Complex Population Structure of Borrelia burgdorferi in Southeastern and South Central Canada as Revealed by Phylogeographic Analysis

Abstract

Lyme disease, caused by the bacterium Borrelia burgdorferi sensu stricto, is an emerging zoonotic disease in Canada and is vectored by the blacklegged tick, Ixodes scapularis. Here we used Bayesian analyses of sequence types (STs), determined by multilocus sequence typing (MLST), to investigate the phylogeography of B. burgdorferi populations in southern Canada and the United States by analyzing MLST data from 564 B. burgdorferi-positive samples collected during surveillance. A total of 107 Canadian samples from field sites were characterized as part of this study, and these data were combined with existing MLST data for samples from the United States and Canada. Only 17% of STs were common between both countries, while 49% occurred only in the United States, and 34% occurred only in Canada. However, STs in southeastern Ontario and southwestern Quebec were typically identical to those in the northeastern United States, suggesting a recent introduction into this region from the United States. In contrast, STs in other locations in Canada (the Maritimes; Long Point, Ontario; and southeastern Manitoba) were frequently unique to those locations but were putative descendants of STs previously found in the United States. The picture in Canada is consistent with relatively recent introductions from multiple refugial populations in the United States. These data thus point to a geographic pattern of populations of B. burgdorferi in North America that may be more complex than simply comprising northeastern, midwestern, and Californian groups. We speculate that this reflects the complex ecology and spatial distribution of key reservoir hosts.

INTRODUCTION

Lyme disease, caused by the spirochete Borrelia burgdorferi sensu stricto (henceforth called B. burgdorferi), is continuing to emerge in the United States and is now emerging in southeastern and south central Canada due to the northward expansion of the range of the tick vector Ixodes scapularis (1). Recent studies have emphasized the scope of genetic diversity of B. burgdorferi (2–4). For members of the B. burgdorferi sensu lato complex (to which B. burgdorferi belongs), diversity is likely to reflect a combination of historic patterns of geographic dispersion and associations or coevolution with different reservoir host species (4, 5). There are three recognized risk areas for Lyme disease in the United States: the Northeast (NE), the upper Midwest (MW), and the West (particularly California). Genetic differentiation of B. burgdorferi from these three regions has been noted and is likely due to geographic isolation by landscape features (4, 6–9). These populations have undergone recent expansions, but the available evidence also points to phylogenetically deeper and more complex patterns of expansion, contraction, and population mixing in the ancient past. Nevertheless, to date, studies have suggested an apparent spread from the Northeast through the Midwest to California (4, 6).

In order to recreate how B. burgdorferi populations in North America have expanded and contracted in the recent and deep past, it is necessary to first generate data concerning contemporary phylogeographic patterns. These data could then give insight into the ecological conditions underlying current population expansions, which in turn can help in the development of new management strategies. Given that past rapid climate changes are thought to have been key drivers of changes in population size and gene flow mediated by population movements (10) and that current rapid warming is thought to be driving range changes of I. scapularis and B. burgdorferi (11, 12), it is important to understand how current climate change and range spread may impact the diversity of B. burgdorferi.

Study of the diversity of B. burgdorferi in North America is of immediate diagnostic and clinical utility, with the recognition that the consequences of infection (mild self-limiting cutaneous or severe systemically disseminated infections) may differ with the infecting strain (13), and the strain may interact with patient genetic heterogeneity in determining clinical outcomes (14). Furthermore, strains vary in their capacity to elicit antibody responses in early infection that are detectable by current gold-standard serological tests (15). Recent studies of B. burgdorferi diversity by multilocus sequence typing (MLST), which uses housekeeping genes with neutral variation, suggested that lineages determined from these housekeeping genes predicted pathogenicity better than outer surface proteins expressed at the point of infection (13). This raises the hypothesis that pathogenicity in humans is a B. burgdorferi phenotype with possible origins in adaptation to different host species and geographic locations and may therefore be predictable.

We have previously analyzed, by MLST, the diversity of B. burgdorferi detected in ticks, which were collected from humans and domesticated pets in a passive tick surveillance program in Canada, and compared this diversity with that found in the United States (4, 16). This analysis suggested that the general longitudinal pattern of differentiation of strains occurring in the northeastern versus the upper midwestern United States is reflected in strains seen in Canada, with some possible skewing of diversity due to founder events as B. burgdorferi invades (16, 17). However, while many of the ticks collected in this study were likely from Canada-resident I. scapularis populations, some were likely not, having been dispersed from the United States by migratory birds (4, 16). To get a better picture of the strain structure in Canada-resident B. burgdorferi transmission cycles, we performed MLST analysis of B. burgdorferi in tick samples collected during active field surveillance in locations where B. burgdorferi is known to be locally transmitted by self-sustaining, reproducing I. scapularis populations in Canada and compared these sequences to those previously obtained in the United States.

