INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Infection with pathogenic species of the Borrelia burgdorferi sensu lato complex can lead to Lyme disease, an infection characterized by highly variable multisystem clinical manifestations (30, 31). Early studies suggested that plasmid-encoded proteins play an important role in Borrelia pathogenesis (26, 28), and in recent years it has been demonstrated that they also play an important role in immune evasion (37, 43). In addition, several plasmid-borne genes have been demonstrated to be up-regulated during infection, supporting a functional role for the proteins in the mammalian environment (1, 11, 39). The ability to define a correlation between specific plasmids and infectivity has until recently been complicated by the inherent instability of the Borrelia plasmids and the variation in plasmid content among isolates (3, 9, 17, 18, 20, 25, 28, 29, 40, 41). Prior to the determination of the complete genome sequence and plasmid content of B. burgdorferi, these features made it difficult to interpret and compare much of the earlier data. Determination of the genome sequence of B. burgdorferi B31MI and other analyses revealed the existence of both linear and circular comigrating plasmids of between 24 and 56 kb that could not be distinguished by agarose gel electrophoresis (5, 10, 15, 35). There are also several distinct yet closely related circular plasmids (designated cp) of approximately 32 kb (cp32) (33, 35) and multiple linear plasmids (designated lp) of approximately 28 kb (lp28s). The cp32s have extended regions of homology, while the lp28s have significantly less identity. In addition to comigrating plasmids, the genome also contains an extraordinary number of plasmid-carried paralogous gene families (n = 175). This genetic redundancy, which could allow for functional complementation of plasmid-encoded proteins, coupled with the inherent instability of the Borrelia plasmids has made it difficult to pinpoint one or more specific plasmids as essential for infectivity (4, 9, 17, 28, 37, 38, 43). However, two recent independent studies by Labandeira-Rey and Skare and by Purser and Norris have provided evidence that lp25 and lp28-1 may be necessary for full virulence (16, 22). These studies clearly demonstrate that the loss of specific plasmids is one mechanism by which decreased virulence can occur. It should also be noted that in a recent study by Siebers et al. (27) it was demonstrated that clonal variations in pathogenicity arise after infection and subsequent subsurface plating of the Lyme disease spirochetes. However, Siebers et al. utilized B. burgdorferi ZS7, an isolate of undefined genomic composition, and did not determine the plasmid contents of the clones with different pathogenicities. Hence, the basis for the loss of infectivity in the clones analyzed was not determined.

In this report, we have continued to assess putative correlations between plasmid content and infectivity and have employed clonal populations of defined genetic composition to achieve this goal. By defining the plasmid compositions of a series of clones derived from an infectious clone of B. burgdorferi B31MI and then testing the infectivities of these clones in C3H-HeJ mice, we demonstrate that loss of infectivity can occur in isolates that still retain lp25 and lp28-1. While lp25 and lp28-1 appear to be important for infectivity, the data presented here demonstrate that mechanisms in addition to the loss of these plasmids can also lead to the loss of infectivity.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Generation of postinfection isogeneic clones of B. burgdorferi B31MI. All analyses were conducted using the cloned isolate B. burgdorferi B31MI (kindly provided by Mark Hanson at MedImmune Inc., Gaithersburg, Md.), designated the parental clone (pc), and clonal populations derived from it. This isolate was selected for analysis because the sequence of its entire genome is now available (http://www.tigr.org) (15). Upon receipt of the isolate, its infectivity was confirmed using the C3H-HeJ murine model for Lyme disease as previously described (37). In brief, 10³ spirochetes were needle inoculated intradermally between the shoulder blades of three 6-week-old C3H-HeJ mice. After 4 weeks, 1-mm-diameter ear punch biopsy specimens were collected and placed in BSK-H complete medium (Sigma) containing antibiotics (phosphomycin, 20 µg ml¹; rifampin, 50 µg ml¹; amphotericin, 2.5 µg ml¹; Sigma). After the infection was confirmed, it was allowed to persist for a total of 3 months. At this point, ear punch biopsy specimens were again collected and placed in medium for cultivation as described above. The spirochetes were maintained in this medium until growth became evident as determined by dark-field microscopy and were then subsurface plated (37). Colonies were evident after approximately 2 weeks, and well-isolated colonies were selected for analysis. These populations are henceforth referred to as postinfection clonal populations.

