MLST of housekeeping genes captures geographic population structure and suggests a European origin of Borrelia burgdorferi

Abstract

Lyme borreliosis, caused by the tick-borne bacterium Borrelia burgdorferi, has become the most common vector-borne disease in North America over the last three decades. To understand the dynamics of the epizootic spread and to predict the evolutionary trajectories of B. burgdorferi, accurate information on the population structure and the evolutionary relationships of the pathogen is crucial. We, therefore, developed a multilocus sequence typing (MLST) scheme for B. burgdorferi based on eight chromosomal housekeeping genes. We validated the MLST scheme on B. burgdorferi specimens from North America and Europe, comprising both cultured isolates and infected ticks. These data were compared with sequences for the commonly used genetic markers rrs-rrlA intergenic spacer (IGS) and the gene encoding the outer surface protein C (ospC). The study demonstrates that the concatenated sequences of the housekeeping genes of B. burgdorferi provide highly resolved phylogenetic signals and that the housekeeping genes evolve differently compared with the IGS locus and ospC. Using sequence data, the study reveals that North American and European populations of B. burgdorferi correspond to genetically distinct populations. Importantly, the MLST data suggest that B. burgdorferi originated in Europe rather than in North America as proposed previously.

Lyme borreliosis is the most prevalent vector-borne disease in the temperate zone of the Northern Hemisphere. It is a tick-borne bacterial zoonosis, with hard ticks of the genus Ixodes acting as vectors (1) and various vertebrate species serving as reservoir hosts (2–6). At present, Lyme borreliosis spirochetes constitute a group of 13 named species (7–18). Several of these, namely Borrelia burgdorferi, Borrelia afzelii, Borrelia garinii, and Borrelia spielmanii, are associated with disease in humans (8, 19–22). B. burgdorferi (sometimes referred to as B. burgdorferi sensu stricto) occurs in both Europe and North America (23). In the United States, Lyme disease is caused by B. burgdorferi only (24).

Despite almost three decades of research and control efforts, the numbers of new human cases of Lyme borreliosis in the United States continue to increase (24). It is likely that this reflects increasing population sizes and geographic ranges of pathogenic genotypes of B. burgdorferi (25) because it is known that populations of the principal vector tick in eastern North America, Ixodes scapularis, have been spreading from past refuges into new woodland habitats (26–34). In addition, it is likely that the wide host range of B. burgdorferi has been facilitating its epidemic dispersal in the northeastern United States (2, 3). There are three main regions in the United States where B. burgdorferi is now prevalent: the Northeast, the upper Midwest, and northern coastal California. B. burgdorferi is relatively rare in Europe compared with the northeastern United States and has not been recorded in Asia (4, 35–37).

The genome of B. burgdorferi is remarkable among bacteria, in that it consists of a linear chromosome, which encodes ≈50% of the predicted proteome, and a large number of linear and circular plasmids (38, 39). Both chromosomal and plasmid-located loci have been used for typing B. burgdorferi, such as the chromosomal rrs-rrlA intergenic spacer (IGS) (40, 41), the gene encoding the membrane protein p66 (42), or the plasmid-located, hypervariable gene encoding the immunodominant outer surface protein C (ospC) (25, 43).

Unambiguous genotyping systems are key to describing epidemiological and ecological patterns and illuminating the evolutionary processes that shape microbial populations. Several recent studies have used sequence information of multiple loci to characterize Lyme borreliosis spirochetes (15, 17, 41, 44, 45). However, these typing approaches deviate from typical multilocus sequence typing (MLST) schemes or multilocus sequence analysis (MLSA) (46) developed for other microbial pathogens in that different categories of loci were combined, such as hypervariable genes encoding outer surface proteins, conserved housekeeping genes, or noncoding loci.

Most MLST/MLSA schemes are based on housekeeping genes, which are subject to purifying selection and slow evolution, and the variation within these genes is nearly neutral (47). Although there are normally fewer polymorphic sites in individual housekeeping genes compared with hypervariable genes, using the combined sequences of multiple housekeeping genes has been shown to provide high discriminatory power while retaining signatures of longer-term evolutionary relationships or clonal stability (47–50) (see also www.mlst.net). Furthermore, analyses of multiple loci can buffer against potentially skewed evolutionary pictures obtained by single-locus analyses (49, 51). Despite their power, typical MLST/MLSA schemes have not yet been applied as tools in population or landscape genetics studies of vector-borne pathogens.

