Geographic Uniformity of the Lyme Disease Spirochete (Borrelia burgdorferi) and Its Shared History With Tick Vector (Ixodes scapularis) in the Northeastern United States

Corresponding author: Daniel E. Dykhuizen, State University of New York, Stony Brook, NY 11794-5245., dandyk{at}life.bio.sunysb.edu (E-mail)

Communicating editor: S. YOKOYAMA

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Over 80% of reported cases of Lyme disease in the United States occur in coastal regions of northeastern and mid-Atlantic states. The genetic structure of the Lyme disease spirochete (Borrelia burgdorferi) and its main tick vector (Ixodes scapularis) was studied concurrently and comparatively by sampling natural populations of I. scapularis ticks along the East Coast from 1996 to 1998. Borrelia is genetically highly diverse at the outer surface protein ospC. Since Borrelia is highly clonal, the ospC alleles can be used to define clones. A newly designed reverse line blotting (RLB) assay shows that up to 10 Borrelia clones can infect a single tick. The clone frequencies in Borrelia populations are the same across the Northeast. On the other hand, I. scapularis populations show strong regional divergence (among northeastern, mid-Atlantic, and southern states) as well as local differentiation. The high genetic diversity within Borrelia populations and the disparity in the genetic structure between Borrelia and its tick vector are likely consequences of strong balancing selection on local Borrelia clones. Demographically, both Borrelia and I. scapularis populations in the Northeast show the characteristics of a species that has recently expanded from a population bottleneck. Major geological and ecological events, such as the last glacial maximum (18,000 years ago) and the modern-day expansion of tick habitats, are likely causes of the observed "founder effects" for the two organisms in the Northeast. We therefore conclude that the genetic structure of B. burgdorferi has been intimately shaped by the natural history of its main vector, the northern lineage of I. scapularis ticks.

THE evolutionary history of a species is in essence a coevolutionary process of species interactions (THOMPSON 1999 ) and this is particularly true of obligatory parasites (PRICE 1980 ). Our study infers the evolutionary history and population dynamics of Borrelia burgdorferi, the agent of Lyme disease, and its main tick vector, Ixodes scapularis (Acari: Ixodidae), from a comparative analysis of biogeographic patterns of the two species along the East Coast of the United States.

Lyme disease (also called Lyme borreliosis; NADELMAN and WORMSER 1998 ) is the most prevalent arthropod-borne disease in Europe and the United States (BARANTON et al. 1998 ; ORLOSKI et al. 2000 ). In the United States, >80% of Lyme disease cases occurred in the northeastern and mid-Atlantic seaboards (ORLOSKI et al. 2000 ). B. burgdorferi (sensu stricto), one of 10 recognized genomic species in the species complex B. burgdorferi sensu lato distributed globally (BARANTON et al. 1992 ; CANICA et al. 1993 ; KAWABATA et al. 1993 ; BALMELLI and PIFFARETTI 1996 ; LE FLECHE et al. 1997 ; WANG et al. 1997 ; POSTIC et al. 1998 ), is the predominant pathogenic species in North America (MATHIESEN et al. 1997 ; RAS et al. 1997 ). The blacklegged tick I. scapularis (KEIRANS et al. 1996 ) is the primary vector of B. burgdorferi in North America (DENNIS et al. 1998 ). The same tick species is also a vector of many other tick-borne diseases such as human babesiosis (SPIELMAN et al. 1985 ), human granulocytic and monocytic ehrlichiosis (SCHAUBER et al. 1998 ), and a tick-borne encephalitis (TELFORD et al. 1997 ).

The genome of B. burgdorferi has been sequenced (FRASER et al. 1997 ; CASJENS et al. 2000 ) and shown to comprise a linear chromosome of 0.91 Mb and at least 21 linear and circular plasmids with combined size of >0.5 Mb. B. burgdorferi is basically clonal, showing very little evidence for recombination or transfer of plasmids (DYKHUIZEN et al. 1993 ; BALMELLI and PIFFARETTI 1996 ). What lateral transfer there is seems to be solely small piece (<1 kb) transfer (DYKHUIZEN and BARANTON 2001 ).

Being a vector-borne, obligate parasite, B. burgdorferi undergoes severe population bottlenecks during both the host-to-tick and tick-to-host transmissions (for reviews of the transmission cycle see LANE et al. 1991 ; SONENSHINE 1993 ). Thus, B. burgdorferi should have a small effective population size and an evolutionary history closely associated with that of I. scapularis. Evolutionary relationships between the parasite and its vector may be revealed only through comparative genetic analysis of local populations of the two organisms. In a previous population genetic study of B. burgdorferi, little genetic differentiation was found among samples collected from different locations on eastern Long Island, New York (QIU et al. 1997 ). The present study includes an analysis of the tick populations so that the population structures of the parasite and its vector can be compared. In addition, this study samples ticks and their infecting Borrelia along the East Coast from Massachusetts to South Carolina, giving a much-expanded survey. Previous surveys of B. burgdorferi and I. scapularis in North America (e.g., RICH et al. 1995 ; NORRIS et al. 1996 , NORRIS et al. 1999 ; MATHIESEN et al. 1997 ) have examined the two organisms separately and have not analyzed within- and among-population variation.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Study sites, tick processing, and determination of Borrelia infection:
Blacklegged ticks (I. scapularis) were collected from vegetation by flagging during their host-seeking seasons (in the fall for adult ticks and early June for nymphs) during 1997–1998. The study area includes multiple sites in eastern Long Island, New York (Fig 1A) and 12 sites along the East Coast (Fig 1B). Samples of other tick species (I. Pacificus and Dermacentor andersoni ) were collected from western North America and used as outgroups in phylogenetic analysis. The sources of the ticks are listed in Table 1. In the laboratory, frozen ticks were bisected individually. Total DNA was extracted using Chelex as described previously (QIU et al. 1997 ). A nested-PCR procedure that targets the outer surface protein A locus (ospA) was used to determine whether the tick was infected with Borrelia (GUTTMAN et al. 1996 ).

View larger version (26K): In this window In a new window Download PPT slide   Figure 1. Tick collection sites. (A) Long Island, New York; (B) East Coast.