MATERIALS AND METHODS

Samples used in the study.

The 107 Canadian samples characterized by MLST as part of this study were B. burgdorferi positive by PCR (16) and were recovered from ticks obtained by drag sampling in areas in Canada where B. burgdorferi is endemic or by collection from wild rodent hosts. Areas of endemicity are defined as locations where B. burgdorferi is being transmitted among wild-animal reservoirs by reproducing populations of I. scapularis ticks (Table 1 and Fig. 1). The methodology of field sample collection was previously described (3). The samples were collected mostly contemporaneously (2006 for samples from the Maritime Provinces [MR] and Manitoba [MB], 2007 to 2010 for samples from Quebec, and 2010 for samples from eastern Ontario), but the oldest samples were those archived from collections conducted in 2001 at Long Point, Ontario (ONLP). DNA was extracted from ticks and screened for B. burgdorferi infection by PCR, as previously described (18). PCR-positive ticks were then used for MLST analysis, as previously described (19). Briefly, MLST was conducted by nested PCR for each of the eight housekeeping genes (clpA, clpX, nifS, pepX, pyrG, recG, rplB, and uvrA), using HotStarTaq (Qiagen, Germany) as previously described (3). PCR fragments were sequenced in the forward and reverse directions and manually compared by using DNASTAR (Lasergene). Sequences of new sequence types (STs) and new alleles were submitted to (and are available from) the mlst.net database (http://borrelia.mlst.net/). To reduce the likelihood of amplification of sequences from different strains coinfecting samples, we prescreened samples for mixed infections by amplifying the chromosomal rrs-rrlA (16S-23S) intergenic spacer (IGS) region, as previously described (3), and then sequencing the amplicons. Any rrs-rrlA amplicons that revealed ambiguous (i.e., two or more equally plausible bases at one position) bases or sequences upon examination of sequence traces were considered to suggest that the samples contained possible mixed infections and were not subject to MLST analysis. The rrs-rrlA sequence was chosen for this purpose because it is one of the most variable sequences of B. burgdorferi used for strain analysis (20). Also, any samples that yielded any other ambiguous sequences of the housekeeping genes were not used for analysis. This led to the rejection of 60 samples (48 on the basis of IGS sequences and 12 on the basis of housekeeping gene sequences).

TABLE 1.

Field sites in Canada where ticks were collected from the environment or captured rodents at locations where tick populations have been established

Province	Location	No. of sites	No. of samples	Type of samplea
MB	Lake of the Woods, northwest shore	1	32	Questing ticks (19 AM, 13 AF)
ON	Long Point Provincial Park	1	7	Questing ticks (1 AM, 6 AF)
ON	Thousand Islands region	1	15	Questing ticks (4 AM, 4 AF, 7 N)
QC	Sites in Montérégie	11	35	16 questing ticks (4 AM, 11 AF, 1 N)
				12 ticks from rodents (9 N, 3 L)
				7 ticks from deer (1 AM, 6 AF)
NS	Lunenburg	1	22	15 questing ticks (10 AM, 3 AF, 2 N)
				7 ticks from rodents (7 L)

^aAM, adult male; AF, adult female; N, nymph; L, larva.

FIG 1.

Locations where B. burgdorferi samples used in this study were collected and the geographic distribution of different MLST STs of B. burgdorferi used in this study. The colored points indicate locations where all samples analyzed in the study were collected, while red arrows indicate the locations of field sites where new samples in this study were obtained. The different-colored points correspond to STs found in different geographic regions. Colors: cyan, STs found only in the Maritimes (MR); orange, STs found only at Long Point, Ontario (ONLP); brown, STs found only in Manitoba (MB); blue, STs found across the Northeast (including Quebec, the Thousand Islands region of Ontario, the Maritimes, and the northeastern United States) (NE); green, STs found in the midwestern United States (MW); yellow, STs found in northeastern and midwestern locations of the United States and Canada (NE+MW); red, STs found only in California. (Maps were created using ArcGIS 10.1.)

