An experimental chain of infection reveals that distinct Borrelia burgdorferi populations are selected in arthropod and mammalian hosts

Authors

  • Jeffrey R. Ryan,

    1. Department of Entomology, College of Agriculture and Life Sciences, North Carolina State University, Raleigh, NC 27606, USA.,
    Search for more papers by this author
    • Present address: Department of Entomology, Division of Communicable Diseases and Immunology, Walter Reed Army Institute of Research, Washington, DC 20307, USA

  • Jay F. Levine,

    1. Department of Microbiology, Parasitology and Pathology, College of Veterinary Medicine, North Carolina State University, 4700 Hillsborough Street, Raleigh, NC 27606, USA.,
    Search for more papers by this author
  • Charles S. Apperson,

    1. Department of Entomology, College of Agriculture and Life Sciences, North Carolina State University, Raleigh, NC 27606, USA.,
    Search for more papers by this author
  • Lori Lubke,

    1. Bacterial Pathogenesis Section, Microscopy Branch, Rocky Mountain Laboratories, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Hamilton, MT 59840, USA.,
    Search for more papers by this author
  • Robert A. Wirtz,

    1. Department of Entomology, Division of Communicable Diseases and Immunology, Walter Reed Army Institute of Research, Washington, DC 20307-5100, USA.
    Search for more papers by this author
    • Present address: Entomology Branch, Centers for Disease Control and Prevention, Atlanta, GA 30341, USA

  • Patricia A. Spears,

    1. Department of Microbiology, Parasitology and Pathology, College of Veterinary Medicine, North Carolina State University, 4700 Hillsborough Street, Raleigh, NC 27606, USA.,
    Search for more papers by this author
  • Paul E. Orndorff

    1. Department of Microbiology, Parasitology and Pathology, College of Veterinary Medicine, North Carolina State University, 4700 Hillsborough Street, Raleigh, NC 27606, USA.,
    Search for more papers by this author

Paul E. Orndorff E-mail paul_orndorff@ncsu.edu; Tel. (919) 829 4207; Fax (919) 829 4455.

Abstract

The prokaryotic, spirochaetal microorganism Borrelia burgdorferi is the causative agent of Lyme disease, an arthropod-borne disease of a variety of vertebrates and the most prevalent arthropod-borne disease of humans in the United States. In order to understand better the normal life cycle of B. burgdorferi, an experimental chain of infection was devised that involved multiple sequential arthropod and mammalian passages. By examining populations of B. burgdorferi emerging from different points in this infectious chain, we demonstrate that selection of B. burgdorferi populations peculiar to arthropod or vertebrate hosts is a property of at least one of the two ecologically distinct strains we examined. Distinct B. burgdorferi populations were identified using an antigenic profile, defined by a set of monoclonal antibodies to eight B. burgdorferi antigens, and a plasmid profile, defined by the naturally occurring plasmids in the starting clonal populations. These two profiles constituted the phenotypical signature of the population. In the strain exhibiting selection in the different hosts, transition from one host to another produced a striking series of alternating phenotypical signatures down the chain of infection. At the molecular level, the alternating signatures were manifested as a reciprocal relationship between the expression of certain antigenic forms of outer surface protein (Osp) B and OspC. In the case of OspC, the antigenic changes could be correlated to the presence of one of two distinctly different alleles of the ospC gene in a full-length and presumably transcriptionally active state. In the case of OspB, two alleles were again identified. However, their differences were minor and their relationship to OspB antigenic variation more complicated. In addition to the reciprocating changes in the antigenic profile, a reciprocating change in the size (probably the multimeric state) of a 9.0 kbp supercoiled plasmid was also noted. Selection of distinct populations in the tick may be responsible for the microorganism's ability to infect a wide range of vertebrate hosts efficiently, in that the tick might provide selective pressure for the elimination of the population selected in the previous host.

Introduction

Most protozoal and other eukaryotic parasites require at least two host species to propagate themselves. In each host, obligate parasitic differentiation takes place to: (i) allow survival in that host (Vickerman, 1978; Gunderson et al., 1987); and (ii) become infective for the next host (Harwood and James, 1979). In contrast, no recognizable differentiation (e.g. sporulation) of arthropod-borne bacterial pathogens (prokaryotic parasites) has been observed in their hosts. Nevertheless, it is clear from many studies on prokaryotic parasites that a number of bacterial genes are required for success in the vector (McDonough et al., 1993; Hinnebusch et al., 1996) and that a variety of adaptive responses come into play (Schwan et al., 1995; Schwan and Hinnebusch, 1998). These responses typically involve an alteration in the expression of certain genes that are controlled at the transcriptional level by environmental signals characteristic of the mammalian host, vector or in vitro growth (Champion et al., 1994; Suk et al., 1995; Stevenson et al., 1995; Tilly et al., 1997). However, these changes in gene expression are adaptive in the sense that they are temporary: a new environment induces a new set of transcription modifiers. Whereas such adaptive changes undoubtedly contribute to the parasite's efficient use of available resources, it is unclear how bacterial parasites could avoid selection (inheritable, genotypical changes) for optimal growth during many generations in such widely different hosts.

We have been interested in the transmission of the prokaryotic microorganism Borrelia burgdorferi, the spirochaete associated with Lyme borreliosis (Burgdorfer et al., 1982; Johnson et al., 1984). This microorganism has a wide range of vertebrate hosts including mammal, avian and reptile species (Anderson, 1987; Barbour and Fish, 1993; Levin et al., 1996) and is transmitted by the bite of infected ticks of the Ixodes persulcatus group (Barbour and Fish, 1993). Transmission is sustained by transfer of the spirochaete from vector to reservoir host and then back to the tick vector. As the environments of the tick mid-gut and the tissues of their vertebrate hosts vary widely (Burgdorfer et al., 1991), we wondered if distinct B. burgdorferi populations were being selected in widely divergent environments and, if so, how this affected subsequent sequential passage. To examine this, we devised an experimental transmission cycle, spanning 14 months, in which we tested two ecologically diverse strains of B. burgdorferi for their response to the alternating selective pressures of mammalian and arthropod host. In one of the strains we tested, no inheritable phenotypical changes could be attributed to host-specific selective pressures by our screening methods. However, in the other strain, we found that there was a well-defined selection exerted in each host producing stable, distinct populations peculiar to arthropod or vertebrate host.

