Leptospira isolates

A globally representative collection of the 20 Leptospira species known at the advent of this project was analyzed here, and provided by members of the leptospirosis research community (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Leptospiral species metadata and genome statistics.

https://doi.org/10.1371/journal.pntd.0004403.t001

DNA preparation from isolates

A standard operating procedure was established for all contributing laboratories to follow in preparing DNA for whole genome sequencing. Leptospira were considered to be like Gram-negative bacteria for the purpose of DNA extraction because of the presence of lipopolysaccharide and a thin peptidoglycan cell wall. Either ~10¹² bacterial cells or 30 mL of the densest possible culture of Leptospira (in EMJH medium) were centrifuged and the pellet resuspended in 180 μl Buffer ATL (all buffer abbreviations are according to the manufacturer and related specifically to components of the kit) (Qiagen Tissue Kit, Valencia, CA, USA), and then purified according to the protocol “Purification of Total DNA from Animal Tissues (Spin-Column Protocol),” including the use of proteinase K, thorough vortexing throughout the procedure, and incubated at 56°C until the cells were completely lysed, according to the manufacturer’s instructions. Lysis was usually complete in 1–3 hr. Before adding Buffer AL, 10 μl of RNAse cocktail was added (mixture of two highly purified ribonucleases, RNase A (500 U/ml) and RNase T1 (20,000 U/ml); Ambion, Life Technologies, Carlsbad, CA), and incubated for 30–60 min at 37°C. After vortexing for 15 s, Buffer AL was added to the sample, which was again mixed thoroughly by vortexing, followed by adding ethanol (96–100%), and mixed again by vortexing. It was considered essential that the sample, Buffer AL, and ethanol were mixed immediately and thoroughly by vortexing or pipetting to yield a homogeneous solution. After this point samples were handled with large bore, genomic DNA-compatible tips. The mixture from step 3 (including any precipitate) was pipetted into the DNeasy Mini spin column placed in a 2 ml collection tube (provided), centrifuged at 6000 x g for 1 min and the flow-through discarded. The DNeasy Mini spin column was placed in a new 2 ml collection tube, and washed with 500 μl Buffer AW1. The DNeasy Mini spin column was placed in a new 2 ml collection tube, washed with 500 μl Buffer AW2, and centrifuge for 3 min at 20,000 x g (14,000 rpm) to dry the DNeasy membrane; remaining ethanol was considered to interfere with sequencing reactions and processing. The DNeasy Mini spin column was eluted twice by adding 2 x 100 μl Buffer AE directly onto the DNeasy membrane, incubating at room temperature for 1 min, and then centrifuged for 1 min at 6000 x g (8000 rpm). DNA was shipped on dry ice to JCVI for sequencing. Quality control included certification of intact, high molecular weight DNA and required 15–30 μg for fragment libraries, complementarily documented by agarose gel image containing a DNA Mass Ladder, OD260/280 determination, and an estimated DNA concentration from a fluorometric assay (SYBR Green, Quant-IT PicoGreen dsDNA Assay Kits).

Genome sequencing, draft assembly and annotation

The genomes of 17 Leptospira species (the whole genome sequences of the remaining 3 species studied in the present analysis, L. interrogans serovar Copenhageni strain L1-130 and L. biflexa serovar Patoc Strain Patoc I, and L. licerasiae serovar Varillal strain VAR010, already were published [33, 37, 38]) were sequenced at JCVI by a combination of Illumina HiSeq (2x100 bp) and 454 FLX Titanium. Briefly, paired-end libraries were constructed for each sequencing technology from random nebulized genomic DNA in the 300–800 bp (Illumina) and 2–3 kb (454) size ranges. Sequence reads were generated with a target average read depth of ~ 20–30 fold (454) and ~60-fold (Illumina) coverage. Sequences for all 18 strains were assembled using the Celera Assembler version 6.1 [38], and ordered using NUCmer [39] to align the contigs to the best-matching closed Leptospira reference genome. All 18 new genome sequences underwent manual gap closure to elevate the genome status to improved high-quality draft (Table 1). Contigs were annotated for protein- and RNA-encoding features using the JCVI automated annotation pipeline essentially as described [40–43] except HMMs were run using HMMER3 [44].

Phylogenetic analysis

16S rRNA trees were generated by first creating a multiple sequence alignment to the bacterial 16S rRNA reference alignment using Ribosomal Database Project release 10 (RDP-X) [45]. The aligned FASTA sequences were downloaded and trimmed to remove gapped columns using Belvu (v2.31) [46]. Based on the alignment, a bootstrapped Maximum-likelihood tree was subsequently inferred using phylipFasta, an in-house wrapper script [47] for the Phylip program [48, 49].

SecY trees were created by first aligning secY nucleotide sequences using Clustal Omega [50] with 100 combined guide-tree/HMM iterations. The multiple sequence alignment was trimmed to remove gapped columns and a bootstrapped Maximum-likelihood tree was inferred as was done for the 16S rRNA trees.

The nucleotide sequences of 7 MLST housekeeping genes were extracted from the 20 genomes, and sequence types (STs) assigned using the MLST website (http://leptospira.mlst.net/) [51]. A multiple sequence alignment of the concatenated sequences of 7 MLST loci was performed using ClustalW implemented in MEGA version 5 [52]. A maximum likelihood tree was re-constructed using an algorithm implemented in PhyML version 3.0.1 [53]. The model of sequence evolution used was the generalized time-reversible (GTR) model with gamma-distributed rate variation. The CLC Main Work Bench version 7.0 was used to edit and display the tree (Qiagen, USA).

Universal protein marker trees were constructed using a set of 39 proteins that are universally conserved among bacteria and produce monophyletic phylogenies, suggesting that they undergo minimal horizontal transfer (S1 Table) [54–56]. Protein sequences were aligned using ClustalW (v1.83) [57] using default settings. The alignment was trimmed to remove gapped columns using trimAl (v1.2r59) using–nogaps option and–fasta output option [58]. Aligned and trimmed predicted amino acid sequences of each species were concatenated as described previously [40], in the following order: AspS, FusA, GyrB, InfB, LepA, LeuS, PyrG, RplA, RplB, RplC, RplE, RplF, RplK, RplM, RplN, RplO, RplP, RplR, RplV RpoA, RpoB, RpsB, RpsC, RpsD, RpsE, RpsG, RpsH, RpsI, RpsK, RpsL, RpsM, RpsO, RpsQ, SecY, SerS, TopA, TsaD, Tuf, and YchF. The resulting alignment of 11241 amino acids was used to generate a Maximum-likelihood tree from 100 bootstrapped replicates using raxmlFasta, an in-house wrapper script for the raxmlHPC (v7.0.4) [59].

A pan-genome tree was constructed using the mean of the BLASTP Score Ratio (BSR) as described previously [60]. The PanOCT output file 100_pairwise_BSR_distance_matrix_phylip.txt was used as input for Neighbor [48, 49] to build an unrooted UPGMA Neighbor-Joining tree. This PanOCT output file is a Phylip-style distance matrix derived from the pairwise mean BSR of core proteins present in 100% of genomes where a single value is presented for each pair of genomes in the pan-genome.

In silico DNA-DNA hybridization

Genome relatedness among Leptospira strains was determined pairwise from fully or partially sequenced genomes using the genome-to-genome distances (GGD) calculator (S2 Table) [61]. This analysis was complemented by in silico DNA-DNA hybridization as previously reported [11].

Pan-genome analysis

Clusters of orthologous proteins were generated using version ver3_13 of PanOCT [62]. Since PanOCT does not place paralogs into its ortholog clusters, but does produce a paralogs.txt file that specifies which clusters are paralogs, an in-house PERL script, paralogs_matchtable.pl, was created to merge paralogous clusters. This approach was necessary because analysis of core and novel genes has historically been defined for clusters containing all paralogs [63–68]. The R script, compute_pangenome.R, from Park et al. [67] and paralog_matchtable.pl were used to construct the pan-genome, core and novel genes plot. We initially chose not to compute permutations in genome order for the reasons described in [69]. As a consequence of a lack of permutations, compute_pangenome.R was modified to load in a defined genome order of addition.

Orthologous protein content was compared and illustrated in a Venn diagram that was constructed using output from an in-house PERL script, create_meta_groupings.pl that uses output from PanOCT and a file that describes how genomes are to be grouped. Genomes were grouped by whether they are infectious (group I or group II), non-infectious (saprophytic), or an outgroup (Leptonema illini). Since there were multiple genomes per group (except for Leptonema), clusters were counted if there was a majority (50%), all-but-one, or all protein members from a particular group or groups. Clusters not matching these criteria were not counted.

Metabolic reconstructions

Four draft metabolic network reconstructions were created for representative species chosen from pathogenic, intermediate and saprophytic clades including the following: L. interrogans (serovar Copenhageni strain L1-130), L. licerasiae (serovar Varillal strain VAR010), L. biflexa (serovar Patoc strain Patoc 1 (Ames)) and L. kmetyi (serovar Malaysia strain Bejo-Iso9). L. kmetyi was included in these comparisons because in addition to having recently been reported to infect humans in the Caribbean islands [70, 71], initial genome examination suggested that L. kmetyi could belong to a transitional group between the group I and group II pathogens and distinct from the other group I pathogens, here represented by L. interrogans. The reconstructions were built using the ModelSeed framework [72].

