OspC Phylogenetic Analyses Support the Feasibility of a Broadly Protective Polyvalent Chimeric Lyme Disease Vaccine

Abstract

Using available Borrelia outer surface protein C (OspC) sequences, a phylogenetic analysis was undertaken to delineate the number of antigenic domains required for inclusion in a broadly protective, chimeric, OspC-based Lyme disease vaccine. The data indicate that approximately 34 would be required and that an OspC-based vaccinogen is feasible.

Lyme disease is the leading arthropod-borne disease in North America and Europe. It is caused by the spirochete species Borrelia burgdorferi (North America and Europe), Borrelia garinii (Europe), and Borrelia afzelii (Europe). A Lyme disease vaccine is not currently available for use in humans. The outer surface protein A (OspA)-based LymeRix vaccine, introduced in 1998, was pulled from the market in 2002 due to poor sales, possibly traced to concerns over vaccine-induced arthritis (21). There is a pressing need for the development of an effective, safe, and broadly protective Lyme disease vaccine (29, 40).

The expression profile (38), immunodominance (11, 42), and critical functional role of OspC during transmission from the tick (34) and during mammalian infection (18, 30, 43-45) have made it an attractive candidate vaccinogen (12). OspC is an ∼22-kDa protein encoded on a 26-kb circular plasmid that is apparently essential, since it is universal among Lyme disease spirochete isolates (9, 26). While ospC exhibits significant sequence diversity, it is genetically stable during infection (20, 41). Phylogenetic analyses conducted to date have delineated phyletic clusters that have been termed OspC types, which are designated by a letter (23, 39, 48). OspC sequence identity is typically greater than 97% within a type but less than 80% between types. It is evident that OspC types have not arisen as a result of geographic isolation, since within tight geographic regions, strains have been identified that express different OspC types (1, 2, 11, 33, 48). The multiplicity of OspC types appears to be maintained by balancing selection, where maintenance of multiple alleles in the population is more advantageous than the directional selection of a single, most fit allele (5, 48). The selective force is not known with certainty but may be related to immune evasion or to an association between OspC type and infectivity in various mammalian hosts (4, 5).

Although vaccination of mice with OspC protein has been demonstrated to be protective, the protection is strain specific (6, 16, 17, 32, 37, 46). This observation suggests that the protective epitopes likely reside within OspC type-specific domains. Consistent with this, we previously demonstrated that the early murine humoral response is type specific and that the immunodominant epitopes of OspC that are presented during early infection in mice and humans are located within hypervariable domains presented at the surface of the OspC dimer (7, 11, 15, 22). By necessity, a broadly protective Lyme disease vaccine must be polyvalent. It has been reported that a polyvalent vaccine formulation consisting of full-length OspC proteins of 14 types has been tested; however, the formulation was too reactogenic and a balance in the antibody response to each OspC type was not obtained (19). The goal of our efforts is to develop a chimeric polyvalent vaccine construct that consists of the type-specific, protective epitopes of those OspC types associated with invasive infection in humans. As proof of principle, we previously developed a construct that incorporates epitopes from OspC types A, B, K, and D (11-13). This construct elicited bactericidal antibody against spirochetes expressing each of the OspC types incorporated into the vaccinogen in a complement-dependent manner (12).

To facilitate the further development of a broadly protective chimeric construct, we have conducted phylogenetic analyses of OspC sequences available in the NCBI databases. The segment of OspC analyzed spanned residues 20 through 200 (using numbering for the B31MI sequence). Shorter sequences were excluded from these analyses, leaving sequences from 280 Borrelia strains available for analysis. The OspC type designation of each sequence was determined through alignment (PAM40 scoring matrix) and pairwise identity matrix analysis. Consistent with earlier studies, sequences that exhibited 95% or greater sequence identity were considered to belong to the same OspC type (3, 48) (Fig. 1). A clear bimodal distribution of sequence comparisons, with a mean sequence identity of 65% between differing OspC type sequences and >97% identity within types was observed. In addition to the 21 types described by Wang et al. (48), 17 additional clusters were defined. We did not assign OspC type designations to clusters that included fewer than 3 sequences, nor did we include in our analyses previously defined OspC types with fewer than three sequences (i.e., types C, G, J, and O to T). OspC types not previously assigned a letter designation were named based on a prototype strain contained within the cluster.

FIG. 1.

Distribution of pairwise comparisons of OspC protein sequence identity. OspC protein sequences from 280 Borrelia isolates were aligned with Clustal and scored with a PAM40 scoring matrix, and pairwise percent identities were calculated. The histogram interval is 1%, and no comparisons were <50% identical.