PCR analyses. PCR analyses were performed using isolated DNA or DNA released from lysed cells as a template. When lysed cells were used as a template, 1 ml of culture (~2.0 × 10⁸ cells) was pelleted (14,500 × g; 5 min), and the cells were resuspended in phosphate-buffered saline (pH 7.4) and enumerated by dark-field microscopy. The cells were pelleted, resuspended in sterile water (5 × 10⁶ cells/µl), and boiled for 5 min; the cellular debris was removed by centrifugation, and the supernatant was transferred to a new tube for use as the template in PCR analyses.

Southern hybridization analyses. Isolated genomic DNA from each clonal population, obtained as previously described (18), was digested with HaeIII (standard conditions) and fractionated by electrophoresis in 0.8% agarose gels. The DNA was stained with ethidium bromide and transferred onto Hybond N membranes by vacuum blotting, using the VacuGene system (Pharmacia) and protocols provided by the manufacturer. The DNA was fixed to the membrane by UV irradiation using the Bio-Rad GS Gene Linker. Hybridization conditions, washes, and oligonucleotide probe labeling were as previously described (6). All oligonucleotide probes and primers are described in Table 1.

Analysis of the infectivity of B. burgdorferi clones in mice. The infectivities of B. burgdorferi B31MI clones were tested using the murine model for Lyme disease infection. The infective potential of B. burgdorferi B31MIpc has been previously established, and we have experienced a 100% success rate in the recovery of spirochetes from experimentally infected C3H-HeJ mice (n = 30) by cultivation of ear punch biopsy specimens (37). In the first round of experiments designed to test for infectivity, an inoculum of ~500 spirochetes was used. In the second round of experiments, three mice were inoculated with 10⁷ spirochetes each. Ear punch biopsy specimens were collected at 4 and 12 weeks postinoculation and placed in BSK-H complete medium containing antibiotics. After 12 weeks, the mice were sacrificed and the spleens, kidneys, and blood were collected and placed individually in BSK-H complete medium.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Analysis of the cp32 plasmid family in pre- and postinfection clonal populations derived from B. burgdorferi B31MIpc. We began our assessment of a possible correlation between specific plasmids and infectivity by determining the cp32 plasmid compositions of several clonal populations derived from B. burgdorferi B31MIpc. We initially focused on the cp32a, since these plasmids carry several gene families suggested to contribute to pathogenicity (21, 32, 36, 37, 42). C3H-HeJ mice were infected for 3 months, and then postinfection clonal populations were collected (n = 100). The composition of the cp32 plasmid family was determined for several clones by using a hybridization approach. The preinfection clonal population was confirmed to harbor seven cp32s (Table 2). All oligonucleotide probes used in this report were designed based on the B. burgdorferi B31MI genome sequence (9). The uhb(+) oligonucleotide targets a segment of the cp32s that is referred to as the upstream homology box (UHB) element (19). This conserved element resides upstream of three different lipoprotein-encoding gene families (ospE, ospF, and family 163) that are carried by the cp32s (and lp56) (2, 4, 15, 19, 35, 38). Note that the designation "family 163" was devised by The Institute for Genomic Research (15). Using the known cp32 sequences for B. burgdorferi B31MI, the sizes of HaeIII restriction fragments derived from each plasmid that carry target sites for each probe could be deduced (Table 3), thereby allowing the determination of which cp32s were present or absent. All clones were found to carry several uhb(+)-hybridizing fragments, with 7 of 20 clones exhibiting hybridization profiles identical to that of B. burgdorferi B31MIpc (Fig. 1 shows representative data). cp32-4 and cp32-3 were present in all clones, while others lacked cp32-6, cp32-9, and/or cp32-7. The loss of these plasmids was confirmed through additional hybridization analyses using oligonucleotides targeting other regions of the plasmids. For example, the loss of cp32-6 was confirmed using a probe targeting the BBM38 gene (Fig. 2). In all other cases, these additional analyses corroborated the absence of these plasmids from certain clones. In some additional cases, plasmid-specific probes were required to ascertain the presence or absence of specific plasmids. For example, both cp32-8 and cp32-1 are predicted to yield HaeIII restriction fragments of 2,048 bp that possess binding sites for the uhb(+) probe; hence, to differentiate between these plasmids, additional probes were required. A plasmid-specific probe was required for cp32-4 as well. The BBL, BBP, and ospF-R oligonucleotides are specific for plasmids cp32-8, cp32-1, and cp32-4, respectively. Hybridization analyses revealed that these plasmids were carried by all clones. The results of the hybridization analyses are summarized in Table 2.

