A Single Multilocus Sequence Typing (MLST) Scheme for Seven Pathogenic Leptospira Species

1Leptospirosis is a global zoonosis caused by pathogenic members of the Leptospira genus –. These are maintained by chronic carrier mammalian hosts (e.g. rats and dogs), and are typically transmitted to humans via water contaminated with urine or by direct contact with reservoir hosts –. Leptospirosis can be caused by 8 pathogenic Leptospira species –, one of which (L. alstonii) is very rarely isolated with only two strains deposited in reference collections to date. A further five so-called ‘intermediate’ Leptospira species cause less severe disease –, and there are six non-pathogenic species,,. In addition, L. kmetyi clusters with the pathogenic species in a comparison of 16S rRNA gene sequence but has only been isolated from the environment thus far, and its ability to cause disease is uncertain. Three species (L. interrogans, L. kirschneri and L. borgpetersenii) cause the majority of human leptospirosis worldwide, –, although clinical manifestations of leptospirosis caused by different pathogenic species are often indistinguishable. It is important, therefore, that typing methodologies incorporate all major pathogenic species, not least as a means of species identification,.

2Multilocus sequence typing (MLST) is the genotyping method of choice for many bacterial pathogens, but the genetic divergence between species of pathogenic Leptospira complicates the development of a single MLST scheme for all of the pathogenic species. The availability of whole genome sequence for L. interrogans facilitated the development by Thaipadungpanit et al. of a MLST scheme that was supported by a public MLST website and database for this species and the closely related L. kirschneri. The primers used for this scheme worked variably for the other species and it was not recommended for use beyond L. interrogans and L. kirschneri. The subsequent publication of the whole genome sequence for L. borgpetersenii, and access to unpublished sequence data for other species including L. weilii, has provided the opportunity to now extend this scheme.

3The objective of this study was to expand the Leptospira MLST scheme to the seven pathogenic species associated with the overwhelming majority of disease (L. interrogans, L. kirschneri, L. borgpetersenii, L. noguchii, L. santarosai, L. weilii and L. alexanderi).

4The 327 Leptospira spp. used in this study are listed in Table S1. Genomic DNA was extracted as described previously, with the exception of isolates from the National Center for Emerging and Zoonotic Infectious Diseases, Centers for Diseases Control and Prevention, USA, which were extracted with the addition of dimethyl sulfoxide. For isolates for which the species was unknown, identification was undertaken by amplification and sequencing of the near full-length 16S rRNA gene (rrs), as described previously. Sequence data for rrs has been deposited in GenBank (see Table S2 for accession numbers). The serovars of isolates from recent human leptospirosis cases in Thailand, Laos and Sri Lanka were determined by the WHO/FAO/OIE Collaborating Center for Reference and Research on Leptospirosis, Queensland Health Forensic and Scientific Services, Brisbane, Australia using the cross-agglutinin absorption test (CAAT). The serovars of contemporary isolates from China were determined using CAAT at the China Medical Culture Collection of National Institutes for Food and Drug Control, China.

5Ethical approvals to obtain contemporary clinical isolates of Leptospira spp. are as follows. Isolates from patients in northeast Thailand between October 2000 and December 2006 were obtained during a prospective study of acute febrile illness, which was approved by the Ethical Committee of the Ministry of Public Health, Royal Government of Thailand. Isolates from patients in Laos between August 2006 and October 2009 were obtained during a prospective fever study approved by the Oxford Tropical Research Ethics Committee (OXTREC) and the National Ethics Committee for Health Research (NECHR), Ministry of Health, Government of Lao PDR. Isolates obtained from patients in Sri Lanka between June 2006 and May 2007 were obtained during a prospective fever study approved by the Research Ethics Committee of the Liverpool School of Tropical Medicine, UK and the Ethical Review Committee of the Faculty of Medicine of the University of Kelaniya, Sri Lanka. Written informed consent was obtained from all subjects except where not required (not required for data analyzed anonymously in Laos PDR at the time of the study). Other clinical isolates were obtained from the China Medical Culture Collection of National Institutes for Food and Drug Control, China, and the National Center for Emerging and Zoonotic Infectious Diseases, Centers for Diseases Control and Prevention, USA. All data were analyzed anonymously.