The COBRApy toolbox [73] was used to perform Flux Balance Analysis (FBA) [74] simulations and constraint-based analyses using the gurobi linear programming solver [75]. The constraint-based model consists of an S matrix composed of distinct metabolites and reactions including exchange and biomass reactions (S3 Table). Each of the reactions has an upper and lower bound on the flux it can carry. Reversible reactions have an upper bound of 1,000 mmol gDW⁻¹h⁻¹ and a lower bound of −1,000 mmol gDW⁻¹ h⁻¹, making them practically unconstrained, while irreversible reactions have a lower bound of zero. By default, the biomass reaction was set as the objective to be maximized. The exchange reactions that allow for extracellular metabolites to pass in and out of the system were defined such that a positive flux indicates flow out. The GapFind MILP algorithm [76] encoded in the COBRApy Toolbox was performed for the models unable to grow in minimal medium (biomass objective function equal to zero), to find exchange reactions allowing for in vitro growth, indicating strain-specific auxotrophies.

An in silico minimal medium was constructed that supported growth for all of the Leptospira models, consisting of trace elements (magnesium, manganese zinc, sulfate, calcium, copper, phosphate, cobalt, chlorine, potassium, ferrous and ferric iron and ammonia), water, oxygen, heme, CO₂, the vitamins thiamine (B1), folate and menaquinone, glycerol and the fatty acids lauric acid, stearic acid, and decanoate and aminoethanol, meso-2,6, diaminopimelate, as well a variety of amino acids (glutamate, aspartate, tyrosine, phenylalanine, asparagine (S3 Table).

Genomic location and genetic organization of vitamin B12 biosynthesis -related gene clusters in Leptospira

PanOCT data were used to identify the cob I/III and cob II gene clusters in infectious and non-infectious Leptospira (when present), which encode proteins predicted to participate in B12 transport or synthesis. To determine the genomic locations of the btuB and cob II and I/III clusters, a custom Postgresql database was created using the annotated genomes of 20 species. Orthologs were identified with blastp (-F “m S” -s T) and conserved genomic neighborhood using the Prokaryotic Sequence homology Analysis Tool (PSAT) [77].

General genomic descriptions

The genomes of 17 Leptospira spp. isolates were newly determined for this study, representing 8 pathogenic, 4 intermediate and 5 saprophytic clades and used in pan-genomic comparative analyses along with the previously reported genomes of L. interrogans serovar Copenhageni strain Fiocruz L1-130 [30], L. biflexa serovar Patoc strain Patoc I (Paris) [33] and L. licerasiae serovar Varillal strain VAR010^T [32]. Thirteen of these isolates were sequenced to a genome finishing status [78, 79] of "Improved High-Quality Draft" (IHQD) and 5 to a status a “High-Quality Draft" (HQD) (Table 1). To achieve a genome finishing status of IHQD, manual finishing was conducted consisting of contig sequence extension, sequence gap closure, and PCR to link physical ends. On average, the genomes assembled into 36 contigs [range 4 (L. kmetyi) to 89 (L. vanthielii)], with an average genome size of 4.26 Mbp in length [range 3.89 Mbp (L. borgpetersenii to 4.71 Mbp (L. noguchii)] at an average of 59.7-fold sequence coverage. The average G+C% was 40.7% [range 35.5% (L. noguchii) to 45.6% (L. wolfii)]. These genomes were predicted to encode an average of 4,197 protein-coding sequences per genome [range 3,932 (L. terpstrae) to 4,582 (L. alexanderi)].

Phylogenetic analysis of the Leptospira Genus to determine evolutionary relatedness

Twenty genome sequences (17 new, 3 previously published) of isolates representing 20 of the 22 known Leptospira spp. (Table 1) were used to determine phylogenetic distances between species; two recently reported species (L. idonii [80] and L. mayottensis [81]) are not included here. Phylogenetic relationships among all Leptospira species were analyzed in five independent ways (Fig 1): A, a core set of 39 concatenated genes coding for housekeeping proteins (universal markers); B, a pan-genus set of 1350 proteins; C, 3) multilocus sequence typing (MLST)[51]; D, 16S rRNA (highly conserved); and E, secY (highly variable)[70]. Leptonema illini was used as the outgroup for all analyses. Each approach yielded different nodes and branches of the species, except for 16S rRNA sequences, for which deduced phylogeny did not discriminate between L. meyeri and L. yanagawae. Additionally, these five approaches revealed three clades correctly clustering members of the nine pathogenic (group I) and five intermediately pathogenic (group II) and six non-pathogenic (saprophytic) species. Only the trees based on secY, the universal markers and the pan-genome clearly separated the closely related pathogenic species L. interrogans, L. kirschneri and L. noguchii from the other 6 pathogenic species. As expected, phylogenetic positions shifted to some extent between the single locus-based analyses and became more consistent using the multi-locus approaches. The following pairs of species showed close relationships: the pathogenic species (except in the secY tree) L. interrogans and L. kirschneri, L alexanderi and L. weilii; the group II species, L. inadai and L. broomii; and the non-pathogenic (except 16S and secY trees) L. wolbachii and L. vanthielii.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Phylogenetic analysis of Leptospira species.

Consensus maximum-likelihood trees are depicted using multiple alignments of 16S rRNA (A), secY (B), MLST (C) and 39 concatenated protein data sets (D). The numbers along the branches denote percent occurrence of nodes among 100 bootstrap replicates. A pan-genome tree was generated based on the mean of the BLASTP score ratio of core 1135 proteins (E). The scale bar represents the number of nucleotide (A-C), amino acid (D & E) substitutions.

https://doi.org/10.1371/journal.pntd.0004403.g001

Furthermore, the genome relatedness between pairs of representative strains of each of the 20 species from fully or partially sequenced genomes confirmed the genetic relatedness among Leptospira species as established by DDH (Fig 1/ S2 Table).

Pan-genome analysis

The pan-genome is defined as a core set of genes shared by all isolates plus a variable set of genes shared by a subset of isolates, and strain-specific or novel genes. Based on these 20 representative genomes and raw PanOCT output, the size of the core- and pan-genome was determined to contain 1,764 and 17,477 genes, respectively; however, with paralogs collapsed, the size of the core- and pan-genome was determined to be 1,592 and 13,822 genes, respectively (S1A Fig). The number of species-specific or novel genes ranged from 233 to 892 (S1A Fig) for each new genome added. After the addition of the third genome (L. noguchii), the size of the core gene set plateaued, while the pan-genome continued to rise (S1A Fig).

To determine whether the Leptospira pan-genome is open or closed (as defined below, the number of new genes identified (i.e., unique or strain-specific genes) for each genome added was determined and fit to a power law function (n = κN^-α) as described previously [63]. Conceptually, a pan-genome is closed when sequencing the genomes of additional isolates fails to increase gene number (i.e., the entire gene repertoire has been discovered) [82]. The exponent (α) indicates whether the pan-genome is open (α ≤ 1) or closed (α > 1) [83]. Using this equation, the pan-genome of Leptospira was inferred to be open (α = 0.49 ± 0.02) (S1B Fig). From an exponential decay function, the number of new genes predicted for each genome (species) added was extrapolated and calculated to be 409 ± 12 on average (S1B Fig).

The distribution of protein clusters representing gene families among the three groups (pathogenic, intermediate, saprophytic) is depicted in a Venn diagram (Fig 2A). Because there were multiple genomes per group (except for the Leptonema outgroup), clusters were counted if there was a majority (50%), all-but-one, or all protein cluster members from a particular group or groups. Focusing on the majority criteria, pathogens and intermediates had nearly equal numbers of group-specific genes (416 and 424, respectively), and the highest number of shared genes between two groups (369). Binary comparisons of pathogens and intermediates with saprophytes revealed just 52, and 78 genes shared, respectively. When comparing only Leptospira-specific genes, the core genome was comprised of 737 genes, with the majority of genes being shared with Leptonema illini. Closer examination of species-specific genes showed that pathogenic Leptospira have more species-specific genes on average (637±129) than do intermediates (418±126) or saprophytes (321±90). L. noguchii sv. Panama str. CZ 214T had the greatest number of species-specific genes among species compared in this study. To understand the function of genes shared among infectious Leptospira, the distribution of protein functions was examined for clusters shared among infectious and non-infectious Leptospira (Fig 2B). The only functional category dominated by pathogenic Leptospira was “mobile and extrachromosomal elements.” The functional categories that stood out most among genes shared between pathogens and intermediates was "biosynthesis of cofactors, prosthetic groups, and carriers" and "fatty acid and phospholipid metabolism." Saprophyte-specific genes dominated 10 of the 16 functional role categories, many of which were involved in central intermediary and energy metabolism, gene regulation, signal transduction, protein fate, cell envelope, and transport functions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Pan-genomic comparisons of 20 Leptospira species.

Panel A: Orthologous protein clusters were binned, counted and placed into a Venn diagram by whether clusters contained proteins from genomes in each of the three Leptospira groups: pathogenic (A), intermediate (B), saprophytic (C) and the Leptonema outgroup (D). Clusters were counted if there was a majority (50%), all-but-one, or all protein members from a particular group or groups (separated by colons). Singleton clusters, representing species-specific or strain-specific genes are noted in circles surrounding the Venn diagram. Clusters not matching any of these criteria or containing at least one protein from another group were considered as ambiguous groupings. The Venn diagram is not to scale. Panel B: Protein clusters unique to pathogenic, intermediate, and saprophytic groups or shared only between pathogenic and intermediate groups were counted by main functional role categories. See key for group colors.

https://doi.org/10.1371/journal.pntd.0004403.g002

Protein secretion systems

All Leptospira clades are predicted to have type II protein secretion systems but do not appear to contain a type III secretion system. Lipoproteins in Leptospira have particular pathogenetic significance because of their potential as vaccine targets and virulence factors involved in host-pathogen interactions [84–93].

