Leptospirosis is a zoonotic disease caused by pathogenic Leptospira spp. and is considered an emerging global public health problem [1], [2]. Furthermore, the impact of leptospirosis has increased, particularly in poverty stricken regions of the world, due to the high mortality (>50%) associated with the recent increase of severe pulmonary haemorrhage syndrome (SPHS) in patients with severe leptospirosis [3], [4]. Based on serology, Leptospira spp. are traditionally classified into 29 serogroups and over 300 serovars [5]–[7]. More recently, genetic methods have attempted to replace the traditional classification methods and DNA-DNA hybridization studies have identified 20 Leptospira spp. to date [5], [8]–[11]. Several typing methods have been employed to classify isolates with differing degrees of success [12]. However, a major limitation is the lack of correlation between the serologic and genotypic classification methods [5], [12], [13].

Multilocus sequence typing (MLST) was originally developed for bacteria using Neisseria meningitidis isolates [14] and, so far, it has been successfully applied to over 30 bacteria [15], [16]. In the field of leptospirosis, efforts to develop a typing method have focused on MLST [17], [18]. MLST allows the adoption of a universal format for a particular bacterial species and permits the sequence data generated to be easily exchanged over the Internet. Traditionally, the loci chosen for MLST analyses are based on 6–10 housekeeping genes that are under selection for metabolic functionality [15]. Since this group is comprised of slowly evolving genes they are likely to be more conserved and stable within a particular species [19]. Unfortunately it has not been possible to identify a set of housekeeping genes with universal applicability to all bacterial pathogens. Rather, MLST loci are chosen empirically and evaluated for each individual pathogen [15]. Ahmed and colleagues presented the first MLST scheme based on loci from four housekeeping genes and two genes encoding outer-membrane proteins for typing L. alexanderi, L. borgpetersenii, L. interrogans, L. kirschneri, L. noguchii and L. santarosai isolates [18]. An alternative MLST scheme using loci from seven housekeeping genes was used to type L. interrogans and L. kirschneri isolates and is available on the Internet (http://leptospira.mlst.net/). The database contains 109 sequence types (ST) and sequences from 263 isolates at time of writing [17]. Although this evidently moved the field forward, a limitation of this database is that it only applies to isolates from two Leptospira species, L. interrogans and L. kirschneri. The ideal MLST scheme should be valid for all Leptospira spp. or at least the pathogenic species [20], and provide discrimination beyond the species level [21].

High-resolution typing, such as that required during outbreak investigations, usually requires the inclusion of genes with greater diversity, e.g. antigen genes, rather than housekeeping genes [15]. The objective of this study was to carry out a bioinformatics-based analysis of Leptospira genes available in GenBank to identify potential targets for improved Leptospira discrimination. The genes ligB, secY, lipL41 and rpoB were identified as potential genes for use in an improved typing scheme.

The MLST scheme proposed by Thaipadungpanit and colleagues is based on loci from seven housekeeping genes and identified 109 unique ST among 263 isolates from either L. interrogans or L. kirschneri [17]. This represented a major advance in the molecular epidemiology of Leptospira isolates. Unfortunately, as noted by the authors themselves, a major limitation of this scheme is that it does not allow for the inclusion of the other common pathogens associated with human leptospirosis [20]. Previously, it was shown that the antigen encoding genes ligB, secY, lipL41 and the rpoB gene are potentially useful for the molecular discrimination of Leptospira strains and that they can be readily amplified from all known pathogenic leptospires [23]–[25], [29], [30]. Furthermore, the authors of the original bacterial MLST scheme recommend the inclusion of loci from antigen coding genes to improve discrimination, especially during outbreak investigations [31].