6Sequence types and the associated sequence data for the seven MLST loci were downloaded from the Leptospira MLST website (http://leptospira.mlst.net) for 199 isolates. MLST was performed using the original scheme for an additional 52 Leptospira isolates that had not been genotyped previously. These included 38 L. interrogans isolates from Laos (n = 11), Malaysia (n = 2), Sri Lanka (n = 14), Japan (n = 1) the Philippines (n = 2), or China (n = 8), and 14 L. kirschneri isolates from Japan (n = 2) or reference collections (n = 12). Alleles were assigned and sequence types defined using the MLST website. The 251 isolates typed by the original scheme are detailed in Table S1.

7Six of the seven original MLST loci (glmU, pntA, sucA, tpiA, pfkB and mreA) were retained in the modified scheme. New primers were designed for these six loci to take into account the known genetic variability at the primer binding sites of L. interrogans, L. borgpetersenii and L. weilii, but the regions of the genes used to define the alleles were unchanged. The new primers corresponded to regions of these genes that were conserved in five Leptospira spp. genome sequences, and were designed using PrimerSelect software (DNASTAR Inc., Wisconsin, USA). Four of these genomes were from the NCBI database (L. interrogans serovar Lai strain 56601 (NC_004342) and serovar Copenhageni strain Fiocruz L1-130 (NC_005823); L. borgpetersenii serovar Hardjo-bovis strain L550 (NC_008508) and strain JB197 (NC_008510) ), and one unpublished L. weilii genome was acquired from the Wellcome Trust Sanger Institute (European Nucleotide Archive (ENA) accession number ERS002113). The seventh locus, fadD, was replaced with caiB, a gene present on chromosome I that encodes carnitine dehydratase belonging to the CoA transferase family III. The primers used for the new PCR scheme are shown in Table 1.

8Amplification and sequencing of caiB was performed for the 251 isolates genotyped by the original MLST scheme. All seven loci were sequenced for a further 74 isolates belonging to species that cannot be genotyped using the original scheme: L. alexanderi (n = 5), L. borgpetersenii (n = 34), L. noguchii (n = 9), L. santarosai (n = 10), and L. weilii (n = 16) (Table S1). Twelve of these were from recent cases of human leptospirosis in Laos (L. borgpetersenii [n = 2], L. weilii [n = 4]), Thailand (L. borgpetersenii [n = 4]) or Sri Lanka (L. borgpetersenii [n = 2]); and six were L. borgpetersenii isolates from recently trapped rodents in Japan. In addition, two L. borgpetersenii isolates were included based on in silico MLST using whole genome sequences from NCBI database (NC_008508 and NC_008510).

9PCR reactions were performed in a volume of 25 µl containing 1.5–3.5 mM MgCl2 (Table 1), 200 µM dNTP (Roche, USA), 1.25 unit Taq DNA Polymerase (Roche, USA), 5 pmol of each forward and reverse primer (Table 1), and approximately 50 ng of Leptospira DNA. Amplifications were performed with the following conditions: one cycle of 95°C for 2 minutes, 30 cycles of 95°C for 10 seconds, 46°C for 15 seconds, and 72°C for 30 seconds, followed by a final period of 72°C for 7 minutes. PCR products were sequenced by Macrogen Inc. (Seoul, Korea), the sequences trimmed to the correct length (Table 1), and edited using SeqMan software (DNASTAR Inc., USA). Allele numbers were assigned to each unique caiB locus, together with new alleles for novel sequences at the six other loci. Allelic profiles (in the order glmU-pntA-sucA-tpiA-pfkB-mreA-caiB) were used to assign sequence types (STs) to all isolates. Wherever possible, the original ST number for a given isolate was preserved in the modified scheme.

