INTRODUCTION

Top Abstract Introduction Materials and Methods Results and Discussion References

Borrelia burgdorferi, the most common arthropod vector-borne pathogen in North America, is maintained in a complex enzootic cycle involving Ixodes scapularis and a mammalian host, typically rodents (28, 40). B. burgdorferi is distinguished from other prokaryotes by the extraordinary complexity of its genome, which is composed of a linear chromosome and a variable plasmid complement consisting of as many as 21 linear and circular plasmids (6, 18, 20). Two lines of evidence implicate plasmid-encoded proteins in the physiological and antigenic changes which underlie the Lyme disease spirochete's adaptation to different host environments as well as virulence expression within the mammalian host. First, most of the differentially expressed borrelial genes identified to date are plasmid encoded (2, 3, 11, 12, 19, 26, 39, 43, 48, 51, 53, 58). Second, loss of plasmids during continuous in vitro cultivation results in diminished infectivity (35, 42, 61, 63). B. burgdorferi plasmids are also of interest because they contain multiple paralogous genes which appear to have undergone extensive recombination during their evolution (2, 5, 8, 10, 15, 20, 39, 45, 49, 64).

The publication of the near-complete genome of a B. burgdorferi B31 clone (20) was a major advance in our understanding of this pathogen's complex molecular and evolutionary biology. The entire genome of this strain was not published, however, due to the difficulties associated with the sequencing and assembly of a family of highly homologous 32-kb circular plasmids (10, 52). Recently, complete sequences for seven of these plasmids have become available from the B. burgdorferi genome database (BBGD; http://www.tigr.org). Each contains a large number of well-conserved open reading frames (ORFs) as well as two variable loci with genes encoding differentially expressed, polymorphic lipoproteins, some of which have been shown to be surface exposed (2, 3, 13, 19, 27, 32, 39, 49, 51, 53-55). Truncated forms of these supercoiled plasmids (designated cp18s) have been described for the N40 and 297 strains but not fully characterized (2, 50).

B. burgdorferi initially was thought to be a single species. In recent years, however, it has become apparent that B. burgdorferi sensu lato is composed of multiple distinct species and genomic groups (4, 24, 31, 59) and that this heterogeneity may relate to differences in disease manifestations (57). Although less marked, genetic diversity also has been detected among the B. burgdorferi sensu stricto isolates which cause Lyme disease in North America (30, 33), and recent findings suggest that this heterogeneity also translates into differences in invasive potential (29, 44, 60). An important implication of these findings is that genetic characterization of multiple B. burgdorferi isolates may be needed in order to understand the relationship(s) between genomic content, borrelial virulence, and disease pathogenesis. To this end, we have extended our previous studies of cp32/18-encoded, differentially expressed lipoproteins (1-3, 39, 62) by comprehensively analyzing the cp32/18 family of plasmids in the 297 strain, a human cerebrospinal fluid isolate (46). In addition to providing evidence for two different types of recombination events involving cp32/18 plasmids, these findings underscore the need for genetic analysis of diverse borrelial isolates in order to elucidate the Lyme disease spirochete's complex parasitic strategies.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results and Discussion References

Bacterial strains and media. Virulent low-passage B. burgdorferi 297, the source of DNA template for all PCR experiments, was passaged and maintained as previously described (1). Electrocompetent Escherichia coli DH5 and DH10B (Gibco/BRL Life Technologies Inc., Gaithersburg, Md.) cells were used for all transformations. Strains and transformants were grown on tryptone-yeast agar or broth supplemented with the appropriate antibiotic.