Cold single-strand conformation polymorphism analysis of tick mitochondrial 16S rRNA genes:
Tick mitochondrial 16S ribosomal RNA genes were amplified using nested PCR. A 460-bp fragment was amplified for 15 cycles using primers 16S + 1 (primer sequences are listed in Table 2) and 16S - 1 in a 25-µl reaction mixture containing 5 µl of the tick DNA extract, 200 µM of each dNTP, 50 mM KCl, 10 mM Tris-HCl (pH 8.3), 2.5 mM MgCl₂, 0.1 µM of each external primer, and 1 unit of recombinant Taq polymerase (Life Technologies, Rockville, MD). The mixture was first heated to 96° for 1 min and then amplified for 15 cycles at 94° for 1 min, 48° for 1 min, and 72° for 1.5 min in a PTC-100 thermal cycler (MJ Research, Watertown, MA). Two microliters of this first amplification was used as a template for the second round of PCR. In the second round, a 300-bp portion of the 460-bp fragment was amplified using the same forward primer 16S + 1 and a reverse primer 16S - 2 in a 50-µl mixture containing 200 µM each dNTP, 50 mM KCl, 10 mM Tris-HCl (pH 8.3), 2.5 mM MgCl₂, 0.5 µM each primer, and 1.25 units of recombinant Taq polymerase. The amplification was run for 35 cycles at 94° for 1 min, 54° for 1 min, and 72° for 1 min. The final PCR products were electrophoresed on a 2% agarose gel (NuSieve:SeaKem 2:1; FMC BioProducts, Rockland, ME) to view the results. These same PCR primers have been used to amplify mitochondrial DNA from 36 species of hard- and soft-body ticks (BLACK and PIESMAN 1994 ), including I. scapularis ticks collected from eastern United States (NORRIS et al. 1996 ).

The PCR-amplified 300-bp fragment of 16S rDNA was subjected to cold single-strand conformation polymorphism (SSCP) analysis (HONGYO et al. 1993 ) as modified by GUTTMAN et al. 1996 to distinguish alleles. Common alleles were included as standards in each gel to identify alleles. Fragments of the 16S gene amplified with the primer pair of 16 + 2 and 16 - 1 were unsuitable for SSCP analysis because they did not form well-defined bands.

Unique alleles, as recognized by distinct SSCP mobility patterns, were reamplified from the first amplification for an additional 30 cycles using 2 µl of the product of the first PCR reaction as template and primers 16S + 1 and 16 - 1. The 460-bp PCR product was purified using the Prep-A-Gene kit (Bio-Rad, Richmond, CA) according to the protocol recommended by the manufacturer. Each unique allele was then cycle sequenced from both directions using the PCR primers on a ABI 373S DNA sequencer (Perkin-Elmer, Norwalk, CT). The sequences were assembled and edited using the software Sequencher (Gene Code, Ann Arbor, MI). A total of 25 haplotypes were identified and their GenBank accession nos. are AF309008, AF309009, AF309010, AF309011, AF309012, AF309013, AF309014, AF309015, AF309016, AF309017, AF309018, AF309019, AF309020, AF309021, AF309022, AF309023, AF309024, AF309025, AF309026, AF309027, AF309028, AF309029, AF309030, AF309031, AF309032.

Cold SSCP analysis of ospA variation:
The genetic diversity of Borrelia was analyzed at the ospA locus (a single-copy gene located on a linear plasmid, lp54; see FRASER et al. 1997 ) for each infected tick. A PCR-amplified 345-bp fragment of ospA (from base 220 to 564) was subjected to cold SSCP analysis as described above. (In this article, Borrelia sequences are numbered according to ospA or ospC sequences of strain B31.) Four SSCP mobility classes (five sequence haplotypes) of the amplified ospA fragments had previously been identified in Borrelia from eastern Long Island (GUTTMAN et al. 1996 ; QIU et al. 1997 ) and were used as controls for SSCP mobility classes. Any new mobility patterns indicating the presence of new ospA sequences were subsequently sequenced. GenBank accession numbers for the four ospA mobility classes (five haplotypes MC1a, MC1b, MC2, MC3, and MC4) and a new mobility class (MC5) are AF309002, AF309003, AF309004, AF309005, AF309006, AF309007.

Reverse line blotting assay of ospC variation:
We also measured the genetic diversity of Borrelia populations using the outer surface protein locus, ospC (a single-copy gene located on a circular plasmid, cp26; see FRASER et al. 1997 ). Because of difficulties in interpreting the SSCP results when ticks are infected with multiple strains of B. burgdorferi , a method using reverse line blotting (SAIKI et al. 1989 ; KAUFHOLD et al. 1994 ; RIJPKEMA et al. 1995 ; KURTENBACH et al. 1998 ) was designed for the direct detection and counting of ospC alleles in tick samples. The alleles of ospC of B. burgdorferi form "major sequence groups" (WANG et al. 1999 ). Sequences between major groups are 10% different from each other while sequences within groups are 1% different from each other (WANG et al. 1999 ). So far 21 major ospC groups (designated A through U) have been identified (SEINOST et al. 1999 ; WANG et al. 1999 ). The advantages of reverse line blotting (RLB) over SSCP for the assay of ospC variation in natural population samples can be specified as the following:

1. RLB is more sensitive. The average number of ospC alleles in an infected tick was 1.5 using the SSCP (WANG et al. 1999 ), compared to 2.6 using RLB assays (see RESULTS).

2. Results of the RLB assay were generally less ambiguous since some ospC alleles show similar mobility patterns on a SSCP gel.

3. Individual bands tend to smear on SSCP gels when there is a high multiplicity of types.

4. The current RLB design identifies major groups but not sequence variation within a major group, while SSCP reveals DNA sequence change regardless of the number of bases changed. In other words, SSCP does not distinguish major groups from variation within these groups.

RLB has the following disadvantages: (1) Major groups, like group C, which are recombinants of other groups, can easily be misclassified (see below); and (2) unknown groups can escape detection (the universal probe can indicate a missing group only when that group is the only clone in a tick). Thus, the presence of some sequence groups like group C will sometimes have to be confirmed using other methods.