An unusual sec system in Leptospira

Leptospira contains the Sec system for signal-peptide-containing proteins and signal peptidase to remove the signal peptide at the time of secretion. Genes encoding the signal recognition particle (SRP) protein Ffh and receptor protein FtsY were not found in any Leptospira genome, nor was the SRP structural RNA. Generally, the lack of SRP and its receptor is unusual in bacteria, although the system is missing in the genus Dehalococcoides and also, apparently, in the uncultivated marine lineage SAR86 [94]. The narrow, elongated spiral shape limits the distance a ribosome can be from the Leptospira plasma membrane and may obviate the need for translation arrest by SRP. However, in looking for novel features near Sec system genes in Leptospira showed a novel gene inserted between the normally consecutive genes for Sec system proteins SecY and YajC, encoding a non-globular protein with an N-terminal signal peptide and a transmembrane segment towards the C-terminus, with the majority of residues in between consisting of low complexity, poorly conserved sequence especially rich in Lys, Glu, and Asn. No homologs to this low-complexity protein occur outside the Leptospira genus. We postulate that this novel gene could be involved in protein secretion.

Unusual sec-independent (twin-arginine) translocation system

Twin-arginine translocation (TAT) in prokaryotes allows completion of complete protein folding prior to Sec-independent secretion through the plasma membrane. Except in the halophiles, where high salt outside the cell explains the need for folding prior to export, TAT substrates tend to be redox cofactor-binding proteins [95]. These proteins fold and bind their cofactors before crossing the membrane. The tatA and tatC gene-encoded components of the translocase are evident in Leptospira genomes, but the twin-arginine signal itself proved elusive. The TIGRFAMs collection [96] hidden Markov model (HMM) TIGR01409 finds no sequence scoring near the trusted cutoff in any species of Leptospira, nor in Leptonema illini. However, alignments of full-length homologs in Leptospira to recognizable TAT translocation substrates from other lineages could be extended into the N-terminal signal region. Such alignments often showed a Lys-Arg dipeptide in the Leptospira sequence aligned to the Arg-Arg motif of recognizable TAT signal sequences. This observation triggered a review of all candidate families of TAT translocation substrates in Leptospira, and produced iterative refinement of the lineage-specific TAT signal, and a catalog of TAT substrates.

Eleven protein families were confirmed as TAT substrates by multiple criteria, including strong conservation of the putative TAT signal within the protein family, alignment to non-spirochete homologs that extended N-terminally into the TAT signal region, and strong sequence similarity of the putative TAT signal motif, usually RKxFL, across the different Leptospira putative TAT translocation substrate families. A continuous 18-residue stretch from each protein in each of these was used to construct a seed alignment of Leptospira TAT signal sequences, including the modified Twin-Arg motif and the adjacent hydrophobic region. Comparative analysis predicted a conserved TAT signal sequence in all Leptospira [97] (Fig 3). Eleven protein families comprising the defining alignment and for two additional families are strong candidates for TAT-dependent, Sec-independent translocation (S4 Table). Only one of the 13 families, the PhoX alkaline phosphatase family, was observed to be largely restricted to pathogenic species of Leptospira [98].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. TAT signal sequence in Leptospira sp.

The X-axis shows position in an ungapped alignment. The Y-axis shows information content, measured in bits.

https://doi.org/10.1371/journal.pntd.0004403.g003

Inspecting the family of LIC_10874 (a 4Fe-4S dicluster domain protein family of LIC_10874) within and outside the genus Leptospira demonstrated conservation of the putative TAT signal in both, and the substitution in Leptospira of the second Arg by Lys, as in other families. This family is notable, however, because in multiple species from phylogenetically distant clades, translation start sites can be assigned with high confidence, and the TAT signal begins rather far (some 50 residues) from that start. Member sequences in this family all share a well-conserved prefix domain, ~50 amino acids in length, between the start of translation, and star of the recognizable TAT signal.

An unexpected feature in the Leptospira TAT system cassette is a probable serine phosphatase encoded next to tatC (family TIGR04400), which either overlaps it or is present within five base pairs in all 20 leptospiral species examined. It is not known whether this putative phosphatase is involved in Sec-independent translocation per se, rather than in its regulation, or in some unrelated process.

Lipoprotein secretion

Cleavage of pro-lipoproteins by the type II signal peptidase occurs within a short lipobox sequence, which includes the invariant cysteine that is targeted for covalent modification with lipids. In E. coli lipoproteins, the -1 position immediately preceding the peptidase cleavage site is highly conserved, being occupied by the small nonpolar amino acids Ala or Gly in the vast majority of cases [99]. A previous analysis noted somewhat larger residues, such as Asn, Ser, and Cys, at the -1 position of experimentally-verified spirochete lipoproteins [100]. Examination of L. interrogans lipoprotein orthologs in saprophytic and intermediate species revealed a number of unexpected amino acids at the -1 position (S5 Table). For example, the bulky amino acid Tyr was found in the -1 position in LipL21 in all intermediate Leptospira spp., whereas those from pathogenic and saprophytic species possess typical -1 residues Ala and Ser, respectively. Conversely, some lipoproteins of saprophytic or intermediate species with expected amino acids at the -1 position were found to have orthologs in L. interrogans with variant amino acids at the -1 position (S5 Table). For example, the outer membrane lipoprotein Loa22 of saprophytic species has the allowed residue Asn at -1 while its orthologs in pathogenic and intermediate species has Leu or Phe at the -1 position. Similarly, LIC11088 orthologs in most intermediate and saprophytic species possess permitted residues at -1, whereas the pathogens have Gln or charged residues. Thus, the availability of genome sequences from across the genus Leptospira has confirmed the much higher flexibility in the leptospiral lipobox and is anticipated to lead to redefinition of the pan-leptospiral lipobox to accommodate increased amino acid flexibility at the -1 position.

The substrate specificities of the first two enzymes in lipoprotein biogenesis, prolipoprotein diacylglyceryl transferase (Lgt) and signal peptidase II (Lsp), are likely to be influenced by amino acids at the -1 position relative to the lipoprotein cleavage site [101]. For this reason, these enzymes would be expected to possess novel structural properties that allow recognition of an expanded set of residues at the -1 position of the lipobox. Consistent with this notion, Lgt orthologs of all 20 Leptospira strains lack the signature sequence that defines most Lgt proteins (Prosite accession PS01311) [102]. Interestingly, Leptonema illini, the bacterium most closely related to Leptospira among sequenced organisms, harbors two Lgt paralogs: one quite close in sequence to leptospiral Lgt orthologs and another with signature sequence similar to Lgt of all other organisms. This arrangement suggests duplication of lgt in an ancestor common to Leptospira and Leptonema, with subsequent loss of the latter and functional divergence of the former to accommodate bulkier -1 lipobox residues. Similarly, Lsp of Leptospira species possesses an extra 22 or 24 residues at a position corresponding to a location within the second periplasmic loop of E. coli Lsp [103], which is missing from the Lsp sequence of other bacteria with well-characterized lipoproteins, including the spirochetes B. burgdorferi and T. pallidum. The sequence features of leptospiral Lgt and Lsp suggest the presence of novel structural features at the active sites of these enzymes consistent with variability at the -1 position of the leptospiral lipobox.

Metabolic reconstructions

In silico, genomically-based metabolic network reconstructions were created for four representative Leptospira species occupying different clades: L. interrogans and L. kmetyi (pathogen; group I), L. licerasiae (intermediate pathogen; group II) and L. biflexa (non-pathogenic). L. kmetyi was chosen for analysis because preliminary genomic inspection suggested unusual features of this species with regard to pathogenesis-related genes (vide infra). The base in silico media and default computational bounds (S3 Table) represent every compound allowed to enter the system for cellular uptake to allow all models to produce biomass. Removal of some of these compounds leads to species-specific growth (a unique model-predicted auxotrophy). The use of a steady-state assumption does not allow the flux balance analysis to take into account specific concentrations of a given metabolite but, rather, the predicted rate of uptake, secretion or transformation. The default uptake bounds for each metabolite are provided (S3 Table). Negative bounds represent entry of the metabolite into the extra-cellular compartment where they can then be consumed by the model. The bounds are only constraints on the maximum rate of consumption of a given compound. The actual rate of consumption is predicted by the model. Units are in mmol gDW-1 hr-1. Analogously to core and pan genomes, the reaction content of each model can be used to construct core and pan metabolic networks. The core network consists of those reactions occurring in all representatives of Leptospira, while the pan network consists of all reactions that can potentially occur in any individual Leptospira species (Fig 4A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Core and pan-metabolic capabilities of the Leptospira genus.

The core and pan metabolic content was determined for genome-scale metabolic models of 4 different Leptospira species. A) Core content, illustrated by the intersection of the Venn diagram, shared with all species. The pan content consists of all content in any model and includes the core content. The Venn diagram is not to scale. B) Classification of reactions in the core and pan reactomes by metabolic subsystem.

https://doi.org/10.1371/journal.pntd.0004403.g004

Major differences in the metabolic networks between the four species arose in amino acid metabolism, biosynthesis of cofactors and vitamins and carbohydrate metabolism (Fig 4B). These large groups were further divided into specific pathways, specifically those of porphyrin metabolism, folate metabolism, starch and sucrose metabolism, as well as phenylalanine and tyrosine metabolism. A large difference was observed in porphyrin metabolism, specifically for the biosynthesis of cobalamin (vitamin B12) in L. interrogans (see below).