Nucleotide sequencing. Greater than 85% of the cp18-2 sequence was derived by analyzing overlapping genomic clones from a Lambda Zap II library (Stratagene, La Jolla, Calif.) (3, 39). Oligonucleotides used as probes for library screening or as primers for PCR amplification of duplex probes are shown in Table 1. The region of cp18-2 extending from mlp9 to p21 (Fig. 1) was amplified by long-distance PCR (GeneAmp XL-PCR kit; Perkin-Elmer Corp., Branchburg, N.J.) using the forward and reverse primers mlp 5'-cons and p21-R+C, respectively (2). The PCR parameters consisted of a preincubation at 94°C for 3 min to fully denature the template DNA, 35 cycles of 94°C for 1 min and 58°C for 8 min, followed by a final incubation at 72°C for 10 min; PCR products were gel purified using the Geneclean glass milk procedure according to the instructions of the manufacturer (Bio101 Inc., Vista, Calif.). Nucleotide sequencing was performed with an Applied Biosystems Inc. model 373A automated DNA sequencer and PRISM ready reaction DyeDeoxy terminator cycle sequencing kits as instructed by the manufacturer (Applied Biosystems, Inc., Foster City, Calif.).

View larger version (7K): [in this window] [in a new window]   FIG. 1.   Schematic linear representation of cp18-2. ORFs are represented as open arrows. Striped arrows represent ORFs belonging to paralogous families in the B31 strain; the boxed numbers above each arrow indicate the Borrelia paralogous family to which the ORF belongs (BBGD). Black arrows designate the previously characterized, differentially expressed p21 and Mlp9 lipoproteins. Predicted leader peptides (LP) are represented by curved lines; IRs are indicated as solid arrowheads. The names of previously characterized proteins are given above the corresponding ORFs. Asterisks indicate possible pseudogenes. The approximate distances in base pairs are indicated below.

PCR amplification of the truncated portion of cp18-2. Long-distance PCR to obtain the deleted portion of cp18-2 was performed as described above, using the forward and reverse primers listed in Table 1. The products were gel purified for sequencing as previously described.

Linkage of elpA1, elpA2, bbk2.10, and bbk2.11 with mlp genes and analysis of downstream variable regions. Long-distance PCR to link elpA1, elpA2, bbk2.10, ospF, and bbk2.11 with mlp genes was performed as described above, using the forward and reverse primers listed in Table 1. The variable regions downstream from elpA1, elpA2, elpB1, ospF, bbk2.10, and bbk2.11 were obtained either as genomic clones or by PCR amplification using the forward and reverse primers listed in Table 1. PCR products were cloned into the pCR2.1-TOPO vector (Invitrogen, Carlsbad, Calif.) and sequenced as described above.

Computer analysis. Routine sequence analysis and multiple sequence alignments were performed using the MacVector version 6.5 software (Oxford Molecular Group Inc., Campbell, Calif.). ORFs were analyzed for potential signal peptides and cellular localization using PSORT (34). Transmembrane domains were predicted with the TMpred program (23). To generate an unrooted phylogenetic tree, pairwise sequence alignments were first generated using the ClustalW program (56). Sequence alignments were then analyzed using the Protdist and Neighbor programs included in the PHYLIP program suite (17). The resulting neighbor-joining tree was viewed using TreeView, version 1.5 (36). Binary comparisons with randomizations were generated using the GAP program from the Genetics Computer Group (Madison, Wis.) sequence analysis software package, version 10.0. Shuffle scores were calculated by subtracting the average quality score based on 100 randomizations from the quality score of the pairwise alignment; the adjusted quality score of the pairwise alignment then was divided by the standard deviation of the average quality score of randomizations. Scores are expressed as the number of standard deviations, with scores 9 indicating statistically significant alignments (41).

Nucleotide sequence accession numbers. The entire sequence of 297 cp18-2 was submitted to GenBank under accession no. AF169008. Revised sequences for the variable ospE/ospF/elp loci have GenBank accession no. U18292 (bbk2.10), U30617 (bbk2.11), U19754 (ospF), AF023852 (ospE/elpB1), AF023853 (p21/elpB2), AF077602 (elpA1), and AF077603 (elpA2). Accession numbers for genes encoded within the 2.9 loci can be found in references 39 and 62.

