Whole genome sequencing of the fish pathogen Francisella noatunensis subsp. orientalis Toba04 gives novel insights into Francisella evolution and pathogenecity

Genome features

Francisella became pathogenic long before it became pathogenic to mammals

Several studies have addressed the molecular phylogeny of Francisella species and subspecies[6, 10] and we conducted a whole genome phylogenetic analysis based on all available Francisella genomes [Figure1B and a tree based on the core set of genes present in the Francisella species [Figure1A. We were able to find all the core genes from F. noatunensis subsp. noatunensis in the incomplete assembly from our sequence data and included that strain only to build the tree. This analysis confirms previous studies that there are two main groups of Francisella, the F. tularensis subspecies group and the F. philomiragia subspecies group, where the fish parasitic species belong to the latter. The phylogenetic distance between the species in the “Philomiragia group” is larger than between the species in the Tularensis group. However, the distance between the two groups is quite large. Hence, neither this nor other studies show a close phylogenetic link between the different pathogenic species within the Francisella genus. It appears likely that the mammalian pathogenic species and the fish pathogenic species result from (at least two) independent events of a free-living Francisella ancestral species acquiring a pathogenic lifestyle, and the phylogeny analysis is consistent with the fish pathogenic Francisella being result of an ancient event resulting in the genome containing no IS elements. This result is also supported by the recent publication conducting evolutionary analysis between fish and mammalian pathogens[4].

The lack of IS elements in the genome of F. noatunensis subsp. orientalis Toba04 suggests that its acquisition of a pathogenic lifestyle happened a long time ago. Moran and Plague found that pathogens or symbionts that have recently adopted an obligate host association have numerous IS elements while ancient obligate host associations most often have no IS elements[13]. This suggests that the present Francisella strain pathogenic to tilapia results from an ancient event where free-living Francisella-like bacteria infected fish (possibly tilapia) and underwent a period of gene decay and genome rearrangement where IS elements may have played a major role. As observed by Moran and Plague, bacteria that have ancient obligate host associations, lack IS-elements, and the present data on F. noatunensis subsp. orientalis Toba04 suggests that it falls into this category. This is also supported by the results of our analyses of phylogeny and pseudogenes content summarized below.

The two fish infecting Francisella does not form a monophyletic group (Figure1A) and this is an indication that they can have become parasitic at two independent events. This further indicates that they have lost there IS-elements independently and this view is supported by the fact that the two fish parasitic bacteria have different re-arrangements in their genomes.

Francisella species have highly rearranged genomes

A hallmark between sequenced Francisella genomes is a high degree of genomic rearrangements. These rearrangements are believed to be created by IS-elements and the elements are typically flanked by pseudogenes[8–10]. In addition, it has been reported that rearrangements may have been created by recombination near rRNAs[14]. Comparing F. noatunensis subsp. orientalis Toba04 with the F. philomiragia subsp. philomiragia genome revealed several rearrangements (Figure2) but there are no IS elements present in F. noatunensis subsp. orientalis Toba04. Even in the recently published paper where they have sequenced another F. noatunensis subsp. orientalis strain found weak evidence of a single IS element[4]. This opens for several different scenarios that could create the observed rearrangements seen in the tilapia parasite. One possibility is that the rearrangements are created by recombination near the rRNA genes. However, only 3 among the 4 rRNA genes in F. noatunensis subsp. orientalis Toba04 are close to the breakpoint of just one block indicating that these rRNAs are not responsible for the other rearrangements observed between the two species (Figure2). Another possibility is that the fish parasitic Francisella subspecies have had IS elements previously but have lost them and that the rearrangements have taken place prior to the loss of the IS elements. If the parasitic life-style in the fish infecting Francisella species is ancient this could also explain the lack of IS-elements in the genome (see above).

