22.4.8

The data collected was used to visualize the variation and displacement of bacterial chromosome regions with respect to adjacent strains (Fig 3). Within the LF genogroup, strains Psal-009 and Psal-011 were double-checked, but a random rearrangement can be seen, without any salvageable pattern, beyond the convergence of the islands in the region close to 500 kb. On the other hand, more evident trends are observed in the EM genogroup, such as the maintenance of islands 2 and 3 among all strains, and in strain Psal-072, the incorporation of elements from island 6 in Psal-069 and part of the migration of elements in island 3 of the same strain. Another variation consists of island 8 in Psal-001 that constantly migrates to island 9, conserving, in turn, island 10, where this region, between islands 9 and 10, is probably relevant in the pathogenic process. Nonetheless, in the same genogroup, islands from the original strain disappeared in neighboring strains, such as islands 6 and island 5 in Psal-005, but reappeared in Psal-072, demonstrating how EM genogroup rearranges effectively. Another interesting pattern observed was how Psal-001 strain, being an original strain in EM, visually exhibited organization and rearrangements more alike to Psal-011. In addition to the above, when observing the islands in Psal-001, most of them presented similarities with Psal-011, unlike island 5, allowing us to suggest that Psal-011 could present a pathogeny more like Psal-001 than LF strains.

Figure 3 - Integrative synteny of pathogenicity islands in genomes with respect to original strains LF-89 and Psal-001, comprising a range between 0 and 3M base pairs

Note: older strains with pathogenicity islands for each genogroup; sequential numbers were used; for adjacent strains, the previous number was preserved if only one island with former content was present; the apostrophe (') was added if the previous island occurred in more than one island in a more recent strain; new islands unrelated to previous strains were identified with a new number

DOI:10.18454/jbg.2023.22.4.9

Synteny of bacterial chromosomes showed patterns, such as strains similar to each other from gene rearrangements perspective. However, unique rearrangements were also observed in Psal-006b and Psal-009, which exhibited the greatest variations. The difference between Psal-001 and Psal-011 was the same as that between Psal-001 and Psal-005, allowing us to propose that Psal-011 may have a similar expression profile to Psal-005 relative to its own genogroup. There is also a possibility of a large inversion in a region of about 1 Mb in size in Psal-009 with Psal-011, and Psal-069 with Psal-072. Moreover, it is relevant to highlight the fact that Psal-011 and Psal-009 are considered genogroup EM strains with pan-genome analysis criteria

[52]

.

Due to the aforementioned, the need for identifying virulence factors among high- and low-virulence strains is reaffirmed through the use of genomic and proteomic tools

The EM genogroup presents similar islands between strains, in a greater proportion than the LF genogroup, with a lower variability. Among the islands observed, a low number of virulent factors was detected, these factors could accumulate in adjacent regions, leaving the islands as gateways for the pathogenic process. On the metabolic, resistance and symbiosis factors, despite presenting low identities, the genes themselves have the same names, reaffirming the type of non-virulent factor in genomic islands. Further studies consider the intervention of these islands and the impact on pathogenicity, in addition to validating the resistance, metabolic and symbiosis islands, regarding their role pertinent to the name of the subtype.