	RESULTS AND DISCUSSION

Top Abstract Introduction Materials and Methods Results and Discussion References

Strategy for molecular characterization of seven cp32/18 plasmids in B. burgdorferi 297. The availability of complete sequences for seven cp32 plasmids from B. burgdorferi B31 (from the BBGD; see above), in concert with our own previous studies (2, 3, 39, 62), enabled us to devise a strategy for a molecular and evolutionary characterization of seven homologous plasmids in B. burgdorferi 297. The strategy was based on the assumption that, as in strain B31, each plasmid consists of large, highly conserved stretches with two interspersed variable regions. We initiated this effort by determining the entire sequence of a truncated family member, using a combination of overlapping genomic library clones and long-distance PCR products. In addition to its smaller size, this plasmid, designated cp18-2 (2), was an attractive candidate for complete sequence analysis for two reasons: (i) its variable loci encode lipoproteins (p21/ElpB2 and Mlp9, respectively) which are selectively expressed during mammalian infection (1, 2, 62) and (ii) loss of this plasmid occurs early during in vitro passage and correlates with attenuation of virulence in the murine model of Lyme disease (D. R. Akins, M. J. Caimano, and J. D. Radolf, unpublished data). Next, to confirm the overall similarity of cp32 plasmids in B. burgdorferi B31 and 297, we amplified from other strain 297 cp32 plasmids the fragment presumed to have been deleted from cp18-2. The two unique, variable lipoprotein-encoding loci in each plasmid were then fully sequenced and linked by long-distance PCR. Last, these results enabled us to assess how recombination has influenced the evolution of individual lipoprotein families as well as the variable regions within the plasmids.

Complete sequence of cp18-2. (i) Content and organization. A linear representation of cp18-2 is shown in Fig. 1. The plasmid contains a total of 21,170 bp, 86% of which is coding sequence. cp18-2 is very similar in overall gene content and organization to its untruncated B. burgdorferi B31 counterparts. This was expected given the high degree of intra- and interstrain similarity already demonstrated for this family of plasmids. Following the convention previously established by Fraser et al. (20); see also the BBGD website), the 31 ORFs were numbered sequentially; the truncated ORF immediately following the deletion junction (described in detail below) was arbitrarily designated ORF01. Interestingly, all but three ORFs are transcribed in the same direction. The salient features of the 31 genes and corresponding polypeptides are summarized in Table 2. Two genes (ORF04 and ORF07), in addition to ORF01, lack identifiable Shine-Dalgarno sequences and therefore may not be translated. All of the genes except ORF16 (rep- in the 2.9 locus [39] could be assigned to a paralogous family from B. burgdorferi B31 (Fig. 1 and Table 2). Paralogs for ORF16 are present in the B31 cp32s but were not annotated (BBGD). The reason for this omission is unclear given that at least one of the rep- genes in B. burgdorferi 297 has been shown to be transcribed during in vitro growth (39).

(ii) Genomic distribution of paralogous ORFs. The B. burgdorferi genome, and its plasmid components in particular, has undergone extensive duplications and rearrangements throughout its evolution (5, 8, 15, 20, 39, 45, 49, 64). Lacking the genomic sequence for strain 297, we used the B31 sequence to examine the genomic distribution of cp18-2 paralogos. This analysis revealed that cp18-2 is a mosaic or patchwork consisting of small, plasmid-specific regions and a number of paralogous sequences distributed to various extents throughout the genome (Fig. 2 and Table 2). The widely distributed ORFs can be grouped into five categories. The first contains two genes (ORF19 and ORF22) with chromosomal, linear, and circular plasmid-encoded paralogs. The second contains two genes (ORF20 and ORF21) with paralogs present on nearly all of the B31 plasmids but not the chromosome. The third contains several ORFs with paralogs located on cp9 and the cp32s. The fourth contains members of 10 paralogous families also represented only on lp54, seven of which (ORF04 to ORF10) form a contiguous stretch on both plasmids (Fig. 2 and Table 2). The fifth group is heterogeneous and contains paralogs present on one or more plasmids (linear and/or circular) other than cp9 and lp54.