Results

The experimental transmission cycle

The two B. burgdorferi strains initially examined were isolated from distinctly different hosts: (i) a raccoon (Procyon lotor ); and (ii) a male, black-legged tick (Ixodes scapularis). A colony-purified (clonal) population (Kurtti et al., 1987) of each strain [P235.1 (raccoon derived) and P64.1 (tick derived)] was used to initiate a chain of infection that was designed to simulate two complete transmission cycles. This chain of infection used Ixodes scapularis ticks and two different rodent species; Oryzomys palustris, the marsh rice rat, and Peromyscus leucopus, the white-footed mouse, in the pattern: mammal 1 (O. palustris) → tick → mammal 2 (P. leucopus) → tick (Fig. 1). Two different mammalian hosts were used because we thought initially that B. burgdorferi populations peculiar to a particular mammalian species might be distinguishable. However, such differences between mammals, if they occurred, were undetectable using our methods.

Figure 1.

. The experimental chain of infection. The diagram shows the original two B. burgdorferi strains used and their sequential passage through two mammalian hosts via infection and reinfection of ticks. At each link in the chain, some of the infected hosts were removed, and primary cultures (i.e. not clonal) of B. burgdorferi were obtained as described in the text. The remainder were used to provide infective hosts to carry on the chain. The primary cultures were examined for their antigenic and plasmid profile. Also, two isolated clones were obtained from each primary culture and additionally examined for their antigenic (Fig. 2) and plasmid (Fig. 3) profile. As mentioned in the text, primary and clonal populations showed identical profiles. Representative clonal profiles from each link in the chain are shown in Figs 2 and 3.

In order to define distinct B. burgdorferi populations, we took advantage of the natural antigenic profile provided by eight antigens. Also, we used the natural plasmid profile, as defined by the number, size and sequence similarity of B. burgdorferi plasmids resident in our parental clones (Mayer, 1988).

B. burgdorferi populations emerging from the different hosts were examined using the following procedure. At every ‘link’ in the chain (the point at which the host changed), an average of eight indiscriminately chosen hosts were removed, and primary (broth) B. burgdorferi cultures were obtained from tissue samples of each host. The remainder of the host population was used to provide infected mammals (or infected ticks) to carry on the chain. The antigenic and plasmid profile of each of the eight primary B. burgdorferi cultures was examined (examples shown in Fig. 2 and Fig. 3 respectively). The striking uniformity of the antigenic and plasmid profiles in populations from random hosts at a given link was characteristic. Two clonal isolates were obtained from of the primary cultures at each link (Figs 2 and 3). In every instance, the primary cultures and the derived clonal isolates had the same antigenic and plasmid profile. This result showed that the profiles of the primary cultures were not generated as a multiclonal composite. Rather, the primary culture profiles were generated by one predominant population. One clonal isolate from each B. burgdorferi strain obtained at each link is shown in subsequent figures that illustrate steps in the infection chain.

Figure 2.

. A. An example of an SDS–PAGE protein profile from the parental cloned B. burgdorferi strain P235.1 (lane P) and primary (uncloned) cultures from five infected mice (P. leucopus) (lanes 1–5). Two individual clonal isolates from mouse 1 are shown (lanes 1.3 and 1.5). B. An example of a Western blot of an SDS–PAGE gel produced in parallel to (A) above and developed using monoclonal antibodies to the proteins indicated to the right of the blot (a complete description of the source of monoclonal antibodies is provided in Experimental procedures). The full profile of this isolate (replicate) during the chain of infection is shown in Fig. 9. Protein molecular weights were estimated using standards supplied by Bio-Rad Laboratories. Molecular weights (in thousands of kilodaltons) are shown on the left.

Figure 3.

. An ethidium bromide-stained agarose gel showing plasmid profiles from the parental clone of B. burgdorferi strain P235.1 (lane P) and primary cultures from four mice (P. leucopus) (lanes 1–4). Profiles of individual clonal isolates from the primary culture from mouse 4 are shown (lanes 4.1 and 4.4). The position of linear molecular weight size markers is shown on the left of the figure.

The chain of infection was replicated in parallel three times for both strains examined. For strain P64.1, we found that the antigenic and plasmid profile did not change at each link (data not shown). We could not confirm that this lack of change indicated a lack of host-associated selection or that the particular phenotypical characters that we were examining did not belie a population shift in this strain. In view of these uncertainties, we did not pursue this isolate further and confine our description to strain P235.1, which showed well-defined, host-associated population shifts in all replicates. In two replicates of strain P235.1, the antigenic and plasmid profiles were identical, and only one is shown here. In the third, the pattern of the antigenic profile differed slightly from the other two. This difference is shown and contrasted with the other two profiles at the end of the Results section.

Antigenic profile

The antigenic profile of P235.1 was constructed by Western blotting using a mixture of 10 monoclonal antibodies to 10 well-characterized B. burgdorferi immunogens (Craven et al., 1996; see Experimental procedures). The profile of strain P235.1 (Fig. 4A) showed a distinctive change at each link in the chain, producing isolates with one of two different antigenic profiles, one associated with mammals and one with ticks. In particular, we found that reacting epitopes of outer surface protein C (OspC) were detected in all isolates recovered from mammals (regardless of the two different mammalian species tested), whereas reacting epitopes of OspB were not detected in isolates from mammals (arrows in Fig. 4A, lanes M1 and M2). The opposite was seen in isolates from ticks (arrows in Fig. 4A, lanes T1 and T2). In the instances in which OspB or OspC were undetected, this was not because of a lack of synthesis of these two antigens. Polyclonal antibody to OspB (Fig. 4B) and OspC indicated their presence (Fig. 4C). Evidently, isolates that were unreactive to the monoclonal antibody produced a different antigenic version of OspB or OspC.

Figure 4.