Of 280 analyzed sequences, 202 were assigned to OspC types, all of which were from Lyme disease-causing species. The remaining 78 sequences were not assigned to an OspC type, as they did not occur in a cluster of three or more sequences of >95% identity. Twenty-seven of the unassigned sequences were from Borrelia species not typically associated with Lyme disease, including B. bissettii, B. japonica, B. andersonii, B. tanukii, and B. valaisiana. A summary of the geographic and biological origin of the isolate from which each OspC sequence was obtained is indicated in Table 1. In an analysis of type-assigned sequences, the majority of B. burgdorferi isolates were from North America (80%), with lesser numbers from Europe (16%) and Asia (4%). Fifty-three percent of B. burgdorferi, 48% of B. afzelii, and 79% of B. garinii OspC sequences originated from isolates collected from humans. It is noteworthy that the B. garinii human isolates were predominantly of cerebrospinal fluid (CSF) origin (68%), whereas B. afzelii isolates were predominantly from the skin (83%). In contrast, B. burgdorferi isolates were recovered from human skin (51%), plasma (30%), and CSF (19%). These findings are in agreement with the known patterns of disease caused by these organisms and indicate that the sample of OspC sequences assessed in this report is representative of the true population of Lyme disease spirochetes.

TABLE 1.

Species, geographical, and biological isolation data for assigned OspC types

OspC type	No. of isolates from:
	Borrelia species				Geographic region					Biological source
	B. burgdorferi	B. garinii	B. afzelii	Other species	North America	Europe	Russia	Asia	Not provided	Human skin	Human blood	Human CSF	Other animal	Tick	Not provided
Assigned
A (B31)a	38				25	9		4		7	9	6	6	7	3
B (LDP73)	13				6	5			2	5	1	1		4	2
D (LDP116)	3				3						1			2
E (N40)	7				7					3				4
F (PAd)	4				4					1				3
H (LDS101)	3				3					2				1
I (HB19)	7				7					3	2	1		1
K (LDP74)	11				11					3	2	1		5
L (SI1)	6				5	1							5	1
M (B356)	3				3					1			2
N (LDP63)	7				7					1	2	1	1	2
U (148)	3				3					1				2
OspCt-72a	4				4					1			1	2
OspCt-Szid	3					3				1		1		1
OspCt-Smar		4				4						1		3
OspCt-PLi		9				9				5	1	3
OspCt-H13		3				3				1		1		1
OspCt-PFiM		6				6				1		3		2
OspCt-PMit		3				3				1		1		1
OspCt-PKi		5				5				3		2
OspCt-PBes		7				6		1		3		3		1
OspCt-HT22		4					3	1						4
OspCt-PHez		5				5						4		1
OspCt-PWa		17				17				1		16
OspCt-Pko			11			7		4		5		1	2	3
OspCt-PLj7			6			5			1	3	1			1	1
OspCt-VS461			3			2			1	1				1	1
OspCt-DK15			3			1		2		1				2
OspCt-HT25			4				1	3						4
Unassigned
B. burgdorferi	8				6	1	1			1	1		1	5
B. garinii		22				7	4	10	1	4		2		15	1
B. afzelii			21			10	9	1	1	7			3	10	1
B. bissettii				8	8								6	2
B. japonica				1				1							1
B. andersonii				6	6								1	5
B. tanukii				1				1					1
B. valaisiana				8		4		4					3	5
Borrelia sp.				3	3								1	2
Total	120	85	48	27	111	113	18	32	6	66	20	48	33	103	10

^aFor letter-assigned OspC types, an example strain is given in parentheses.

To facilitate further phylogenetic analyses, the set of sequences analyzed was reduced to 74 by eliminating identical sequences. These sequences were then aligned and analyzed using the Phylip (v. 3.66) phylogenetics package with bootstrapping (n = 1,000). Distances were calculated for the regions spanning amino acids 20 to 200, 20 to 130, and 131 to 200 using the Dayhoff PAM matrix, and trees were created by neighbor joining. The B. hermsii OspC ortholog (Vtp) sequence served as an out-group (27, 31). A consensus tree was generated by majority rule (50% cutoff for group inclusion). Distances were recalculated for the consensus tree by the maximum likelihood method under the Dayhoff PAM model (Fig. 2). The consensus trees generated with the amino acids 20 to 200 segment of OspC were well supported at the terminal nodes, with all determined OspC types clustering as expected. While several of the deeper branches were less supported by the bootstrap analyses (Fig. 2A), this is not unexpected since the extended regions of identity among the sequences makes their phylogenetic differentiation subtle. Consensus trees generated using the amino acids 20 to 200 and 20 to 130 segments of OspC exhibited similar phylogenetic clustering (Fig. 2A and B) with clustering patterns consistent with species identity. However, the consensus tree generated using amino acids 131 to 200 (Fig. 2C) yielded significantly different clustering patterns that were not strongly supported by bootstrap analyses. This observation is consistent with the hypothesis that recombination between short segments of the ospC gene has occurred between strains of differing OspC types. Evidence for recombination of short segments of ospC between OspC types can be seen in specific sequences. For example, sequences of the B. afzelii OspC type, PLj7, have regions within the amino acid 20 to 130 domain that are identical to that seen in B. garinii OspC sequences that form the Pki cluster. In the amino acids 131 to 200 region of PLj7, the hypervariable loop 5 and loop 6 regions have motifs identical to those seen in B. burgdorferi OspC types F and M, respectively. Further evidence for recombination was noted during a broader assessment of regional similarities in DNA sequence by bootscanning (SimPlot v. 3.5.1) (25). In bootscanning, potential recombination is assessed by generation of phylogenetic trees (Kimura model, Ts/Tv ratio = 2.0, neighbor joining) using sequence segments contained within a sliding window (40-base window, 10-base step interval). The trees are bootstrapped (n = 100), and the number of permuted trees supporting sequence grouping within that window is reported. Evidence of recombination is typically considered to be supported when >70% of permuted trees support a particular grouping of sequences. In the case presented in Fig. 3, the PLj7 sequence was scanned for possible areas of recombination with OspC types Pki, F, and M, as described above. Peaks representing possible recombination between types PLj7 and Pki occur in regions predicted to encode the unstructured N terminus, alpha helices 1 and 4, and loop 4. There was evidence of possible recombination with type M at alpha helix 3 and at the junction of loop 6 and alpha helix 5, and with type F at alpha helices 3 and 5. Possible recombination was also found between numerous other OspC types (data not shown). The mechanisms associated for genetic transfer in the Borrelia, particularly involving the ospC-bearing cp26 plasmid, have not been clearly delineated (3, 8, 10, 14, 24, 28, 35, 36, 47).