View larger version (53K): [in this window] [in a new window]   FIG. 1.   Restriction fragment length polymorphism pattern analysis of the cp32 family of plasmids. DNA was isolated from each clone, digested with HaeIII, fractionated by agarose gel electrophoresis, transferred onto Hybond N membrane by vacuum blotting, and hybridized with the 32P-labeled uhb(+) oligonucleotide as described in the text. The clones analyzed are indicated above each lane, and the plasmid from which each hybridizing restriction fragment was derived is indicated on the right. The migration positions for the size standards are indicated between the panels. pc indicates DNA from the parental clone, B31MI.

View larger version (27K): [in this window] [in a new window]   FIG. 2.   Confirmation through Southern hybridization analyses of the loss of cp32-6 from clones. HaeIII-digested DNA from each clone (indicated above each lane) was fractionated and blotted as described in the legend to Fig. 1 and then hybridized with the BBM38- targeting cp32-6 probe. pc indicates DNA from the parental clone, B31MI.

Analysis of infectivities of clonal populations derived from B. burgdorferi B31MIpc that differ in their cp32 plasmid compositions. To determine if the observed variation in cp32 composition influences the abilities of clonal populations to establish infection in mice, clones lacking one or more cp32s were needle inoculated into C3H-HeJ mice. Initial experiments were performed using an inoculum of 10³ spirochetes for the parental clone and clones 17, 24, and 53. At 1 and 3 months postinoculation, ear punch biopsy specimens and blood were recovered and placed in complete BSK-H medium to allow cultivation of the spirochetes. Positive cultures were obtained for clones 53 and pc from both the 1- and 3-month time points, while all other cultures were negative. To allow for the possibility that the kinetics of growth might be slower for some clones, the cultures were allowed to persist for 2 months. However, even after this time only clones 53 and pc were culture positive. The experiment was repeated using three mice per clone, an inoculum of 10⁷ spirochetes, and a broader range of clones (clones pc, 9, 14, 24, 29, 36, 37, and 53). At 1 month postinoculation, positive cultures were obtained for clones pc, 29, 36, 37, and 53 but not clones 9, 14, 17, and 24. At 3 months postinoculation, the mice were sacrificed, and ear punch biopsy specimens, blood, spleens, and kidneys were collected. Clones pc 29, 36, 37, and 53 were culture positive, whereas clones 9, 14, 17, and 24 were culture negative for all tissues tested.