To contribute to the understanding of the emergence of Lyme borreliosis, we aimed to test whether B. burgdorferi is structured geographically and to infer the evolutionary origin of this bacterial species. We, therefore, have developed a MLST scheme for B. burgdorferi based on housekeeping genes to characterize B. burgdorferi populations at the different phylogenetic levels required for evolutionary, epidemiological, and population genetics analyses. We show that North American and European populations of B. burgdorferi constitute distinct lineages and that the housekeeping genes evolve differently compared with the IGS locus and ospC. Importantly, phylogenetic analyses of the concatenated housekeeping genes suggest that B. burgdorferi originated in Europe.

Results

B. burgdorferi Samples and Housekeeping Genes Analyzed.

Specimens of B. burgdorferi from the Northeast and Midwest United States, California, and Europe that were analyzed by MLST are listed in supporting information (SI) Table S1. These included field-collected infected ticks and isolates from patients and ticks. The characteristics of the eight housekeeping genes selected, i.e., clpA, clpX, nifS, pepX, pyrG, recG, rplB, and uvrA, such as GC content (10) or dN/dS ratio, are shown in Table S2.

MLST of B. burgdorferi and Comparison with the Major IGS Genotypes and ospC Major Groups.

In total, 33 sequence types (STs) were defined among the 64 samples of B. burgdorferi analyzed (Table 1 and Table S3). Most notably, strains from North America did not share any STs with those from Europe (Fig. S1, Table S1, and Table S3). A permutation test using the allelic profiles of B. burgdorferi (Table S3) provided significant evidence (P < 0.003) that the B. burgdorferi populations from Europe and the United States constitute distinct lineages. The distribution of the dissimilarity statistic is shown in Fig. S2. However, STs 20–23 from Europe (IGS genotype 1) shared alleles with North American strains (ST1 and ST2, IGS genotype 1) for several housekeeping genes (clpX, pepX, pyrG, and rplB) (Fig. S1 and Table S3). The same European STs shared an uvrA allele with ST14 from the United States, which has been typed as IGS genotype 6.

Table 1.

STs, IGS types, and ospC major groups of B. burgdorferi

Strain	MLST ST	IGS type*	ospC group†
B31, BL206, B515, 16812UT	1	1	A
Ca4, Ca5, Ca6	2	1	A
297, B504, 498801UT	3	2	K
B509	4	2	H
Ca_WTB27	5	2	A
Ca92-0953	6	2	H
MR623, B373	7	3	B
B156	8	4	F
MR661, MR654, 15506UT	9	4	N
Ca_WTB32	10	4	F
JD1, BL538, BL515, 114311UT	11	5	C
BL522, B356	12	6	M
Ca92-1337	13	6	M
MR616, 15912UT	14	6	G
B500	15	7	I
B331, B361	16	7	I
Ca92-1096	17	7	I
B485, MR607, MR662	18	8	U
N40, B418, MR640, B348, 15903UT	19	9	E
IPT2, IPT69, IPT191	20	1	B
IPT23, IPT190	20	1	M1
20604LT	20	1	ND
22521LT	20	ND	ND
IPT19	21	1	Q
21509LT	21	ND	ND
IPT135	22	1	Q
IPT137	23	1	B
IPT39	24	5	S
IPT58	24	5	L
IPT193, IPT198	25	10	N1
NE49	26	10	Q
Z41293	27	11	O1
Z41493	28	12	N1
47703UT	29	2	L
51405UT	30	6	B
48102UT	31	9	E
519014UT	32	2	H
20111LT	33	5	ND

ND, not done because of the lack of PCR products.

*According to the method described by Bunikis et al. (41).

^†According to the method described by Wang et al. (43) and Seinost et al. (52).

Cultured isolates of B. burgdorferi representing the previously described nine major IGS genotypes (41) were included in this study. Some isolates belonging to the same major IGS genotypes (e.g., 8 and 9) were found to have identical alleles for all eight housekeeping genes (Table 1 and Table S3), pointing to some clonal frame within B. burgdorferi. Among the European strains, three (NE49, IPT193, and IPT198) represented IGS genotype 10 (41), and two (Z41293 and Z41493) were found to be new IGS genotypes. Here these were termed IGS genotypes 11 and 12 (Table 1 and Table S4).