View larger version (40K): [in this window] [in a new window]   FIG. 2.   Distribution of cp18-2 paralogous families within the B. burgdorferi B31 genome.

(iii) Homologies with other proteins. Database searches revealed that only three ORFs matched proteins with known functions. Two of these (ORF11 and ORF12; previously designated blyA and blyB, respectively) are both cp32/18 and B. burgdorferi specific and encode the hemolytic activity described by Guina and Oliver (21). The third gene (ORF21; previously designated orfC [64]) belongs to a paralogous family (family 49) whose members recently were shown by Stevenson et al. (49) to be related to the ParA/SopA protein involved in the partitioning of low-copy-number plasmids between daughter cells following cell division (22). As with the B31 paralogs, ORF21 contains the ATP-1 and ATP-2 motifs required for binding and hydrolysis of ATP as well as two motifs postulated to be involved in the binding of accessory proteins or in membrane attachment during partitioning (data not shown) (49). Stevenson and coworkers (49) also proposed that sequence variation among ParA/SopA homologs prevents incompatibility of cp32 homologs within individual borrelial strains; studies are currently under way to examine the sequence heterogeneity of the strain 297 paralogs. In par-mediated plasmid partitioning, ParA/SopA interacts with a protein, ParB/SopB, which complexes to a cis-acting site, parS/sopC, near the origin of plasmid replication (22). The applicability of this scenario to the partitioning of borrelial plasmids, however, is uncertain given that directed searches failed to identify a homolog for ParB or a site resembling parS/sopC in cp18-2 or, for that matter, in the B. burgdorferi B31 genome. If these functions are encoded on cp18-2, the relevant sequences evidently lack detectable homologies.

(iv) Export signals and transmembrane domains. The 31 ORFs were analyzed for N-terminal export signals, transmembrane domains, and cellular location, using the PSORT and TMpred algorithms. A summary of these data are included in Table 2. PSORT identified N-terminal export signals on eight polypeptides (ORF11 [blyA], ORF13 [orfC], ORF14 [orfD], ORF16 [rep-], ORF17 [mlp9], ORF28 [p21], ORF29 [elpB2], and ORF30), seven of which have been characterized previously (2, 21, 39, 62). It is noteworthy that PSORT failed to recognize the signal peptidase II (SPase II) cleavage sites for two of the three lipoproteins (Mlp9 and ElpB2). Although lipid modification of ElpB2 has not been confirmed experimentally, an alkaline phosphatase fusion containing the Mlp9 leader peptide is lipidated when expressed in E. coli (39). Of the five nonlipoproteins with predicted leader peptides, only ORF14 and ORF16 were predicted to have SPase I cleavage sites. In both cases, however, these predictions are questionable. The putative leader sequence of ORF16 has three prolines within the hydrophobic core region and a predicted cleavage site 30 amino acids downstream from the end of the hydrophobic stretch. With respect to ORF14, previous analyses using Triton X-114 phase partitioning and in vivo lipid labeling of the recombinant protein expressed in E. coli indicated that the protein's leader peptide also is not cleaved (39). TMpred identified potential internal transmembrane domains, predictive of a cytoplasmic membrane location, in 10 ORFs, while one (ORF16) was predicted to have two internal transmembrane domains in addition to its leader peptide. PSORT, in contrast, failed to identify the putative transmembrane domains for 6 of these 10 polypeptides, placing them instead within the cytoplasmic compartment. These discrepancies probably reflect the relatively low TMpred scores (<1,000) for all seven membrane-spanning regions. The two proteins predicted by both PSORT and TMpred to possess internal transmembrane domains were ORF15 and ORF23. These two polypeptides are members of the family (family 80) of borrelia direct repeat (Bdr) proteins (formerly rep⁺ within the 2.9 loci [39] and orfE [64]) which characteristically possess an internal, tandemly repeated motif with K-I-D core sequence and C-terminal cytoplasmic membrane anchors (9, 65). The above analyses clearly highlight the need for caution when interpreting computer-based predictions of protein topology and cellular location. Equally important, they indicate that most of the coding capacity of the plasmid is devoted to polypeptides residing within intracellular compartments.