Complete analysis of the plasmid compositions of infectious and noninfectious clonal populations derived from B. burgdorferi B31 MI. To determine if there were additional differences in the linear- and non-cp32 circular-plasmid profiles in the infectious and noninfectious clonal isolates described above, a PCR strategy was employed to test for each plasmid. B. burgdorferi B31MIpc and B. burgdorferi MSK5 served as positive controls to verify that the primers were specific. Amplification was scored as follows: , no amplification; +/, weak amplification; and +, strong amplification. Weak amplification may indicate that certain plasmids have been lost from the majority of the population. Upon PCR, all primer sets yielded amplicons of the expected sizes. Representative PCR results are shown in Fig. 3, and the data are summarized in Table 2. Several clones were found to lack one or more different plasmids. The plasmids found to be lost from one or more clones included lp5, lp28-2, lp28-3, cp9, cp32-6, or cp32-9. In addition, as previously reported, we did not detect plasmids cp32-2 and cp32-5 in B. burgdorferi B31 MIpc or (not surprisingly) in any of its clonal derivatives. Several clones were found to have lost as many as four different plasmids relative to the parental population.

View larger version (62K): [in this window] [in a new window]   FIG. 3.   Determination of the plasmid compositions of B. burgdorferi B31MI clones by using a PCR approach. Isolated genomic DNA served as a template for PCR analyses using plasmid-specific primer sets. Amplification products were fractionated on a 1% agarose gel and stained with ethidium bromide. The gel was then scanned to generate the figure. All methods were as described in the text. On the left, the B. burgdorferi B31MI clones analyzed and the migration positions of the size standards are indicated. The plasmid tested for is indicated above each lane.

View larger version (20K): [in this window] [in a new window]   FIG. 4.   Southern hybridization analysis of lp28-3. HaeIII-digested DNA from each clone (indicated above each lane) was fractionated and blotted as described in the legend to Fig. 1. The blot was hybridized with the lp28-3 targeting probe, BBH08. Molecular standards are indicated on the left. pc, indicates DNA from the parental clone, B31MI.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Several studies have demonstrated that loss of infectivity is associated with plasmid loss in the Lyme disease spirochetes (26, 41) and that this is the primary mechanism associated with the phenomenon. One of the first studies to suggest that such a correlation exists was conducted by Schwan and colleagues (26), where the loss of cp7.6 and lp22 was concomitant with the loss of infectivity. It is unclear if B. burgdorferi B31MI carries homologs of these plasmids, since the sequences for these plasmids were not determined. It is likely that cp7.6 is analogous to cp9.0 (13, 15). Analysis of cp9.0 revealed that it does not appear to encode proteins with homology to proteins known to participate in virulence. In this report, we demonstrate that several clones that have lost this plasmid remain infective. A similar observation has recently been reported by Purser and Norris and by Labandeira-Rey and Skare (16, 22). However, the absence of a strict requirement for cp9.0 should not be interpreted to suggest that cp9.0-encoded proteins are not potentially important in Borrelia pathogenesis. Most of the open reading frames present on cp9.0 are actually members of plasmid-carried paralogous gene families (5, 15), so paralogs encoded by other plasmids might complement lost functions. The identification of a homolog of lp22, a second plasmid linked to virulence by Schwan et al. (26), in B. burgdorferi B31MI is complicated by the presence of numerous linear plasmids in this isolate, ranging from 17 to 30 kb. Other linear plasmids in this size range have also been implicated as necessary for virulence. The loss of lp28.7 coincides with an increase in the 50% infective dose (ID₅₀), (40, 41). This plasmid is likely a homolog of one of the four lp28s present in the B. burgdorferi B31MI genome (8). In general, the complete plasmid contents of the isolates analyzed in these earlier studies were largely unknown, making it difficult to interpret the data and to compare it with the data obtained from other studies that used different isolates. Hence, while a correlation between plasmids and infectivity appeared to exist, the specific plasmids associated with the phenomenon remained elusive.