For many analyses, an additional 4 samples collected during field surveillance in Quebec (3) and 20 samples collected during passive tick surveillance in Canada were also used because they carried novel STs that have to date been found only in Canada (see Table 1 in reference 4). The sequences of these samples were obtained from the mlst.net database.

For phylogenetic analyses, Canadian samples were combined with an additional 453 samples from the United States (all that were available at the time of analysis), the sequences of which were also obtained from the mlst.net database (http://borrelia.mlst.net/).

Nucleotide sequence analysis.

New sequences obtained in Canada were compiled in Lasergene (DNAStar, Madison, WI, USA), and these STs and their individual alleles were compared against existing STs and alleles in the mlst.net database by using the Single Locus Sequence Query and Allelic Profile Query functions. New alleles and STs were allocated identification numbers according to the protocol of the mlst.net database. The geographical distribution of STs was visualized using ArcGIS version 10.2 (ESRI). Population diversity (number of STs per sample) in each study site (in Canada) or region (in the United States) was calculated by dividing the number of individual STs at a site or region by the number of samples collected at that site/region. The diversity of STs in different sites and regions was identified in this process for comparison of populations and mapping.

Regarding the frequency of any new STs discovered in Canadian locations/regions, our null hypothesis was that the prevalence of newly discovered STs in the Canadian samples was not significantly different from the upper 95% confidence interval (CI) for the prevalence in the regions of the United States to the south that are the most likely source populations for B. burgdorferi carried northwards by migratory birds or other hosts (the upper Midwest for samples from Manitoba and Long Point and the Northeast for samples from eastern Ontario, Quebec, and the Maritimes). Therefore, the prevalence of any novel STs in Canadian locations was compared to the prevalence in the NE (0/363; 95% CI = 0 to 0.01) or the MW (0/62; 95% CI = 0 to 0.06) by Fisher's exact test.

Genetic diversity, phylogenetic relationships, and population structure.

We analyzed the genetic diversity and population structure of B. burgdorferi strains from Canada and compared the results from Canadian samples with those from samples from the United States, using a number of analyses. Phylogenetic relationships were reconstructed by using MrBayes v3.2.1 software (21), in which Markov chain Monte Carlo (MCMC) samplings were run for 500,000 generations, with trees being sampled every 1,000th generation (22).

Pairwise F_ST values were calculated for the B. burgdorferi populations in different regions/locations by using the ARLEQUIN 3.1 program (23), with 100 permutations being run to assess the significance of the F_ST value. The level of significance was altered from a P value of <0.05 by Bonferroni correction to a P value of <0.001 to account for multiple pairwise comparisons.

Allelic profiles were analyzed by using eBURST (24) and global optimal eBURST (goeBURST) (25). eBURST is based on a simple model of clonal expansion and divergence and provides a convenient method to establish relationships of descent for bacterial populations. goeBURST allows a global optimization procedure (instead of local optimization), an extended set of tiebreak rules, and improved graphical representation of clonal complexes, including double-locus variants (DLVs) and triple-locus variants (TLVs). Both algorithms are tailored for the use of MLST data and cluster STs as disjointed tree collections based on a set of hierarchical rules related to the number of single-locus variants (SLVs), DLVs (eBURST), and TLVs (goeBURST). The minimum number of identical loci for group definition was set to 5, and the minimum count of SLVs for subgroup definition was set to 0. The same samples were used in goeBURST to obtain a graphical display of clonal complexes. The Minimum Spanning Tree extension of PHYLOVIZ V1.0 (26) was used to visualize the possible evolutionary relationship between STs according to their allelic profiles in the goeBURST diagram. A bootstrap procedure implemented in eBURST gave statistical confidence to the assignment of clonal complex founders, which were inferred as the ST within a clonal complex that had the highest number of single-locus variants.

The population structure of the different STs identified across the United States and Canada was computed with Bayesian Analysis of Population Structure (BAPS) version 6.0 (27), using clustering with a linked locus module and codon model as recommended for MLST data. In this process, mixture analysis was performed with K values from 2 to 20, and optimal partitions were identified by the maximum log marginal likelihood value.

Investigation of the degree of recombination among B. burgdorferi strains and admixture among populations.

The relative contribution of recombination (r) and mutation (m) to variation among the sequences was estimated by the r/m ratio calculated with ClonalFrame software v1.1 (28). The r/m value was obtained with 50,000 burn-in iterations, followed by 50,000 MCMC iterations and a thinning interval of 100 iterations before recording the parameter values for the posterior sample. The initial value for m was Watterson's theta value calculated for the sample by using DnaSP5 (29).