Design of ospC allele-specific probes: Among the 21 major ospC groups of B. burgdorferi identified (SEINOST et al. 1999 ; WANG et al. 1999 ), four major groups (P, Q, R, and S) are found only in Europe and therefore were not included in this study. Groups L and O were also excluded in this study for lack of pure cultures of these two clones to be used as controls. Oligonucleotide probes specific for most major groups were designed according to sequences in a hypervariable gene region (bases 237–285, Fig 2). Sequences of groups C and I are similar in this region and were distinguished using probes targeted at another region. A conserved 27-bp sequence (bases 70–96) shared by all alleles of ospC of B. burgdorferi was used as a control. Nucleotide sequences used in the RLB assay are listed in Table 2.

View larger version (38K): In this window In a new window Download PPT slide   Figure 2. The hypervariable region of ospC sequences used for designing oligonucleotide probes for the reverse line blotting (RLB) assay (B31 numbering). The probes used are boxed.

Fluorescein-labeled PCR amplification: A portion of the ospC gene was PCR amplified directly from infected ticks using a pair of primers OC6 (+) and OC602 (-) (WANG et al. 1999 ), one of which (OC6+) was commercially modified with fluorescein at the 5' end. The 50-µl reaction mixture contained 10 µl of tick extract, 5 µl of 10x PCR buffer, 2.5 mM of MgCl₂, 200 µM of each dNTP, 1 µM of each primer, and 2.5 units of Taq polymerase. After initial denaturation at 96° for 1 min, the mixture was incubated for 45 cycles at 94° for 30 sec, 54° for 30 sec, and 72° for 1 min in a PTC-100 thermal cycler. The PCR product was examined afterward for purity and DNA concentration by electrophoresis on a 2% agarose gel. Fluorescein-labeled ospC genes were also amplified from cloned ospC preparation (WANG et al. 1999 ) or pure cultures of Borrelia strains to serve as positive controls in the RLB.

Tailing of oligonucleotide probes with poly(dT): Short stretches of oligonucleotides usually do not contain a sufficient number of thymine (T) residues to bind to a nylon membrane when activated by UV light (SAIKI et al. 1989 ). Thus, between 400 and 600 thymine residues were added to the 3' end of each oligonucleotide. The oligonucleotides were commercially synthesized and cartridge purified. After being reconstituted in water (not TE buffer, which inhibits terminal transferase activity), probes were tailed at 37° for 1 hr in a 50-µl mixture, containing 200 pmol oligonucleotides, 10 µl 5x reaction buffer (supplied by the manufacturer), 8 µl of 10 mM dTTP (nonsodium form, since sodium is an inhibitor of terminal transferase), and 4 units of terminal deoxynucleotidyl transferase [TdT (added last); Gibco BRL/Life Technologies]. The length of poly(dT) tails was examined by electrophoresis on a precast 20% TBE polyacrylamide gel (Novex, San Diego).

Membrane preparation: The tailed oligonucleotide probes were fixed onto a 15 x 15-cm nylon membrane (Stratagene, La Jolla, CA), using a MiniSlot 30 manifold (Immunetics, Cambridge, MA). The probes (40 pmol or 10 µl from 50-µl tailing product) were diluted in 2 ml of TE buffer and evenly applied into individual slots (0.8 mm x 13 cm) created by the manifold. The wet membrane was immediately UV crosslinked (125 mJ), briefly rinsed in water, and stored at 4° in 2x SSC. Once prepared, the membrane can be reused (up to 10 times) by stripping the membrane between uses (BOEHRINGER MANNHEIM 1995 ).

Checkerboard hybridization assays: Up to 1350 individual hybridization assays (30 probes x 45 samples) can be performed simultaneously using a checkerboard hybridization system (Checkerboard System 1; Immunetics; SOCRANSKY et al. 1994 ). The membrane was loaded into a 45-channel miniblotter. The lines in the miniblotter were perpendicular to the lines of the fixed probes. The channels were first filled with 100 µl of a hybridization solution (DIG Easy Hyb; Boehringer Mannheim). The hybridization solution contains 0.1 mg/ml Poly(A) (Sigma, St. Louis) to prevent nonspecific binding to the long poly(dT) tail of the probe. The miniblotter was incubated for 2 hr at 42° on a rocking platform. This solution was then aspirated and immediately replaced by 100 µl of hybridization solution containing also 1–10 µl (depending on concentration of amplified ospC genes) of denatured PCR products. The PCR products were denatured by heating in a boiling water bath for 10 min and chilling immediately on ice. The miniblotter was incubated again for 2 hr at 42° on a rocking platform. Subsequently, the channels were aspirated and washed twice at room temperature for 5 min with 2x SSC-0.1% SDS to remove unbound PCR products. The membrane was further washed stringently twice at 50° for 15 min with 0.1x SSC-0.1% SDS. Chemiluminescent detection was subsequently performed using a procedure suggested by the supplier (Boehringer Mannheim).

Statistical methods:
Gene frequency estimation: The frequencies of the SSCP alleles of ospA in a Borrelia population were directly estimated from the occurrence of individual mobility classes in an SSCP assay. In the checkerboard hybridization assay of ospC variation using RLB, a positive signal produced on the membrane indicated the presence of a major ospC group in the tick. Numbers of positive signals were used as absolute gene frequencies of individual alleles in a population. The direct counting method (for both SSCP alleles and ospC major sequence groups) tends to underestimate the frequent alleles and overestimate rare ones, biasing toward an evenness in allele frequency distribution. RANNALA et al. 2000 (see also ANDERSON and SCHEET 2001 ) developed an unbiased maximum-likelihood method for estimating gene frequencies in bacterial populations using molecular detection. For simplicity, the direct counting method was used in this study because the bias is very small and does not alter results of the analysis.

Tests of natural selection: The Ewens-Watterson-Slatkin test of neutrality (EWENS 1972 ; WATTERSON 1978 ; SLATKIN 1994 ), implemented in the computer program Arlequin Ver 1.1 (SCHNEIDER et al. 1997 ), was used for testing the presence of natural selection in maintaining the genetic diversity within Borrelia populations (on the basis of ospA and ospC frequencies) and tick populations (on the basis of frequencies of mitochondrial haplotypes).