The samples analyzed in the present study represent 14 previously described ospC major groups (6, 43, 52). According to the definition introduced by Wang et al. (43), alleles of ospC differ by >8% between ospC major groups but are <2% different within an ospC major group. In this study three novel ospC major groups were detected among the European isolates (termed M1, N1, and O1; Table 1), owing to 8–16% dissimilarity to their closest neighbors in the phylogenetic trees. For a few infected ticks from Latvia, it was not possible to amplify the IGS locus or ospC (Table 1).

As reported earlier (43), some ospC major groups were found for both New World and Old World strains of B. burgdorferi. For example, ospC major group B was found for ST7 (North American strains B373 and MR623), ST20, and ST23 (European strains IPT2, IPT69, IPT191, and IPT137) (Table 1). Furthermore, strains IPT58 (Europe) and 47703UT (United States) possess almost identical ospC alleles with only a 1-nt difference in 484 bp analyzed. Both housekeeping gene sequences and IGS sequences of IPT58 cluster with other European strains (Figs. 1, 2, and 3).

Fig. 1.

Bayesian phylogenetic inference of IGS sequences of B. burgdorferi. Posterior probability values of clades are provided. Symbols refer to the major IGS genotypes as defined by Bunikis et al. (41). Samples 21509LT and 22521LT were excluded because of the lack of PCR products. Non-color-coded strains are from the Northeast and Midwest United States. Yellow, strains from California; blue, strains from Europe. The branch length of the outgroup B. garinii is not according to scale (indicated by slashes). (Scale bar: 10% divergence.)

Fig. 2.

Bayesian phylogenetic inference of ospC gene sequences of B. burgdorferi. Posterior probability values of clades are provided. Symbols refer to the major IGS genotypes as defined by Bunikis et al. (41). Samples 21509LT, 22521LT, 20604LT, and 20111LT are missing because of the lack of PCR products. Non-color-coded strains are from the Northeast and Midwest United States. Yellow, strains from California; blue, strains from Europe. (Scale bar: 10% divergence.)

Fig. 3.

Bayesian phylogenetic inference of concatenated sequences of the housekeeping genes (clpA, clpX, nifS, pepX, pyrG, recG, rplB, and uvrA) of B. burgdorferi. Posterior probability values of clades are provided. Symbols refer to the major IGS genotypes as defined by Bunikis et al. (41). Non-color-coded strains are from the Northeast and Midwest United States. Yellow, strains from California; blue, strains from Europe. The branch length of the outgroup B. garinii is not according to scale (indicated by slashes). (Scale bar: 1% divergence.)

Phylogenetic Analyses.

For the concatenated housekeeping genes, trees were constructed by using MrBayes software (53) (Fig. 3) and the neighbor joining method in MEGA 3.1 (54) (Fig. S3A). The trees were rooted with sequences of B. garinii strain PBi from Europe as an outgroup. The two MLST trees obtained by Bayesian inference and the neighbor joining method were highly similar and reasonably well supported at deep branches and at terminal nodes. In contrast to highly recombinogenic bacterial species, such as Streptococcus pneumoniae (55), where the phylogenetic relationships cannot be inferred reliably, a robust intraspecific phylogenetic signal was found to be present in B. burgdorferi. Except for strains NE49, IPT193, IPT198, Z41293, and Z41493, the European strains formed two subclades among North American strains, but, consistent with the geographical distribution of the STs, North American and European populations of B. burgdorferi were placed in distinct terminal nodes. Furthermore, strains from California always branched off from samples derived from the Northeast/Midwest United States (Fig. 3).

The tree based on the IGS locus also revealed patterns of clustering being broadly consistent with geographic source. However, not all spirochetal samples from California and the Northeast/Midwest United States were distinguished in the IGS tree (Fig. 1, i.e., ca92-1096). The topologies of the MLST tree and the IGS tree differ in that several samples representing major IGS genotypes (i.e., IGS genotypes 2, 5, and 6) were divided and placed in different clusters in the MLST tree. A partition homogeneity test (56) using MLST and IGS sequences showed that the sum of the tree length of 99 of 100 random partitions was longer than the sum of the tree length for the original MLST–IGS partition. This confirms that the MLST tree presented here is significantly different from the IGS tree (P = 0.01).

The European samples NE49, IPT193, IPT198, Z41293, and Z41493 branched off deeply from the other strains with high posterior probability in both the IGS tree and the MLST tree (Figs. 1 and 3). This divergence was also found for individual housekeeping genes (Fig. S3 B–I). The basal position of this clade in the MLST tree suggests that these European strains are most closely related to the common ancestor of the strains analyzed in this study.