To estimate the contribution of admixture to genetic variation among BAPS groups, admixture analysis was conducted with BAPS 6.0 using the following parameters: a minimum population size of 3, 100 iterations used to estimate the admixture coefficient for individuals, 200 reference individuals from each population, and 20 iterations used to estimate the admixture coefficient for reference individuals. Gene flow among the populations was plotted in BAPS 6.0, which uses a model-based representation of the molecular variability of populations and their affinities toward each other (30).

RESULTS

Nucleotide sequence analysis.

A total of 131 samples from Canada were used in the analysis, of which 111 were collected from field sites where B. burgdorferi is now endemic. Sequences of four of these samples collected in the field were available from a previous study (3). The 107 samples characterized by MLST as part of this study comprise 39 STs, 21 of which were novel (new STs were assigned the numbers 225 and 519 to 538) (see File S1 in the supplemental material). Two novel alleles were identified, both from ticks collected at Long Point, Ontario. These alleles corresponded to recG allele 167 for ST522 and clpA allele 182 for ST524. There were 20 samples from Canada collected during passive surveillance that had STs found only in Canada.

Genetic diversity and geographic distribution of STs.

The 564 B. burgdorferi samples used for analyses in this study were divided into 111 STs. Of the STs already in the mlst.net database, 18 STs were unique to California, 27 STs were unique to the Midwest (6 of which have been found in “midwestern” Canada [14]), 29 STs were unique to but widespread in the Northeast (including 14 found in eastern Canada, from eastern Ontario to the Maritimes [14]), and 6 STs occurred in both the Midwest and the Northeast (Table 2 and Fig. 1). In the samples characterized for the first time in this study, 39 STs were found. Twenty-one STs were unique to Canada, of which 4 STs were found only in the Maritimes (among 22 samples from Lunenburg, Nova Scotia); 5 STs were found only at Long Point, Ontario (from 7 samples); and 11 STs occurred only at the site in southeastern Manitoba (from 32 samples). In contrast, only one new ST was found in the samples from southern Quebec (from 35 samples), and no new STs were found in the samples from the Ontario Thousand Islands site (from 15 samples). In total, including STs identified in previous studies, 54 STs are unique to the United States, 38 STs are unique to Canada, and only 19 STs are common to both countries. Thus, there were STs that occur across wide regions (those occurring in California and those occurring across the northeastern United States, the upper midwestern United States, or both the northeastern and upper midwestern United States and Canada) and STs that that to date have been found only in specific sites in Canada (those at Long Point, Ontario; Lunenburg, Nova Scotia; and the field site in Manitoba) (Fig. 1 [note that this map also shows Canada-specific STs from samples obtained during passive surveillance]).

TABLE 2.

New and total STs in samples collected from sites in Canada^a

Site of sample collection	Total no. of samples	Total no. of sites	Total no. of STsb	No. of STs per sample	No. of STs unique to location	No. (proportion) of samples carrying unique STs	No. of unique STs per sample
Field sites in Canada
MR	22	1	12	0.54	4	4 (0.19)†	0.18
QC	35	11	10	0.29	1	1 (0.03)	0.03
ONTH	15	1	7	0.47	0	0	0
QCTH	50	14	12	0.24	1	1 (0.02)	0.02
“NE” Canada	72	15	17 (8, 3, 0)	0.24	5	5 (0.07)	0.07
ONLP	7	1	7	1	5	5 (0.71)†	0.71
MB	32	1	19	0.59	11	13 (0.41)†	0.34
“MW” Canada	39	2	24 (0, 1, 6)	0.61	16	18 (0.46)	0.41
Sites in USA from mlst.net database
NE	363	30	29	0.08
MW	62	23	30	0.48
California	28	23	18	0.64

^aData on STs from the United States already in the mlst.net database are shown by geographic region for comparison. MR, Atlantic Maritime Provinces; QC, Quebec; ONTH, eastern Ontario's Thousand Islands; QCTH, Quebec and Thousand Islands combined; ONLP, Long Point, Ontario; MB, Manitoba; NE, northeastern United States; MW, midwestern United States. “NE” Canada comprises data from the Atlantic Maritime Provinces, Quebec, and eastern Ontario's Thousand Islands combined, while “MW” Canada comprises data from Long Point, Ontario, and Manitoba combined. † indicates that the prevalence of samples carrying unique STs was significantly (P < 0.001) greater than the upper 95% confidence interval for their possible prevalence in source locations in the United States.