In this study, we have focused our analyses on clonal populations derived from B. burgdorferi B31MI. The genome sequence for this isolate has recently been determined, allowing these studies to be conducted with an isolate of known genetic composition (15). Other recent studies of this topic have also exploited the genome sequence in their analyses of plasmid composition and infectivity (16, 22). The studies described here began with a particular interest in the possible role of the cp32 plasmid family in infectivity and virulence. Analyses conducted by our laboratory and others have demonstrated that several of the genes carried by these plasmids are differentially expressed during infection (1, 12, 36, 39; J. V. McDowell, S.-Y. Sung, G. Price, and R. T. Marconi, submitted for publication) and that some may play a role in immune evasion (37). An initial assessment of cp32 compositions in different B31MI-derived clones led to the identification of several with differing cp32 restriction fragment length polymorphism patterns. To determine if these variants were infective, we employed the C3H-HeJ mouse model. Initial analyses utilized low inoculating doses of ~500 spirochetes (intradermal inoculation), and infectivity was assessed by the ability to cultivate the bacteria from ear punch biopsy specimens. We found that several of the clones were not infective. One possibility was that these particular clones might simply have altered ID₅₀s. To maximize the potential establishment of infection, we used a high inoculating dose of ~10⁷ spirochetes. As before, the same results were obtained for each clone. In addition, cultures of other tissues and organs from the mice that were culture negative with the ear punch biopsy specimens were also culture negative. Hence, the phenomenon that we observed was that the clones were either infectious or noninfectious. The use of two different inoculum sizes (500 and 10⁷ spirochetes) that are at opposite ends of the scale but yielded the same results demonstrates that the molecular events that occurred in these clones did not simply alter the ID₅₀. It is important to note that the 10⁷-spirochete inoculating dose had no discernible negative impact on any of the inoculated mice. In assessing these data, a clear-cut correlation between the loss of any one specific cp32 and the loss of infectivity was not discernible. Recent studies conducted by others have reached similar conclusions (16, 22). From this, we conclude that a complete set of all seven cp32s is not required for infectivity, maintenance of infection, or in vitro cultivation. The apparent lack of a requirement for a complete set of cp32s is consistent with the extensive variation in composition that has been demonstrated for this plasmid family in a wide range of isolates (19, 38). However, in view of the extensive genetic redundancy among the cp32s, it is perhaps not surprising that the loss of a subset of these genetic elements would have miminal impact on infectivity, virulence, or the ability to grow in vitro, since the proteins encoded by the multiple cp32s may complement each other. It should be noted that lp56 also contains an integrated cp32 and could serve to complement functions lost through the loss of cp32s (8). Interestingly, all Lyme disease isolates analyzed, as well as relapsing fever isolates, carry at least some cp32s (6, 7, 19, 23, 24, 34, 38). This observation suggests that maintenance of at least a subset of these plasmids is a necessary biological feature at the genus-wide level. It is important to note that the clones analyzed in this report were only briefly cultivated postinfection, suggesting that the changes in cp32 plasmids occurred during infection. Alternatively, the plasmids could have been lost during the postinfection cultivation period. Several observations argue against this. We have found that the cp32s of Borrelia garinii Pbi were retained after 311 in vitro passages (~4 years of continuous cultivation) (R. T. Marconi, unpublished data). Furthermore, to our knowledge, the loss of cp32s specifically during in vitro cultivation has not been reported. Hence, the cp32s may actually be less stable during infection than they are during in vitro cultivation. Studies by Eggers and Samuels suggest that the cp32s may actually encode prophage (14). It is possible that the mammalian environment stimulates bacteriophage transduction or plasmid transfer during infection, leading to changes in cp32 composition.