The pseudogenes of F. noatunensis subsp. orientalis Toba04 are old

We identified 252 putative pseudogenes in the F. noatunensis subsp. orientalis Toba04, a number comparable with the other pathogenic Francisella species (see Table1). We have manually curated the pseudogenes using the Artemis tool[15] aiming to avoid problems due to mutation error in homopolymer tracts often observed on pyrosequencing data. Since rearrangement is believed to be a major mechanism for generating pseudogenes in Francisella species one would expect that pseudogenes in F. noatunensis subsp. orientalis Toba04 then would be older compared to those in F. tularensis strains which is still under-going genomic rearrangement and genome decay. In order to access the age of the pseudogenes in the sequenced Francisella genomes we counted the number of inactivating mutations in the different species. The analysis shows that F. noatunensis subsp. orientalis Toba04 pseudogenes have significantly more inactivating mutations than the other pathogenic Francisella strains (Figure3). If we assume a similar mutation rate in the different species, this indicates that the pseudogenes in F. noatunensis subsp. orientalis Toba04 on average are older than the pseudogenes in the mammalian pathogenic Francisella strains (or alternatively posses a higher mutation rate). The high number of pseudogenes identified in the pathogenic Francisella species (see Table1) is typical for bacteria associated with eukaryotic hosts[16]. It has been suggested that pseudogenes may be quickly removed from bacterial genomes (Kuo and Ochman 2010) and the majority of pseudogenes will only possess 1 inactivation mutation. The data obtained from the fish pathogenic F. noatunensis subsp. orientalis Toba04 clearly demonstrate a higher proportion of pseudogenes with multiple inactivation mutations (see Figure3) pointing towards older pseudogenes in this species.

Old pseudogenes under neutral evolution in F. noatunensis subsp. orientalis Toba04

Several recent papers have addressed the extinction dynamics of pseudogenes in bacterial pathogens. Kuo and Ochman found that in Salmonella, there appears to be positive selection for deletion events leading to rapid removal of pseudogenes[17]. We performed two analyses with the pseudogenes present in F. noatunensis subsp. orientalis Toba 04, F. tularensis subsp. tularensis SCHUs4 and F. tularensis subsp. holarctica OSU18 to investigate the frequency and nature of pseudogenes in these genomes. For every pseudogene we counted the number of inactivating mutations and also calculated the nucleotide substitution rate between the pseudogenes and homologous gene. Given that the pseudogenes are not under selection (neutral evolution), one would expect a linear line in the log-normal graph. To some approximation, this is what we observe for all strains expect for F. noatunensis subsp. orientalis Toba04 where we see fewer than expected pseudogenes with one inactivating mutation and also a larger than expected number with many (7 or more) inactivating mutations (Figure3). The reason for this deviation may be that F. noatunensis subsp. orientalis Toba04 has evolved more extensively after its transformation into an intracellular pathogen so that most pseudogenisation events are old. This is also consistent with old genomic rearrangements (happening before the IS elements were lost) since pseudogenisation is often a consequence of rearrangements. It may also be that Francisella noatunensis subsp. orientalis Toba04 already lost all its non-essential genes some time ago so that losses of additional genes are deleterious. This would be consistent with the data for the other strains analysed if these have become pathogenic (and metabolically incompetent) more recently and the process of genome decay is still on-going. In the nucleotide substitution rate calculation analysis we found large number of pseudogenes at the rate of 0.4-0.5 (Figure4) for F. noatunensis subsp. orientalis Toba04. This is also in agreement with the suggestion that the pseudogenisation has happened a long time ago in the Tilapia parasite. In none of the strains analyzed we see the pattern reported by Kuo and Ochman for Francisella indicating positive selection for pseudogenes deletion. This may relate to the difference in lifestyle between Salmonella and Francisella pathogens. The recent publication from Sjödin et al., also showed that Francisella noatunensis subspecies has evolved well before Francisella tularensis subspecies and Francisella noatunensis subsp. orientalis has no IS elements and the same scenario might have appeared in those strains as well[4].