^bNumbers in parentheses indicate the numbers of STs that were previously found in the northeastern, the midwestern, and both the northeastern and midwestern United States, respectively.

In the samples from the United States, the level of ST diversity (the number of STs per sample) was higher in the Californian samples (0.64; 18/28 samples) than in samples from the upper Midwest (0.48; 30/62) and lowest in samples from the Northeast (0.08; 29/363) (Table 2). Canadian samples from Manitoba and Long Point in Ontario, locations which correspond to the same longitude as the upper midwestern United States, also showed a higher number of STs/sample (0.61; 24/39) than did samples from Ontario's Thousand Islands, Quebec, and Lunenburg, Nova Scotia (0.24; 17/72), which correspond to the same longitude as the northeastern United States (Table 2). STs from Lunenburg, Nova Scotia, were, however, more diverse than the combined samples from eastern Ontario and southern Quebec, with 0.54 and 0.24 STs per sample, respectively (Table 2). The prevalence of STs unique to the Manitoba site and to Long Point, Ontario (13/32 and 5/7 samples, respectively), was significantly higher than the upper 95% CI for their possible prevalence in the upper midwestern United States (P < 0.001 for both). The prevalence of STs unique to Lunenburg, Nova Scotia (4/22 samples), was significantly higher than the upper 95% CI for their possible prevalence in the northeastern United States (P < 0.001). However, the prevalence of the ST unique to Quebec (1/35) was not significantly higher than the upper 95% CI for its possible prevalence in the northeastern United States (P > 0.08).

STs previously recorded in the northeastern United States that were also found in Canadian sites were found only in eastern Canada (from Thousand Islands, Ontario, eastwards). STs previously recorded in the upper midwestern United States that were also found in Canada were found only in the more western Canadian sites (Long Point, Ontario, and the Manitoba site). On the basis of these observations of differences in STs among sites and regions, below, we keep the notation NE for STs found in the northeastern United States (although some STs that are found in the northeastern United States are also found in eastern parts of Canada) and MW for STs found in the upper midwestern United States (although some STs that are found in the upper midwestern United States are also found in western Ontario and Manitoba). For Canada, we consider the following sites/regions as comprising different ranges of STs: the site in Manitoba (MB), Long Point, Ontario (ONLP), and the Maritimes (MR). Given the similarity in STs and in their diversity, STs from the Thousand Islands region of Ontario and the sites in southern Quebec were considered one group, QCTH.

Phylogenetic relationships among STs.

In the phylogenetic tree, each clade frequently comprised members of each broad geographic region of North America (Fig. 2), suggesting ancient genetic signatures. For example, while the STs from California remained absent in all other regions, some of them (ST2 and ST403) are closely related genetically to ST1 from the northeastern United States and are in the same clade. New MB, ONLP, and MR STs were spread over different clades in the phylogeny. The new QCTH ST (ST519 from sites in Quebec) was most closely related to ST316 from, and unique to, MR (Fig. 2). Pairwise F_ST values (Table 3) supported moderate genetic differentiation and population structuring among the different geographic regions in the United States, and also, MB STs showed the highest value (F_ST = 0.19164) for genetic differentiation from the NE STs (Table 3).

FIG 2.

Bayesian phylogenetic tree for the 111 STs of B. burgdorferi. STs are color coded according to their geographic location. Cyan, Maritime Provinces; orange, Long Point, Ontario; brown, Manitoba; blue, northeastern United States, eastern Ontario, and southwestern Quebec; green, midwestern United States; yellow, STs found in both the northeastern and midwestern United States; red, California. Outgroups used to root the tree were Borrelia californiensis, B. andersonii, and B. bissettii. Posterior probabilities of >70% are shown beside the nodes. The bar corresponds to the number of substitutions per unit branch length. Membership of STs in one of 13 elucidated BAPS groups is also indicated on the tree.

TABLE 3.

Matrix of pairwise F_ST values of the STs in different geographic regions of North America^a

Region	Pairwise FST value
	NE	MW	QCTH	MR	LP	MB	California
NE	0
MW	0.10094
QCTH	0.08202	0.03131
MR	0.05961	0.02552	0
ONLP	0.07614	0.01173	0.00618	0.02804
MB	0.19164	0.0652	0.10063	0.08337	0.07865
California	0.0758	0.06103	0	0.02246	0.03457	0.08927	0

^aValues in boldface type are significant at a threshold for significance of α = 0.0011.