10Sequence alignment, the identification of polymorphic sites, and the construction of phylogenetic trees were performed using MEGA version 5.0. The rate of substitution at synonymous and non-synonymous sites (dN and dS, respectively) was determined using a modified Nei-Gojobori method. Neighbor-joining trees were constructed from concatenated sequences of MLST loci using Kimura's two-parameter model. Maximum likelihood trees were constructed from concatenated sequences of MLST loci using an algorithm implemented in PhyML version 3.0.1. The model of sequence evolution used was the generalized time-reversible (GTR) model with gamma-distributed rate variation. The model parameters were adjusted as follows: transition/transversion ratio was fixed to 4.0, and the gamma shape parameter accounting for rate variation among sites and proportion of invariant sites was optimized. We set the program to search the tree using the Nearest Neighbor Interchange (NNI) method and chose BioNJ as an initial tree. The MEGA program was used to display and edit the tree. Discriminatory ability (D value) and 95% confidence intervals (CI) were estimated as described previously,. Relatedness between STs was analyzed based on allelic profiles using eBURST version 3,. Neighbor-joining and maximum likelihood trees were also constructed for 100 isolates using the near full length rrs gene corresponding to positions 2417657 to 2418837 in the genome of L. interrogans serovar Lai strain 56601 (GenBank accession number NC_004342.2).

11The modified MLST scheme was applied to 327 isolates representing seven pathogenic Leptospira spp. (Table S1). A total of 190 different STs were resolved. The data for these can be accessed at the modified MLST website (http://leptospira.mlst.net/). The discriminatory ability for different species ranged from 0.5 ST per isolate for L. interrogans and L. borgpetersenii to 1.0 ST per isolate for L. alexanderi, L. noguchii and L. santarosai (Table 2). The number of alleles per locus ranged from 51 (caiB) to 70 (pfkB) (Table 2). Two L. borgpetersenii isolates from mice trapped in Japan, both of which were assigned as ST 197, contained a non-standard length allele (caiB 51) due to a 78 bp deletion (position 13–90) in caiB. We excluded these two isolates from the phylogenetic analysis. The species of the remaining 325 isolates were as follows: L. interrogans (n = 222), L. borgpetersenii (n = 34), L. kirschneri (n = 29), L. weilii (n = 16), L. santarosai (n = 10), L. nogouchii (n = 9), L. alexanderi (n = 5). The dN/dS ratio for each locus was less than one for all seven species, indicating a lack of positive selection (Table S3).

12Neighbor-joining and maximum-likelihood trees were constructed from concatenated sequences of the seven loci for 325 isolates. In general, phylogenetic analysis of the seven concatenated genes strongly supported current species assignments, although there was significant subdivision within the L. weilii population combined with a close relationship between L. weilii and L. alexanderi (Figure 1A–D). This was further explored in an analysis that considered the contribution of each individual locus to the phylogeny. Seven maximum likelihood trees were constructed, one for each of the seven MLST loci (Figure 2). All seven trees confirmed the close genetic relatedness between L. weilii and L. alexanderi. The tpiA tree was notable for a subdivision of L. weilii that was distinct and genetically distant from other L. weilii strains as well as from L. alexanderi and L. borgpetersenii. This outlying group contained seven L. weilii isolates with a highly divergent tpiA (allele 51). Constructing a tree using the concatenated sequences of all loci except tpiA resulted in all of the L. weilii isolates falling into only two lineages, and resolved the apparently polyphyletic nature of this species (Figure S1).

13eBURST was used to identify groups of related STs (clonal complexes, CCs) among the 111 STs of L. interrogans (there were too few STs for the other species to use this procedure). This demonstrated two CCs, the putative founders of which were ST 37 and ST 12 respectively, and are thus named CC 37 and CC 12 (Figure 3). CC 37 contained 11 STs representing 22 isolates that originated from a wide geographic region including Thailand, Laos, Sri Lanka, China, Indonesia, Japan, Australia, Brazil, Jamaica and the Netherlands. CC 12 contained 8 STs representing 23 isolates from Asia (Sri Lanka, China, Malaysia and Thailand). Isolates belonging to the same CCs clustered together in the same branch of the phylogenetic tree (Figure S2).