Comparative analysis of F noatunensis subsp. orientalis Toba04 with other available Francisella genomes

We performed a comparison between the Francisella genomes (Table1) in order to find shared and unique genes for each of the genomes. For this analysis we grouped similar genes shared between the species into ortholog clusters and the gene numbers given below refer to the number of ortholog clusters. We identified a set of 1,077 genes present in all the analyzed genomes – representing a potential core set of genes for Francisella (Figure5). As expected, the free-living metabolically competent strains had the highest number of unique genes although the fish and mammalian parasites each have 305 and 233 unique genes among the Francisella genomes analyzed. This indicates that the pathogenic strains not only have lost genes after turning parasitic but they have also acquired genes. Whether gene loss or acquisition has happened before or after this transition cannot be ruled out from the available data. Gene acquisition is likely to happen by horizontal gene transfer. For 42 genes unique to F. noatunensis subsp. orientalis Toba04 and for 30 genes unique to F. tularensis subsp. tularensis we found significant homology (more than 30% identity over at least 50% of the gene length) to genes in Genbank from bacteria species likely to be present in the same environment as the respective Francisella strains (Additional file1: Tables S1 and S2). It is not unlikely that most of the other acquired genes in each of these pathogenic strains may be a result of horizontal gene transfer.

Intact unique genes present in free-living Francisella genomes

We identified 522 genes present only in the metabolically competent genomes. The assumption is that these genes have been present also in the ancestors of the analyzed pathogenic strains but lost after they became parasites. In agreement with previous reports, we find that many of the 522 genes are involved in metabolism, intracellular transport and amino acid biosynthesis[8–10]. In addition we identified 17 membrane-associated proteins and 62 proteins annotated with signal-peptide (3 also annotated as membrane associated) which might play an important role in immunity during host parasite interactions, which are present only in metabolically competent genomes. Genes like the capsule polysaccharide biosynthesis protein required for immunity to destroy foreign antigens are present only in the metabolically competent genomes. We used DAVID[18] to identify over-represented functional terms (including Gene Ontology terms) among the unique genes from the metabolically competent species(compared to their frequency in the complete Francisella philomiragia genome). Functional terms related to DNA (integration, replication, binding, and metabolism), amino acid metabolism, transporters, membrane-association and signal peptides are among the most significant terms (Additional file1: Figure S3). The present analysis narrows down the list of genes unique to free-living Francisella. Taken together, these results are consistent with earlier analyses and indicate that pathogenic Francisella strains have lost a substantial number of genes (512), and many of these have functions related to metabolism and DNA. This indicates that the parasitic Francisella species share many features in their adaptation to a parasitic lifestyle and that the fish pathogen may serve as a valuable model to study features of the mammalian parasitic Francisella.

Intact unique genes present in the fish parasite F. noatunensis subsp. orientalisToba04

We found 305 genes unique for the F. noatunensis subsp. orientalis genome. Performing BLASTp search against non-redundant protein database we were able find significant matches for 260 genes which are not present in any of the other Francisella strains. Among these genes a putative lipopolysaccharide biosynthesis membrane protein is present. It is important for the virulent causing serious diseases in human and animals[19, 20]. The other genes represent transporters, transferases, purine biosynthesis, thiamine biosynthesis, oxidation reduction, catalytic activities and hypothetical proteins and could be important for the pathogenecity of the strain.

Intact unique genes present in the human parasitic Francisella genomes

The human parasitic Francisella genomes contain 233 genes that are not present in the other Francisella genomes studied. We compared these 233 genes with all genes in the non-redundant protein database using BLAST to find homologs in other mammalian strains and found 162 unique genes that are shared between F. tularensis subsp. tularensis, F. tularensis subsp. holarctica, and F. tularensis subsp. mediasiatica (Additional file1: Table S3). There are 63 genes which are present only in F. tularensis subsp. holarctica strain, all of them hypothetical proteins. In F. tularensis subsp. mediasiatica 11 genes are unique and includes a protein involved in the type III restriction-modification system (FTM_0875), part of the defence mechanism against foreign DNA[21], where as F. tularensis subsp. tularensis has 44 unique genes and all of them hypothetical proteins. The unique proteins present in this strain might be related to the highly virulent nature in this strain although functions of these proteins are unknown. The other 44 genes are shared between these three strains. 30 of these genes have homologs in other bacteria and their presence in the Francisella genomes may be due to horizontal gene transfer. These genes represent functions like ATP binding, transferase activities and lipopolysaccharides biosynthesis and 11 of them are hypothetical proteins.