(v) Inverted repeats. Flanking the stretch from ORF18 to ORF24 are 185- and 191-bp inverted repeats (IRs) which are 42% identical to the 183- and 184-bp IRs on cp8.3 of B. burgdorferi Ip21 (15) (Fig. 1 and 3). The cp18-2 IRs are similar to their cp8.3 counterparts in that they overlap the Shine-Dalgarno sequences and translational starts of the adjacent genes (Fig. 3). They differ, however, in that they are theoretically unable to form stem-loop structures comparable to those which were proposed to interfere with translation of cp8.3 transcripts (15) (data not shown). Three of the genes (ORF19, ORF20, and ORF22) flanked by the IRs in cp18-2 are paralogs for the genes flanked by the IRs in cp8.3; it is noteworthy that the parA/sopA homolog is one of the two genes missing from this region in cp8.3 (Fig. 3). A survey of the B31 genome revealed that similar IRs are found only on the cp32 family and cp9 and that the cp32 IRs flank the same five ORFs as in cp18-2 whereas the cp9 IRs flank the same three ORFs as in cp8.3. (20), BBGD. Variable combinations of these five ORFs are found in similar configurations on B31 linear plasmids which lack the IRs (Fig. 3). One possible interpretation of these observations is that the IR-flanked genes were acquired initially by cp32/18 and/or cp9 plasmids (perhaps via a transpositional insertion) and that various combinations of these genes subsequently recombined into other plasmids. The absence of the ParA/SopA homologs in cp8.3 and cp9 raises the intriguing possibility that the replication and partitioning mechanisms used by these small plasmids differ from those of the larger supercoiled plasmids. In this regard, it should be noted that plasmids which replicate as rolling circles typically are less than 10 kb (25) and that a B. burgdorferi chromosomal gene (BB0607) encodes a protein related to the DNA helicase involved in rolling circle replication in Bacillus subtilis (37).

View larger version (31K): [in this window] [in a new window]   FIG. 3.   Organization of IRs and/or related flanking ORFs in selected B. burgdorferi plasmids. ORFs are depicted as shaded boxes, with similarly shaded boxes indicating paralogous ORFs; the arrow below each ORF indicates the direction of transcription. Percent similarities between cp18-2 and corresponding paralogs (using amino acid sequences) and IRs are indicated.

Identification of the cp18-2 deletion. Stevenson and coworkers (50) showed that the truncated cp32 homolog in B. burgdorferi N40 lacks an approximate 12.5-kb fragment. However, because their study predated the availability of the B31 cp32 sequences, they were unable to identify the missing ORFs. Early in our sequencing effort, it became apparent that the N40 and 297 truncations differed substantially. This observation prompted us to define the deletion in cp18-2. Using 5' primers specific for each of the separately encoded ospE, ospF, and elp genes and a reverse primer located within ORF01 (Table 1), 9- to 10-kb fragments were amplified from strain 297 DNA (Fig. 4A and data not shown). Additional PCRs using the same 5' primers and nested reverse primers (Table 1) generated smaller fragments of the exact sizes predicted from the 297 and B31 sequences (Fig. 4A and data not shown). Together with the cp18-2 sequence, these results established the high degree of similarity between the strain B31 and 297 cp32s.