Phylogenetic inference: The nucleotide sequences of all distinct haplotypes of the 460 bp of the tick mitochondrial 16S rRNA gene were aligned using CLUSTAL W (THOMPSON et al. 1994 ). Sequences of other haplotypes retrieved from GenBank (RICH et al. 1995 ; NORRIS et al. 1996 ) were included in the alignment. Unrooted minimum spanning trees (MSTs) were constructed manually to show sequence changes among haplotypes. Equal-length MSTs were discriminated using the approach of molecular variance parsimony (NEIGEL and AVISE 1993 ; EXCOFFIER and SMOUSE 1994 ). This method assumes that older alleles tend to be more frequent than new ones and new mutations are more likely to be found in the same population than in different ones. Missing links in the MSTs were filled with hypothetical haplotypes.

Genetic structure and spatial statistics: Genetic differentiation among local Borrelia populations was tested using the G-test of homogeneity of allele frequency distributions, using the biostatistical package BIOM (APPLIED BIOSTATISTICS 1996 ). For contingency tables with a small average number of entries (less than five), a nonparametric exact test (RAYMOND and ROUSSET 1995 ) was used. The population structure of I. scapularis along the East Coast was analyzed using analysis of molecular variance (AMOVA), a general framework for summarizing information on both nucleotide sequence variation and haplotype variation within and among geographic samples (EXCOFFIER et al. 1992 ). Three hierarchical levels were used: (1) within populations, (2) among populations within groups, and (3) among groups. Both the exact tests of population differentiation and AMOVA were performed using the molecular population genetic software package Arlequin Ver 1.1 (SCHNEIDER et al. 1997 ).

Demographic history: The program FLUCTUATE (KUHNER et al. 1998 ) was used to estimate population growth for both the tick and the Borrelia. On the basis of sequences of DNA haplotypes and frequencies of these haplotypes, the program coestimates the current population size (in the form of mutation parameter ) and population growth rate (g, exponential growth rate, with the unit of 1/µ individuals per generation).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

B. burgdorferi population structure:
Infection rate: The infection rate as determined by PCR amplification of the Borrelia ospA gene was 20–80% in the Northeast and 0–7% in North and South Carolina (Table 1). One inland tick population from Pennsylvania (PA) also had low Borrelia infection rate (1 out of 39 ticks or 2.6%). All these adult ticks were collected in the same season (November 1997) and the density of adult I. scapularis ticks was as high in the South as in the Northeast. These results are in agreement with the findings of other field studies (SONENSHINE et al. 1995 ; PIESMAN et al. 1999 ; OLIVER et al. 2000 ).

Genetic diversity at ospA: Four SSCP mobility classes (five sequence haplotypes, MC1a, MC1b, MC2, MC3, and MC4) of ospA were identified in previous surveys of eastern Long Island, New York (GUTTMAN et al. 1996 ; QIU et al. 1997 ). The frequency distributions of these classes were uniform across five populations of B. burgdorferi in adult ticks on eastern Long Island (QIU et al. 1997 ). In this survey we wished to determine if populations along the East Coast of the United States are also uniform.

We determined the ospA alleles of the Borrelia from 109 infected I. scapularis ticks collected from 10 locations along the East Coast. Only one new ospA SSCP mobility class, MC5, was found in this much-expanded biogeographic survey. The new mobility class was observed only once in one of the three infected adult I. scapularis ticks from Pea Island, North Carolina. DNA sequencing revealed that the mobility shift was caused by a single third-base synonymous substitution at the nucleotide position 465 as compared with the B31 type. The creation of MC5 is most parsimoniously interpreted as a recombination between MC1a and MC4, not as a new mutation.

Not only did all but one Borrelia in infected ticks belong to one of the four previously identified mobility classes, but the frequencies of the mobility classes were uniform from Massachusetts to Maryland for 1997 (MA, RI, NY1, NJ, MD1, and MD2; P = 0.1837, Table 3). This geographic uniformity was also found for the 1996 samples from Long Island, New York (LI-96) and from Connecticut (CT; P = 0.7170, Table 3). The frequencies of mobility classes are temporally stable as well in the northeast region from 1995 to 1997 (LI-95, LI-96, CT, and NY1; P = 0.207), with the pattern of MC2 and MC3 relatively common and that of MC1 and MC4 relatively rare (Table 3). However, the frequencies of the mobility classes collected from Long Island, New York during the fall of 1994 were different from the pattern found in the later years (LI-94 vs. LI-95 and LI-96; P < 0.001). This difference between 1994 and the other years was confirmed by analysis of the frequencies of major ospC groups (Table 4).

Multiple infections of ospC clones: A much higher level of DNA sequence diversity exists at ospC than in ospA. Fig 3 shows the chemiluminescence image of an RLB assay for a highly infected population (NY1, from Millbrook, New York). The controls were amplified from cloned ospC alleles and should hybridize with their respective probes only. The off-diagonal signals in the figure were therefore caused by cross-hybridization among alleles. We were able to remove the cross-hybridization of A, G, and J probes by using shorter oligonucleotides as probes (Table 2 shows the revised probes). Since the group C allele is a recombinant of multiple alleles (B, I, and E; see WANG et al. 1999 ), we were not able to design a probe specific for the group C allele. Due to the cross-reactivity among B, C, and I probes, we were unable to estimate the frequencies of groups B, C, and I in some samples (Table 4).

View larger version (31K): In this window In a new window Download PPT slide   Figure 3. Reverse line blotting assay of ospC variation. Oligonucleotide probes specific for ospC major groups (MG-A to MG-N, and OC-All, a probe for all ospC alleles) were UV fixed on the membrane in horizontal lines. Samples of fluorescein-labeled PCR amplicons were applied vertically to the probe lines. Samples include, on the right, PCR products amplified from cloned ospC genes (WANG et al. 1999 ), used as controls for probe specificity. PCR products amplified from individual Borrelia-infected ticks (NY1, from Millbrook, New York) were shown on the left of the membrane, one tick per vertical channel.