. A. Western blot of strain P235.1 chain of infection. Blotting patterns were obtained by subjecting whole-cell lysates from representatives in the experimental chain of infection to SDS–PAGE and subsequently transferring protein bands to nitrocellulose and immunoblotting with the monoclonal antibodies shown. B. Western blot corresponding to (A) only using polyclonal antibody to OspB. C. Western blot corresponding to (A) only using polyclonal antibody to OspC. Nomenclature for the sequential isolates from left to right are as in Fig. 1. P, the parental clone used as inoculum to infect M1, the marsh rice rat (Oryzomys palustris) that was used subsequently to infect T1, the larval Ixodes scapularis tick that was used subsequently to infect M2, the white-footed mouse (Peromyscus leucopus) that was fed upon by T2, the nymphal Ixodes scapularis tick. Reactivity to various monoclonal antibodies is denoted on the right of each blot. Protein molecular weights were estimated using standards supplied by Bio-Rad Laboratories. Molecular mass (in thousands of kilodaltons) is shown on the left. Antigenically reactive OspB and OspC are denoted by arrows.

Sequence profiles of OspC and OspB

To understand further the antigenic variation seen in OspC and OspB proteins of the B. burgdorferi clonal isolates from mammals and ticks, we amplified, via the polymerase chain reaction (PCR), and then subsequently sequenced full-length amplicons of the genes encoding OspC and OspB from each isolate (P, M1, T1, M2, T2; shown in Fig. 4). The alignment of the deduced amino acid sequence of ospB and ospC revealed two allelic forms of each gene (Fig. 5A and B respectively). For both genes, one sequence was specific to mammals and the other to ticks.

Figure 5.

. Amino acid sequence alignments of the genes encoding OspB (A) and OspC (B). The alignments were generated from DNA sequence obtained from amplicons produced using primers for the full-length gene (see Experimental procedures). Nomenclature for the sequential isolates: P, the parental clone used as inoculum to infect M1, the marsh rice rat (Oryzomys palustris) that was subsequently used to infect T1, the larval Ixodes scapularis tick that was subsequently used to infect M2, the white-footed mouse (Peromyscus leucopus) that was fed upon by T2, the nymphal Ixodes scapularis tick. The top line indicates the deduced amino acid sequence of the parental clone (P). Dots represent residues in M1, T1, M2 and T2 that are identical to the parental clone (P). Where the amino acids differ, the different amino acid is indicated. Dashes represent deletions. A. The amino acid sequence alignment of the gene encoding OspB. The residue number refers to the ospB coding region of B. burgdorferi B31 accession number X74808. B. The amino acid sequence alignment of the gene encoding OspC. The residue number refers to the ospC coding region of B. burgdorferi T255 accession number X81524. Solid bars along the top of the ospC sequence indicate sites of annealing of internal primers used as described in the text. The DNA sequences for ospB and ospC have been deposited in GenBank under accession numbers AF077663, AF077664 and AF077661, AF077662 respectively.

For ospB (Fig. 5A), there was no dramatic change in the deduced amino acid (or DNA) sequence at each link in the chain: both alleles closely resembled (> 97% amino acid identity) the sequence of ospB reported by Coleman et al. (1994) (GenBank acession number X74808) and were more than 98% identical to each other. Further, the relationship between a sequence change in ospB and an antigenic change in OspB was not related in a simple way (see Discussion). For these reasons, the molecular characterization of this gene was deferred. For ospC, on the other hand, two dramatically different alleles (67.7% identical) were revealed, with one sequence correlating with the monoclonal reactive form of OspC, and the other sequence correlating with a the lack of reactivity (compare Fig. 5B with Fig. 4).

Each of the two versions of ospC closely resembled previously identified ospC gene sequences. The amino acid sequence of the mammal isolates was 98.9% identical to the ospC gene sequence of B. burgdorferi strain T255 (GenBank accession number X81524) (Jauris-Heipke et al., 1995). Likewise, the amino acid sequence of the tick isolates was 98.3% identical to the ospC gene sequence of B. burgdorferi strain OC8 (GenBank accession number AF029867).

Because of the distinctly different nature of the ospC alleles, we were able to document that the unique portions of both ospC alleles were present in the starting (parental) strain. To do this, we designed PCR primers that took advantage of sequences internal to the full-length genes that were specific to mammal or tick isolates (refer to Fig. 5B). We then determined by PCR amplification that the genetic material for both alleles was present in the parental strain. The last isolate in the chain (the second tick isolate; T2) was also examined. As expected, the parental and tick isolates contained the genetic material for the internal portions of both ospC alleles (Fig. 6). Subsequent DNA sequencing of the amplicons generated by the internal primers confirmed their identity (data not shown).

Figure 6.

. PCR analysis of B. burgdorferi isolates of the parental (P) and the second tick isolate (T2) using three different PCR primer sets. The appropriately sized amplicons are seen in both strains after amplification with tick-specific internal primers (pcr1-t and pcr2-t) and mammal-specific primers (pcr1-m and pcr2-m) (open arrow). Control (full-length) amplification reactions using primers C1 and C2 and resultant full-length amplicons of the appropriate length are also shown (closed arrow). Molecular weights were determined by comparison to a standard of φX DNA digested with HaeIII.