The conversion of static metabolic network reconstructions into computable mathematical models allows computation of phenotypes based on the content of each reconstruction [104]. Thus, the four strain-specific reconstructed networks were converted into genome-scale metabolic models that allow for the computational/simulation prediction of phenotype. This set of genomic scale models (GEMs) allows for a meaningful interpretation of the content of each reconstruction and allows for the prediction of each strain’s different metabolic capabilities [105]. Because reactions belonging to the amino acid metabolism subsystem made up the majority of reactions in the pan-reactome, it was hypothesized that these capabilities may reflect functional differences between different Leptospira species. To test this hypothesis, different minimal media formulations were created in silico and used to test each model’s growth capabilities. The models predicted all of the Leptospira tested to be auxotrophic for the amino acids aspartate, histidine and asparagine as well as vitamins B₁ (thiamin) and vitamin K2 (menaquinone).

Beyond the auxotrophies predicted to be shared by all 4 Leptospira models, potential species-specific auxotrophies for other vitamins and amino acids were also identified. All of the strain-specific models were predicted to be auxotrophic for phenylalanine except for L. interrogans, which was predicted to have the enzyme prephenate dehydrogenase encoded for by novF that converts chorismate to prephanate, a precursor to tyrosine and phenylalanine. The model for L. interrogans lacks prephanate oxidoreductase, which would predict inability to convert prephanate to tyrosine. Only L. kmetyi was found to have the enzymatic machinery capable of synthesizing tyrosine from phenylalanine. Among these representative species of the three clades, only L. interrogans was predicted to be an L-glutamate auxotroph due to the lack of L-glutamate oxidoreductase.

Additional major differences were observed between the pathogenic and the non-pathogen Leptospira. A major difference in the lysine biosynthesis pathway was observed for the models of the pathogens, L. kmetyi and L. interrogans. Only these models possessed the dapABCDE genes required to convert L-aspartate 4-semialdehyde to LL-2,6, diaminopimelate required for peptidoglycan and lysine biosynthesis. Therefore both L. licerasiae and L. biflexa were predicted to be LL-2,6, diaminopimelate auxotrophs. Furthermore, only the pathogens L. interrogans and L. kmetyi possessed a full folate (vitamin B9) biosynthesis pathway using as precursor guanosine 5’-triphosphate. L. biflexa and L. licerasiae could produce vitamin B₂ (riboflavin), but lack the reactions to convert it to folate including dihydroneopterin aldolase encoded for by folB; therefore, the models for L. biflexa and L. licerasiae were folate auxotrophs while the models for L. kmetyi and L. interrogans were not.

Vitamin B12 biosynthesis

The vitamin B12 biosynthesis genes in infectious Leptospira are grouped into two clusters: cob I/III and cob II (Table 2). Though the exact number of reactions for each pathway in Leptospira remains unknown, in Salmonella enterica Typhimurium cob I comprises genes for the biosynthesis of adenosylcobinamide, cob II genes for the synthesis of the lower axial ligand 5,6-dimethylbenzimidazole (DMB) and a third cluster cob III, the nucleotide loop that joins DMB to the corrin ring to complete B12 biosynthesis. In infectious Leptospira, cob II is a five-gene cluster that includes three genes cobTSC that participate in the synthesis of DMB (and two genes that may or may not participate in B12 biosynthesis), suggesting that the first cluster encodes enzymes for the synthesis of adenosylcobinamide guanine diphosphate. The first 12 genes encode enzymes that participate in the synthesis of the corrin ring (cob I) whereas the last five, enzymes for the addition of the nucleotide loop (cob III). Intriguingly, cob III of the infectious species includes a gene cbiZ encoding an enzyme that participates in an alternative cobinamide salvage pathway first described in the archeon, strain Göl [106].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Proteins involved in biosynthesis of vitamin B12 in Leptospira.

https://doi.org/10.1371/journal.pntd.0004403.t002

Cob I/III gene clusters in the sequenced pathogenic Leptospira vary in length, from 16 in L. santarosai Shermani 1342K^T to 19 in L. alexanderi Manhoa 3L60^T and L. borgpetersenii Javanica UI 0993. Presumably owing to repeated in vitro sub-culture in medium containing B12, several genes have been inactivated or deleted in the strains tested. For example, the Javanica UI 0993 cobI/III cluster contains a gene fragment resulting from a premature stop codon in a gene encoding a histidine phosphatase superfamily branch 1 (hps_1) protein present in all other pathogenic Leptospira including other L. borgpetersenii strains (Hardjo L550 and JB197); and, cob I/III in L. alexanderi 3L 60T, contains a disrupted cobyrinic acid a,c-diamide synthase, inactivated by a frame shift mutation. A gene encoding a flavodoxin reductase present in cobI/III of other infectious has been deleted in L. santarosai 1342KT and L. alstoni 80–412. The cob I/III clusters of L. kmetyi (group I) and all group II pathogens contain three genes, two encoding a putative cobalt transporter (cbtBA) and a gene encoding an additional hps_1 protein (Table 2).

Cob II also varies in length among all infectious (pathogenic and intermediately pathogenic) Leptospira, from three genes in L. broomii Hurstbridge 5399 and L. inadai Lyme 10^T to seven in L. noguchii Panama CZ214^T. The genes comprising the cob II cluster in non-pathogenic Leptospira are found in two discrete clusters in non-pathogenic species (e.g., LEPBI_I2857 and LEPBI_I2858, LEPBI_I2938 –LEPBI_I2940) in L. biflexa (Table 2), suggesting that homologs in pathogenic Leptospira were acquired en bloc after the divergence of pathogenic and non-pathogenic Leptospira.

Leptospiral glycobiology: structure and diversity of rfb/O-antigen loci, lipid A, and sialic acid biosynthesis-encoding regions

General features of Leptospira rfb loci.

Lipopolysaccharide (LPS) has long been a major focus of leptospiral microbiology not because of its (low potency) endotoxigenic activities (see below in the lipid A section) and because, notably, leptospiral LPS is the basis for serovar identification and vaccine development [36, 107–109]. Because of the importance of LPS in leptospiral biology, we carried out a comprehensive analysis of the genomic locations, structures and neighborhood of leptospiral rfb loci, also known as the O-antigen loci, in 20 species of Leptospira.

Using previously described leptospiral O-antigen gene clusters as a guide [110–112], we identified and schematized all clades of Leptospira rfb loci (Fig 5; S2 Fig depicts the rfb locus in genomic context). Of the genomes representing 20 Leptospira species, 17 known serovars were compared. The O-antigen biosynthesis gene clusters were located in three different genomic locations and ranged in size from 3,768 bp (L. wolffii sv. Khorat) to 121,402 bp (L. alexanderi sv. Manhao3). This region in L. wolffii sv. Khorat is now the smallest predicted leptospiral rfb biosynthesis cluster, consisting of just 4 genes, replacing the locus of L. licerasiae sv. Varillal [32]. All pathogenic leptospiral species, and the intermediates L. inadai, L. broomii and L. fainei have their rfb loci located in the same genomic position, sandwiched between a copper-binding protein on the left and the ribosomal protein S6 on the right (S2 Fig). The same protein-encoding genes (viz., MarR and DASS) define the start and end of the O-antigen cluster, respectively. Of the rfb loci of pathogenic Leptospira, serovars Manhao 3, Javanica, and Pingchang were most similar in size and gene content (Fig 5). Notably, L. broomii [113] (and L. fainei [114–116], both of which are serovar Hurstbridge, had nearly identical rfb gene clusters, predicting that L. broomii would also be serovar Hurstbridge, and confirmed by serology [113]. The presence of a specific serovar in different species has been previously observed in isolates of both L. interrogans and L. borgpetersenii serovar Hardjo, which have highly similar gene content in their O-antigen biosynthetic loci [117]. The rfb gene cluster of saprophytic Leptospira is downstream of the gene encoding ribosomal protein S6, lacks DASS, and is smaller (median 60,710 bp vs. ~99,520 bp) than pathogenic Leptospira rfb gene clusters (S2 Fig). O-antigen gene loci in serovars Varillal and Khorat are located in a third location, between murC and purK, consistent with a novel branching in the phylogenetic tree.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Structure of Leptospira rfb locus gene clusters.

The rfb region and beginning and ending CDSs (blue) 9 of pathogenic (A), 5 intermediate (B), and 6 saprophytic (C) representative Leptospira species were compared. rfb region CDSs are labeled by locus identifier and colored by functional role categories as noted in the boxed key. Gene symbols, when present, are noted above their respective genes. BLASTP matches between CDSs are colored by protein percent identity (see key).

https://doi.org/10.1371/journal.pntd.0004403.g005

Consistently, the downstream flanking genes in the rfb loci are far more conserved than the upstream genes (Fig 5; S2 Fig). This finding is especially true for the pathogenic serovars and between three of the five intermediate serovars represented. For pathogenic serovars and serovar Lyme, a conserved block of genes is involved in O polysaccharide processing via the Wzy-dependent pathway. This export system was also identified in two saprophytic serovars (e.g., Holland and Codice). Overall, 12 of the representative 20 species genomes encoded the Wzy-dependent system and one genome (L. licerasiae sv. Varillal only encoded a putative flippase (Wzx) with no identifiable Wzy ortholog. There were no orthologs of Wzz, the O-antigen chain length determinant, in the 20 genomes studied. Also conserved in the 3-prime region of only those serovars with the Wzy-dependent pathway is a gene encoding a protein with homology to E. coli WcaJ/S. enterica WbaP, which are members of the PHPT family of polyisoprenyl-phosphate hexose-1-phosphate transferases that function to transfer glycosyl-1-phosphate to a lipid undecaprenol carrier, initiating formation of the O-unit in O-antigen assembly. In L. borgpeterseni serovar Hardjobovis, this protein, encoded by orfH13, is an UND-pp-galactosyltransferase [118].