Most of the infected adult ticks from the Northeast were infected with multiple Borrelia clones (Fig 3). The degree of multiplicity (the average number of distinct ospC clones per infected tick) is particularly high in localities with high Borrelia prevalence. For example, ticks from Millbrook, New York (NY1) had a multiplicity of 3.1 clones per infected tick. Some ticks from that sample (ticks 2 and 32) were infected with at least 8 clones. On the other hand, all of the Borrelia-infected ticks from the southeast were infected with only a single clone (Table 4). Even in the Northeast, fewer nymphal ticks were infected with Borrelia and the multiplicity of infecting clones was lower in infected ticks. Infected nymphal ticks from Long Island were infected with about one-half the number of clones as the adults. To compare the results from an SSCP assay with RLB, the sample (SI-94) used in the original SSCP study (WANG et al. 1999 ) was reanalyzed using RLB. The RLB assay showed a higher sensitivity in detecting the presence of alleles in ticks than did SSCP. An average of 2.6 clones per infected tick was detected using RLB compared to 1.5 clones using SSCP. Three ticks were infected with group N clones, which had not been detected previously using SSCP. Also alleles with minor differences in this population were indistinguishable by RLB, but gave different mobility patterns using SSCP.

The five coastal (from Massachusetts to Maryland) samples (MA, NY1, SI-96, MD1, and MD2) were not significantly different in their frequency distribution of major ospC groups (samples with small numbers, such as RI, NY2, and NJ, were excluded in the test). Major group K, one of the four major ospC groups that cause disseminating Lyme disease in humans (SEINOST et al. 1999 ), is one of the most common clones in all samples (Table 4).

Allele frequency distributions at both ospA and ospC in all the local Borrelia populations were significantly more than expected even from neutral distributions (results of Ewens-Watterson tests, not shown), suggesting balancing selection in the form of frequency-dependent selection. For examples of the actual distributions, see QIU et al. 1997 and WANG et al. 1999 .

Linkage between ospA and ospC: Because of the high degree of clonality of B. burgdorferi (DYKHUIZEN et al. 1993 ; BALMELLI and PIFFARETTI 1996 ), it is not surprising that results of analyses of Borrelia populations (such as tests of population differentiation and Ewens-Watterson tests of neutrality) are similar using either ospA or ospC as the genetic marker, even though the two genes are located on different plasmids. The pattern of linkage between the SSCP alleles of ospA and the major groups of ospC has been expanded from previous work (WANG et al. 1999 ) and is shown in Table 5. These associations of ospA and ospC alleles are inferred from the alleles that are present together in ticks from all samples. Almost all ospC major groups are in strict linkage with only one of the ospA SSCP alleles except groups G and K, which are associated with more than one ospA mobility class.

I. scapularis population structure:
Divergence of northern and southern populations: Cold SSCP analysis of the I. scapularis mitochondrial 16S rRNA gene revealed 10 common haplotypes (named A–J) in tick samples collected from the northeastern United States. Four rare haplotypes, each observed only once, were also identified. SSCP mobility patterns of the tick mitochondrial 16S haplotypes found in the northern populations are shown in Fig 4A. In the Southeast, three haplotypes, M, N, and F, dominated most of the ticks in two samples from Beaufort, South Carolina. The F allele was found in both the Southeast and the Northeast. Haplotypes found in tick samples from mid-Atlantic sites (MD2 from southern Maryland; NC1 and NC2 from North Carolina) carried mostly common alleles (e.g., D and F) of the Northeast. Samples from the mid-Atlantic region, however, had a relatively large proportion of locally distinct haplotypes that were absent from either the northeastern or the South Carolina samples. SSCP mobility patterns of unique haplotypes from the North and South Carolina samples are shown in Fig 4B. A total of 31 distinct I. scapularis haplotypes were identified from this and two other studies (RICH et al. 1995 ; NORRIS et al. 1996 ).

View larger version (95K): In this window In a new window Download PPT slide   Figure 4. SSCP analysis of tick mitochondrial (mt) 16S rRNA variation. A distinct haplotype is identified by a unique mobility pattern of a pair of single SSCP bands (indicated by arrows). Shown are common haplotypes found in the northern samples (A) and the southern samples (B) of I. scapularis.

Local populations of southern I. scapularis (e.g., NC2 and SC) are genetically much more heterogeneous than northern populations (Table 6). Inland I. scapularis populations may also be much more diverse than coastal populations, as indicated by comparing two North Carolina samples (NC1 to NC2, Table 6). Variation of DNA sequences and allele frequencies of these tick samples do not deviate from neutral expectations, as analyzed by Ewens-Watterson tests (results not shown) and TAJIMA's (1989) tests (Table 6).

Minimum spanning tree (Fig 5) and mismatch distributions (Fig 6) clearly show differences in the genetic composition between the northern and southern I. scapularis populations. The minimum spanning tree revealed two major mitochondrial clades, in agreement with the findings of previous studies (RICH et al. 1995 ; NORRIS et al. 1996 ). Clade A is distributed mainly in the North, but also exists in the South (e.g., SC). Clade B is found only in southern populations and not in northern populations. Some haplotypes in the mid-Atlantic and South are genetically intermediate between the typical northern haplotype (F) and the common southern haplotypes (M and N). Northern tick samples typically show a unimodal mismatch distribution, while mid-Atlantic and southern samples show separated peaks (Fig 6). Because of possible population mixture in the South due to recent immigration from the North, the South Carolina sample was analyzed without clade A haplotypes (F and A). Including clade A sequences in the mismatch analysis of the SC sample would result in three peaks. For the mid-Atlantic sample (NC2), however, clade A haplotypes were too numerous and locally distinct to be attributed to immigrants from the North.

View larger version (17K): In this window In a new window Download PPT slide   Figure 5. Minimum spanning tree of mitochondrial 16S rDNA haplotypes of I. scapularis. Bars between two haplotypes indicate number of sequence differences.

View larger version (15K): In this window In a new window Download PPT slide   Figure 6. Pairwise distributions of mt 16S rDNA haplotypes of I. scapularis. Northern, mid-Atlantic, and Southern tick populations are represented by LI2 (from Wildwood State Park, Long Island, New York), NC2 (from Washington County, North Carolina), and SC (from Beaufort County, South Carolina; clade A haplotypes are excluded), respectively.

Temporal stability: Exact tests of population differentiation (RAYMOND and ROUSSET 1995 ) showed no significant temporal shift of genetic composition between pairs of samples collected in different years from the same site. One pair was from the North (Dutchess County, New York: NY3-90 (adults) vs. NY3-95 (nymphs), P = 0.4299) and the other was from the South (Beaufort, South Carolina: SC-97 vs. SC-98, P = 0.5537). Since there is no evidence of change of allele frequencies over these time spans, samples from the same sites were pooled.