Plasmid profile

As with the antigenic pattern, the plasmid profile of strain P235.1 displayed a pattern that was characteristic of mammalian or tick isolates and again showed the same exact recapitulation of the pattern at each successive alternating link (Fig. 7A). This reversible pattern was most evident in the 9.0 kb plasmid band that was present in tick isolates but not in mammalian isolates (arrow in Fig. 7A, lanes T1 and T2), but close examination revealed that the entire plasmid profile was recapitulated. To relate the changes in plasmid profile to the appearance and disappearance of specific plasmids, a probe, derived by nick-translation of the DNA isolated from the 9.0 kb plasmid band, was used. Using this probe in a Southern blot analysis, we found that we were indeed seeing an exact recapitulation of the plasmid profile in each sequential isolate from ticks and mammals with regard to the plasmids sharing similarity with the 9.0 kb plasmid (Fig. 7B). To investigate this further, electron microscopical studies of the material from the hybridizing bands were initiated and revealed that the 9.0 kb band represented a single species of supercoiled plasmid (average contour length 9.0 kb; Fig. 8). Other intermediate hybridizing bands in these lanes were identified as a 9.0 kb open circular species (assumed to be an artifactual form of the 9.0 kb supercoiled plasmid) and a band composed of approximately 20 kb open circular forms. These latter open circular forms might represent artifactual forms of a hypothetical dimeric form of the 9.0 kb plasmid (Fig. 8). One lower hybridizing band (below the 9.0 kb band and migrating with an apparent length of 4.5 kb) may represent a hemimeric form of the 9.0 kb plasmid. (As an aside, the 9.0 kb plasmid may be a dimer of the 4.5 kb plasmid, a suggestion supported by previous work describing a 4.5 kb plasmid from another strain of B. burgdorferi ; Hyde and Johnson, 1988). A 28 kb supercoiled plasmid (top arrow in Fig. 7A and Fig. 8) was associated with all isolates. From the appearance of the blot (Fig. 7B), the 28 kb supercoiled plasmid may represent a trimeric form of the 9.0 kb plasmid, the latter associated exclusively with the tick isolates. However, a 28 kb plasmid band was retained in the tick isolates that showed little or no reactivity in the 9.0 kb probe (compare the band density at 28 kb in Fig. 7A with the probe reactivity of the corresponding band in Fig. 7B). This may simply reflect the presence of two (or more) different supercoiled plasmids of 28 kb; one with similarity to the 9.0 kb plasmid, the other(s) unrelated to the 9.0 kb plasmid. The existence of multiple circular plasmids of approximately 28 kb in B. burgdorferi has been reported previously (Stevenson et al., 1996; Fraser et al., 1997).

Figure 7.

. A. The plasmid profile of strain P235.1 recovered during the chain of infection. B. A Southern blot of strain P235.1 profile using the 9.0 kb plasmid (designated by arrows) as a probe during the chain of infection. Nomenclature for the reisolates tested from left to right are as follows: P, the parental clone used as inoculum to infect M1, the marsh rice rat (Oryzomys palustris) that was subsequently used to infect T1, the larval Ixodes scapularis tick that was subsequently used to infect M2, the white-footed mouse (Peromyscus leucopus) that was fed upon by T2, the nymphal Ixodes scapularis tick. Arrows denote the relationship of plasmid bands to probe-reactive bands using the 9.0 kb plasmid from strain P235.1 as a probe.

Figure 8.

. Electron micrographs of plasmids that relate plasmid size (contour length in kb) and conformation (supercoiled or open circular) to agarose gel bands that showed hybridizing activity with the 9.0 kb supercoiled plasmid. Refer to Fig. 5, lanes T1 or T2 for reference to the plasmid banding pattern. OC, open circular DNA; SC, supercoiled DNA.

Generality of the phenotypic characters selected in each host type

As stated earlier in this section, three independent chains of infection were performed with strain P235.1. In one of the three chains, a somewhat different result was obtained than with the other two in that (i) we could detect a truncated form of OspB (OspB′, as determined using polyclonal OspB antiserum) in the parental population, (bottom arrows in Fig. 9); and (ii) the reciprocating appearance/disappearance of OspB in ticks and mammals (respectively) was not observed in the antigenic profile. Rather, OspB was always present (upper arrows in Fig. 9). Whereas this replicate illustrates that populations being expanded for inoculation and those emerging under selective pressure may not have exactly the same antigenic profile, it strengthens one observation made with the other replicates: that the variants seen to emerge in the first two links in the chain are recapitulated exactly in the subsequent links (compare lanes M1 with M2 and T1 with T2 in Fig. 9). Interestingly, the plasmid profile of this replicate (not shown) was identical to the other two, indicating that specific changes in antigenic profile do not invariably accompany specific changes in plasmid profile.

Figure 9.

. A. Western blots of the third replicate of the P235.1 chain of infection. Blotting patterns were obtained by subjecting whole-cell lysates from representatives in the experimental chain of infection to SDS–PAGE, transferring the protein bands to nitrocellulose and immunoblotting with the monoclonal antibodies shown. B. Western blot corresponding to (A) only using polyclonal antibody to OspC. Nomenclature for the sequential isolates from left to right follow as for Fig. 1. P, the parental clone used as inoculum to infect M1, the marsh rice rat (Oryzomys palustris) that was subsequently used to infect T1, the larval Ixodes scapularis tick that was subsequently used to infect M2, the white-footed mouse (Peromyscus leucopus) that was fed upon by T2, the nymphal Ixodes scapularis tick. Reactivity to various monoclonal antibodies is denoted on the right of each blot. Antigenically reactive OspB and OspC are denoted by arrows. The bottom arrows denote a smaller, yet immunologically reactive, form of OspB (OspB′) as determined using polyclonal OspB antiserum.

Discussion

The host-associated changes noted in strain P235.1 demonstrate that selection in mammals and ticks can give rise to stable, readily distinguishable B. burgdorferi populations. These populations were recapitulated upon sequential passage, with one population being characteristic of ticks and the other of mammals. This finding opens the possibility that selection plays a role in the normal transmission of B. burgdorferi. Whereas we could not document distinct host-associated populations for one of our isolates (P64.1), such documentation may rely heavily upon the choice of phenotypical characters to follow for a particular strain and the stability of the characters in the absence of selection.

Periodic selection is a feature of the pathogenesis of a number of eukaryotic and prokaryotic microorganisms. For example, Neisseria gonorrhea (Seifert, 1996), Trypanosoma brusi (Pays, 1991), Plasmodium falciparum (Reeder and Brown, 1996) and Borrelia hermsii (Restrepo and Barbour, 1994) all exhibit well-documented forms of antigenic variation involving the selection of distinct subpopulations. Further, Zhang et al. (1997) recently demonstrated the existence of a recombination capability in B. burgdorferi similar to that seen in B. hermsii, which promotes extensive antigenic variation of a surface-exposed lipoprotein, VlsE. However, in these cases, the selective pressure is applied by the vertebrate host's immune system. More in line with our observations, Munderloh et al. (1993) noted changes in the plasmid profile of one cloned strain of B. burgdorferi after co-culturing the spirochaetes with a tick cell line and subsequently inoculating them into hamsters. Also, Hu et al. (1992) described antigenic changes associated with seven out of nine B. burgdorferi strains after introduction of the strains into ticks and later reisolation. However, there have been no studies describing the behaviour of B. burgdorferi populations during multiple sequential selection.