14The impact of changing the single locus in the MLST scheme was evaluated by comparing the results of the original and modified schemes for the 251 isolates of L. interrogans and L. kirschneri in the study collection (n = 222 and n = 29, respectively). This comparison included almost all (117/120) unique STs registered previously in the MLST database based on the original scheme. The remaining three STs (STs 26, 33 and 114) were not included in this study because DNA or isolates were not available for the four associated isolates.

15A neighbor-joining tree constructed from the concatenated sequences of the seven loci of the original scheme demonstrated two distinct phylogenetic clusters corresponding to L. interrogans and L. kirschneri, as reported previously. Isolates belonging to the L. interrogans cluster were resolved into 111 STs (0.5 ST per isolate), and those in the L. kirschneri cluster were resolved into 22 STs (0.8 ST per isolate). Comparison of neighbor-joining trees for the two schemes reconstructed from concatenated sequences of seven loci showed these to be highly similar, retaining the clear separation of the two species. The modified scheme resolved five new STs (137, 138, 139, 140, and 141) as a result of sub-division of existing STs, while four original STs (94, 100, 103, 104) were lost through merging of STs (Table S1). The discriminatory power (D value) of the two schemes was not significantly different (original scheme 0.936 (95% CI 0.911–0.961) versus modified scheme 0.937 (95% CI 0.912–0.962). The D value of caiB was significantly higher than that of fadD, the allele which it replaced (fadD, 0.793 (95% CI 0.760–0.826); caiB, 0.876 (95% CI 0.852–0.901; p<0.05)). None of the caiB alleles were shared between L. interrogans and L. kirschneri.

16Phylogenetic analysis has been performed previously for the Leptospira genus based on the rrs gene. We compared the phylogeny between this and MLST for the seven pathogenic Leptospira species included in this study. Sequence of the near-full length rrs gene was available for 100 study isolates, including 83 clinical isolates for which rrs sequencing was performed by us, and 17 sequences that were obtained from GenBank. Of these, ninety-four isolates had been genotyped by MLST and the remaining six isolates (L. santarosai or L. noguchii) were not represented in our collection. Neighbor-joining and maximum-likelihood trees based on rrs are shown in Figures 4A and 4B. The broad phylogenetic structure was similar to that observed using MLST, although several differences were noted. The clustering of L. alexanderi and L. weilii observed by MLST was not replicated by rrs, which assigned the two species to discrete branches. Three contemporary isolates had an ambiguous phylogeny by rrs that was not observed by MLST, as follows. One human isolate (ST 18) from Thailand in the L. interrogans cluster and two isolates (both ST 136) from shrews in Japan residing in the L. kirschneri cluster based on MLST were clustered by rrs into a single branch located between L. interrogans and L. kirschneri. This result suggests that recombination may occur in rrs between the two closely related L. interrogans and L. kirschneri species. Nucleotide distances between species were compared for data derived from MLST and rrs. These ranged from 3.35% to 18.43% for MLST (Table 3), and 0.15% to 1.01% for rrs (Table 4). The nucleotide distance between L. alexanderi and L. weilii was 3.35% by MLST, but only 0.75% by rrs.

17Typing of Leptospira spp. based on serovars is a long established approach for epidemiological surveillance. We examined whether isolates of the same serovar were also closely related by genotype, and vice versa. This was possible for 254/325 isolates that had both a ST and serovar designation, the remaining 71 isolates being of unknown serovar. We first considered the serovars corresponding to the 31 STs represented by at least two isolates. Of these, 13 STs contained isolates belonging to a single serovar and 18 STs contained isolates that were assigned to two or three different serovars (Table S4). We extended this analysis to those isolates belonging to L. interrogans CC 37 or CC 12. The 23 isolates in CC 12 belonged to three different serovars (Pyrogenes, Bangkok and Fugis), and the 22 isolates in CC 37 belonged to nine serovars (Canicola, Gem, Guaratuba, Medanensis, Paidjan, Portlandvere, Pyrogenes, Pomona, and Hebdomadis). We then identified those serovars that contained more than one ST. This was the case for 16 serovars, the number of STs per serovar ranging from 2 to 10 (Table S5). There was a variable degree of relatedness between STs belonging to the same serovar. These observations indicate that serovar is a poor indicator of genetic relatedness and can vary within a given clone or lineage, an observation that is most likely to be due to horizontal gene transfer.