AMOVA: The genetic composition of local tick populations shows enough local similarity that the populations can be subdivided into three localities (Table 7), as revealed by AMOVA (EXCOFFIER et al. 1992 ).

The three-group structure is caused by the genetic distinctness of I. scapularis in the mid-Atlantic region. The M and N haplotypes, which predominated samples from Beaufort, South Carolina, were absent from the tick samples from southern Maryland (MD2) and North Carolina (NC1 and NC2). Ticks from Washington County, North Carolina (NC2) contain mostly typical northern haplotypes (A, D, E, and F) but a large proportion of haplotypes in this sample (12 of 35 ticks or 34%) are distinct and found only in this region (6 O type ticks, 3 K type, 1 L type, and 2 singletons, NC2-22 and NC2-29; Table 6). Haplotype O belongs to clade B while K, L, and NC2-22 and -29 are all clade A (Fig 5), suggesting that these haplotypes had local origins and were not immigrants from either the northern or the southern populations. Moreover, haplotypes found in this region tend to be phylogenetically intermediate between clades A and B. The genetic distinctness of mid-Atlantic I. scapularis (especially inland populations) implies local origin of these populations and limited gene flow between distant tick populations.

Coastal vs. inland populations in North and South Carolina: The dissimilarity in the genetic composition of tick populations over a short distance is most striking between the two samples from North Carolina. The population NC1 was collected from a bird sanctuary on a coastal island (Pea Island National Wildlife Refuge, Cape Hatteras National Seashore, North Carolina) and NC2 from a nearby inland site 80 km away (Washington County, North Carolina). Whereas all but one tick from the coastal site were northern types, the 35 ticks from the inland site were a mixture of northern and southern types. The coastal population was very much like the northeastern populations, implying a possible role of migratory birds in bringing the northern ticks south during the fall. Since no southern haplotypes were found in this coastal site, the long-distance gene flow mediated by migratory birds may be unidirectional (i.e., from north to south only). The ongoing long distance transportation of ticks from north to south by birds was further supported by evidence from the South Carolina samples (SC 97 and 98), collected also from a coastal site (Spring Island, South Carolina). While the stable presence of the most common northern haplotype in these two samples (F, at 20%) may be the result of mitochondrial introgression, the rarer northern haplotypes (A and G) that were found in the sample from November 1998 could be new immigrants from the North. More sampling from coastal and inland sites will be needed to test this hypothesis.

High tick diversity in the New York-Connecticut region: Within the northern tick populations, the samples from New York and Connecticut showed higher levels of genetic diversity. The number of haplotypes in these populations ranged from six to eight (Table 6), while samples from Rhode Island, Massachusetts, New Jersey, Pennsylvania, and northern Maryland showed lower diversities with three to four haplotypes (Table 6). The higher levels of tick genetic diversity in the New York-Connecticut region imply that this region could be the source of the current geographic expansion of the northern ticks.

Comparisons between I. scapularis and B. burgdorferi:
Population structure: Within the range of northern blacklegged ticks, local populations are differentiated. On the other hand, local populations of Borrelia are similar within this range. Noncorrespondence in the level of genetic differentiation among local geographic samples between the northern blacklegged ticks and Borrelia was shown by pairwise F_ST tests (Fig 7). As discussed below, we propose that balancing selection at ospC creates this nonconcordance. The balancing selection counters the effects of drift in B. burgdorferi, but, of course, not in the blacklegged tick. The other possibility for the geographic uniformity of Borrelia is that the uniformity is maintained by a high level of gene flow mediated by some long-ranging reservoir hosts, such as birds.

View larger version (24K): In this window In a new window Download PPT slide   Figure 7. Tests of population differentiation: B. burgdorferi vs. I. scapularis. Shown is the significant genetic structuring of the northern I. scapularis ticks within a small area (A, Long Island) and along the East Coast (B), in comparison to the gene-frequency uniformity of Borrelia populations from the same set of locations. The pairwise FST values were obtained using Arlequin (version 1.1; SCHNEIDER et al. 1997 ; 0.05 < * < 0.01, 0.01 < ** < 0.001, *** < 0.001; unannotated bars are not significant).

Demographic history: Maximum-likelihood tests of population growth (KUHNER et al. 1998 ) were performed on a tick population sample from Long Island (LI2, from Wildwood State Park in 1995), the Borrelia infecting this tick population (data from QIU et al. 1997 ), a mid-Atlantic sample (NC2), and a southern sample (SC; Fig 8). The B. burgdorferi population in the Northeast and the northern I. scapularis share a remarkably similar demographic history, which is very different from the demographic history of the southern population. Clade A haployptes were excluded from the South Carolina sample because these may belong to recent immigrant ticks from the North. Maximum-likelihood estimates of population growth, however, are not significantly different even if clade A haplotypes are included.

View larger version (29K): In this window In a new window Download PPT slide   Figure 8. Tests of population growth. Using the program FLUCTUATE (KUHNER et al. 1998 ), population growth rate (g) and population size () were coestimated for a northern tick sample (LI2), a mid-Atlantic sample (NC2), a southern tick sample (SC, clade A haplotypes excluded), and Borrelia carried by the northern tick sample. The graphs are likelihood curves of population growth rate when the population size estimate is at its maximum-likelihood value. Estimates of population growth rate (g) for the northern tick sample and the Borrelia sample are both significantly greater than zero. Estimate of g for the mid-Atlantic and southern samples are not significantly different from zero. (The 95% confidence range of g is approximated by 2 log-likelihood units on each side of the maximum-likelihood estimates.)

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Contemporary patterns of biogeographic variation of a species are a result of both ongoing ecological and selective processes (such as gene flow and natural selection) and species histories (e.g., range expansion and contraction; TEMPLETON et al. 1995 ). It is therefore often necessary to separate these two sources of biogeographic variation (referred to as "ecological" and "historical" biogeography, respectively; see MORRONE and CRISCI 1995 ) when interpreting observed geographic patterns of a species. In this study, comparison of the population structure of Borrelia and I. scapularis suggests a common demographic history of the parasite and its tick vector along the northeast coast of the United States, resulting presumably from the same historical causes. This study also revealed significant differences in the genetic structures of the parasite and its tick vector, suggesting differential selective pressures on the Borrelia, causing differences in the population structure of the bacteria and tick host.