B. burgdorferi population structure.

The goeBURST algorithm revealed that the B. burgdorferi STs are divided into 18 clonal complexes when only SLVs were included or 16 clonal complexes when DLVs were included as well as 45 singletons (Fig. 3). Except for ST403, ST2, and ST13, which formed a clonal complex with STs from outside California (ST1 and ST12), the remaining Californian STs formed clonal complexes with local STs, or they were singletons. While novel STs from MR were unique to that location, they were mostly SLVs, DLVs, or TLVs of STs found in the northeastern United States. Nearly all STs from field sites and passive surveillance in QCTH were the same as or closely related (SLV) to NE STs. The exception, ST519, was a DLV of an ST found to date only in the Maritimes (see the section above). STs from ONLP formed clonal complexes mostly with NE STs (Fig. 3) but appeared to have ancestry from both NE and MW STs, being SLVs of STs found in the northeastern United States and DLVs or TLVs of STs found in the midwestern United States. STs from MB were divergent from MW STs but were most frequently SLVs, DLVs, or TLVs of STs found in the midwestern United States. While the MB, MR, and ONLP STs diverged from U.S. STs and those from QCTH, they were most closely related to, and often had likely ancestors in, STs found immediately to the south: MB with the MW STs and MR with the NE STs. However, STs from ONLP, which lies on longitude 80°W, which nowadays separates NE and MW populations (14), had elements of relation and ancestry to both MW and NE STs. Inferred founders with >60% bootstrap support were STs from the northeastern United States for two clonal complexes; other inferred founder (or possible founder) STs had <60% bootstrap support, and these also included the NE STs, the MW STs, and both the NE and MW STs (Fig. 3).

FIG 3.

goeBURST network of the 111 STs of B. burgdorferi used or obtained in this study. STs are color coded according to their geographic location. Blue, northeastern United States, Quebec, the Thousand Islands region of Ontario, and the Maritimes; green, the Midwest; yellow, STs occurring in both the Northeast and the Midwest; red, California; cyan, STs occurring only in the Maritimes; orange, STs occurring only at Long Point, Ontario; brown, STs occurring only in Manitoba. Colored lines connecting STs in the network indicate the phylogenetic links between STs and the degree of support: black lines are inferred without tiebreak rules, blue lines are inferred by using tiebreak rule 1 (SLV), yellow lines are inferred by using tiebreaking at the frequency of the loci, and light gray lines are inferred by using tiebreak rule 2 (DLV). Inferred founder STs with >60% bootstrap support are circled in red, and potential founders with 35 to 60% support are circled in orange. TLVs are indicated by dashed lines. The population structure obtained by BAPS analysis is indicated as circles surrounding the clonal complexes and singleton STs.

BAPS analysis best supported the existence of 13 subpopulations for which the log marginal likelihood value was highest. These subpopulations were not geographically structured, which is consistent with the lack of geographic structuring on the basis of phylogeny. This often suggests an ancient population structure and/or relatively high migration rates, although the latter is unlikely for B. burgdorferi in North America (4). BAPS groups 5 and 7 had STs from all seven geographic locations and contained 13 and 17 STs, respectively. BAPS groups 1, 4, and 11 contained STs from 6 geographic regions with 13, 9, and 15 STs, respectively. Groups 6 and 10 had STs from 4 geographic regions and contained 8 and 5 STs, respectively. Groups 2, 8, and 12 contained 3, 6, and 6 STs, respectively, which occurred in 2 to 3 geographic regions. BAPS group population 13 comprised only STs from California (ST398 and ST399) (Fig. 2 and 3). There was general concordance between BAPS groups and clonal complexes revealed by goeBURST. However, two BAPS groups contained members from more than one clonal complex, and singletons were divided among BAPS groups (Fig. 3). Admixture analysis by BAPS revealed significant admixture (P < 0.05).

Investigation of the degree of recombination and admixture analysis.

The r/m value estimated for the sequences was 0.0162 (95% credibility interval, 0.0002 to 0.1497) when the initial m (i.e., Watterson's theta estimated in DnaSP5) was 28.64. This indicated a very low contribution of recombination to variation among sequences (31). Where recombination was found by admixture analysis, it occurred mostly (90% or more) within BAPS groups, as represented in the network by self-looping arrows (Fig. 4).

FIG 4.