18The L. interrogans cluster contained four sub-groups, the largest of which contained 104 different STs comprising isolates with a global distribution. The other three sub-groups contained isolates from the Asia-Pacific region comprising eight STs (Figure S2). The L. kirschneri cluster contained two major sub-groups, the larger of which contained reference isolates originating from Asia, Europe, Africa, North America and South America. The smaller sub-group contained ST 68 (n = 2) and ST 71 (n = 1), all of which were clinical isolates from Thailand. These three isolates were distinct in that they contained a sucA allele (allele 1) that is also found in L. interrogans, the only example of shared alleles between the two species. The L. noguchii cluster containing nine reference isolates were divided into three sub-groups that originated from United States and Central America, Hawaii, and South America, respectively. The L. borgpetersenii cluster contained two sub-groups, the larger of which contained reference isolates from Europe, Asia, and Africa, and recent isolates from human leptospirosis cases in Thailand, Laos and Sri Lanka, and rodents in Japan (n = 6). The smaller cluster contained only reference isolates originating from several continents but not Asia. The L. santarosai cluster contained 10 reference isolates each of a different ST, nine of which were from Central and South America. L. weilii fell into three sub-groups. The first contained reference isolates from China and Indonesia and recent clinical isolates from Laos; the second contained reference isolates from China, Vietnam and Malaysia plus a recent clinical isolate from Laos; and the third contained 4 reference isolates from Australia, Indonesia and Malaysia. All members the first sub-group contained the highly divergent tpiA (allele 51). A single L. alexanderi group contained reference isolates originating from China.

19MLST was performed on 126 new isolates, including contemporary isolates from humans or animals from Laos, Malaysia, the Philippines, Sri Lanka, Japan and China. This allowed us to examine the population of Leptospira across a wider region of Asia than has been performed previously. We noted that the dominant clone defined previously in Thailand (L. interrogans ST 34) was also isolated from humans in neighboring Laos (8/27 [30%] of isolates tested). The closest relative of ST 34 found in this study was the single locus variant ST 112 (Figure 3), which was isolated from a human in India. Most of the L. borgpetersenii isolates associated with human leptospirosis in Thailand, Laos, and Sri Lanka were closely related, with seven out of the eight isolates belonging to ST 143 (n = 4) or ST 144 (n = 3), which are single locus variants. ST 143 was also represented in recent isolates from rats in Japan (n = 3), as well as reference isolates from China (n = 1) and Indonesia (n = 1). These observations indicate that this L. borgpetersenii lineage is widely distributed in Asia, and that the rat is one of the maintenance hosts in this region. A reference isolate from a case of human leptospirosis originating in Sri Lanka in 1964 was also L. borgpetersenii ST 144, indicating the presence of this lineage in Asia for at least half a century. L. borgpetersenii ST 146 was recovered from shrews in Japan in 2011, and a shrew in Czechoslovakia in 1953. These data demonstrate the potential of MLST to define the phylogeography of Leptospira species over time, and to create links with their maintenance host. L. weilii formed a small minority of recent clinical isolates from cases of human leptospirosis in Asia (4/140 [3%]).

20The discrimination provided by MLST allowed us to re-examine the relationship between isolates that are pathogenic for man and their putative maintenance host. For this we made the assumption that the same ST isolated from an animal and a human implied that the strain had crossed the species barrier. This analysis was possible for 316 isolates of known origin, the remainder being of unknown origin (n = 8) or from the environment (n = 1). The 316 isolates originated from humans (n = 230), rodents (n = 48), amphibians (n = 2), and other mammals (n = 36). Several STs contained isolates obtained both from humans and rodents (n = 10), humans and other mammals (n = 4), or humans, rodents and other mammals (n = 1). We found a single instance where an isolate of L. kirschneri ST 122 associated with human leptospirosis was maintained in a rodent. Our analysis provided confirmatory evidence for the role of rodents as major maintenance hosts for L. interrogans, and bovines as major maintenance hosts for L. borgpetersenii and L. weilii. This also demonstrates the potential utility of MLST for the investigation of maintenance hosts during outbreak and other epidemiological investigations.