Geographic uniformity of Borrelia in the Northeast:
This survey of local Borrelia populations along the East Coast showed that the clone frequencies of the Lyme disease bacteria in the northeastern and mid-Atlantic region of the country are uniform. Not only are the frequencies of ospA and ospC alleles the same in all localities (Table 3 and Table 4), but the linkage relationships between the alleles of these two loci are also the same across localities.

Is the geographic uniformity of B. burgdorferi in the Northeast caused more by gene flow or by selection? We think that selection is more important than gene flow in maintaining the constant allele frequencies across Borrelia populations for the following reasons:

1. The gene-frequency uniformity across populations of Borrelia cannot be explained by gene flow via ticks since the tick populations show genetic differentiation (Fig 7). Movements of mammals and migratory birds infected with the spirochete can mediate gene flow among Borrelia (SONENSHINE 1993 ). Long distance flow of Borrelia via migratory birds is used to explain the high genetic diversity of Borrelia in low-density southern populations. However, it is unlikely that the migration of infected small mammals or birds is at a high enough frequency to homogenize the spirochete populations.

2. Despite the geographic uniformity in the frequency distribution of ospA and ospC alleles, the frequency distribution is not temporally stable. Borrelia infecting tick populations from the same location (Shelter Island, New York) collected in 1994 are different from those collected in 1996 as measured at both the ospA and ospC loci (Table 3 and Table 4). This sudden shift in allele frequencies of ospA between 1994 and 1996 happened both on Shelter Island and in Wildwood State Park on Long Island (QIU et al. 1997 ), showing the change cannot be explained by migration, but must be explained by selection.

3. The gene frequency distributions of Borrelia infecting nymphal ticks are significantly different from the distributions infecting the adults from the same cohort (QIU et al. 1997 ). Since larval and nymphal ticks feed on different hosts, this suggests selection is maintaining the gene frequency distributions.

4. The distribution is significantly different from that expected from neutrality, suggesting strong balancing selection. The balancing selection is evidenced by high local nucleotide sequence diversity at loci that are directly involved in establishing infection in mammalian hosts, such as ospC (GILMORE and PIESMAN 2000 ; SCHWAN and PIESMAN 2000 ). The effect of balancing selection is further displayed in the overdispersed allele frequency distribution at many nontargeted loci of selection (e.g., silent variations of ospA SSCP alleles). Thus, we conclude that strong balancing selection caused by the host immune response is the major force maintaining the geographic uniformity of Borrelia across the Northeast.

Divergence among North American Borrelia:
Compared with other Borrelia species found in North America (MATHIESEN et al. 1997 ; POSTIC et al. 1998 ; NORRIS et al. 1999 ), B. burgdorferi shows little genetic variation. The relative genetic homogeneity of B. burgdorferi is in agreement with an evolutionary history of recent population growth and rapid geographic range expansion of this species.

It is possible to learn more about the population biology of B. burgdorferi by cross-species comparisons. NORRIS et al. 1997 , NORRIS et al. 1999 published data on 91 B. bissettii strains from Colorado. Since the same ospA primers were used in both studies, the variation within and between species can be compared. No northeastern ospA types were found in the Colorado study and vice versa. This divergence can be due to either adaptation to different hosts and vector or geographic isolation with the accumulation of neutral differences.

The hypothesis of adaptive divergence between B. burgdorferi and B. bissettii was tested using McDonald-Kreitman tests (MCDONALD and KREITMAN 1991 ; Table 8). Synonymous substitutions outnumber replacement changes in each species (3 to 1 in the northeastern sample and 23 to 5 in the Colorado sample). However, the majority of the fixed differences between the two species are replacement substitutions (15 fixed nonsynonymous substitutions vs. 4 fixed synonymous substitutions). Consequently, the neutrality index (N.I.; Rand AND KANN 1996 ) for ospA is much smaller than one (N.I. = 0.064) and the McDonald-Kreitman test is highly significant (Table 8). The two housekeeping genes (fla and hbb) show no evidence of adaptive substitutions by the same test (Table 8). OspA is directly involved in tick-spirochete interaction (as opposed to OspC, which is more important in spirochete-host interactions) for establishing infection in Ixodes ticks (SCHWAN 1996 ; SCHWAN and PIESMAN 2000 ). These results suggest differential adaptation of ospA to genetically different ticks.

Separate histories of the northern and southern I. scapularis populations:
Results of this study revealed substantial differences in genetic structure, evolutionary history, and epidemiological characters between the northern and southern blacklegged ticks.

First, it can be concluded that the species of I. scapularis is geographically structured across its range along the East Coast. Although the northern and southern blacklegged ticks are taxonomically now described as a single species (OLIVER et al. 1993 ; KEIRANS et al. 1996 ), two distinct mitochondrial lineages were identified from this and other studies (RICH et al. 1995 ; NORRIS et al. 1996 ). One clade is common in the South but not found in the North (the "Southern Clade," NORRIS et al. 1996 ), while the other is mainly found in the North (the "American Clade," NORRIS et al. 1996 ). Northern tick populations consist exclusively of clade A haplotypes, which are all only one or two mutational steps away from the most common haplotype (F), showing a star phylogeny (Fig 5). The average within-population sequence divergence is only 0.2%. Ticks from the South consist of predominantly clade B haplotypes (mostly M and N) and the average within-population sequence diversity is much higher, 1.5% (Fig 5). Studies of hybridization between the northern and southern blacklegged ticks using nuclear loci have so far failed to detect genetic discontinuity (WESSON et al. 1993 ; OLIVER et al. 1993 ; NORRIS et al. 1996 ). Morphological studies, however, showed North-South and East-West differences of blacklegged ticks (HUTCHESON et al. 1995 ; KEIRANS et al. 1996 ). Morphological and mitrochondrial data thus suggest I. scapularis ticks are geographically structured across its range of distribution due to lack of gene flow among populations. The strongest evidence for the lack of gene flow in this study is the extensive local differentiation of ticks collected from inland North Carolina (NC2, Table 6). Nevertheless, the presence of the most common clade A haplotypes in southern populations (F and A, Table 6; RICH et al. 1995 ; NORRIS et al. 1996 ) suggests secondary contact of the two previously separate lineages as a result of recent range expansion of northern ticks.