Gene flow occurring between different identified populations (BAPS groups), computed by using BAPS 6.0 software. The arrows indicate the direction of gene flow, and the value accompanying each arrow represents the estimated average levels of DNA transition as relative gene flow weights among two or more populations. In this network, only significant admixture results (P < 0.05) are shown.

DISCUSSION

Here we analyzed the STs of B. burgdorferi occurring in locations in Canada where I. scapularis tick populations are known to have become established (mostly in recent years), and this is the first cataloguing of STs from these locations. We had expected that we would find a range of STs in each site that originated from source populations in the United States directly to the south (having been carried in by northward-migrating birds or other hosts), with perhaps some skewing of the frequencies due to founder events (17). We did find STs that had been found previously in the United States in each site, and indeed, these STs have been found mostly in locations directly to the south of the Canadian sites, with those in MB and ONLP having been found in the upper midwestern United States and those in sites further east in Canada having been found in the northeastern United States. However, surprisingly, we found that in the whole North American data set, only approximately one-fifth of STs were common to both countries. One-half of the STs occurred only in the United States, and about one-third occurred only in Canada. Below, we discuss how our findings may improve our understanding of the phylogeography of B. burgdorferi in North America and, in the light of our analyses, raise hypotheses for the phylogeographic pattern observed in Canada.

Our study advances the understanding of the phylogeography of North American B. burgdorferi in general through (i) phylogenetic, goeBURST, and BAPS analyses; (ii) estimation of F_ST values among populations; (iii) reassessment of the origin of inferred clonal complex founders; and (iv) investigation of the small amount of variation among the housekeeping genes that was due to recombination. Since the description of B. burgdorferi in the northeastern United States in the 1980s (32), a clear pattern of geographic distribution of vector ticks, Lyme disease cases, and B. burgdorferi has been observed, with conspicuous gaps occurring between the Northeast and Midwest (in the region of Ohio) and California (33–35). This pattern of apparent population structure has been attributed to geographic and other landscape barriers (4). Previous evaluations suggested that there have been multiple continent-wide population expansions (and presumably contractions) with local within-region population expansions in recent time (4, 6). This analysis confirms previous analyses showing that geographical patterns of ST occurrence in Canada and the United States are not represented in clades of the phylogenetic tree, the membership of clonal complexes by goeBURST analysis, and the membership of BAPS population groups: these groups contain STs from multiple geographic locations in most cases. Therefore, the underlying, ancestral genetic pattern is not geographically defined. If clades, clonal complexes, and BAPS groups of North American B. burgdorferi strains are not defined geographically (and assuming that mutations in different locations do not represent homoplasy), then perhaps they are defined ecologically. Previously, it was suggested that clonal complexes represent population expansions associated with introductions from Europe, with founder STs originating in the northeastern United States (6). However, in this study, potential founders were also STs that occurred in both the Northeast and Midwest, suggesting that the origin of the multiple population expansions may be less clearly associated with introductions. An alternative hypothesis for the occurrence of clades and clonal complexes is that they represent broad associations with reservoir host species abundant at the time of past expansions and whose descendants persist today. There is recent evidence that some B. burgdorferi strains may be more efficiently transmitted by some host species (16, 36–38), although there is no complete host specialization as there is for B. burgdorferi sensu lato species in Europe (39), and B. burgdorferi in North America remains a host generalist (40). Even so, F_ST values for comparisons between the northeastern, midwestern, and Californian populations were lower than those among rodent-specialist Borrelia afzelii populations in western Europe, which have a limited capacity for spatial mixing. However, these values were significantly greater than zero, which is not the case for the bird specialist Borrelia garinii in Europe, likely due to the considerable capacity for spatial mixing of bird-borne B. garinii populations (41). Thus, perhaps the North American populations show a pattern of spatial mixing that is at least partly driven by terrestrial host dispersions. Furthermore, we found that part of the genetic variation was associated with within-population horizontal gene transfer and recombination, as detected in previous studies (42). This has been observed to occur during infections of reservoir hosts (43). Almost all of the recombination occurred within BAPS groups, and this finding may support the idea that some degree of host association has driven the North American phylogeographic picture: both donor and recipient strains must have been capable of infecting the same individual host for a recombination event to occur. Further prospective studies to seek associations between host species and particular STs, clonal complexes, or clades are needed to support this hypothesis. Other hypotheses for the occurrence of the clades could, however, include population expansions and contractions associated with past climate changes (i.e., glacial and interglacial conditions [44]) and mass extinctions of vertebrate hosts (45).