One of the striking features of the chain of infection was that the antigenic and plasmid profiles associated with mammal or arthropod P235.1 isolates were recapitulated identically at each alternating link in the chain. This recapitulation may have been caused by independent selection events in each new host that generated populations with virtually identical antigen and plasmid phenotypical signatures peculiar to those selective pressures. That is, variants with a variety of signatures arose constantly, and only two particular ones were selected. At the opposite extreme, it may be that two principal subpopulations were formed early in the chain of infection and that subsequent host-associated changes reflect the expansion and contraction of the siblings of those two populations. This latter explanation assumes that a tick's blood meal is large enough to pass on minor representatives of the host's B. burgdorferi population. From the size estimates of the tick's blood meal, this possibility cannot be ruled out (Piesman et al., 1990).

Neither of the above explanations is particularly compelling, especially as each step in the chain involved multiple hosts and was repeated three times with only a small variation in the antigenic recapitulation pattern of one replicate. Perhaps a more probable explanation for both the striking uniformity seen in the populations emerging from all hosts sampled at a single link and the striking recapitulation of populations at alternating links would be to assume that there is more at work than random selection. For example, a site-specific recombinase may be more active in a tick or a mammal and may tend to promote certain DNA rearrangements. Our observations of the 9.0 kb plasmid rearrangements and the allelic variation in ospB and ospC may be representative of such an activity. If such rearrangements were accompanied by a change in gene expression advantageous in one type of host, selective pressure may be sufficient to generate the repeating pattern we observed. (This assumes that in vitro growth is selectively neutral in order to preserve the genotypical changes seen in the host.) Such environmental modulation of recombinase activity has been proposed for other microorganisms (Gally et al., 1993; Zhao et al., 1997) and provides a testable hypothesis to account for our observations.

At the molecular level, we suspect, although we cannot prove, that the changes in plasmid profile result from rearrangements in the form of multimerization and (or) recombination (collectively referred to hereafter as rearrangements) of the 9.0 kb circular plasmid. The formation of concatenated supercoiled plasmids has been observed previously in B. burgdorferi (Simpson et al., 1990a; Dunn et al., 1994). Also, repetitive sequences have been noted in both linear and circular plasmids of B. burgdorferi (Simpson et al., 1990b; Dunn et al., 1994; Porcella et al., 1996; Zückert and Meyer, 1996; Fraser et al., 1997), providing homologous areas for recombination. Further, Hyde and Johnson (1988) have reported the existence of a stable dimer of a 4.6- kb plasmid whose properties appear to be very similar to our 9.0 kb plasmid. Our 9.0 kb plasmid was similar to an 8.4 kb circular plasmid in B. burgdorferi strain SH-2-82 described earlier by Simpson et al. (1990a). This similarity was determined using a probe (kindly supplied by T. G. Schwan) from the 8.4 kb plasmid (data not shown). The precise relationship of our 9.0 kb plasmid to the 8.4 kb and other interesting and similarly sized B. burgdorferi plasmids (Champion et al., 1994; Dunn et al., 1994; Xu et al., 1996; Fraser et al., 1997), is yet to be determined.

Considering the opportunity for recombination and the different selective pressures encountered, we were concerned that plasmids would be lost during the infection chain. However, we found that all apparent losses may be accounted for by simple plasmid rearrangements. The reason the plasmids do not appear to be lost may stem from the presence of genes required for in vivo growth (in tick and mammal) on the plasmids, thus making plasmid loss lethal (Margolis et al., 1994; Fraser et al., 1997). The factors of plasmid-loss lethality and the propensity for plasmid rearrangement may affect the interpretation of reports of plasmid loss after prolonged zoonotic passage (Golde and Dolan, 1995).

The only pronounced external phenotypical changes that were noted during the chain of infection were in the alternating expression of different antigenic forms of OspB and OspC. As OspC is encoded on a circular plasmid of a size similar to our putative 9.0 kb trimer (28 kb; Marconi et al., 1993; Sadziene et al., 1993), we were careful to examine the possibility that rearrangements of the plasmid encoding OspC were taking place (and thus could be responsible for the variation we saw in different hosts). Our results, using an ospC gene probe (kindly provided by P. Rosa), confirmed the 28 kb plasmid location of ospC, but this plasmid did not show any change in size during the chain of infection (data not shown). However, our evidence from sequence analysis of PCR amplicons from all host isolates indicated the presence of two allelic forms of ospB and ospC.

The two ospB allelic forms differed little and, although they showed the reciprocating changes associated with a particular host, these changes did not match with the reciprocating antigenic changes. That is, they appeared to be out of phase with the antigenic changes. Specifically, antigenic variation of OspB occurred at the link between the parental and the first mammalian isolate, yet no change in ospB sequence is seen. As the antigenic variation seen could not be correlated to a change in DNA sequence of the ospB gene, we speculate that transcriptional or translational modifications may be involved. Additionally, OspB variation was not as consistently noted in replicates of the chain of infection as was OspC (see Fig. 9), suggesting that a more complicated or random mechanism of variation may be responsible for OspB antigenic variation.

For ospC, the presence of one of two distinctly different allelic forms correlated well with the reciprocating antigenic changes. This correlation strongly supports the notion that ospC allelic variation is responsible for OspC antigenic variation. The pronounced internal differences in the two ospC alleles allowed further characterization of the mechanism of their reciprocating appearance in specific hosts. We found that, although full-length amplicons of only one allelic form predominated at each link in the chain of infection, both tick- and mammal-specific internal portions of each allele were present in all isolates examined. The unamplified copy of ospC in each isolate evidently lacked one or both ends needed for primer binding and amplification of the full-length gene. These data support a model of recombinational activation of an incomplete copy of ospC to generate a full-length allelic variant that is then expressed. Such a mechanism is reminiscent of Neisseria gonorrhoeae pilin antigenic variation (Seifert, 1996).