Second, the patterns of the genetic composition of populations suggested that the northern and southern I. scapularis populations have separate and different evolutionary demographic histories. The northern tick population samples showed evidence of exponential increase in effective population size, such as star phylogenies (Fig 5) and unimodal distributions of pairwise nucleotide differences (Fig 6) among haplotypes (SLATKIN and HUDSON 1991 ; ROGERS and HARPENDING 1992 ; HARPENDING et al. 1998 ). Corresponding genetic patterns of southern population samples were bifurcated (NC2) or trifurcated (SC) gene phylogenies (Fig 5), with bimodal mismatch distributions (Fig 6), in agreement with a long evolutionary history of constant and stable population size. The difference in demographic history between the northern and southern populations is further supported by the maximum-likelihood estimates of population growth rates of these tick samples (Fig 8).

Third, the B. burgdorferi infection rate in I. scapularis ticks is markedly different between ticks collected from the Northeast and those from the Southeast (Fig 3). The localization of Lyme disease in the range of the northern lineage of I. scapularis suggests that genetic (as well as ecological) factors involving the ticks may play a part in the concentration of the disease in the North. The low Borrelia infection rate in ticks from the South is generally attributed to the lack of an efficient enzootic cycle of immature ticks and white-footed mice in the South compared to the one in the Northeast (SONENSHINE et al. 1995 ; OLIVER 1996 ; OSTFELD and KEESING 2000 ). The other possible cause of the low rate of Borrelia infection in southern I. scapularis, however, could involve the genetic heterogeneity of southern ticks. PIESMAN and SINSKY 1988 showed that I. scapularis from Alabama were highly efficient in both acquiring and transmitting Borrelia infection, using hamsters as experimental hosts. We observed in this study that ticks with southern (clade B) haplotypes can be naturally infected with B. burgdorferi. However, small differences in vector competence of different tick lineages could lead to large differences in Borrelia population size. On the basis of the geographic correspondence of the northern tick lineage and Lyme disease, and of the concurrence of the Lyme disease epidemic with range expansion of the northern I. scapularis, TELFORD 1998 argued that the species name I. dammini is epidemiologically justified.

Shared vicariant histories between B. burgdorferi and northern I. scapularis:
Both B. burgdorferi and its vector, the northern lineage of I. scapularis, exhibit characteristics of a young evolutionary lineage, such as a low level of neutral genetic polymorphism within populations. Furthermore, statistical inference of population history on the basis of genetic compositions of local population samples suggested that the parasite and its vector share a similar demographic history of recent population growth and geographic expansion (Fig 8). The similar demographic histories between Borrelia and I. scapularis in the North may have the same evolutionary causes.

It has long been observed that species diversity of many animals and plants decreases with higher latitude (LAPORTE 1968 ). Besides ecological factors, the last glacial maximum, which was 18,000 years ago, may have played an important part in creating such latitudinal gradients in species diversity. Fragmentation of a formerly continuous species distribution caused by geological events such as tectonic movements and glaciation is known as the "vicariance hypothesis" in biogeography (NELSON and PLATNICK 1981 ; WILEY 1988 ). A genetic consequence of population contraction during the glacial maxima and population expansion during the intervening warming periods is the loss of genetic variation of populations in the affected regions, relative to the genetic diversity of populations that were not affected (HEWITT 1996 , HEWITT 2000 ). Vicariance due to Pleistocene glaciation events seems to have caused parallel biogeographic patterns of reduced genetic variability in the North for a wide range of animal and plant species in North America (AVISE 1992 ) and Europe (SAGE and WOLFF 1986 ; MAGOULAS et al. 1996 ; MARILA et al. 1996 ). These empirical studies often found short-spanning star phylogenies among haplotypes (differing from each other with one or two nucleotides) within northern populations, as observed for the northern I. scapularis ticks in this study. The New York-Connecticut region is a likely location of refugia populations of tick and Borrelia, suggested by its higher level of haplotype diversity in tick mitochondrial sequences. Northern blacklegged ticks in the Midwest may have originated from different refugia populations, as evidenced by the existence of locally distinct clade A haplotypes (e.g., MN2-1 reported by NORRIS et al. 1996 ).

Conclusion:
This study showed that the population structure of B. burgdorferi, a vector-transmitted parasitic bacterial species, is strongly shaped by the evolutionary history of its vector, the northern lineage of I. scapularis. We saw high infection rates in the Northeast where the genetic variability of the tick is low. There is also strong evidence of recent expansion of both the tick and Borrelia populations in the regions that were covered by glacier 18,000 years ago, as opposed to the tick populations in the South where the populations were not similarly destroyed. However, the population genetics of the northern tick and Borrelia are different, presumably because of the frequency-dependent selection on the Borrelia clones by the mammalian host. We postulate that the epidemic of B. burgdorferi in the Northeast (not only in humans but also in ticks and other mammals) is due to the biotic and genetic deprivation, induced by geological events like glaciations and modern human activities like deforestation and reforestation (BARBOUR and FISH 1993 ). As the ecosystem is reduced in species diversity, the remaining species will lose much of their genetic diversity due to population bottlenecks caused by human interference, followed by rapid expansion of the population. Disease organisms can evolve to specialize on the large number of very similar hosts, increasing virulence. Epidemics will then become both more frequent and more severe for the remaining species including humans (OSTFELD and KEESING 2000 ).

	FOOTNOTES

¹ Present address: Center for Advanced Research in Biotechnology, 9600 Gudelsky Drive, Rockville, MD 20850.

	ACKNOWLEDGMENTS

We thank Edward Bosler, Jason Campbell, Howard Ginsberg, Richard Ostfeld, Bill Golde, Daniel Taub, M. Bushmich, and, in particular, Rob Mactier for helping us to collect ticks. Much of the DNA sequencing was done in John Dunn's lab in the Brookhaven National Lab. The study benefited greatly from guidance and advice from Mike Bell, Dou