The other major pathway of O-antigen polysaccharide biosynthesis is the Wzm/Wzt–encoded or ABC-transporter dependent pathway [119, 120]. Six of the 20 representative genomes encoded orthologs of Wzm and Wzt. Two of the genomes were from intermediate and four were from saprophytic groups, representing 5 known serovars (e.g. Hurstbridge, Saopaulo, Hualin, Hardjo type Went, and Patoc). The genome analysis did not provide a clear indication of the export system used by serovar Khorat. One other known O-antigen biosynthesis pathway the synthase-dependent pathway [119]. BLASTP searches of WbbE and WbbF from the only known example of this pathway, from the plasmid-encoded O:54 antigen of S. enterica serovar Borreze [121], failed to identify any homologs in the representative 20 Leptospira genomes. It is possible that serovar Khorat uses a novel mechanism for O -antigen biosynthesis.

A dTDP-rhamnose biosynthesis gene cluster, encoding rfbABCD was found in the conserved 3-prime end of the predicted O-antigen biosynthetic gene clusters of only pathogenic Leptospira spp. serovar Saopaulo, found in the saprophytic species L. yanagawae, encoded homologs of all four of these genes, but in the order rfbCBAD, where the genes rfbABC appear to have been inverted (Fig 5C). The genes rfbAB and rfbC, were found in a different location with rfbC separated by several genes in serovars Hualin, Hardjo, and Patoc. These same isolates lacked an rfbD homolog. Only rfbAB homologs were identified in serovar Hurstbridge.

Sialic acids as post-translational modifications restricted to pathogenic Leptospira.

Previous studies have demonstrated that pathogenic Leptospira endogenously synthesize Neu5Ac, the most common sialic acid, and that an observed gene fusion event suggested that L. interrogans uses a Neu5Ac biosynthetic pathway that is more similar to that of animals than to other bacteria. Lectin-based affinity purification of NulO-modified molecules, followed by mass spectrometric identification suggested post-translational modification of surface lipoproteins, including the putative virulence factor Loa22 [124, 125]. In the genomes analyzed for this study, 3 of the 9 pathogens had the complete cluster of genes involved in the production of sialic acids; 3 more lacked 1 gene in the cluster (Table 3). L. weilii contains only 2 genes from the cluster (spsE and rfbB3) and L. kirschneri and L. noguchii have only the spsE gene

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Leptospiral Proteins for Sialic Acid Biosynthesis.

https://doi.org/10.1371/journal.pntd.0004403.t003

All genomes, except L. licerasiae and L. wolffii have a N-acetylneuraminic (sialic) acid synthetase (spsE) gene (NP_711790.1). Phylogenetic analysis of this protein shows 2 distinct groups (S3 Fig). The first group contains the proteins from pathogens that contain the whole cluster. These proteins are related to the synthetases involved in the production of legionaminic acids. The second group contains the proteins from the intermediate species, the saprophytes and the pathogens L. kirschneri and L. noguchii. This group of synthetases is related to those producing pseudaminic acids.

The lack of a second sialic acid synthetase (NP_711794.1) in L. kirschneri and L. noguchii differentiates these pathogens from L. interrogans, which does contain this gene. These synthetases contain a phosphatase domain in addition to the NeuB domain, which suggests an animal-like Neu5Ac biosynthetic pathway. The pathogen L. weilli lacks NP_711794.1 but saprophyte L. vanthielli contains a similar synthetase but one that is missing the N-terminal transferase domain present in leptospiral pathogens. Finally, a UDP-N-acetylglucosamine diphosphorylase (NP_714003.1) was found in all leptospiral genomes studied. This gene is not located within the sialic acid gene cluster, and is also annotated as a MobA-like NTP transferase domain, therefore its role in sialic acid biosynthesis is unclear.

The sialic acid biosynthetic genes in leptospiral pathogens have some notable characteristics. L. alexanderi lacks O-acyltransferase (neuD) and this species and L. borgpetersenii have a truncated version of a nucleoside-diphosphate-sugar epimerase (NP_711787.1) (S3 Fig; Table 3). Only L. santarosai has a N-acetylneuraminic (sialic) acid synthetase with a phosphoglycerate dehydrogenase domain. Notably, none of the intermediate or saprophyte species contain the metabolic machinery to synthesize sialic acids, confirming previous suggestions [124].

Leptospiral mobile elements: phage and CRISPR-Cas systems

Phage.

Bacteriophages are abundant biological entities that have significant effects on bacterial evolution. Some estimates suggest that there are approximately ten-fold more phages than bacteria [126, 127]. However, our current knowledge of phages infecting Leptospira spp. is limited. Three distinct L. biflexa phages were isolated from sewage water in Paris. Morphological analysis by electron microscopy revealed that these three phages belong to the Myoviridae family and seem to be morphologically similar with polyhedral heads and contractile tails [128]. One of these phages, the 74-kb LE1 prophage, was shown to replicate as a double-stranded circular replicon in L. biflexa [129]. The genome of LE1 has a GC content of 36%, similar to that of Leptospira spp., and most of the 79 predicted ORFs display no similarity to known ORFs, but 21 ORFs appeared to be organized in clusters that might encode head and tail structural proteins and immunity repressor proteins [130].

Next generation sequencing and refinement of computational methods have allowed comparative genome analysis to discover new prophage and genomic islands [131]. A few phage related genomic islands have thus been characterized in L. interrogans and L. licerasiae [122, 132, 133] and it was previously shown that one of these genomic islands can excise from the L. interrogans chromosome [133].

To determine the distribution of prophages within the Leptospira genus, Phage_Finder [60] was run under both strict (-S) and non-strict modes to identify predicted prophage regions. Phage_Finder predicted a total of 14 major prophage regions across the 20 genomes, most of which were found to be shared between the Leptospira species (Table 4). Among the prophage sequences, the LE1-like prophage is found in many genomes, suggesting that double-stranded DNA tailed phages, which are the most frequently observed phages in bacteria [134], are common phages infecting Leptospira. The presence of numerous phage-associated sequences in the genome of pathogens and intermediates, in comparison to the saprophytes, suggests that phages have played an important part in the evolution of these lineages, as has been experimentally shown in L. biflexa [135].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Predicted Leptospira Prophage Regions.

https://doi.org/10.1371/journal.pntd.0004403.t004

Further experimental studies of Leptospira phages, which would include both electron microscopic visualization and production of phage in vitro, will be important to determine whether recombinant Leptospira phage might be useful for genetic manipulation studies of different Leptospira species, particularly pathogens and intermediates.

CRISPR/Cas systems.

Three described types of CRISPR/Cas systems are common in Leptospira genomes and only found in infectious members of the genus (Table 5): the E. coli (type I-E), DVULG (type I-C), and MYXAN systems [96, 136, 137]. A single sequenced genome, L. inadai serovar Lyme str. 10, has the recently described PreFran type, which has been found in Prevotella and Francisella [96]. Four of the 20 representative strains have components of two CRISPR/Cas types, suggesting that for some isolates there is redundancy in CRISPR/Cas machinery. Surprisingly, half of the 20 representative Leptospira strains contained predicted CRISPR repeats, which were more common in pathogens and intermediates than in saprophytes, which lacked CRISPR systems. In none of the six saprophytes examined—L. biflexa, L. meyeri, L. wolbachii, L. vanthielii, L. yanagawae, and L. terpstrae—were CRISPR/Cas systems detected, suggesting that these species rely on some other mechanism for escaping phage/plasmid attack. In these saprophytes, we were also unable to detect sequences encoding prophage, while CRISPR systems and prophage occurred together in several, but not all, representative pathogenic and intermediate strains.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. CRISPR/Cas Systems Identified in Leptospira Species Representatives.

https://doi.org/10.1371/journal.pntd.0004403.t005

When present, between one and six CRISPR repeat arrays were detected, containing between three and 25 spacer sequences (Table 5). Since CRISPR spacer sequences in other organisms are known to target phage sequences for destruction, we wondered if any of the 239 predicted spacer sequences targeted any of the known Leptospira spp. phage or predicted prophages. A database containing the nucleotide sequences of the 19 predicted prophages from this study plus the LE1 phage and the prophage from Qin et al. [32, 132] was constructed and used to search all 239 predicted spacer sequences using BLAST+ 2.2.30 [138]. Upon filtering the data for matches spanning the entire spacer sequence, with 3 or fewer mismatches and with a bitscore of 30+ revealed three spacer sequences matching two predicted prophage sequences (Fig 6). Two different L. noguchii spacers matched the same predicted L. santarosai prophage (Table 5). One of the same L. noguchii spacers also matched a predicted L. weilii prophage and was also recognized by an L. weilii spacer (Fig 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. CRISPR Spacer Sequences that Recognize Leptospira Predicted Prophages.