Heterogeneity of the ospC gene sequence is well documented (Rosa et al., 1992; Livey et al., 1995). However, it is thought to be caused by lateral gene transfer and evolutionary divergence (Jauris-Heipke et al., 1995). Nevertheless, our results indicate the possibility of an antigenic variation mechanism involving recombination between two (at least) resident copies of the ospC gene. Such a recombinatorial source of variation is not unusual in Borrelia sp. For example, the gene encoding the variable major protein (vmp) of B. hermsii as well as the vmp-like vlsE gene of B. burgdorferi displays antigenic variation resulting from gene recombination (Restrepo and Barbour, 1994; Zhang et al., 1997). Whereas the changes in OspB and OspC antigenic form may play an important role in the mammal–tick transmission process, other unscored genotypical changes may also be involved in determining fitness and in producing the strikingly different phenotypical patterns we observe in different hosts.

Our results have shown that selective pressures change the populations of B. burgdorferi in mammalian and arthropod hosts. This finding is consistent with microbes faced with dramatically different environments. One possible benefit for the microbe in its transmission from vertebrate to vertebrate via a vector would be that the population emerging from a variety of vertebrate hosts would be antigenically ‘reset’ in the tick. For example, a B. burgdorferi population emerging from the reptile Eumeces inexpectatus (five-lined skink) (Levin et al., 1996) and selected in this host would probably be at a disadvantage in directly infecting a mouse or a human. The tick would provide the selective pressure for at least the elimination of the previously selected population. It might also produce a population of microorganisms more generally infectious for vertebrates. In this latter respect, selection would be reminiscent of the differentiation required in eukaryotic parasites to assume a form infective for mammals while in an invertebrate vector.

Experimental procedures

Obtaining clonal isolates of B. burgdorferi and in vitro cultivation

A wild-type spirochaete isolate (C235), was obtained from the blood of a raccoon (Procyon lotor ) that was live-trapped on Marine Corps Base, Camp Lejeune, Onslow County, NC, USA (Ouellette et al., 1997). The other isolate (C64) was obtained from a male I. scapularis tick collected at Pine Island, Currituck County, on the Outer Banks of North Carolina, USA. Both isolates were identified as B. burgdorferi sensu stricto by a species-specific PCR assay (Johnson et al., 1992) using a nested primer for a conserved region of the gene that encodes for flagellin (J. R. Ryan, PhD thesis, North Carolina State University).

The isolates were cloned by a previously described solid plating method (Kurtti et al., 1987; Bundoc and Barbour, 1989). Briefly, Barbour–Stoenner–Kelly (BSK) II medium was modified by the addition of rifampin and phosphomycin (Sigma) and fungizone (Gibco BRL). The medium was warmed to 50°C, and an equal volume of 3% agarose (autoclaved and kept molten at 65°C) was added. The mixture was inverted several times, and 8.5 ml was added to a polystyrene Petri dish. The plates were kept in a candle jar at 35°C overnight before use. Numerous BSK-agarose plates were spread with 100 μl of the spirochaete suspension in liquid BSK II mixture and incubated at 35°C in a candle jar. After 2 weeks, the BSK-agarose plates were examined with a dissecting microscope for the presence of colonies. Single colonies were selected from plates that gave 3–10 distinct colonies and were passed to liquid BSK II medium for expansion.

Protocols for in vitro cultivation have been described previously (Barbour, 1984). Briefly, spirochaetes for needle inoculation and all experimental reisolates were expanded in BSK-H (Sigma) supplemented by the addition of gelatin (7%), rabbit serum (6% v/v) and the antibiotics, fungizone (Gibco BRL) and kanamycin (8.0 μg ml−1; Sigma; Wittenbrink et al., 1994). Cultures were maintained at 35°C and grown to a density of 2 × 108 cells ml−1. Spirochaetal density was determined using a Petroff–Hauser chamber.

Sources of mammalian hosts and ticks

Marsh rice rats (Oryzomys palustris) were obtained from a pathogen-free outbred colony at the Naval Dental Research Laboratory, Great Lakes, NY, USA. White-footed mice (Peromyscus leucopus) were obtained from a pathogen-free outbred colony at Harvard School of Tropical Medicine and Hygiene, Boston, MA, USA. A previously described standard system for infecting I. scapularis nymphs was used (Piesman, 1993). Briefly, ticks were derived from a laboratory-reared colony; the parental ticks originating from Fort Bragg, Cumberland County, NC, USA. Ticks were reared through three successive generations before use and were blood fed on non-infected New Zealand white rabbits. Transovarial transmission of B. burgdorferi is inefficient (Piesman et al., 1986), and any B. burgdorferi infection in the parental generation was eliminated through successive rearings. An indirect immunofluorescence antibody assay (IFA) using a species-specific monoclonal antibody to OspA (H5332) (Barbour and Schrumpf, 1986) was used to confirm that fourth generation larvae were not infected.

Infection by tick bite and reisolation of spirochaetes

Initially, marsh rice rats were infected by needle inoculation. An inoculum of 0.6 ml containing the minimum infectious dose (2 × 104 spirochaetes of C235.1 and 2 × 106 spirochaetes of C64.1, determined previously in an outbred ICR mouse strain) was injected intradermally (Gern et al., 1993). Inoculum concentrations were adjusted by the method of Stoenner (1974) using a Petroff–Hauser cell chamber.

Ticks were infected with B. burgdorferi by feeding larvae on marsh rice rats that had been infected by needle inoculation. Experimental animals were held for 28 days before ticks were placed on the ears of the rats. All animals were injected intramuscularly with a 0.1 ml ketaset–rompum mixture before placing 100 uninfected larvae on the animal's head and ears. The animal cages were set over metal pans filled with 0.25–0.5 inches of water to capture engorged ticks dropping off each animal.