The genomes of the studied fish and human pathogenic genomes all contain genes unique among Francisella species. It seems unlikely that they have all been present in an ancestral Francisella genome and more likely that some genes have been horizontally transferred from other organisms, likely in most cases from environmental bacteria. Most of the genes like lipid biosynthesis, polysaccharides biosynthesis, cold-shock DNA-binding domain-containing protein, membrane protein and other genes in the horizontal gene transfer list could have been useful for survival of the pathogen or be involved in virulence. The hypothetical proteins present in F. tularensis subsp. tularensis could be important to determine the high virulent nature of this species. Having a novel pathogenic Francisella genomic sequence available, we use the opportunity to analyze in a comparative manner with the selected set of Francisella genomes (Table1) with focus on systems that are believed to be important for virulence. The distant evolutionary relationship between the human pathogenic and the fish pathogenic strains, can potentially give some insight into virulence and also genome decay and what is the essential, core, set of genes in Francisella.

Virulence mechanism

Sequencing and assembly

The F. noatunensis subsp. orientalis Toba04 strain was sequenced using 454-pyrosequencing[35] generating 263,717 reads consisting of 56,522,682 bases. These were assembled using Newbler v.2[35]http://454.com/products-solutions/analysis-tools/gs-de-novo-assembler.asp leading to 21 contigs with total length of 1,848,209 bases and N50 of 215,480 base pairs. The gaps were closed by first analysing the contig graph, which presents the connections between contigs, based on the repeat information present in the reads. The edges between the contigs with coverage less than 10 were removed. Further comparative analysis (using BLAST) with fully sequenced strains of F.tularensis subsp. tularensis SCHUs4 was performed, and the edges between contigs which were inconsistent across these strains were removed and incremental assembly using runViewer program present in the Newbler was performed. This led to 8 contigs with total length of 1,847,034 and N50 of 429,132. Later a series of PCR, suggested on the basis of contig graph, were performed to join these contigs and check the correctness of assembly. Several sets of specific primers were designed at the ends of each contig.

GoTaq PCR enzyme (Promega) was used in all amplifications. Two and two primers were combined, in individual PCR reactions, so that all possible combinations of primers were tested (an all against all combination approach). Genomic DNA identical to the DNA used for sequencing was used as template. Fragments appearing as single bands in agarose gel electrophoresis were either purified using ExoSAP-IT (GE Healthcare) and sequenced on both strands using the PCR primers and the BigDye chemistry (Applied Biosystems), or the PCR fragments were cloned into a pCR4_TOPO vector following the supplied instructions (Invitrogen) and several random clones were sequenced using the vector primers. All the sequences were used in the assembly. To check the correctness of assembly 10 pair of primers were designed in one of the contigs and PCR reactions were performed. All reactions gave result as expected. This led to 19 primers out of which 12 gave products (5 were one sided) and 7 failed. This information was fed back into the assembler as fake paired ends and another series of incremental assembly was performed leading to 5 contigs with 1,715,028 bases and N50 of 1,033,009. Another series of PCR based on these results led to 63 primer products and information about repeats and orientation. A straight-forward fake paired-end presentation into the assembler broke up the assembly due to short matches of product in various locations confusing the assembler. So the reassembly was done incrementally in a fashion maintaining the consistency between PCR results and reads based assembly. This lead to 1 big (length 1,857,341) and 3 small scaffolds (total length 8940). The pink line in Figure5 is the scaffold linking while plain black line contig linking. Since all the contigs do not have paired end connection information, they don’t show in the scaffold-graph (including contigs of less than 500 bases), these gaps can also be filled manually (just replacing the gap with contigs based on graph). For example between 14 and 15 we can put in 3, 2, and 5 but Newbler did not have enough information to do it automatically (Additional file1: Figure S2).