To determine if there might be a correlation between the loss of other plasmids and infectivity, we determined the complete plasmid compositions of several clones that were found in the analyses above to be either infectious or noninfectious. lp28-3 was the only plasmid absent from all noninfectious clones. lp28 is one of four similarly sized linear plasmids carried by B. burgdorferi B31MI (15). The specific absence of this plasmid was confirmed by both PCR and Southern hybridization. While we initially interpreted these data to suggest a correlation between Ip28-3 and infectivity, additional analyses of other infectious clones revealed that some of them also lacked lp28-3. Analysis of clones lacking this plasmid revealed that they were infective even when an inoculum of ~500 spirochetes was used. Hence, it is evident that lp28-3 is not strictly required for infectivity and that its loss does not affect the ID₅₀. Two of the lp28-3-minus infectious clones also lacked lp28-2. In a previous analysis, it was demonstrated that the loss of an ~28-kb linear plasmid correlated with an increase in ID₅₀ from 10³ to 10⁵ in the hamster model (41). However, in that study it was not specifically determined which of the four lp28s was lost. In an analysis of linear plasmids, Palmer et al. demonstrated that lp28-1, lp28-3, lp28-4, lp36, and lp54 were carried by all 15 Lyme disease spirochete isolates analyzed. They interpreted this to mean that these plasmids are biologically necessary (20). However, the data presented here demonstrate that the loss of lp28-3 and lp28-2 does not abrogate the ability of B31MI clones to establish and maintain infection. Purser and Norris also reported that lp28-2 (and lp28-4) is not required for infectivity (22). In that analysis, the authors did not test clones lacking lp28-3, since all of the clones in their collection carried the plasmid. It is possible that the decrease in ID₅₀ reported by Xu et al. resulted from the loss of lp28-1 (41).

Evidence has recently been presented which demonstrates that the loss of lp25 and lp28-1 is coincident with a low-infectivity phenotype (16, 22). However, none of the noninfectious clones analyzed as part of our study had lost either of these plasmids. We do not believe this to represent a discrepancy between these reports and the data presented here. Specific plasmids may very well play critical roles in infectivity or virulence, and the data presented by others suggest that lp25 and lp28-1 are in this category. What is striking is that in the three recent reports that have assessed the possible link between specific plasmids and infectivity, the specific plasmids lost by the clonal populations differed widely. We found that most of the clones analyzed in this report had lost lp28-3, while this plasmid was universally present in the clones analyzed by Purser and Norris and Labandeira-Rey and Skare. A second notable difference was in the cp32s. Purser and Norris found that all of the cp32s, with the exception of cp32-3, which was lost from only a single clone (B. burgdorferi 5A3), were retained (note that Labandeira and Skare did not determine the complete composition of the cp32 plasmid family in all clones analyzed). In contrast, we found that six of the eight clones derived from B. burgdorferi B31MI pc had lost one or more cp32s. The data presented here suggest that undefined mechanisms independent of the loss of a specific plasmid can also lead to the loss of infectivity. In addition, it is also evident from the variation noted in this and other recent studies that the dynamics of plasmid loss can vary widely under different experimental conditions (16, 22).

The central point of the analyses presented here is that the loss of infectivity does not in all cases correlate with the loss of a specific plasmid. However, this is not to say that such a correlation does not exist. It remains possible that there is an essential plasmid or combination of plasmids that are required for infectivity and virulence. In view of the genetic redundancy among the plasmids and thus the potential for complementation of function, it would seem most likely that this putative essential plasmid would be one of those that carries genes not harbored by other elements of the genome so that complementation could not occur. Perhaps a possible biological role for the redundancy in the Lyme disease spirochete genome is to shield the bacteria from the potentially adverse consequences associated with the inherent instability of its plasmid-based genome. Studies by others suggest that lp25 and lp28-1 are possible essential plasmids (16, 22). Nonetheless, even though certain plasmids may prove essential, the data presented here demonstrate that other mechanisms independent of plasmid loss appear to also be at play. These may include small-scale mutations or rearrangements in essential genes and/or differences in gene expression patterns among clonal populations. The analyses presented here provide further insight into the fluid nature of the Borrelia genome and the mechanisms associated with loss of infectivity. Furthermore, this study illustrates the complexity of assessing the contribution of the plasmid component of the Borrelia genome to the biology and pathogenesis of the Lyme disease spirochetes.