To test whether the overall differences between the allelic profiles of the European and American strains were due to this diverged European clade, a permutation test was also performed without these strains (i.e., NE49, IPT193, IPT198, Z41293, and Z41493). The results were significant (P < 0.001), showing that the overall difference between the European and American populations was not due to this diverged group (data not shown).

The ospC tree shows a substantially discordant topology compared with the MLST and IGS trees (Fig. 2). For example, the European samples NE49, IPT193, IPT198, Z41293, and Z41493, which cluster together in the MLST tree, are placed in three different clades scattered across the ospC tree. Furthermore, most strains from California were found to cluster together with strains from the Northeast/Midwest United States. Using the ospC sequence of B. garinii as an outgroup resulted in translocation of IPT39 to the base of the tree (Fig. 2).

Discussion

Using a sequence-based approach, this study shows that North American and European B. burgdorferi populations correspond to distinct lineages. Importantly, the study suggests that B. burgdorferi originated in Europe and not in North America as proposed previously based on diversity patterns of ospC (57) or arbitrarily primed PCR and pulsed-field gel electrophoresis (58). Comparison of the phylogenetic trees indicates that the commonly used genetic markers of B. burgdorferi, the IGS locus and ospC, evolve differently compared with the housekeeping genes.

The data presented here using the MLST scheme strongly suggest that it can be used as a powerful means for studying the evolution, epidemiology, and population/landscape genetics of Lyme borreliosis spirochetes. We demonstrate that the approach is capable of (i) unambiguously genotyping B. burgdorferi samples, (ii) establishing evolutionary relationships among the bacterial populations at different levels, (iii) capturing geographic population structure of the agents, and (iv) typing infections directly from tick vectors without the need for prior culturing of the pathogens.

The clinical samples analyzed in this study were chosen to represent all of the nine previously described major IGS genotypes of B. burgdorferi (41). In addition to these clinical samples, the MLST scheme was further validated by using samples of B. burgdorferi isolated from ticks and B. burgdorferi DNA extracted directly from infected ticks collected in different geographic regions. Whereas the geographic origins of clinical isolates are often not possible to determine, isolates from ticks can be referenced geographically, thereby serving as a useful resource for population studies of B. burgdorferi. However, isolation of B. burgdorferi may yield a biased picture toward strains that adapt better to in vitro conditions (59, 60). Therefore, it is most desirable to genotype B. burgdorferi directly from ticks or hosts. In the present study we demonstrate that MLST can be applied equally well to isolates and to ticks infected with B. burgdorferi without prior cultivation of the bacteria.

In this study, different STs were determined for North American and European B. burgdorferi populations, indicating transcontinental diversification. In contrast, no consistent separation of the regional populations was observed at the level of the major IGS genotypes and ospC major groups (Table 1). A signal of geographic resolution similar to that found for the distribution of the STs was captured by the MLST trees, which also discriminated between western and eastern populations of B. burgdorferi from North America. However, the sample size of Californian strains used in the present study was small, and more samples need to be analyzed to confirm this observation.

The finding that the regional populations of B. burgdorferi differ genetically suggests limited migration of the strains between the regions due to barriers. This is corroborated by the distinct distributional ranges of I. scapularis, Ixodes pacificus, and Ixodes ricinus ticks, the principal vectors of Lyme borreliosis in eastern North America, the Pacific region of North America, and Europe, respectively (5). Although no molecular clock for the evolution of the housekeeping genes of B. burgdorferi has yet been established, it is likely that the different populations of B. burgdorferi have been separated for a long time (i.e., probably in the order of magnitude of several million years), because housekeeping genes evolve slowly. It is interesting to note that the European samples NE49, IPT193, IPT198, Z41293, and Z41493 formed the most diverged clade among the B. burgdorferi samples in the MLST tree. The deep branching of this cluster suggests a common ancestor of the B. burgdorferi populations that evolved in Europe. In contrast, based on ospC, it has previously been hypothesized that B. burgdorferi originated in the New World (57, 61). As discussed below, these conflicting scenarios of speciation and origin of B. burgdorferi are likely to be related to different evolutionary pathways of the housekeeping genes and ospC.