Engorged larvae were removed twice daily from the water-containing pan. When not feeding, ticks were maintained at 97–100% relative humidity (Winston and Bates, 1960) in a vial containing a moistened Plaster of Paris/powdered, activated charcoal base (Farrell and Wharton, 1948) and held in an incubator with a 12 h:12 h (light–dark) cycle at 19°C. Some of the ticks were processed immediately to determine percentage infection by indirect immunofluorescence assay (IFA) microscopy with a species-specific monoclonal antibody (H5332), as described by Levine et al. (1989), thus confirming rice rat infection through xenodiagnosis. Other ticks were held till moulting was complete.

A small portion of the infected nymphs were homogenized for strain isolation as described below. The remaining infected nymphs were used to infect experimental animals by placing 10 ticks on the head and ears of an anaesthetized animal. Infected nymphs were allowed to feed to repletion on non-infected white-footed mice (n = 12) and rice rats (n = 12). Twenty-eight days after all the nymphs had dropped off, non-infected nymphs were fed upon the infected rodents. These engorged ticks were allowed to moult to adults, and the proportion infected was determined by IFA. Reisolation of spirochaetes in BSK-H media was made from each step in the chain of infection. Spirochaetes were recovered from rodent ear tissue and urinary bladder by aseptically placing a sample of the tissue in augmented BSK-H media (synonymous with BSK-II; Schwan et al., 1988). Spirochaetes were isolated from freshly moulted ticks by first sterilizing the outer surfaces of the tick and then dissecting and placing a sample of mid-gut in BSK-H media.

All rodents were euthanized by carbon dioxide intoxication, and samples were taken aseptically from ear and urinary bladder tissues and placed in 8 ml culture tubes of liquid BSK-H. Cultures were checked periodically for spirochaetes by darkfield microscopy. Ticks were dissected and examined for infection by IFA (Levine et al., 1989).

Reisolates were selected indiscriminately from each tick and rodent series and recloned by the previously described solid plating method (Kurtti et al., 1987). Five clones were picked from each plate, passed once in BSK-H media and grown to late-log phase. Spirochaetes were concentrated by centrifugation and resuspended in a minimal volume of BSK-H containing 20% glycerol. Spirochaete resuspensions were placed in cryovials and stored at – 80°C until needed. Two clones from each series were selected randomly from freezer stocks and included in comparative analyses of experimental replicates to parental and original cloned strains (see Figs 2 and 3).

Spirochaetes used for molecular comparison met the following criteria: (i) the mammal reisolates selected for sampling came from the urinary bladder; (ii) only first passage reisolates were used for the comparisons; (iii) antibiotics other than kanamycin and fungizone were not used to modify the media; and (iv) all reisolates were checked for contaminants by plating samples on the following four non-specific media and looking for growth (blood agar plates, trypticase soy agar slants, Sabouraud-dextrose agar slants, thioglycollate tubes). The same lots of commercially procured BSK-H media and rabbit serum were used to reduce variation in the protocol.

SDS–PAGE and Western blot analyses

Protein profiles of the pre- and post-infection reisolates were analysed using SDS–PAGE and Western blotting with a monoclonal antibody (mAb) ladder. Spirochaete cultures were pelleted by centrifugation at 10 240 g for 20 min at 4°C and washed three times with phosphate-buffered saline/5 mM MgCl2 (PBS/M). The final pellet was resuspended in a minimal volume of PBS/M to give an optical density reading of 2.0 at 600 nm in a spectrophotometer. Whole-cell lysates were prepared by mixing equal volumes of the adjusted cellular suspension with 2 × SDS–PAGE sample buffer and boiling the mixture for 5 min.

Whole-cell lysate samples were run in a discontinuous SDS–PAGE system using a vertical slab gel electrophoresis system (Bio-Rad). The stacking and resolving gels were 4.0% and 12.5% bis-acrylamide respectively (Laemmli, 1970). Gels were subjected to electrophoresis at a constant current of 15 mA until the dye front was 1 cm from the end of the plate. Protein bands from the gel were transferred to nitrocellulose (Towbin et al., 1979) and immunoblotted with the monoclonal antibody mixture to 10 immunogenic and common proteins found in B. burgdorferi. Of the 10, eight reacted with our B. burgdorferi isolates. The mAb mixture was developed by the Centers for Disease Control and Prevention (Craven et al., 1996) for diagnostics in determining human infections and contains the following mAbs specific to the following proteins and generated by the indicated contributors: 181.1, 93 000 MW (protoplasmic cylinder), B. Luft; CB312, 72 000 MW (DNA k), J. Benach; 149, 66 000 MW (GroEL), B. Luft; H9724, 41 000 MW (flagellin), A. Barbour; FIBA6E11, 39 000 MW (in periplasmic sheath), T. Schwan; 84C, 34 000 MW (outer surface protein B), D. Thomas; H5332, 31 000 MW (outer surface protein A), A. Barbour; H1C8, 29 000 MW (outer surface protein D), A. Barbour; 4B8F4, 23 000 MW (outer surface protein C), S. Padula; and CB625, 22 000 MW (in periplasmic space), J. Benach (Johnson et al., 1995).

DNA purification and characterization

One millilitre of bacterial suspension (approximately 108 cells) was subjected to the Qiagen plasmid extraction kit (tip 20) protocol to obtain plasmid DNA. The eluted plasmid fractions were concentrated by ethanol precipitation and applied to a 0.4% agarose gel (Schwan et al., 1988). Gel electrophoresis was performed in 40 mM Tris-acetate–2 mM EDTA (TAE) buffer at 50 V for 10 min and at 12 V for 16 h. Restriction fragments of whole λ DNA (Sigma) were used as size markers. The DNA in the gel was then stained with ethidium bromide and examined with an ultraviolet transilluminator.

Electron microscopy of plasmid DNA

Excised gel slices of low melting point agarose were subjected to beta-Agarase I (New England BioLabs) digestion in accordance with the manufacturer's instructions. The DNA in the agarase reaction mixture was prepared for electron microscopy using the Kleinschmidt aqueous spreading technique (Garon, 1986). Each mounted sample was viewed in a JEOL 100B transmission electron microscope at 8000 × magnification. Plasmid DNA molecules were sized by measuring contour length intervals using pBR322 as a calibrating control. Contour length measurements were measured with a Numonics graphic calculator interfaced to a Tektronix 4052A computer.