To determine the potential benefits of using ligB, secY, rpoB and lipL41 loci in an MLST typing scheme focused on strain differentiation we identified the corresponding sequences in a reference collection containing 38 Leptospira strains (Table 1). The overall level of pairwise identity ranged from 82–92% among the individual loci while the intraspecies identity was even higher (Figure 1). Furthermore, when the sequences were concatenated to create a superlocus for each strain and analysed, the overall pairwise identity was >90%. Following the modelling of phylogenetic trees and in agreement with previous studies, two distinct evolutionary branches were observed, the first contained L. kirschneri, L. interrogans and L. noguchii strains and the second the L. borgpetersenii, L. santarosai and L. weilii strains (Figure 4 and Figure S2) [23], [29]. There was some evidence that serovars from the same serogroup clustered together, serogroups: Icterohaemorrhagiae: L. interrogans Icterohaemorrhagiae RGA and Copenhageni Fiocruz L1-130; Australis: L. interrogans Australis Ballico and Muenchen Muenchen C90; and Javanica: L. borgpetersenii Javanica Veldrat Batavia 46, Poi Poi and Ceylonica Piyasena (Figure 2). This may indicate homoplasy (similarity due to convergent evolution) of the genes from these serovars. This analysis showed a separation of serovar Bratislava from the other L. interrogans serovars (Figure 2). Alignment analyses demonstrated this is probably due to the high similarity of the L. interrogans Jez Bratislava strain rpoB gene sequence with those from the L. borgpetersenii strains (data not shown). Despite this, serovar Bratislava was correctly located within the L. interrogans clade. We did not determine however, whether this was due to sequence mosaicism or horizontal gene transfer. The phylogenetic organisation of the Leptospira genus based on the superlocus supports the theory [29] that L. interrogans is more recently evolved from L. kirschneri, more recent evolutionary subdivisions resulted in the separation of L. borgpetersenii followed by L. santarosai and L. weilii clades.

Analysis of the discriminatory power of the ligB-secY-rpoB-lipL41 superlocus found that the L. interrogans, L. noguchii, L. santarosai and L. weilii strains could be separated into individual ST. However, among the L. kirschneri and L. borgpetersenii strains, two could not be resolved by the superlocus. Serological analysis of the two L. borgpetersenii strains indicated that both belonged to serogroup Sejroe serovar Hardjo-bovis and their genomes were found to be highly conserved. Yet they are distinct clonal subtypes, both strains established chronic infections in cattle yet differed in their ability to cause lethal infections in hamsters [32]. Despite this no specific polymorphisms were observed in either of the L. borgpetersenii L550 or JB197 strains. These polymorphic regions are normally useful for surveillance purposes, to monitor outbreaks or for epidemiological studies. Thirteen serovars, out of 38 in this study, exhibited unique polymorphic sites. These findings highlight the efficiency of the proposed ligB-secY-rpoB-lipL41 superlocus to discriminate Leptospira strains. The study previously performed by Ahmed and colleagues [18] was the pioneer in the use of a concatenated superlocus to discriminate among the Leptospira species. However, this work was intended as a step towards the study of pathogen evolution rather than strain discrimination. Inclusion of low polymorphic genes such as rrs2 and lipL32 used by Ahmed et al., [18] did not contribute to enhance the discriminative power of the MLST scheme presented here. In the present work, we included some of those sequences in combination with recently sequenced polymorphic genes to increase the resolution, and observed the occurrence of discrimination to the subspecies level. Although a reduced number of strains and isolates were included in our study we believe the proposed superlocus presents a solid basis for discriminating within large panels of Leptospira strains and isolates.

An analysis of the ligB, secY, rpoB and lipL41 loci found that the ligB locus was the most conserved with 15 alleles, followed by secY with 16 alleles, rpoB with 19 alleles and lipL41 with 28 alleles out of a potential 38 (Figure 3 and Figure S1). Following analysis of the 263 isolates (109 ST) contained in the Leptospira MLST database (leptospira.mlst.net), 16 strains (corresponding to 13 ST) were identified as having the corresponding ligB, secY, rpoB and lipL41 loci sequences available (Figure 4). To determine the utility of the loci proposed in this study the relevant sequence for each individual locus was concatenated to the glmU, pntA, sucA, fadD, tpiA, pfkB and mreA superlocus of each strain (Figures S2 B–E). All possible variables were evaluated in order to identify the most useful additional loci. The combination of the original superlocus together with the rpoB and the lipL41 loci was found to be the simplest superlocus that could discriminate between all 16 of the strains (Figure S2 F). This is in agreement with the ability of the superlocus determined by Ahmed and colleagues [18], which includes the lipL41 locus, to discriminate the Leptospira spp. in study.

The phylogenetic analysis of our sequences showed a great diversity of ST and no clustering, due to the use of epidemiologically unrelated strains. Thus, when the two new loci sequences were concatenated to the original ST sequences we observed the complete discrimination of the strains, although our adapted scheme remains limited to L. interrogans and L. kirschneri. We recommend that the rpoB and lipL41 loci be evaluated in existing or future MLST schemes to enhance their typing power during outbreak investigations.