Annotation

We used the prokaryotic annotation pipeline from TIGR to annotate the genome: Glimmer 3[36] was used to predict the genes in the genome. NcRNAdb[37] and RNammer[38] was used to predict 23s and 16s RNA genes. tRNAscan-SE[39] was used to predict tRNAs present in the genome. We used TMHMM[40] to predict transmembrane helices and SignalP[41] to predict signal peptides. The ORFs predicted were compared against NCBI’s non-redundant database using BlastX. Protein coding genes from predicted results were curated manually using Artemis[15]. All the predicted proteins were searched against NCBI’s COG database (Clusters of Orthologous Groups of proteins based on phylogenetic classification of proteins encoded in complete genomes) to find the protein family. Cognitor was used to predict the COGs for each protein[42]. EC numbers were assigned to the proteins using the BRENDA database[43]. Interpro-Scan was used to add domain based annotation of the proteins[44]. CGview was used to make circular genome plot[45].

IS elements and genome rearrangement plot

IS-finder[46] web server was used to find the IS-.elements. We used BLASTx[47] to compare against the IS elements database and the results were manually checked. Nucmer in Mummer 3.0[48] package was used to prepare the comparison plot between F. noatunensis subsp. orientalis and F. philomiragia subsp. philomiragia. To plot F. noatunensis subsp. orientalis’s pseudogenes against F. philomiragia subsp.philomiragia genome, we compared pseudogenes against F. philomiragia subspecies genome and the gene coordinates were extracted from BLASTp results using in-house Perl program. We used that coordinates to map the pseudogenes in the comparison plot.

Mutation count

To count the number of mutation in the pseudogenes in Francisella subspecies, we compared all the pseudogenes present in the genomes against the NCBI’s non-redundant database. From the Blast result we calculated the possible inactivating mutations including insertions, substitutions, premature stop codons and point mutations using a Perl program written in-house. The log-normal graphs were made using MS-Excel.

Maximum likelihood distance

We calculated the maximum likelihood distance between the pseudogenes present in Francisella species and its 1^st homologous hit from BLASTp result. We used biodist module in bio++ package to calculate the maximum likelihood distance values using L95 nucleotide substitution model.

Phylogenetic tree

Mauve 2.3.1[49] was used to prepare the whole genome alignment between all the Francisella species. The same software was also used to show the rearrangement between F. noatunensis subsp. orientalis and F. philomiragia subsp. philomiragia. The location of rRNAs and IS-elements were marked manually. The genome alignment was loaded in to the MEGA4[50] for editing and subsequently the Neighbor-joining method was used to predict the tree with 1000 replicates for the bootstrap value calculation. We used the same procedure for the phylogenetic tree predicted using core gene sets. Incomplete assembly of F. noatunensis subsp. noatunensis was compared against protein sequences of F. noatunensis subsp. orientalis to find core set of genes commonly present in Francisella species.

Metabolic pathways

Pathway tools software[51] was used to predict the metabolic pathway for F. noatunensis subsp. orientalis. We also predicted metabolic pathway for F. philomiragia subsp. philomiragia to use it as reference for F. noatunensis subsp. orientalis and comparative analysis. The Pathologic module was used to predict the pathways for both the genomes. The function Overviews->highlight->species comparison were used to compare pathways between genomes.

Genome comparison

Protein sequences from F. tularensis subsp. tularensis SCHUs4, F. tularensis subsp. holarctica OSU18, F. tularensis subsp. mediasiatica FSC 147, F. tularensis subsp. novicida U112, F. philomiragia subsp. philomiragia ATCC 25017 and F. noatunensis subsp. orientalis genomes were extracted from GenBank file. We used BLASTp[47] to compare the protein sequences against themselves to find unique and the genes which are shared between the genomes. We calculated two values to classify if two proteins from different species are same: (i) we calculated the score bit percentage from the top homologous hit. We divided the top hit’s score bit value by the values from subsequent hits from other Francisella species. The score bit percentage should be >=65% of top homologous hit’s score bit to classify the same two proteins. (ii) We calculated the alignment length percentage between the query and subject. The alignment length percentage is calculated from the length of the subject protein and length of the alignment. It should be >=75% between query and subject to classify two proteins as the same. We used the same method to cluster the proteins in to ortholog groups.