View larger version (26K): [in this window] [in a new window]   FIG. 4.   The cp18-2 deletion. (A) cp32 ORFs predicted to be deleted from cp18-2. Small arrows designate the nested primer pairs used to confirm the similarity between the cp18-2 deletion in a full-length 297 cp32 (represented by cp32-3) and the corresponding segments from the B31 homologs (represented by cp32-7). (B) The deletions in cp18-2 and cp18 of N40 are shown alongside a representative B31 plasmid (cp32-7).

Evolutionary analysis of the cp32/18-specific, variable loci. Sequence divergence among cp32/18 plasmids can be attributed primarily to variable regions defined by the genes encoding the Mlp (formerly 2.9-LP [39]) and OspE/OspF/Elp lipoprotein antigens (2). A characterization of these plasmid-specific loci, therefore, was essential for definition of the B. burgdorferi 297 homologs. Sequences for the loci encoding the Mlp lipoproteins were presented in two prior studies (39, 62). In a third report (2), we provided partial sequences for the seven OspE/OspF/Elp-encoding regions and established linkage between the loci containing the ospF, p21/elpB2, and ospE/elpB1 genes and 2.9 loci containing the mlp8, mlp9, and mlp10 genes, respectively (2). To complete this phase of the study, we sequenced all remaining gaps and linked the four unpaired ospE/ospF/elp loci with mlp loci by long-distance PCR.

View larger version (35K): [in this window] [in a new window]   FIG. 5.   Schematic representation of the variable loci in B. burgdorferi 297 (A) and B31 (B). Similarly shaded boxes are used to indicate paralogous ORFs; the arrow below each ORF indicates the direction of transcription. In panel A, paralogs of cp18-2 ORF30 and ORF31 are designated 30 and 31; nonparalogous ORFs are indicated by asterisks.

View larger version (15K): [in this window] [in a new window]   FIG. 6.   Unrooted neighbor-joining phylograms of OspE, OspF, and Elp (A) and Mlp (B) homologs in B. burgdorferi 297 and B31 cp32/18 plasmids. Scale bars represent the base pair substitutions per site.

View larger version (95K): [in this window] [in a new window]   FIG. 7.   Sequence relatedness among the Mlp lipoproteins. The sequences of class I (A), class II (B), and the entire complement of proteins (C) were aligned using the ClustalW algorithm. Shaded residues represent identical or conserved matches.

Summary and conclusions. In the present study, we exploited existing sequence information and the highly conserved nature of the cp32/18 plasmids in order to characterize seven homologs from B. burgdorferi 297. Though clearly sequence intensive, a major advantage of our strategy was that it enabled us to delineate the principal features of an entire family of plasmids without having to completely sequence each member. Inherent in this approach, however, is the need to extrapolate from heterologous genomic sequences. Nevertheless, we believe that any potential inaccuracies due to genomic polymorphisms between strains 297 and B31 are unlikely to alter our major findings and are easily counterbalanced by the plethora of new genetic information pertaining to a highly virulent, neurotropic clinical isolate (46). The need for such information is highlighted by our finding, using both needle and tick inoculation in mouse infectivity tests, that the B31 clone used for the genomic sequence is less virulent than either the parental B31 isolate or B. burgdorferi 297 (M. J. Caimano, D. R. Akins, S. K. Wikel, and J. D. Radolf, unpublished data). While these differences could reflect the loss of unrelated sequences during cloning, it is our working hypothesis that the greater virulence of B. burgdorferi 297 reflects, at least in part, the greater sequence diversity of its cp32/18 plasmids. Pending the development of efficient mutagenesis techniques for virulent borrelial isolates, it may not be possible to directly evaluate this conjecture in the near future. Nevertheless, the work reported here sets the stage for a variety of studies ranging from expression mapping and cellular localization of the cp32/18-encoded gene products in spirochetes cultivated under various in vitro and in vivo conditions to molecular epidemiological correlations of spirochetal genetic diversity and disease expression.