21This study has described the successful modification of a Leptospira MLST scheme to allow the characterization of isolates of seven pathogenic Leptospira species. The original and modified schemes provide similar levels of resolution, and the modified scheme presents considerable advantages for the scientific community with minimal negative impact. We contacted all of the investigators who had deposited MLST data into the public database, and were able to include all but 4 isolates (old STs 26, 33 and 114) in the modified scheme. The current MLST website and database (http://leptospira.mlst.net/) has been updated to reflect the new improved scheme, and the old scheme will remain available via the website for a period of 6 months and then removed to avoid confusion.

22The modified MLST scheme assigned isolates to distinct clades with 100% accuracy and can, therefore, be used to assign both species and genotype to isolates of the seven Leptospira species tested here. Furthermore, we were able to identify relatedness between STs in specific geographic regions. For example, a clonal complex of L. interrogans (CC 12) was identified that contained recent isolates from several countries in southeast Asia, and genetic relatedness was identified between isolates from Australia and Papua New Guinea. Although the number of Leptospira isolates that have undergone MLST is relatively small to date, this suggests that the scheme is capable of describing the molecular epidemiology of the main pathogenic species within the genus on a global scale. There is notable genetic diversity both within a given species and between species, indicating that MLST would also be predicted to be capable of demonstrating transmission pathways of specific clones between maintenance hosts and man.

23Several other multilocus sequence based typing schemes for Leptospira spp. have been reported in the literature. The most comparable scheme to the one described in our study is that described by Leon et al., which is also based on seven housekeeping genes and is restricted to L. interrogans and L. kirschneri. The species identity of 1/51 isolates tested was apparently misclassified by MLST, although misidentification of the isolate was also considered a possibility. This scheme is not supported by a public database and website and does not extend to other pathogenic species. A six locus scheme described by Ahmed et al. included rrs2 (one of two 16S rRNA genes that would be predicted to be highly conserved), three housekeeping genes, and two genes encoding surface expressed proteins (lipL41 and lipL32), albeit that these appeared evolutionary neutral. Discrepancy between serovar identity and speciation was evident for 2/120 isolates in its first description, with multiple examples of apparent serovar misclassification or mislabeling in a more extensive validation of the scheme using 271 isolates. For example, the largest cluster of L. interrogans also contained three other Leptospira species. This scheme is not associated with a public database and website although sequences for the 271 strains can be downloaded from GenBank, and phylogenetic analysis of new sequences using this scheme requires their download and offline analysis. A further scheme consisting of four loci including genes encoding surface expressed protein has been evaluated in silico for 38 pathogenic Leptospira strains, which demonstrated correct species assignment for 34 of these.

24The finding in this study that serovar is a poor indicator of genetic relatedness and can vary within a given clone or lineage is most readily explained by horizontal gene transfer of genes encoding the surface determinants that confer serovar designation. Serovar variation is thought to relate to differences in the LPS O-antigen encoded by the rfb gene cluster, although the precise nature of this variation is not fully defined. A comparison of the sequence of the rfb cluster in five isolates all of different serovar (four L. interrogans and one L. borgpetersenii) demonstrated the presence of a set of genes that were present in all isolates, together with several genes that were variably present. In addition, a study that used comparative genomic hybridization to compare the gene content of L. interrogans serovar Lai strain Lai with eleven L. interrogans belonging to different serovars reported notable divergence in the rfb cluster. Large-scale whole genome sequencing projects that are currently underway are likely to shed important new evidence on the genetic structure and putative stability of the regions involved in serovar designation.

25In summary, the modified MLST scheme described here proved highly discriminatory for seven pathogenic species of Leptospira, providing both isolate characterization and robust assignment to species in addition to phylogenetic evidence for the relatedness between the species. Crucially, this scheme is also supported by a public website. We recommend that this MLST scheme and the updated website at http://leptospira.mlst.net/ should be used for the future characterization of Leptospira isolates.