The shapes of the MLST, IGS, and ospC trees as determined in this study differ. First, the topologies of the trees are discordant, with the ospC tree being most different. This may indicate that ospC has undergone recombination in the past (61). Second, the ospC tree is characterized by deeper branching than the MLST and IGS trees, suggesting longer-term evolutionary relationships among the ospC major groups or, alternatively, faster evolution. The lack of clear geographic structure of B. burgdorferi in North America when using ospC as a marker is consistent with earlier studies of B. burgdorferi undertaken in the northeastern United States (25) and may be related to the evolution of the ospC major groups in ancestral spirochete populations (61). Balancing selection could have maintained such an ancient polymorphism of ospC in the wake of past population bottlenecks, as proposed previously (25, 43, 61). In two studies, the European strains NE49, Z41293, and Z41493 have been shown to be borderline as far as the species status is concerned (15, 23). While these strains form distinct clades in the MLST and IGS trees showing significantly increased branch length, they were scattered across the ospC tree. The misplacement of these European strains in the ospC tree provides indisputable evidence that ospC has been evolving differently from the IGS locus and the eight housekeeping genes.

Balancing selection is a form of frequency-dependent selection and should result in a dN/dS ratio of >1 for genes under immune selection. However, we and others (43) have found an overall dN/dS ratio of <1 for ospC, a gene encoding an immunodominant outer surface lipoprotein of B. burgdorferi. Sliding window analysis (data not shown) (43) and a “sitewise likelihood-ratio” method (62) showed that different parts of ospC display different dN/dS ratios (Table S5). This indicates that some regions of the gene are under positive immune selection whereas others are more conserved because of functional constraints. This is consistent with recent findings that ospC encodes a lipoprotein that is essential for colonization of the tick's salivary glands (63, 64) and/or early infection of the vertebrate host (65, 66).

Previous epidemiological and theoretical work suggests that clones and clonal complexes, as demarcated by MLST/MLSA of housekeeping genes, are remarkably congruent with pathotypes and ecotypes in several bacterial species (67). Thus, MLST may also have the power to identify ecotypes within B. burgdorferi a priori of determining their ecological niches. This is supported by data from a recent experimental study on the transmissibility of different isolates of B. burgdorferi (i.e., BL206 and B348) to ticks by white-footed mice, an important reservoir host in the northeastern United States (68). However, further transmission studies are needed to clarify whether the STs, clones, or clusters identified within B. burgdorferi by MLST do consistently correspond to ecotypes that are selectively maintained or, alternatively, whether the lineages defined by MLST evolved through genetic drift (3, 36).

A few other studies have used multiple loci to type Lyme borreliosis spirochetes or to delineate new species within this group as an alternative approach to whole DNA–DNA hybridization (15, 17, 41, 44, 45). These schemes combine different categories of loci that are likely to evolve at different rates. However, using loci that evolve at different rates and under different tree structures may pose problems in inferring phylogenetic trees (69, 70). To provide a genotyping method that can also be used to confidently infer evolutionary pathways, we developed a MLST scheme for B. burgdorferi that is based exclusively on housekeeping genes that display an appropriate window of no more than 5% allelic divergence among different strains (71). For relatively clonal bacteria, such as B. burgdorferi (72), it is widely accepted that evolutionary relationships can be inferred by using multiple housekeeping genes (73).

Taken together, the study suggests that B. burgdorferi originated in Europe but that this species has been prevalent in North America for a long time. It is likely that the surfacing of clinical cases of Lyme borreliosis in North America three decades ago represents the reemergence of this tick-borne pathogen out of refuges in which it has been persisting throughout the post-Columbian settlements and the industrial revolution (3). Future landscape genetic studies of B. burgdorferi using MLST as a tool will unveil whether signatures of population bottlenecks, population expansions, and geographic dispersal are present in the housekeeping genes.

Materials and Methods

B. burgdorferi DNA was extracted from cultured isolates and directly from infected ticks. Primers were designed to conserved regions of eight housekeeping genes (Table S6; see also www.mlst.net). These genes as well as the IGS and ospC (6, 40, 41) were amplified by nested or seminested PCR followed by DNA sequencing. Sequences of these loci were analyzed for characteristics such as GC content or dN/dS ratio. Sequences of the housekeeping genes were assigned allele numbers according to the MLST web site hosted at Imperial College London (www.mlst.net), and STs were defined. Phylogenetic trees of the individual and concatenated sequences of the housekeeping genes and individual IGS and ospC sequences were constructed by using MrBayes software (53) and the neighbor joining method (54). To compare the tree topologies, a partition homogeneity test (56), implemented in PAUP 4.0, was applied. A permutation test was performed to analyze the distributions of the allelic profiles of B. burgdorferi from Europe and North America. Details can be found in SI Materials and Methods.

GenBank database accession numbers of all IGS and ospC sequences used in this study are listed in Table S7.