Southern blot analysis

Equal amounts of DNA were first subjected to electrophoresis in 0.4% agarose gels with TAE buffer. The gel was then stained with ethidium bromide and the DNA transferred to Gene Screen Plus membranes (Dupont, NEN Research Products) without further manipulation according to the method of Southern (1975). Membranes were prepared for DNA probe hybridization as described previously (Schwan et al., 1989). The probe DNA was prepared from a 9.0 kb supercoiled plasmid band by labelling with [α-32P]-dCTP using nick translation (Maniatis et al., 1982). Membranes were hybridized and washed at high stringency as described previously (Schwan et al., 1989). Kodak X-Omat film was exposed to the membranes at – 70°C with an intensifying screen and developed with a Kodak X-Omat M20 processor.

Sequence analysis of the ospB and ospC genes

The coding regions of the ospB and ospC genes were amplified by the PCR containing 200 μM dATP, 200 μM dCTP, 200 μM dGTP, 200 μM dTTP, 1 × PCR buffer (10 mM Tris, pH 8.3, 50 mM KCl, 0.001% BSA, 1.5 mM MgCl), 1.25 U of AmpliTaq Gold DNA polymerase (Perkin-Elmer) and 1 μM each primer. The primers used for the amplification of the full-length ospC gene were C1 (5′-ATG AAA AAG AAT ACA TTA AGT GCG; 1–24) and C2 (5′-TTA AGG TTT TTT TGG ACT TTC TGC; 633–610). The primers used for amplification of the full-length ospB gene were B1 (5′-ATG AGA TTA TTA ATA GGA TTT GCT TTA GC; 1–29) and B2 (5′-TTT TAA AGC GTT TTT AAG CTC TGA AAG; 888–862). The target DNA consisted of 5 μl of frozen B. burgdorferi in BSK-H containing 20% glycerol. Reaction mixtures were incubated at 95°C for 12 min before cycling at 95°C for 1 min, 50°C for 1.5 min and 72°C for 2 min for 30 cycles, followed by extension at 72°C for 7 min. Another 1.25 U of AmpliTaq Gold was added, and the reactions were cycled under the same conditions for an additional 30 cycles. An aliquot was visualized by agarose gel electrophoresis and ethidium bromide staining for the presence of a single band. The sample was then purified through a Centricon-100 (Amicon) as recommended by the manufacturer.

Sequence analysis of the ospC and ospB amplicons was performed at the UNC-CH Automated DNA Sequencing Facility on a model 377 DNA sequencer (Perkin-Elmer, Applied Biosystems Division) using the ABI Prism dye terminator cycle sequencing ready reaction kit with AmpliTaq DNA polymerase, FS (Perkin-Elmer, Applied Biosystems Division). Sequencing of the ospC amplicon was performed with the amplification primers C1 and C2. Sequencing of the ospB amplicon was performed with the amplification primer B2 as well as sequencing primers B1.3 (5′-AAA AGG TTG TTG AGT CAA TTG GTT C; 54–77), B1.4 (5′-AGT GGA AAT TAA AGA AGG TAC TGT TAC TC; 606–634) and B2.2 (5′-CTT AAA TCA TAT TTG CCG GAG C; 239–218). Subsequent analysis was performed using the Multiple Alignment Construction and Analysis Workbench (MACAW) program (Schuler et al., 1991).

PCR analysis

The internal variable region of the ospC genes was amplified by the PCR containing 200 μM dATP, 200 μM dCTP, 200 μM dGTP, 200 μM dTTP, 1× PCR buffer (10 mM Tris, pH 8.3, 50 mM KCl, 0.001% BSA, 1.5 mM MgCl), 1.25 U of AmpliTaq Gold DNA polymerase (Perkin-Elmer) and 1 μM each primer. The primers used for the amplification of the mammal-specific variable portion of the gene were pcr1-m (5′-AAG CAG GTG GTA CTT TAG GTA GC; 242–264) and pcr2-m (5′-TCA TCA TCA CTA GCA TCC TGT TTG; 461–438). The primers used for the amplification of the tick-specific variable portion of the gene were pcr1-t (5′-AGC AAA ATG GTT TGG GTG CT; 242–264) and pcr2-t (5′-AGT AGC ATT ACC AGC AGC CAC T; 461–438). The target DNA consisted of 5 μl of frozen B. burgdorferi in BSK-H containing 20% glycerol. Reaction mixtures were incubated at 95°C for 12 min before cycling at 95°C for 1 min, 50°C for 1.5 min and 72°C for 2 min for 30 cycles, followed by extension at 72°C for 7 min. Another 1.25 U of AmpliTaq Gold DNA polymerase (Perkin-Elmer) was added, and the reactions were cycled under the same conditions for an additional 30 cycles. An aliquot was visualized by agarose gel electrophoresis and ethidium bromide staining.

Footnotes

  1. Present address: Department of Entomology, Division of Communicable Diseases and Immunology, Walter Reed Army Institute of Research, Washington, DC 20307, USA

  2. Present address: Entomology Branch, Centers for Disease Control and Prevention, Atlanta, GA 30341, USA

Acknowledgements

We thank Tom Schwan and Merry Schrumpf of the National Institutes of Health, Rocky Mountain Laboratories for providing assistance in Southern blot analyses. We acknowledge May Chu, Barbara Johnson and Joe Piesman of the Centers for Disease Control (CDC), Division of Vector-Borne Infectious Diseases Laboratory for providing the monoclonal antibodies used throughout this study and providing facilities and assistance in preliminary studies. We also thank Christine Happ (CDC) for kindly assisting in rodent infectivity trials. Finally, our thanks to Craig Altier, John Horton and Terri Hamrick for their comments and suggestions. Funds to support this work were provided by the Centers for Disease Control and Prevention and the State of North Carolina.

Ancillary