The CRISPR sequences are shown, which correspond to specific prophage accession numbers as listed in Table 4.

https://doi.org/10.1371/journal.pntd.0004403.g006

Virulence and Survival Mechanisms

Resistance to oxidative stress.

Three enzyme systems have conventionally been associated with the ability of pathogenic bacteria to defend against host-derived oxidative stress-related mediators such as hydrogen peroxide and superoxide radicals: catalases, peroxidases and superoxide dismutase. While catalases generate water and oxygen from H₂0₂, peroxidases generate water and an oxygen radical.

KatA and another but uncharacterized predicted catalase ortholog (LEP1GSC062_4039) were only found in pathogenic Leptospira, suggesting that these enzymes play an important role for Leptospira living within the mammalian host (Table 6). Conversely, superoxide dismutase was not found in pathogenic Leptospira but only in saprophytic Leptospira, suggesting either that pathogenic Leptospira are not exposed to oxygen radicals in the environment, or, more likely, that this clade of Leptospira has developed alternative ways to detoxify oxygen radicals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Proteins involved in the resistance to oxidative stress.

https://doi.org/10.1371/journal.pntd.0004403.t006

PF07598 paralogous gene family.

Previous work identified a group I-specific family of proteins corresponding to Pfam model PF07598 [149] that was expressed in vivo in a hamster model of acute leptospirosis [149], and expanded in strains, e.g., Copenhageni and Lai, that commonly cause severe disease, suggesting that these proteins contribute to Leptospira virulence. These prior studies focused on finding PF07598 orthologs to the L. interogans Lai attenuated strain in the 20 representative Leptospira spp., but did not look for strain-specific homologs that match PF07598 HMM. To identify novel PF07598 family members, we identified clusters of protein orthologs from our pan-genome run that matched PF07598 above trusted cut-offs; this analysis also included matches within the genomes of previously sequenced strains L. borgpetersenii Hardjo L550 and JB197 (and previously annotated as conserved hypothetical proteins). At least 26 distinct orthologs ranging in length from 47 (LEP1GSC049_1303 unique to L. kirschneri) amino acids to 651 (LEP1GSC193_2756; L. alstoni 80–412) were identified (Table 7; only homologs longer than 200 amino acids shown). As previously reported, L. santarosai 1342KT contains two distinct homologs, while L. kirschneri 3522 CT at least 15 and L. noguchii CZ214 at least 14 including 5 (LEP1GSC059_0232, LEP1GSC059_3018, LEP1GSC059_3019, LEP1GSC059_3599 and LEP1GSC059_3600) without an apparent ortholog in any of the other strains tested (Table 7). L. borgpetersenii Javanica (4 total) and the previously sequenced genomes of two Hardjo strains (3) contain two orthologs in common, while Javanica has two distinct copies (LEP1GSC103_4030 and LEP1GSC103_0672) not present in Hardjo and both Hardjo strains share an ortholog (LBJ_1339 and LBL_1564) not present in Javanica. In addition, L. interrogans L1-130 contains an ortholog (LIC_10639) shared with L. noguchii CZ214T and L. kirschneri 3522 CT not present in L. interrogans 56601 consistent with the hypothesis that serovar Lai has lost this ortholog.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Distribution of the PF07598 paralogous gene family in Leptospira.

https://doi.org/10.1371/journal.pntd.0004403.t007

To better understand the evolution of this paralogous gene family, a phylogenetic tree of all PF07598 members detected in infectious Leptospira was constructed using homologs longer than 200 amino acids shown (Fig 7A). A complex web of lineage specific gene duplications and loss was revealed. For example, as highlighted in (Fig 7A), successive gene duplications and subsequent gene loss have led to four distinct clusters containing proteins in L. interrogans, L. kirschneri and L. noguchii PF07598 family members. The initial duplication event led to divergence of the first orthologous group comprising LEP1GSC059_0224 (L. noguchii) and LIC_12985 and LA0591 (L. interrogans) and apparent loss of the corresponding ortholog in L. kirschneri (*). The second, an L. interrogans-specific event (**), led to the divergence of two orthologous groups comprised of LA_0589, LIC_12986 and LEP1GSC049_3370, and LA_3388, LIC_10778 and LEP1GSC049_0186, respectively. L. interrogans Lai 56601 has seemingly lost an ortholog belonging to a group containing LIC10639 and LEP1GSC049_1381 comprised of proteins belonging to L. interrogans Copenhageni and L. kirschneri, respectively; and L. kirschneri 3522CT, L. alexanderi and both L. borgpetersenii Hardjo strains seem to have lost an ortholog present in the other pathogenic strains (***). This pattern of species and serovar specific gene duplication and gene deletion occurs throughout the tree. K-means clustering with Kendall rank correlation grouped the L. interrogans Lai orthologs into three clusters comprising family members with >90% inclusion probability: LA_1400 and LA_1402 (cluster A); LA_0589, LA_0591, LA_0835 and LA_3388 (cluster B); and LA_0620, LA_0769, LA_0934, LA_2628 LA_3271 and LA_3490 (cluster C) (Fig 7B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Phylogenetic Relationship of PF07598 Paralogous Family in Leptospira.

(A) Unrooted bootstrapped phylogenetic tree; (*) Gene duplication event; (**) gene duplication event; (***) gene deletion. (B) Principal components analysis was used to arrange PF07598 family members. Color legend indicates the PF07598 family members from specific serovars depicted as diamonds. Arrowheads indicate L. noguchii-specific orthologs. Only PF07598 family members longer than 200 amino acids are included in the analysis. Clusters (A, B and C) were defined by K-means clustering with Kendall rank correlation.

https://doi.org/10.1371/journal.pntd.0004403.g007

Gene regulation and sensory transduction

Two Component Systems (TCS).

TCSs are the predominant molecular switches controlling signaling events in bacteria. Typically, TCSs consist of a sensor histidine kinase (HK) and an effector response regulator (RR). A single polypeptide merging both components results in hybrid histidine kinases (HHKs). HKs and RRs are usually found adjacent to each other in the genome. Orphan TCS proteins are unpaired HKs/RRs, which work with their cognate partners that are far apart in the genome. In addition, multistep phosphorelays may include intermediate histidine phosphotransferase proteins (Hpt), adding further complexity to TCS networks. Pathogenic, intermediate and saprophytic Leptospira species encode an unusually large and diverse set of TCSs, including orphan HKs and RRs, HHKs and Hpts in addition to classical paired HK/RR systems (Table 8 and S13 Table). Of note, more than 60% of the TCS genes found in Leptospira genomes encode non-classical orphan HK, orphan RR, HHK and Hpt proteins (Table 8). Overall, pathogenic species had the lowest average number of TCS genes (76), while saprophytic ones had the largest (102) (Table 8). Genome size-normalized TCS data revealed that pathogenic Leptospira species have roughly 35% less TCS genes in comparison to intermediate and saprophytic species (S6 Fig and Table 8). Additionally, pathogenic species had a proportionally lower number of strain-specific TCS genes compared to intermediate and saprophytic species (S7 Fig and S14 Table). We also identified a core set of 16 TCS genes shared among all the Leptospira genomes being analyzed (S7 Fig). Half of these TCS genes, conserved among all Leptospira species, were orphan HK/RRs (S15 Table). Taking into account their high conservation throughout the species, irrespective of saprophytic or pathogenic mode of lives, this core set of TCSs probably regulates pivotal cellular pathways in Leptospira.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Summary of two component systems identified in 20 Leptospira genomes.

https://doi.org/10.1371/journal.pntd.0004403.t008

Although there were 15 TCS genes conserved among pathogen and intermediate species, we did not identify genes that were shared between saprophyte:pathogen or saprophyte: intermediate species (S7 Fig). This finding is in agreement with the previous observation that intermediate Leptospira spp. are more closely related to pathogens rather than to saprophytes, and that the gene order is more conserved in pathogenic and intermediate species [32]. Among the 15 TCS genes found in both pathogenic and intermediate species, 5 were orphan RRs, 4 were orphan HKs, 2 were HHKs, and only 1 was a classical HK:RR pair (S13 Table).

It is worth noting that L. kmetyi and L. alstoni were outliers with respect to being pathogenic species that harbored the largest number of species-specific TCS genes (27 and 17, respectively) and the largest overall number of TCSs within this species group (S7 Fig and S14 Table). In contrast, L. alexanderi did not contain any species-specific TCS genes (S7 Fig and S14 Table). The seven TCS genes present in all pathogenic species (S16 Table) may represent moieties involved in common signaling pathways and may play a role in host colonization and pathogenetic mechanisms.

Lipopolysaccharide rfb biosynthetic loci in 20 Leptospira species

A remarkable feature of Leptospira—one that has often dominated the study of Leptospira over the past century—is the serologically-determined variety of Leptospira serovars. Serovar identification based on serological agglutination tests previously was the basis of Leptospira taxonomy, and continues to be important for understanding leptospirosis epidemiology related to human and animal infection and mammalian reservoir hosts. The present comparative genome analysis of 20 Leptospira species, as well as hundreds of additional Leptospira whole genome sequences not yet completely analyzed, demonstrates genetic determinants of the antigenic carbohydrates (putatively O-antigens). Therefore molecular analysis of genes in the rfb O-antigen locus rather than serological tests will be the future basis of serovar identification.