FIG. 2.

Consensus phylogenetic trees of representative OspC protein sequences. OspC sequences spanning amino acids 20 to 200 (A), 20 to 130 (B), and 131 to 200 (C) were bootstrapped (n = 1,000), distances calculated, and neighbor joining trees created and reconciled to a consensus tree. The Vtp sequence of B. hermsii served as an out-group. Labels indicate the species as B. burgdorferi (Bb), B. garinii (Bg), or B. afzelii (Ba), the isolate strain designation, the assigned OspC type (boldface type), and the number of identical OspC sequences from other strains represented by this single sequence (in parentheses). Bootstrap support is shown at all nodes that differentiate between OspC types.

FIG. 3.

Bootscan analysis of the OspC type PLj7 (B. afzelii) ospC sequence. The PLj7 sequence was analyzed by comparison with OspC types Pki (B. garinii; solid black line), F (B. burgdorferi; solid gray line), and M (B. burgdorferi; dashed black line). The bootscan window was 40 bases, with a 10-base step. Comparison was by strict consensus with 100 bootstrap replicates. Nucleotide numbering (bottom) begins at the signal sequence, and the bootscan included nucleotides 58 to 600. The graph shows only those peaks where greater than 50% of the permuted trees support a sequence grouping within the window. The 70% level considered to represent possible recombination is indicated by a dashed line. For reference, the secondary structure of the encoded protein (based on the B31MI crystal structure [22]) is denoted at the top as alpha helices (numbered in white boxes), loops (numbered in gray boxes), and small beta strands (black boxes).

The evidence that OspC variability occurs by exchange between existing OspC types rather than by hypermutation provides evidence that there is a limit to the number of OspC type-specific epitopes required for inclusion in a broadly protective vaccinogen. Since currently mapped linear epitopes are all contained in the C-terminal region of OspC (amino acids 131 to 200), it is possible to define a theoretical number of antigenic domains required for a chimeric vaccinogen. By inspecting this region in the 74 representative sequences described above, the number of unique epitope-containing regions can be reduced to 34 by elimination of sequences that are either identical or have only a single amino acid change (Fig. 4). It is likely that this number can be further restricted by epitope mapping since some epitopes may convey protection against two or more OspC types. Further reduction in the required number of antigenic domains could also come from consideration of only those OspC types associated with human disease or, more specifically, with invasive human disease (1, 11, 39). One theoretical concern with vaccination against a subset of OspC epitopes is the potential to drive selection toward types not included in the vaccinogen, thus increasing the fraction of the population bearing those rare types. However, as humans are incidental hosts, it is unlikely that vaccination will significantly alter the population distribution of strains expressing specific OspC types in the tick vector or mammalian reservoirs.

FIG. 4.

Alignment of the epitope-containing region of OspC protein sequences from all OspC types defined in this study. All sequences within OspC types that differ by more than one amino acid are indicated by a representative sequence. The identity threshold for shading is 80%. Secondary structural alpha helices and loops (corresponding to the B31 structure) are shown below the alignment (22).

In summary, the extensive nature of the OspC database has allowed thorough analyses to be conducted which have defined new OspC types and provided information regarding their frequency of isolation and association with human disease. The data suggest that the number of OspC epitope-containing sequences required for inclusion in a broadly protective chimeric vaccinogen is limited and that the development of a chimeric vaccinogen is feasible.