Clustering analysis using goeBURST showed that STs unique to Canadian locations most likely had common ancestors with STs in U.S. regions immediately to the south of where they were found. A number of findings suggested that the B. burgdorferi populations in some of the Canadian locations (MB, ONLP, and MR) had characteristics that made them distinct from the known U.S. populations. First, the high proportion of unique STs in these Canadian locations, and the proportions of samples in each location that carried unique STs, precluded the idea that novel STs were found merely by chance due to extra field sampling effort increasing the likelihood of finding additional rare STs. Second, F_ST values suggested moderate differentiation of the MB B. burgdorferi population from northeastern U.S. populations, and the F_ST value was higher than that for comparisons between U.S. populations for which barriers to gene flow have already been identified (4). Third, the ST diversity (as revealed by the number of STs/sample) in Canadian sites was greater than that in the corresponding regions (those to the south of them) in the United States. Together, these analyses suggest that it may not currently be possible to imply simple processes of invasion of strains occurring in MB, ONLP, and MR from the currently known B. burgdorferi populations in the northeastern and upper midwestern regions of the United States. In contrast, however, STs from southern Quebec and eastern Ontario were almost all the same as those known to occur in the northeastern United States, which suggests that B. burgdorferi strains in these regions are direct invaders from the northeastern United States.

One hypothesis for the ST diversity observed in our study is that the MLST method used produces spurious STs by random or unpredictable amplification by PCR of different loci in samples carrying mixed-strain infections that were not detected by examination of traces. Although this cannot be ruled out in all cases, a prospective study showed that this was a very unlikely cause of the occurrence of new STs that were recombinations of previously identified alleles of MLST loci (see Files S2 and S3 in the supplemental material). A second hypothesis is that the B. burgdorferi populations in the Maritimes and western Ontario and Manitoba comprise refugial populations. From previous studies on refugial populations in other species, the expected observation would be that refugial populations comprise divergent strains that share one or a few ancestors (46). While there was some evidence of this pattern in the Manitoba site (BAPS group G1) (Fig. 3), it was clear that STs in each location (MB, ONLP, and MR) come from multiple populations and multiple parts of the phylogenetic tree and are derived from different ancestors. In fact, the diversity of strains in the Manitoba site was among the greatest of any location/region. Furthermore, while the I. scapularis population in Long Point could be refugial (47), local history suggests that ticks were only recently introduced into the sites in Manitoba and Nova Scotia (L. R. Lindsay, unpublished data). Therefore, the pattern of STs seen here would be more consistent with B. burgdorferi populations in these locations having been recently introduced from multiple populations in the United States. If so, then as the STs have not been discovered in the United States, these STs may have originated in refugial source populations in the United States close to or bordering Canada where the ecology and diversity of B. burgdorferi have only recently begun to be explored (e.g., see references 48 and 49). This in turn suggests that the population structure of B. burgdorferi in North America is more complex than currently thought. We speculate that such a complex pattern may have arisen due to population expansions from postglacial refugia in northern regions of the United States that occurred with landscape change over the last century (50), combined with complex patterns of spatial mixing of B. burgdorferi strains. The equally complex phylogeographic patterns of key reservoir hosts such as Peromyscus species (51, 52) may have promoted differentiation in refugia that is now being amplified as changing ecological conditions increasingly support expansions of B. burgdorferi populations (53). The temporal aspects of sample collection, which spanned 2006 to 2010 for most sites but 2001 for one site, were not addressed in our study because of the assumption that such short time scales would not normally impact interpretations of sequences subject to the very low rates of mutation and recombination expected of housekeeping genes (19). However, it would be prudent in the future to investigate this assumption and explore rates of mutation and recombination in zones of emergence.

The main significance of our findings for Canada is that many STs at the western and eastern edges of the range of I. scapularis are different from those in the United States, while STs in Quebec and eastern Ontario are mostly the same as those already found in the northeastern United States. The ecological origins and consequences for pathogenicity of this pattern need to be further investigated. Our conclusion at present is that the apparent differentiation of populations of B. burgdorferi in Canada is most likely due to the importation of STs from refugia further south in the United States that have not been explored to date. However, further elucidation of the phylogeography of B. burgdorferi in North America and Canada, and of its ecological drivers, awaits more comprehensive and wider coverage of field sampling, culture, and isolation of strains; detailed analysis by whole-genome sequencing; and a better understanding of the mechanisms of B. burgdorferi dispersion.