Leptospiral LPS has long been known to be non-endotoxigenic, the basis for which is thought to be due to unusual modifications of the core Lipid A component of LPS resulting in altered Toll-like receptor-mediated innate immune responses [161–164]. Leptospiral serovar, determined by LPS antigenicity, seems to have some association with mammalian host predilection—for example, serovar Copenhageni with Rattus spp., serovar Canicola with dogs [1, 2, 165]—but a causal, mechanistic role here remains to be experimentally demonstrated. Indirect evidence from other microbial systems suggests the speculative hypothesis that the diversity of leptospiral LPS may be driven by environmental ecology-mediated selection pressures [166, 167], such as has been reported in Salmonella spp. which has many (>2000) serotypes). The genus Leptospira has been reported to contain more than 300 serovars [1, 2, 165], the basis for which remains essentially unexplored.

Metabolic reconstructions

Here we report the first genomically-predicted metabolic network analysis [168, 169] of Leptospira, comparing members of the pathogen, intermediate pathogen and saprophyte clades. These large-scale reconstructions allow classification of the conserved metabolic capabilities (core metabolic network) and the unique metabolic capabilities (pan metabolic network). These reconstructions can be further converted to metabolic models of metabolism to probe metabolic capabilities computationally.

The most striking differences between infectious/pathogenic Leptospira and non-pathogens arose in porphyrin and vitamin biosynthetic capabilities. L. interrogans was shown to have a nearly complete vitamin B12 biosynthetic pathway that enables de novo B12 synthesis from an L-glutamate precursor, while L. biflexa completely lacked this pathway. Only pathogenic Leptospira—L. interrogans and L. kmetyi—were predicted to have a full folate biosynthetic pathway. These differences in biosynthetic capabilities may allow such pathogens to survive in nutrient-limited niches within the mammalian host. These observations are consistent with previous observations that found that L. interrogans serovar Canicola can grow in vitro in the absence of B12 but not B₁ [170], but in contrast with observations of others who concluded that pathogenic Leptospira could grow in the absence of B₁ but not B12 [171].

An open question in Leptospira biology is why L. interrogans grows more slowly than do intermediate pathogens and saprophytes, such as L. licerasiae and L. biflexa, which grow rapidly in defined EMJH medium [38]. The metabolic network model of L. interrogans was shown to lack L-glutamate oxidoreductase, an enzyme involved in recruiting ammonia as a nitrogen source [172, 173]), predicting a lower growth yield compared to the other Leptospira models in our in-silico minimal media analysis. The model of L. biflexa predicted the greatest yield with this reaction because this Leptospira contains the L-aspartate ammonia-lyase reaction allowing it to convert L-aspartate into fumarate and ammonia, in addition to using this component solely for biomass generation. These observations hint at one possible solution to the question of different growth rates, but model-guided experimentation is required to validate this prediction. Predictions made using these metabolic networks depend on an accurate reactome [174], and must be validated experimentally. If the models reported here are further curated and experimentally validated they would be the first such mode for a pathogenic spirochete. Such an approach will yield fundamental insights into diverse metabolic capabilities of this diverse phylum, including Borrelia spp., which cause Lyme borreliosis and relapsing fever, and Treponema spp., which cause syphilis, yaws, periodontitis and other diseases.

Vitamin autotrophy

Vitamin B12 (cobalamin) is the largest and most complex of natural organometallic cofactors and coenzymes, its de novo synthesis requiring ~30 energetically costly enzymatic steps [175]. Mammals have evolved highly complex, regulated mechanisms to absorb, transport and store cobalamin [176], reminiscent of the baroque processes involved in iron uptake, transport and storage in humans [177] [178]. We found that the genetic machinery for B12 autotrophy is found in infectious but not saprophytic Leptospira, leading us to speculate that such autotrophy allows Leptospira to infect mammals in the face of B12 sequestration by the host.

All Leptospira survive in the external environment, but our analysis predicts that only pathogenic strains make cobalamin de novo from L-glutamate, suggesting that this process is critical in vivo. Important in this context, cobalamin absorption and utilization in mammals is mediated by an elaborate set of carrier proteins, receptors and transporters [176] that generally are presumed necessary to process and protect this very large molecule. Considering the mechanistic details of cobalamin handling in mammals and the B12 autotrophy of infectious Leptospira, we hypothesize that mammalian B12 systems deprive invasive microbes of cobalamin, akin to the role of iron absorption, transport and sequestration known to sequester iron from pathogens, which have evolved siderophore mechanisms to acquire iron in vivo in mammalian hosts. Comparative analysis of de novo cobalamin biosynthesis in Leptospira predicts that infectious Leptospira are autotrophic for synthesizing this compound while saprophytes are auxotrophic, and suggests lines of experimentation to explore further the details cobalamin biosynthesis in Leptospira. The significance of the absence of complete cobalamin biosynthetic pathways in some group I Leptospira remains unclear.

Detoxification of reactive oxygen species suggests resistance to host defense and differences in ecological niche

Previous comparative biochemical studies of spirochetes demonstrated catalase activity only in pathogenic Leptospira (all of which were previously classified inclusively as L. interrogans), and superoxide dismutase activity only in the saprophytic L. biflexa; peroxidase activity was present in both clades [179]. An important finding in our comparative genome analysis was that the Leptospira catalases, KatA and the putative catalase ortholog (LEP1GSC062_4039), were only found in pathogenic Leptospira, while the single leptospiral superoxide dismutase gene sod was found only in saprophytic Leptospira. Catalase has classically been associated with resistance to phagocyte-produced oxidative burst-mediated killing of pathogens (viz. hydrogen peroxide), typically intracellularly after phagocytosis. The presence of catalases only in pathogenic Leptospira suggests the testable hypothesis that this enzyme class may be involved in intracellular resistance to intracellular host cell killing, following on the published observations of pathogenic Leptospira within phagolysosomes [180, 181]; whether pathogenic Leptospira survive and proliferate in this subcellular compartment has not been conclusively demonstrated. Conversely, the absence of sod in pathogenic Leptospira suggests that this clade occupies an environmental niche not exposed to oxidative radicals. Presumably the observation that sod is restricted to saprophytes suggests that this clade is exposed to a different context in which oxidative radicals are found in the environment.

Adhesion

Extracellular matrix (ECM)-binding proteins potentially mediate Leptospira adhesion to mammalian host cells, movement of Leptospira in and through tissue interstitium, and generally promoting invasion/colonization processes [37, 182–195]. A diverse array of ECM-binding proteins has been identified suggesting a redundancy of adhesion molecules that are probably part of invasion strategies of Leptospira. Indeed, many putative adhesins are multifunctional as they bind plasminogen and generate plasmin [185, 196, 197], increasing proteolytic processes associated with infection, or they could participate in immune evasion strategies by interacting with complement regulators (see below). Moreover, several of these proteins are recognized by human leptospirosis serum samples indicating their expression during infection. A caveat is that almost all these putative adhesins have been identified based on binding studies with recombinant proteins. Although site-directed mutagenesis of pathogenic Leptospira spp remains difficult, relatively straightforward methods have been developed for functional analysis of putative adhesins genes through gain-of-function studies in L. biflexa [92] [198][199]. Many putative adhesins are present in saprophytic Leptospira but the presence of DNA sequences does not mean that the proteins are expressed. Although non-pathogenic Leptospira species may encode the genes for putative adhesins, transcripts of ompL1 and the protein itself, for example, have not been detected in L. biflexa serovar Patoc (190). In any event, adhesion to environmental biotic or abiotic structures may be part of the biology of saprophytic Leptospira but such a concept remains speculative at this time [200].

Many adhesins are multipurpose proteins as they bind PLG and produce PLA. The generation of fully active PLA, aside from its other functions, also contributes to the enhanced degradation of complement components. Indeed, it has been shown that in Leptospira, PLA decreases C3b and human IgG deposition, most probably through their degradation, thereby hampering opsonization, restricting complement antibacterial functions [201]. Another mechanism of complement evasion is through the acquisition of host regulators of complement activation. Surface microbial proteins that bind to complement inhibitors and activate them permit pathogens to inhibit the complement response on the bacteria [202]. Binding of Leptospira to factor H (FH), factor H-like protein (FHL-1) and C4 binding protein (C4BP) has been reported [203, 204] and several complement regulators-binding proteins have been identified [205–211]. Lsa23 is an example of multifunctional protein capable of binding ECM, PLG/PLA and complement regulators and should play a role in leptospiral virulence [211].

Immune evasion via proteolysis of complement

The protein family includes several metalloprotease members that are considered virulence factors in several pathogens [212]. Aureolysin, a zinc-dependent metalloprotease of S. aureus, acts in synergy with host regulators to inactivate C3 thus potentially inactivating host immune response [213]. Indeed, in the case of Leptospira, thermolysins were only found in pathogenic strains. Recently, experimental evidence of the sequence LIC13322 encoding a putative thermolysin on the direct degradation of complement factors has been reported, suggesting its role in immune evasion by pathogenic leptospiral strains [214]. Thus, it seems that pathogenic Leptospira, like other successful pathogens, utilize at least two strategies to circumvent the complement system: acquisition of host complement inhibitors and degradation of complement components, either thorough PLG/PLA generation or by the presence of bacterial proteases.

Sialic acids

We show here a distinctive presence of the sialic acid cluster in most pathogenic Leptospira species and notably absent from intermediately pathogenic and saprophytic Leptospira species. This fact suggests a role for sialic acids as virulence determinants. The data shown are consistent with previously published data in which L. interrogans and L. alexanderi produced di-acetylated nonulosonic acids and L. santarosai did not produce identifiable nonulosonic acid species, and this observation could be associated with some species missing particular enzymes of the pathway [215].

The pathogens that contain the whole sialic acid biosynthesis cluster have N-acetylneuraminic acid synthetases predicted to produce legionaminic acids, while other species have a cluster predicted to produce pseudaminic acid. Pseudaminic acid has been shown to be required for flagella biogenesis in Campylobacter spp. and Helicobacter spp. and function as virulence factors [216]. Leptospiral flagella are located between the inner membrane (IM) and outer membrane (OM) and drive motility [217]. The presence of a pseudaminic acid pathway in all species suggests glycosylation could have a similar role in flagella biogenesis.

Legionaminic acid has been associated with virulence in Legionella pneumophila and Campylobacter coli where its cell surface location seems to be involved in adhesion, cell-cell interaction and immune evasion [218]. We hypothesize that similar mechanisms occur in pathogenic leptospires.

PF07598 paralogous gene family

A novel gene family—first identified in a pathogenomic screen of L. interrogans serovar Lai [219]—was found to have orthologs present in pathogenic Leptospira but not in intermediate or saprophytic Leptospira. In L. interrogars serovar Lai, these genes were reported to be upregulated in vivo [219] and here we report that the numbers of paralogs varies among the pathogens, with the L. interrogans, L. kirschneri and L. noguchi having the most. These observations suggest that the PF07598 genes contribute to leptospiral virulence but the mechanism(s) by which they do so remain to be elucidated; no functional annotations for the PF07598 gene family is yet possible. Experimental studies of this gene family will likely provide insight into leptospirosis pathogenesis.

Motility and chemotaxis

Although motility is essential for pathogenesis of Leptospira [150, 151], all Leptospira spirochetes, including those belonging to intermediate and saprophytic species, are motile. Consistent with this observation, we found that pathogenic, intermediate and saprophytic species of Leptospira have all the genes necessary to assemble a functional flagellar apparatus. Furthermore, the flagella-encoding genes are highly conserved within the genus, indicating that pathogenic and non-pathogenic Leptospira do not differ significantly with respect to their flagella apparatus and structure.

In contrast to what we found with regard to motility, there was high diversity with respect to predicted amino acid sequence identity among genes encoding chemotaxis proteins from different species of Leptospira. Furthermore, not all chemotaxis proteins are present in all the species, which corresponds to different chemotactic behaviors observed in pathogenic and saprophytic Leptospira [220]. The majority of the diversity among the chemotaxis proteins was observed in MCPs, trans-membrane sensor proteins that trigger the intracellular signal transduction in bacterial chemotaxis [221], and located at the cell poles near the basal body and flagellar motor as other chemotaxis proteins [217, 222]. This finding, together with the observation that there is a higher diversity among genes associated with the basal body and that asymmetrical periplasmic flagellar rotation occurs with the interaction of basal body and chemotaxis proteins, suggests that the sensing and chemotactic response regulated by this proteins may impact their survival in specific environments, including their ability to infect a mammalian host.

Gene regulation

Transcriptional regulation is the most common form of regulation in bacteria, often via a specific transcriptional regulator (activator or repressor) or an alternative sigma factor. In this study, we found that Leptospira have only three types of alternative sigma (σ) factors (σ⁵⁴, σ^F, σ^E), which is fewer than E. coli which have genome sizes similar to those of Leptospira. Our analysis revealed that there is a distinct difference in the σ⁵⁴ regulatory networks between pathogenic and saprophytic Leptospira species. All pathogenic species have two σ⁵⁴ regulatory networks, LepA-σ⁵⁴ and LepB-σ⁵⁴, while saprophytic Leptospira only have LepA-σ⁵⁴. Historically, σ⁵⁴ is known to be involved in nitrogen assimilation, and now is well recognized to regulate diverse functions in response to various stimuli [223]. In addition, σ⁵⁴ has also been shown to be essential for infection in some pathogenic bacteria. For example, in another spirochetal pathogen, B. burgdorferi, the EBP activator, Rrp2, and σ⁴, controls production of RpoS which in turn, governs expression of many virulence factors important for mammalian infection such as OspC [224, 225]. Thus, it is logical to postulate that LepB-σ⁵⁴, is involved in survival in natural environments for both pathogenic and saprophytic Leptospira, while LepA-σ⁵⁴ plays a role in survival in the host for pathogenic species. This hypothesis merits experimental testing.

Extracytoplasmic function σ factors (ECF σ factors) are the most diverse alternative σ factors found in many bacteria [226, 227]. Many bacteria contain multiple ECF σ factors. For example, Pseudomonas aeruginosa has more than 19 ECF σ factors [226]. Based on sequence analysis, ECF σ factors have been grouped into over 40 classes [226]. Our analyses showed that Leptospira have 5–10 ECF σ factors, and pathogenic Leptospira have 5 more ECFσ factors than saprophytic species (S11 Table), which is consistent with the more complex life cycle of pathogenic species than of saprophytic species. All Leptospira spp. have one copy of ECF31 and ECF43 with unknown functions. Pathogenic Leptospira have additional 5 unclassified ECFσ factors. Saprophytic Leptospira have one copy of ECF41 and ECF42 that are not found in pathogenic species. Although functions of ECF41 and ECF42-type remain unknown, one report showed that one of the ECF41 σ factors, SigJ in Mycobacterium tuberculosis, is involved in resistance to hydrogen peroxide [228]. It is unclear whether ECF41 in saprophytic Leptospira (LEPBI_I1070) has a similar function as SigJ, and if so, how it contributes to the survival of saprophytic Leptospira in the environment.

The activity of ECF σ factors is often regulated by an anti-σ factor, a transmembrane protein that binds and inhibits the activity of ECF σ [229]. Cleavage of anti-σ factor by proteases leads to release and activate σ^E. Extracellular signals regulate this intramembrane proteolysis often via an anti-anti-σ factor (or called anti-σ antagonist). Our analyses revealed that both pathogenic and saprophytic Leptospira have more than 30 σ^E regulators. Among them, 17 are only found in pathogenic/intermediate Leptospira, while 19 are found solely in saprophytic Leptospira (S11 Table). These differences likely reflect the variety of signals sensed by pathogenic and saprophytic Leptospira.

In addition to alternative σ factors, both pathogenic and saprophytic Leptospira species have many putative transcriptional regulators, far more than is found in other pathogenic spirochetes such as Borrelia burgdorferi and Treponema pallidum [230, 231]. Our initial analyses of transcriptional regulators among Leptospira species did not yield a distinct pattern of correlation with pathogenicity. Further in silico and experimental analyses to confirm the prediction and more importantly, to determine their regulatory role in Leptospira, is needed.

Sensory transduction

Leptospira species have a high number of two-component sensory systems (TCSs) (70–100) compared to Borrelia, Treponema and Bradyspira (6–20). The number of TCS genes found in a particular Leptospira species strongly correlated with the likely diversity of ecological niches that the species encounters, a phenomenon observed in other bacteria [203]. The lower number of TCSs found in pathogenic species may be linked to the process of host adaptation, whereas the larger numbers of unique TCSs in intermediates and, even more in saprophytes, may be instrumental for sensing and adapting to a more diverse range of environmental conditions. Regardless, almost all of the pathogenic species encode more than 70 TCS genes, indicating that the Leptospira pathogen requires a highly complex network of signaling processes for its life cycle. Interestingly, the proportion of TCS genes encoding orphan HK/RR and HHK proteins are higher in Leptospira (> 60%) than in other bacteria where orphan TCS proteins are unusual. These findings suggest that branched signaling pathways may be relevant in this genus and could confer added physiological advantages to Leptospira under specific circumstances.

One limitation in the present analytical approach is the difficulty in defining a robust and confident automatic method to segregate orthologous clusters among all Leptospira strains, and especially those that relate to TCS function. As in most in silico analyses, further biochemical experiments are needed to confirm the role of the various TCS categories identified in this study. Another limitation of this cross-species comparative analysis is that differences in serovars/strains with Leptospira species were not studied; such analyses will be a future priority given the strength of the approach and the depth of existing data. This in silico approach will also not identify novel virulence factors nor mechanisms of pathogenesis based on sequence analysis alone.

In summary, the large-scale comparative genomic analysis of 20 Leptospira species has provided broad insights into how infectious members of this genus acquired the genes necessary to acquire pathogenicity and virulence, placing these species within a definitive phylogeny. Novel, Leptospira species-specific genes and gene families were identified. Genomically-based metabolic reconstruction predictions predict novel adaptation of infectious Leptospira to mammals (summarized in Table 9), including sialic acid biosynthesis, pathogen-specific porphyrin metabolism and the first-time demonstration of riboswitch-regulated cobalamin (B12) autotrophy as a bacterial virulence factor. Only pathogenic Leptospira contain CRISPR/Cas systems, suggesting not only a potential mechanism for this clade’s refractoriness to gene targeting but also possible novel means to be able to genetically modify pathogenic Leptospira. Whether restriction modification systems might contribute to gene targeting has yet to be analyzed in detail, but the publicly available whole genome data sets provided in support of the present work will be contribute to carrying out such analyses. A novel virulence-related genes/gene family epitomized by the PF07598 group of paralogs suggests adaptation and diversification of this protein family within the pathogenic clade. Identifying large scale changes in infectious (pathogenic and intermediate pathogenic) as compared to non-infectious Leptospira has yield large-scale, novel insights into the evolution of a bacterial pathogen, provides the basis for new directions in leptospirosis pathogenesis research. It also makes novel genomic and pathogenomic contributions to the field of bacterial pathogenesis, which is of general interest.